Coastal Tectonics
Geological Society Special Publications Series Editors
A. J. FLEET R. E. HOLDSWORTH A. C. MORTON M. S. STOKER
It is recommended that reference to all or part of this book should be made in one of the following ways. STEWART, I. S. & VITA-FINZI, C. (eds) 1998. Coastal Tectonics. Geological Society, London, Special Publications, 146. CHAPPELL, J., OTA, Y. & CAMPBELL, C. 1998. Decoupling post-glacial tectonism and eustasy at Huon Peninsula, Papua New Guinea. In: STEWART, I. S. & V~TA-FINZI, C. (eds) 1998. Coastal Tectonics. Geological Society, London, Special Publications, 146, 31-40.
GEOLOGICAL SOCIETY SPECIAL PUBLICATION NO. 146
Coastal Tectonics
EDITED BY
I A I N S. S T E W A R T Brunel University, UK AND
CLAUDIO VITA-FINZI University College London, UK
1998 Published by The Geological Society London
THE GEOLOGICAL SOCIETY The Society was founded in 1807 as The Geological Society of London and is the oldest geological society in the world. It received its Royal Charter in 1825 for the purpose of 'investigating the mineral structure of the Earth'. The Society is Britain's national society for geology with a membership of around 8500. It has countrywide coverage and approximately 1500 members reside overseas. The Society is responsible for all aspects of the geological sciences including professional matters. The Society has its own publishing house, which produces the Society's international journals, books and maps, and which acts as the European distributor for publications of the American Association of Petroleum Geologists, SEPM and the Geological Society of America. Fellowship is open to those holding a recognized honours degree in geology or cognate subject and who have at least two years' relevant postgraduate experience, or who have not less than six years' relevant experience in geology or a cognate subject. A Fellow who has not less than five years' relevant postgraduate experience in the practice of geology may apply for validation and, subject to approval, may be able to use the designatory letters C Geol (Chartered Geologist). Further information about the Society is available from the Membership Manager, The Geological Society, Burlington House, Piccadilly, London WlV 0JU, UK. The Society is a Registered Charity, No. 210161. Published by The Geological Society from: The Geological Society Publishing House Unit 7, BrassmiU Enterprise Centre Brassmill Lane Bath BA1 3JN UK (Orders: Tel. 01225 445046 Fax 01225 442836)
Distributors
USA AAPG Bookstore PO Box 979 Tulsa OK 74101-0979 USA (Orders: Tel. (918) 584-2555 Fax (918) 560-2652)
First published 1998
Australia The publishers make no representation, express or implied, with regard to the accuracy of the information contained in this book and cannot accept any legal responsibility for any errors or omissions that may be made. 9 The Geological Society 1998. All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with the provisions of the Copyright Licensing Agency, 90 Tottenham Court Road, London WlP 9HE. Users registered with the Copyright Clearance Center, 27 Congress Street, Salem, MA 01970, USA: the item-fee code for this publication is 0305-8719/98/$10.00.
Australian Mineral Foundation 63 Conyngham Street Glenside South Australia 5065 Australia (Orders: Tel. (08) 379-0444 Fax (08) 379-4634)
India Affiliated East-West Press PVT Ltd G- 1/ 16 Ansari Road New Delhi 110 002 India (Orders: Tel. (11) 327-9113 Fax (11) 326-0538)
Japan British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library. ISBN 1-86239-024-X
Typeset by Aarontype Ltd, Unit 47, Easton Business Centre, Felix Road, Bristol BS5 0HE, UK Printed by The Alden Press, Osney Mead, Oxford, UK
Kanda Book Trading Co. Cityhouse Tama 204 Tsurumaki 1-3-10 Tama-Shi Tokyo 0206-0034 Japan (Orders: Tel. (0423) 57-7650 Fax (0423) 57-7651)
Contents
Preface PELTIER, W. R. Global glacial adjustment and coastal tectonics CHAPPELL,J., OTA, Y. • CAMPBELL,C. Decoupling post-glacial tectonism and eustasy at Huon Peninsula, Papua New Guinea SOTER, S. Holocene uplift and subsidence of the Helike Delta, Gulf of Corinth, Greece TRECKER, M. A., GURROLA, L. D. & KELLER, E. A. Oxygen isotope correlation of marine terraces and uplift of the Mesa Hills, Santa Barbara, California, USA BORDONI, P. & VALENSISE, G. Deformation of the 125 ka marine terrace in Italy: tectonic implications CORNET, Y. 8z DEMOULIN, A. Neotectonic implications of a lineament-coplanarity analysis in Southern Calabria, Italy FLEMMING, N. C. Archaeological evidence for vertical tectonic movement on the continental shelf during the Palaeolithic, Neolithic and Bronze Age periods GALILI, E. & SHARVIT,J. Ancient coastal installations and the tectonic stability of the Israeli coast in historical times FOULGER, G. R. & HOFTON, M. A. Regional vertical motion in Iceland 1987-1992, determined using GPS surveying ORME, A. R. Late Quaternary tectonism along the Pacific coast of the Californias: a contrast in style THACKRAY, G. D. Convergent-margin deformation of Pleistocene strata on the Olympic Coast of Washington, USA MERRITTS, D., EBY, R., HARRIS, R., EDWARDS, R. L. & CHENG, H. Variable rates of Late Quaternary surface uplift along the Banda Arc-Australian plate collision zone, eastern Indonesia REYSS, J. L., PIRAZZOLI, P. A., HAGHIPOUR, A., HATTIe, C. & FONTUGNE, M. Quaternary marine terraces and tectonic uplift rates on the south coast of Iran MEGHRAOUI, M.; OUTTANI, F., CHOUKRI, A. & FRIZONDE LAMOTTE, O. Coastal tectonics across the South Atlas Thrust Front and the Agadir Active Zone, Morocco MURRAY-WALLACE, C. V., BELPERIO, A. P. & CANN, J. H. Quaternary neotectonism and intra-plate volcanism: the Coorong to Mount Gambier Coastal Plain, southeastern Australia: a review NUNN, P. Late Cenozoic emergence of the islands of the northern Lau-Colville Ridge, southwest Pacific BEZERRA, F. H. R., LIMA-FILHO, F. P., AMARAL, R. F., CALDAS, L. H. O. & COSTA-NETO, L. X. Holocene coastal tectonics in NE Brazil BILHAM, R. Slip parameters for the Rann of Kachchh, India, I6 June 1819, earthquake, quantified from contemporary accounts MCNEILL, L. C., GOLDFINGER, C., YEATS, R. S. & KULM, L. D. The effects of upper plate deformation on records of prehistoric Cascadia subduction zone earthquakes DOMINEY-HOWES, D., DAWSON, A. & SMITH, D. Late Holocene coastal tectonics at Falasama, western Crete, (Greece): a sedimentary study GOFF, J. R., CROZIER, M., SUTHERLAND, V., COCHRAN, U. & SHANE, P. Possible tsunami deposits from the 1855 earthquake, North Island, New Zealand Index
vii 1 31 41 57 71 111 129 147 165 179 199 213
225 239 255
269 279 295 319 341 351 373
Preface At first glance, coastal tectonics is as redundant a category as inland tectonics, for the shoreline does not necessarily coincide with a distinctive geodynamic environment. What prompted the international conference on the subject that led to this book was the Editors' conviction that coasts favour the study of active tectonics (a) by providing a reference d a t u m - namely sea l e v e l - against which deformation can be measured and (b) by supplying datable material and environmental clues with which the progress of deformation can be traced. As a bonus we have coasts which temporarily coincide with a tectonic boundary or major structure and wash it clean for our inspection. Consider plate boundaries such as those of the western Americas where subduction and transform displacement are now operating, or the extensional coasts of the Gulf of Corinth where normal faulting will perpetuate tectonic conditions on the coast for some time to come. There are also countless locations, notably oceanic islands, which are tectonic at one remove, as their uplift or subsidence reflects the dynamic behaviour of the lithosphere elsewhere. There was a further question to be resolved. The original plan had been to focus on Late Quaternary coastal tectonics, but this soon emerged as unnecessarily restricting: why 'late', and why Quaternary, when many active coasts began to deform in the Tertiary or even earlier, and when much illuminating work depends on the evidence of seismology and geodesy? We have an excellent precedent for our title: that of the survey by Ken Lajoie (1986) that did much to define the scope and procedures of tectonic investigations on coasts. Lajoie opened his discussion by observing that between one third and one half of the Earth's marine coastlines lie along or near tectonically active plate boundaries. By implication he was emphasizing mechanism rather than narrative, and that was our intention when we organized a conference around the application of high-resolution coastal chronologies to the testing and refining of crustal models at local, regional and global scales. The papers that follow (which include seven that were solicited after the meeting) have accordingly been grouped into sections which deal in turn with the extraction of tectonic data from the many kinds of noise in the coastal record and with their bearing on the analysis of interplate and intraplate tectonics and the construction of earthquake sequences. Lack of space meant that some themes, such as salt tectonics, receive little mention; conversely, areas which have attracted investigation from various viewpoints, notably coastal California, are discussed in more than one paper. Of the many possible dating methods, the emphasis is on radiocarbon and U-series techniques, but one of the papers reviews the potential value of stable isotopes in the correlation of marine terraces, two are primarily concerned with archaeological indicators of tectonic displacement, and another exploits historical records which have long lain unread in the archives; Strombus bubonius emerges reinvigorated in its new tectonic role. Some persistent geodynamic problems are at best highlighted by the work reported in this book, notably the distinction between seismic and aseismic contributions to net tectonic strain; the distinction between stable and unstable coastlines, however, emerges as unhelpful.
Vlll
We thank the following, as well as a few others who wished to remain anonymous, for advice and help with the meeting and with reviewing the manuscripts: F. A. Aberg, N. N. Ambraseys, R. Armijo, K. Berryman, M. Berberian, A. L. Bloom, D. Q. Bowen, J. Coleman, P. E. F. Collins, A. B. Cundy, A. Dawson, M. Evron, R. W. Fairbridge, N. C. Flemming, G. R. Foulger, A. Hull, M. Ivanovich, H. Kelsey, A. J. Long, J. M. McArthur, M. Meghraoui, K. Morris, W. Murphy, D. Neev, A. R. Nelson, P. D. Nunn, J.-L. Ortlieb, Y. Ota, C. Pain, P . A. Pirazzoli, J. P. Platt, G. Roberts, F. Sigmundsson, P. Stewart, F. W. Taylor, A. B. Watts, M. Weinstein-Evron and C. Zazo. We are grateful to A. Hills for editorial assistance. Iain Stewart & Claudio Vita-Finzi
Global glacial isostatic adjustment and coastal tectonics W . R. P E L T I E R
Department of Physics, University of Toronto, Toronto, Ont., Canada M5S 1A7 (e-mail:
[email protected]) Abstract: A global and gravitationally self-consistent model of the process of glacial isostatic adjustment (GIA) has been developed that extremely well reconciles the vast majority of available records of Holocene relative sea-level history, not only from sites that were ice covered at last glacial maximum (LGM) but also from sites that are well removed from such locations. There do exist, however, data that have been construed to constitute a significant challenge to this theory, namely, the long records of relative sea-level history derived on the basis of U/Th-dated coral sequences from the Huon Peninsula of Papua New Guinea and from Tahiti in the central equatorial Pacific Ocean. Following a review of the theoretical model and a discussion of the extent to which it is able to successfully reconcile a very wide range of Holocene shoreline observations, the discussion focuses upon the interpretation of these very important and interesting records, which are subject to different levels and types of tectonic contamination. These analyses suggest that existing estimates of the levels of Holocene tectonic contamination at both locations may require revision. In this context, it is suggested that the global model of the GIA process is sufficiently accurate that the magnitude and form of local tectonic effects during the Holocene period might be sensibly estimated by simply subtracting the GIA prediction for a given site from the observed variation of relative sea level.
The late Pleistocene cycle of glaciation and deglaciation, which has been the dominant contributor to climate system variability for the last 900 000 years of Earth history, is indelibly recorded in the geological record of relative sea-level (rsl) change. As each of these 100000 year quasi-periodic cycles of ice-sheet advance and retreat involved a fall and subsequent rise of eustatic sea level of c. 120 m, it is hardly surprising that the record of these events should be of such high quality. The best proxy recordings of this glacial cycle, from a long timescale perspective, undoubtedly consist of those based upon oxygen isotopic measurements made on the tests of Foraminifera extracted from deep-sea sedimentary cores. Shackleton (1967) demonstrated that the records thereby derived on the basis of benthic species provided a high-quality proxy for the amount of land ice that existed on the continents at the time in the past represented by the depth in the core at which the isotopic measurement was made. It is, of course, on the basis of records of this kind that the important role played by orbital insolation variations in driving the ice-age cycle was first clearly established (Hays et al. 1976). Although the linkage between orbital insolation forcing and ice volume response is not nearly so direct as Milankovitch had envisioned (e.g. see Tarasov & Peltier (1997) for a recent discussion) it was nevertheless clear on the basis of such data that the small changes in the effective intensity of the Sun, caused by temporal variations of the geometric properties of the Earth's orbit, were able to induce significant cryospheric response.
Of primary interest in the present context will be the variations of rsl that are associated with the most recent deglaciation event, which began subsequent to last glacial maximum (LGM) 21000 sidereal years ago. Even though this event had essentially ended by c. 4000 years ago, rsl continues to change owing to this cause (by rsl, in all that follows, I will imply sea level measured with respect to the surface of the solid Earth). This lingering memory of the deglaciation process is essentially a consequence of the fact that the Earth's shape is continuing to deform because of the shift in surface mass load that occurred during deglaciation as the vast Laurentide, Northwest European and southern hemisphere ice complexes disintegrated and the meltwater thereby produced was added to the ocean basins. This continuing deformation is a consequence of the very high value of the effective viscosity of the Earth's mantle, which governs the timescale of the return to gravitational equilibrium of the ice-solid earthocean system subsequent to deglaciation. Because this continuing relaxation of shape depends so strongly upon mantle viscosity, observations of the process may be employed to infer this Earth property and its variation with depth. That such inferences provide information of fundamental importance will be clear by virtue of the fact that knowledge of the steady-state creep resistance of the mantle is required in the construction of mantle convection models of the process of continental drift and sea-floor spreading. In the discussion of these ideas to be presented in what follows, I will begin with a brief
PELTIER,W. R. 1998. Global glacial isostatic adjustment and coastal tectonics. In: STEWART,I. S. & VITA-FINZI, C. (eds) Coastal Tectonics. Geological Society, London, Special Publications, 146, 1-29.
2
W . R . PELTIER
review of the theoretical structure of the global model of the glacial isostatic adjustment (GIA) process that has been under continuous refinement at Toronto for some time. The origins of this model lie in the analysis presented by Peltier (1974) of the viscoelastic response of spherically symmetrical models of the planet to variations of surface mass load. Using the impulse response Green function for the perturbation of the surface gravitational potential derived for such models by Peltier & Andrews (1976), Farrell & Clark (1976) discussed the primitive form of a 'Sea-level equation' that could be employed to predict the variations of rsl that should occur as a result of the combined influence of the deformation of the solid Earth caused by the changing surface load and the deformation of the geoid (the surface of constant gravitational potential that is coincident with mean sea level (msl) over the oceans). This equation was constructed by analogy with that introduced by Platzman (1971) to describe the influence of the elastic yielding of the sea floor onto the ocean tides. In the studies by Clark et al. (1978) and Peltier et al. (1978) this equation was more accurately expressed and solved for the realistic model of northern hemisphere deglaciation that had been produced by Peltier & Andrews (1976) and designated ICE-1. Further refinements to the theoretical structure that were thereafter introduced included the additional mathematical analysis required to calculate the rotational response to the glaciation--deglaciation process (Peltier 1982; Wu & Peltier 1984), analysis of the free air gravity anomalies associated with this dynamical forcing (Wu & Peltier 1983; Mitrovica & Peltier 1989; Peltier et al. 1992), more accurate spectral methods for the solution of the sea-level equation itself (Mitrovica & Peltier 1991) and development of a technique with which one might incorporate into the solution the full influence of time dependence of the coastline (Peltier 1994) and of the (rather less important) feedback of the changing rotational state of the planet onto sea-level history itself (Peltier 1998a, b). Various parts of this theoretical structure have been subsequently reproduced by others and the complete structure now serves as basis for the continuing international effort to fully understand the GIA process. These contributions from workers outside the Toronto group include those by Lambeck et al. (1990), who employed a much simplified version of the sea-level equation to investigate the postglacial rebound of Fennoscandia; Lambeck et al. (e.g. 1996), who performed a detailed series of analyses of postglacial rsl histories of the British Isles; Han & Wahr (1995), who have
rederived the viscoelastic normal mode formalism of Peltier (1976, 1985); and Fang & Hager (1995), who have also invested effort to understand the rudiments of the normal mode theory. The implications of this work to our understanding of mantle rheology have also been thoughtfully addressed in the recent literature by Karato & Wu (1993). In the following section of this paper I briefly review the structure of this formal theory of the GIA process. I will then focus upon the problem of tuning the viscosity profile of the model, and will examine the extent to which the theory is able to accurately reconcile a globally distributed set of rsl histories obtained on the basis of a4C dating of various rsl indicators. Given the rather good fit to such data that the model delivers, further analyses are devoted to the investigation of observations that have been suggested to disagree profoundly with the theoretical predictions, Arguments are presented to the effect that these concerns are not particularly well founded, and conclusions are offered.
The global theory of GIA and rsl change The record of sea-level history that is contained in the geological record is a recording of the level of the sea relative to the deforming surface of the solid Earth. It is this fact which makes the interpretation of the record as challenging as it so clearly is. If we define this rsl history to be S(0, A, t), with 0 and A latitude and longitude, respectively, and t time, then we might usefully express rsl history in the following schematic fashion:
s(o, A, t) : c(o, A, t)[G(0, ~, t) - R(0, ;~, t)] (1) in which C(O, A, t) is the so-called 'ocean function', which equals zero over land or land-locked water and unity over the surface of the global ocean. In equation (1), G(O, A, t) is the geoid of classical geodesy which is defined by the surface of constant gravitational potential that is coincident with msl over the oceans and R(O, A, t) is the local radius of the solid Earth. To predict the function S(O, A, t) we are therefore obliged to develop a theory on the basis of which we may compute the triplet of functions (C, G, R). The key ingredient of such a theory, as previously mentioned, was provided by Peltier (1974) who developed a mathematical structure with which one could calculate both G and R assuming C to be fixed to the present-day ocean function. That analysis, which was based upon the application of first-order perturbation theory, led to
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS
3
S(O, A, t) and to compare these predictions with geologically inferred rsl histories. In this process we would construct the function A~(t) such as to ensure conservation of mass by insisting that
the re-expression of equation (i) in the form:
s(o,A,t) = c(o,;~,t)
• { I dt' J. J dfl L(O', t')
I
c pwX(t ) df~
• +
A~(t) } g
= [ .w{f (2)
in which ~2 is the surface of the Earth, L is the history of variations of surface mass load (mass per unit area) that occur as a result of the glaciation-deglaciation process, q~L and F L are respectively viscoelastic surface load Green functions for the gravitational potential perturbation and radial displacement, and the function A~(t) is constructed so as to ensure that the variation of surface mass load is mass conserving in the sense that only the mass of water produced by melting (or accreting) continental ice appears in (or disappears from) the oceans. The argument "7 in these Green functions is simply the angular separation between source point (0',A') and field point (0, A), a spatial dependence which results from the assumption that the Earth model of interest is spherically symmetrical in its physical properties. From a mathematical perspective the right-hand side of equation (2) is a (triple) convolution integral. The theory required to construct q5L and F L was presented by Peltier (1974) and Peltier & Andrews (1976), and requires knowledge only of the radial viscoelastic structure of the planet. From a technical perspective the challenge posed by equation (2) arises because of the composite property of the surface load L. This may be made explicit by expanding it in the form
c(0, A, t) =
pr1(o, ~, t) + pwS(O, A, t)
(3)
in which pI and pw are the densities of ice and water, respectively, and I and S are respectively ice and water 'thickness'. We consider L to be positive when the net mass per unit area is increasing and negative where it is decreasing. Clearly, when equation (3) is inserted in equation (2) the resulting equation will be seen to constitute an integral equation for the rsl history S(O, A, t), as S now appears not only on the left-hand side but also under the triple convolution integral on the right-hand side. Given an assumed history of ice-sheet thickness variations I(O,A,t) and a radial viscoelastic structure for an assumed spherically symmetrical model of the planetary interior, we could proceed to solve this integral equation to predict
d,' J. J
t')
A ,b ( t_____). ) x [G(t - t') - R(t - t')]~df~ + pwA g J = -Mj(t)
(4)
in which the integral on the left-hand side is the mass that has been added to the oceans by time t, which must equal the mass of water produced by melting ice, here defined as Mi(t). The negative sign affixed to this function on the righthand side of equation (4) is employed to indicate explicitly that Mr(t) itself is negative when this component of the surface mass load is being removed from the surface. By defining {I)(t)
MI(t)
1
g
pwA(t)
A(t)
x ( I r .dt' I~J d~2'L(O',~',t ')
(5)
• [a(t- t')- R(t- t')]~ /
o
in which A(t) is the surface area of the oceans at time t and ()o indicates integration over the oceans, we will then ensure that solutions to equation (2) conserve mass. Because of the form of the A~(t)/g correction to equation (2), one should not expect that the amount of sea-level rise that occurs far from the ice sheets will be well approximated by the first term on the righthand side of equation (5). Although it is precisely the form of the sealevel equation (2) that I have employed as basis for most of my work on the GIA process, there are in fact two potentially relevant physical effects that are not included in this version of the theory. These are, respectively, the influence of the time dependence of the ocean function C(O, A, t) and the feedback onto sea level of the changing rotational state of the planet. I will return to a discussion of the important former effect below. The latter effect turns out to be small but it is difficult to be certain that this is so without actually doing the calculation. To accomplish this we proceed iteratively by first solving equation (2) to determine the global rsl history S by assuming C to be constant and
4
W . R . PELTIER
including a model of the history of ice-sheet loading and unloading before LGM (this may be constructed by employing the SPECMAP 6180 record of Imbrie et al. (1984)). Given the complete history of surface mass loading L, we then solve the Euler equation d dt (gij 6dj) + Eijk6djgkltdl = 0
(6)
in which Jij is the moment of inertia tensor of the planet, wj are the components of its angular velocity vector and eijk is the Levi-Cevita alternating tensor. Assuming a biaxial model for the undeformed shape of the planet (see Peltier & Jiang (1996a) and Peltier (1997) for the complete but unnecessary triaxial theory), highly accurate solutions to equation (6) may be constructed by employing the standard perturbation expansion:
As discussed in detail by Peltier (1982) and Wu & Peltier (1984), equations (8) may be solved most efficiently by using Laplace transform techniques to determine the wi(t) once L(O, A, t) has been fully determined by solving the sea-level equation (2). Given the solution to equation (8) we may simply incorporate the influence of the changing rotation into equation (2) by extending it as:
S(O, A, t ) = C(O, A, t)
{I'
dt'
dgt'
--00
• [L(O', A', t')G~(% t - t') +
9R(0', ~', t')6~(% t- t')]
+
A~(t) } g
(lO)
~Oi = ~(6ij + mi)
J i j = Iij, i r j (7)
Jn = A + Ill J22 = A nt- 122 ,]33 : C -Jr- 133
in which (A, A, C) are the principle moments of inertia, f~ is the angular velocity of the unperturbed Earth, and Iig and mi are assumed small fluctuations away from the unperturbed basic state. On substitution of equation (7) into equation (6) and dropping all terms of higher order than first in the fluctuations we obtain the following decoupled system of equations for polar motion and rotation, respectively: i . -
-
O-r
m + m = ~
(8a)
(8b)
/'~/3 = 9'II3
in which the so-called excitation functions are and t~3, err = ( C - A ) f ~ / A is the Chandler wobble frequency of the rigid Earth, m = mi + im2, ~ = 91 + i92, i = x/Z]- and the ~i are, with the dot indicating time differentiation, respectively, 113
91 - ( C 123
92-(C_A) 93--
/33
C
I23
+ f~(C- A~
(9a)
/r13
fl(C-A)
(9b)
in which the Green function G~ = [0(% t - t')/g - 1-'(% t - t')] is the same kernel as in equation (2), 9g(0 ', A', t') is the variation of the centrifugal potential because of the changing rotational state which, following Dahlen (1976), may be written (to first order in perturbation theory, to be consistent with the approximation employed to solve equation (6)), as +1
ff~R = 900 Yoo(O, A) + Z
92m Y2m(O' "~)
(11)
m=--I
where 900 = ~w3(t )fla 2
92o = - ~w3(t)i2a 2V/4-/5 92-1 -- (Wl -- iw2)(f]a2/2)V/2/15 92+1 = (wl + iw2)(f~aE/Z)x/~/15 and the tidal-loading Green function G~ is expressed in terms of tidal Love numbers h~ and klx as 1
a~(% t) = g ,=o
[1 + kT(t) -- hT(t)]P1 (cos "7)
02) just as the surface-loading Green function G~ is expressed in terms of surface load Love numbers h) and k L as oo
G~(7, t) = a Z [l + k~(t) - hL(t)]PI(cos 7) me l=o
(9c)
(13)
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS Both sets of viscoelastic Love numbers are calculated using the theoretical ideas and methods developed by Peltier (1974, 1976, 1985). There will be no purpose served by reviewing these technical details here. A further aspect of theory that will be of particular interest, however, is that required to incorporate the full impact of the changing coastline that occurs as land becomes inundated by the sea as sea levels rise because of ice-sheet melting and as land that was once ice covered rises out of the sea as a result of the process of postglacial crustal 'rebound'. To understand how these additional influences may be incorporated into the theory it is helpful to begin by noting that both forms of the 'sea-level equation', namely, those represented by equations (2) and (10), are constructs of first-order perturbation theory that deliver solutions for the history of rsl change with respect to an unspecified and thus arbitrary datum. It is precisely this arbitrariness that may be exploited so as to incorporate the full influence of ocean function time dependence. We simply fix this datum by determining a timeindependent field T'(O, 4) such that
5
solution S(0, A, t), we then determine a new T'(O, A) and thus a new C = C2(0, A, t) using equation (14). We continue this iterative process until the solution for C(O, A, t) converges, which typically occurs in just a few iterations. Very recently, a further refinement of this theory has been developed which has both increased the accuracy of the computation of palaeotopography as defined in equation (14) and improved the understanding of mass balance when this is examined from the perspective of the eustatic sea-level rise expected on the basis of the total ice melt and the net sea-level rise that is predicted by solving equation (2) or (10). This involves a subtle aspect of the theory that has not been explored until recently (Peltier 1998c) which is as follows. We consider the evolution of rsl Sis(P, A, t) at a point on the landscape that is ice covered at L G M but which later comes to be inundated by the sea. Here I employ the subscript IS to denote inland sea. At such points the time series for Sis that is delivered by solving equation (2) or (10) has the following mathematical form: SIs(P, A, t > tD) : ASIs(0, A ) H ( t - tD)
S(O, A, t p ) + T'(O, 4 ) = Tp(0, A) + S~s(O, A, t >_ tD) in which S is a solution to either equation (2) or (10), tp is the present time and Tp(O, A) is the present-day topography of the planet with respect to sea level determined, say, by the ETOPO5 model (or some other higher-resolution model if one is available). If we then construct a time-dependent topography for t h e planet by computing
T(O, A, t) = S(O, A, t) + [Tp(0, A) - S(O, A, tp)] and correct this by adding to T(O, A, t) the thickness of ice 1(8, A, t) to obtain
PT(O, A, t ) = T(O, A, t)+ I(0, A, t) = S(O, A, t) + T'(O, A) + I(0, A, t) (14) it will be clear that where T + I is positive there is (perhaps ice-covered) land that stands above sea level and that where T + I is negative there is ocean. We may then define a 'first estimate' of the time-dependent ocean function as the function C I(o, A, t) that is unity wherever T + I is negative. Given this first estimate we then return to equation (2) or (10) and solve it again incorporating this form of the time dependence C(O, A, t) -- C l(0, A, t). Given the new of
05)
in which to(O, A) is the time of 'deglaciation' when the sea first occupies the region at latitude 0 and longitude A, ASIs < 0 is a spontaneous fall of sea level that is delivered by the solution of equation (2) or (10) at the instant tD, H(t - tD) is the Heaviside step function (+ 1 for t _> to and zero for t < tD) and S]s(O, A, t >__tD) is a function that vanishes at t = tD and thereafter decreases with time so as to represent the fall of rsl that occurs in the inland sea as a consequence of postglacial rebound of the crust. Solutions to equation (2) or (10) deliver an abrupt fall of sea level at t = tD because at that instant there is a marked difference in the gravitational potential between such locations and that which defines the surface of the exterior ocean with this potential being higher than the surface ocean because of the influence of glacial rebound. Solutions to equation (2) or (10) therefore deliver a sharp drop of sea level at the instant of deglaciation t = tD even though the region is being inundated. This is a source of mass to the exterior ocean and constitutes an additional removal of load from the deglaciating region. As this removal of load can only represent, in fact, an additional removal of ice, it is clear that ASIs is actually to be associated with an 'implicit' component of the ice unloading history. I call this 'implicit ice' and its contribution to the loading history, expressed in
6
W . R . PELTIER
terms of an equivalent ice thickness, may be computed from the expression
L(O, A, t)= pIIEx(O, A, t)+ pwASisH(t - tD) = p, I/EX(0, A, t)+ Pw A S I s H ( t - tD)I Pl L J
= pi[I~x(O, A, t)+ I~M(O, A, t)] (16a) in which IEx(O, A, t) is the ice thickness history determined by tuning the model (see below) to fit rsl observations from the ice-covered region, and I~M(O, A, t) is the implicit ice that was also removed to deliver the mass to the exterior ocean that is obtained from the solution of equation (2) or (10). In this solution the ice load IIM 'masquerades' as a fall of sea level. A second contribution to the implicit component of the ice load is connected to the field S(O, A, t) in equation (15). It will be clear that, for rsl to be able to fall in regions that were ice covered at LGM but which are inundated by the sea at t = tD, there must be water in the icecovered region subsequent to tD! Now, in the solution of equation (2) or (10), the unloading of these regions is fully accounted for, partly by IEx and partly by I~M. What is not accounted for, in this equation that derives from the application of first-order perturbation theory, is the net mass of water that fills the depression of the surface that exists at t = tD. This water must, of course, also be delivered by the ice that disappeared from the surface as inundation occurred. This additional contribution to the net implicit ice applies no load to the surface.at t = tD, only subsequently as rebound occurs, but it does 22.5
I
contribute to the thickness of ice that must have existed on the surface at LGM and thus to the palaeotopography defined in equation (14). It may also be computed very accurately as follows. We simply compute the present altitude of the marine limit with respect to msl as a function of geographical position, say ML(O, A, tp), and to this we add the present-day bathymetry, say D(O, A, tp), to obtain, in ice thickness equivalent form, this second contribution to implicit ice as
I~M =PW[ML(O, A, tp)+ D(O, A, tp)] (16b) p~
In computing the full palaeotopography in equation (14) we must therefore also include this second contribution from implicit ice to obtain
PT(O, A, t ) = T(O, A, t)+ IEX(0, A, t) +
:IM(0, ~, t) + I2IM(O, A, t)
Examples of the results obtained from this refinement of the palaeotopography calculation will be provided in what follows. Before employing this theory to illustrate in detail the extent to which postglacial rsl histories may be explained in terms of it, it will be useful to first illustrate the general forms that such solutions possess. To this end I will immediately consider the nature of the solution obtained when the radial variation of mantle viscosity is fixed to that of model VM2 shown in Fig. 1. The origins of this model will be described in the next section. With the radial elastic structure of the model also fixed to that of the Preliminary Reference Earth Model (PREM; Dziewonski &
I
t
I
g 21.5 '~
..........
~7~f
............................
21.0
20.5
20.0
., I
3400
4000
I
I
i
4600
5200
5800
6400
Radius (km) Fig. 1. The viscosity models VM1, VM2 and VM3, which are discussed in detail in the later sections of the text.
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS
7
Fig. 2. Time slices through the ice-thickness maps that constitute the ICE-4G model of deglaciation (note that these are slightly modified from the fields derived by Peltier (1994, 1996), primarily by an increase of the ice thickness over the Laurentide complex that was centred on Hudson Bay).
8
W . R . PELTIER
Anderson 1981), solutions will be discussed for the ICE-4G deglaciation model of Peltier (1994), examples of the northern hemisphere isopacks for which are shown in Fig. 2. Notable in this figure are the extensive North American and Northwest European ice complexes that existed at LGM in which the thicknesses of the continental ice sheets, were typically of order 4 km. In the southern hemisphere component of ICE4G there was also significantly more ice at LGM than at present over West Antarctica and also significant ice cover over Western Patagonia. In Fig. 3 1 show the present-day predicted rate of rsl rise for these choices of the input fields and for both the version of the sea-level equation that excludes the influence of rotational feedback (equation (2)) and that which includes this effect (equation (10)). The top and middle plates of Fig. 3 show these respective solutions, whereas the bottom plate displays the difference between them. Evident upon inspection of these illustrative results is that the influence of rotational feedback upon this characterization of the sea-level response to deglaciation is extremely weak and is strongly dominated by the degree two and order one pattern that is forced entirely (see equation (11)) by the polar motion component of the rotational response to deglaciation. That the influence of rotational feedback is weak not only from the perspective of the present-day rate of rsl rise driven by the GIA process but also from the perspective of the complete history of rsl change is demonstrated in Fig. 4, where I have compared predicted and observed rsl histories at six different locations. The first two, from Barbados and the Huon Peninsula, are U/Th-dated coral records that will figure prominently in what is to follow. The remaining four are from sites that are as close as possible to the extrema of the degree two and order one pattern that characterizes the contribution to the rsl record by rotational feedback. Even at these locations, where the influence of the feedback is most intense, it is clearly extremely weak and negligible for most purposes. This contradicts the claim to the contrary made recently by Bills & James (1996). One final aspect of the general form of the solution upon which I will comment here concerns the time-dependent topography of the planet with respect to sea level that develops as a result of the deglaciation process. Figure 5 illustrates the northern hemisphere component of this field determined for the ICE-4G (VM2) model by executing the steps described in equations (14)-(16). Evident by inspection of this figure will be the vast land bridges that existed at LGM in both the present-day Bering
Strait, which was then entirely dry land (the continent of 'Beringia'), and the present-day English Channel. At that time, most of the present-day Indonesian Archipelago was dry land, and a vast land bridge also connected Australia to Papua New Guinea. These aspects of the global 'topographically self-consistent' solution to the sea-level equation were first described by Peltier (1994). Figure 6 illustrates the time dependence of the coastline in several of these locations in terms of what I have previously called 'inundation maps', on the basis of which one may infer at a glance the time at which a particular land bridge first became impassible. Also of note in the second plate of Fig. 6, which simply portrays the regions of the Aegean Archipelago and Mediterranean Sea that are predicted to have been dry land at LGM but which are now beneath the sea, is that there are many candidates for Atlantis!
Tuning the model parameters Although some evidence has already been presented to the effect that the ICE-4G (VM2) model successfully fits a considerable range of rsl observations, it will prove useful, before providing a more systematic demonstration of this fact, to discuss the procedure that has been followed to arrive at the two required input components of this model. These components consist respectively of the radial viscosity profile (VM2) and the deglaciation history (ICE-4G). As the model of deglaciation history has been discussed at length by Peltier (1994, 1996), I will focus herein on the viscosity structure. Focusing then upon the radial profile of mantle viscosity, the profile VM2, or rather the family of VM2-1ike profiles, has been inferred (Peltier 1996, 1998b; Peltier & Jiang 1996b, 1997) through application of a formal Bayesian inversion procedure based upon the use of the simple VM 1 profile shown in Fig. 1 as starting model. As a first estimate of the viscosity profile, VM 1 has several properties which strongly suggest it to be highly appropriate. Of these, the most important will be clear on the basis of inspection of Fig. 7, which shows a sequence of predictions of the non-tidal acceleration of planetary rotation (represented by )2, which is the time rate of change of the degree two axial component of the gravitational potential field of the planet) and the speed of polar wander as a function of the viscosity of the lower mantle with the upper mantle and transition-zone viscosity fixed to 1021Pas. In each of these calculations the lithospheric thickness has been held fixed to
G L A C I A L ISOSTATIC A D J U S T M E N T A N D C O A S T A L T E C T O N I C S
9
Fig. 3. Predictions of the present-day rate of rsl rise using the VM2 viscosity model and the ICE-4G deglaciation history. Results are shown for analyses performed that both exclude (top plate) and include (central plate) the influence of rotational feedback and (bottom plate) the difference between these predictions, which isolates the influence of the changing rotation alone.
10
W. R. PELTIER 11018 H U O N PEN. P A P U A N G
5004 BARBADOS
0
I .
.1
. ** r |"
' -21
3.5
' ~
0
I
i
-7
I
I
0
.
.
.
' J[
o
-14
.
.........................................
"~-~l
4055 C L I N T O N CT
-~4
'-~1
11002 TOKYO
BAY JAP
'
10
§
................................................
-21
i
-14
,I -7
,
0
I
-21
0
o ' -21
-14
0
Time (ka BP)
-7
L-
~-
-7
-14
0
11083 A D E L A I D E S A U S T
5038 RIO C O L O R A D O A R G E N T I N A
'1
o. . . . . . . .
' -21
o T ....... i
'
-14
,I
-7
i
i
0
Time (ka BP)
Fig. 4, Relative sea-level histories at six significant locations illustrating the negligibly weak impact of rotational feedback upon the Holocene record. The Barbados and Huon Peninsula sites are the two primary locations from which coral-based records are available. For the latter site only the raw data are shown, uncorrected for any influence of tectonic uplift. The final four sites are located as close as possible to the centres of the degree two and order one structure that characterizes the contribution of rotational feedback on rsl history. It is at these locations that the influence of this feedback is maximum. The continuous curve in each frame represents the rsl history that includes rotational feedback whereas the dashed curve is that obtained excluding this effect. The inset in each plate represents the difference between these two histories on a scale that ranges between - 3 . 5 m and +3.5m. Both sets of calculations were performed with the ICE-4G (VM2) model. 120.6km. In Fig. 7a and b the range of the observed values of these properties of Earth's present-day rotational state (see Peltier & Jiang (1996b) and Peltier (1997), for a full discussion) is shown as the hatched region. Inspection of these
results will demonstrate that both of these rotational observables are fitted by precisely the same viscosity model, namely that labelled VM1 in Fig. 1, as the value of lower-mantle viscosity preferred by both observations is seen to be
G L A C I A L ISOSTATIC A D J U S T M E N T A N D COASTAL T E C T O N I C S
11
Fig. 5. Time-dependent topography of the northern hemisphere of the planet from LGM to present according to the ICE-4G (VM2) model of the GIA process. These results include the contributions from both explicit and implicit ice.
12
W. R. PELTIER
Fig. 6. Inundation maps for the Bering Strait, Australia-Papua New Guinea and the Indonesian Archipelago. Also shown is a map centred on the Greek archipelago showing the regions that would have been dry land at LGM (shown as beige) but which are today sea covered.
2 • 10 21 Pas. Because these rotational observables depend upon entirely independent components of the moment of inertia tensor (see equations (9)) it is highly unlikely, in my view, that they could be reconciled by precisely the same model of the radial viscoelastic structure if both observables were not primarily controlled by the GIA process. Furthermore, these rotational data are sensitive essentially to the average value of the viscosity from the base of the lithosphere to the core-mantle boundary (cmb). This may be seen by computing the Frrchet kernel FKR(r) for either of the rotational data, as this appears in the expression for the perturbation of an arbitrary measure of the response 6R
that is induced by a perturbation in the viscosity model 6 logl0 u(r) as 6R =
r~FKR(r)6 logl0u(r) dr
(17)
in which b and a are the radii of the crab and the Earth, respectively. The Fr~chet kernels for each of the rotational data, computed at the model VM1, are shown in Fig. 8c. Because these functional derivatives for both )2 (J2) and polar wander speed (PW) are very slowly varying functions of radius r it is clear from equation (17) that these data determine essentially the average value of u(r) through the
G L A C I A L ISOSTATIC A D J U S T M E N T A N D COASTAL TECTONICS
13
Fig. 7. (a))2 as a function of lower-mantle viscosity//LM when the upper-mantle and transition zone viscosity is fixed to the value rUM = 1.0 • 1021Pa s and the lithospheric thickness is L = 120.6 km. Results are shown for both of the models of glaciation history for which inertia perturbations are shown in Fig. 3. (b) Polar wander speed as a function of lower-mantle viscosity VLM, with other parameters as in (a).
mantle. Because VM1 fits both observations it seems clear that this m o d e l has the correct value o f this Earth property. However, it is also clear that this m o d e l does n o t provide an acceptable fit to all data related to the G I A process. A n extremely i m p o r t a n t example of a dataset that is n o t fitted by the VM1 m o d e l consists of the relaxation spectrum for F e n n o s c a n d i a n r e b o u n d originally inferred
by M c C o n n e l l (1968). His analysis of the strandline data for the post-glacial recovery o f this region led h i m to infer the variation o f relaxation time (shown as its inverse) as a function of spherical h a r m o n i c degree I shown in Fig. 9, in which the star symbol plotted adjacent to the low-degree asymptote at a relaxation time near 4600 years represents the relaxation time inferred by a M o n t e Carlo derived exponential fit to
14
W. R. PELTIER 0.0 G r-
$ -5.0
•
PW J - ' ~ "
n=15
n=25 -lO.O BATHURST
0
n=40
-15.0
Co) -20.0 I 0.5
I 0.8
I 1.0
(c) ! 0.8
-
0.5
1.0
0.5'
! 0.8
1.0
D i m e n s i o n l e s s radius Fig. 8. Fr6chet derivatives for a representative set of the data related to the GIA process. (a) shows a sequence of kernels for the inverse relaxation times of a number of spherical harmonic degrees of the McConnell relaxation spectrum based upon the analytical formula of Peltier (1976). (b) shows Frchet derivatives for the site-specific relaxation times at sites near the centre of Laurentide rebound (Bathurst Inlet; this is actually a high Arctic site) and at the center for Fennoscandian rebound (the Angerman River site). (e) shows kernels for the non-tidal acceleration of rotation ()2) and polar wander speed (PW) that were determined numerically using the procedure embodied in equations (Sa) and (8b) of the text. Inspection of this suite of kernels, all of which were computed on model VMI, which is employed as starting model in the Bayesian inversions, demonstrates that the observables whose sensitivity to viscosity variations they represent offer the potential of significant resolution from the Earth's surface to the crab. the rsl record at Angerman River, which is located near the centre of Fennoscandian rebound. Also shown in this figure are the theoretically predicted spectra for the VMI, VM2 and VM3 viscosity models shown in Fig. 1. Inspection will show that the relaxation spectrum predicted by VM 1 is such that relaxation times are too long at all spherical harmonic degrees, implying that upper-mantle and transition-zone viscosity is too high. That it is in fact the viscosity over this range of depths to which the Fennoscandian relaxation spectrum is sensitive is demonstrated in Fig. 8a, which shows Fr6chet kernels for these relaxation time data for several values of the spherical harmonic degree, computed on the basis of the exact mathematical formula for them given by Peltier (1976). It will be clear by inspecting the spectrum for the VM2 model, also shown in Fig. 9, that the softer upper mantle and transition zone that characterize VM2 allow this model to fit the McConnell data extremely well.
The final set of data employed in the formal Bayesian construction of VM2 consists of a set of 21 relaxation times inferred on the basis of Monte Carlo fits of an assumed exponential uplift curve to individual rsl histories at sites that were once ice covered and at which the rsl records are distinctly exponential in form. Fifteen of these sites are in Canada and six in Sweden and Norway. These locations and the data from them are discussed in detail by Peltier (1996, 1998b). Examples of the Fr6chet derivatives for such site-specific relaxation time data are shown in Fig. 8b for the Angerman River site in Sweden discussed previously and for the Bathurst Inlet site in the Canadian high Arctic. Inspection of these functions demonstrates that the rsl data controlled by the post-glacial recovery of the Laurentian platform are most sensitive to the viscosity in the upper part of the lower mantle, whereas those from Fennoscandia are most sensitive to shallower transition-zone structure.
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAE TECTONICS
1.2
I
I 7
I
I
15
I
it t o five beaches
1.0
Fit 0.8
Lo
six beaches
#t Angermanland (raw data)
,e,,,i
s
0.6 V
O
U
~
.
~
.
,/"
-el
0.4 ,-..4
0.2
_~"'~*~.~__ - - .~- - - -_.=.~..~..~-~ """
cD
VM1
M. & e'. 1 9 9 7
0.o -0.2 .0
I
I
I
I
1.2
1,4
1.6
1.8
.
2.0
Fig. 9. The relaxation spectrum for Fennoscandian rebound of McConnell (1968) based upon both five and six strandline inversions. Also shown as the star symbol adjacent to the low wavenumber asymptote of the spectrum is the inverse relaxation time inferred from Monte Carlo fit to the rsl history at the Angerman River location, which is located near the centre of Fennoscandian rebound. Theoretical predictions of the spectrum are shown for models VM1, VM2 and VM3 in Fig. 1 as well as for model MF of Mitrovica & Forte (1997).
On the basis of this discussion it should now be clear why the formal Bayesian inversion of the totality of the above-described data deliver VM2 when VM1 is employed as the initial estimate. Because VM1 is too stiff in the upper mantle and transition zone to fit the McConnell spectrum, the viscosity in this region must be reduced. However, this reduction reduces the mean viscosity of the mantle, which is unacceptable to the rotational data. The latter data therefore require that the lower-mantle viscosity be increased to restore the mean value of viscosity to that in VM1. This adjustment occurs primarily in the lower part of the lower mantle because model VM1 fits the site-specific relaxation time data from the Hudson Bay region very well, meaning that the viscosity in the upper part of the lower mantle, to which these data are most sensitive (see Fig. 8), is held fixed near that in VM1, namely 2 x 102a Pa s. It will serve no useful purpose here to review the formal mathematical procedure employed in the Bayesian inversion which delivers VM2 from the VM1 first guess. The interested reader will find detailed discussions of this procedure in the studies by Tarantolla & Valette (1982, 1984), Jackson & Matsu'ura (1985) and Backus (1988). It will be useful, however, to end this discussion of the procedure employed to deduce the radial viscosity profile by illustrating the range of VM2-type models that may be derived by
applying variations on the basic procedure. Examples of such profiles are shown in Fig. 10a and b. In Fig. 10a the two versions of VM2 differ from one another only because the site-specific relaxation time data employed in the inversion are deduced from the envelope sampled form of the rsl curves as in the archive of Tushingham & Peltier (1992) or from the raw age-height pairs directly. The former procedure leads to the version of VM2 shown as the heavy continuous line in Fig. 10a, whereas the latter procedure leads to the thin continuous line. These two variations of VM2 are clearly very close to one another. In Fig. 10b the raw data based version of VM2 from Fig. 10a is compared with a new version in which the forward predictions of the site-specific relaxation times were made using the version of the sea-level equation in which the full influence of time-dependent ocean function was included, an influence that was neglected in the inversions for which results are shown in Fig. 10a. Inspection of Fig. 10b will show that in this most accurate of the inversions the viscosity in the upper part of the lower mantle is somewhat elevated above 2 x 1021pas but only very slightly. In this model the ratio of the viscosity in the upper part of the lower mantle to that in the upper mantle and transition zone is approximately five. It will be useful to end this section with a very brief discussion of the relation between models in
16
W. R. P E L T I E R 22.5
I
I
I
I
I
I
l
22.0
21.5 VM1 21.0
20.5
- raw data
r~
(a) I
20.0 ~ I.,=4
o oO g
Eo0
22.5
22.0
o
-
VM2 C(O
21.5
VM2
21.0
L-----------"
-
f
20.5 (b) raw site data 20.0 3400
I
4000
I 4600
l 5200
I 5800
6400
Radius ( k m ) Fig. 10. Viscosity profiles determined by simultaneous formal Bayesian inversion of the Fennoscandian relaxation spectrum of McConnell (1968), the site-specific relaxation times from 23 ice-covered sites in Canada and Fennoscandia, and the non-tidal acceleration of the rate of axial rotation. In (a) the dashed line is the VM 1 viscosity profile employed as starting model in the inversion process, and two versions of the final model are shown as the dark and light continuous lines, respectively, these being distinct versions of VM2. The former of these two inferred models has been obtained using site-specific relaxation times obtained from fits to the envelope sampled data compiled by Tushingham & Peltier (1992), whereas the latter was obtained on the basis of sitespecific relaxation times deduced from the raw data themselves. In (b), where the dashed line again indicates VMI, the two versions shown are the final VM2 model (that based upon use of the raw data to determine the sitespecific relaxation times) and a further version in which the starting model predictions were made with the version of the model that included the full influence of time dependence of the ocean function. Incorporation of the latter effect in the forward model slightly decreases the forward predictions for the site-specific relaxation times and therefore slightly increases the inferred viscosity in the upper part of the lower mantle, essentially back to the value of 2 • 1021 Pa s that is characteristic of the starting model VM1 in this region. the class V M 2 to those which have recently been a d v o c a t e d by other workers. The closest o f these other models to V M 2 is that derived by L a m b e c k et al. (1990) by trial-and-error fits to a set o f rsl curves from F e n n o s c a n d i a . Their m o d e l is iden-
tical to V M 2 t h r o u g h o u t the u p p e r m a n t l e and transition zone, approximately a factor o f two higher in viscosity in the u p p e r part o f the lowerm a n t l e (4.5 • 1021 Pa s), and essentially equal in the lower part o f the lower m a n t l e (in fact, these
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS workers quoted a range of allowed lower-mantle viscosities of (2-7) x 1021 Pa s). In the analyses presented herein, it is the data from the Hudson Bay region of Canada that collectively require the viscosity to be somewhat lower in the upper part of the lower mantle. Mitrovica & Peltier (1995) suggested that the site-specific rsl data from the Hudson Bay region exhibit a significant spread in relaxation times ranging from a low near 2000 years (Ottawa Islands) to a high near 7600 years (Richmond Gulf). Advocates of higher values of viscosity in the upper part of the lower mantle than that in VM2 (e.g. Mitrovica 1996; Forte & Mitrovica 1996; Mitrovica & Forte 1997; Simons & Hager 1997; to be referred to collectively in what follows as MFSH) have focused entirely
1.2
I
17
upon the Richmond Gulf record and ignore all the rest, an approach which introduces significant bias as there is no reason to believe that the Richmond Gulf record is superior to any of the others. It has, in fact, become clear in the course of recent analyses (Peltier 1998c) that the previously published high estimate of the relaxation time at Richmond Gulf (Mitrovica & Peltier 1995) is in error. Rather than being near 7600 years as suggested by Mitrovica & Peltier (1995), it is in fact best estimated as 3400 4- 400 years on the basis of a complete reanalysis of all available 14C data from the southeast Hudson Bay region when these are properly transformed onto the sidereal timescale using the Calib. 3.0 software of Stuiver & Reimer (1993). The VM2 model
I
I
I
a) 1.0
Fit to six b e a c h e s *
0.8
Angermanland
v~
(raw data)
,if,' ,'
IQ
0.6
noo
'
'
'
'
'
Depth
o
/
..... ._.._.._. .........
0.4 i
m o.2
Q)
I
o.o E-4
1.2
1.(
*.=4
I
I
1.6
1.B
I
I
2.0
i
L2
I
I
(11
b)
0
n~ 1.0
I Fit to six b e a c h e s
N ~
I
1.4 ,-,
*
0.0
Angermanlemd
(raw data)
~" ~ q t . ~ _ ' ~ "~1 I]--' ~ L li
~
4) m 0.6
,~
191
t* i
eoo
*
I
J
'
Depth
o
,..4 0 4 *
,9. . . . ~ . . ~ . . - . I : ' . . . . - L - . . . 4 - ... . - . . - .I. T . . . -' . . ~ - - - ~t " "
0.2
0.0 1,0
I 1.2
! 1.4
I, 1.6
I 1.8
2.0
Logto(degree) Fig. ll. (a) The inverse relaxation time spectrum for Fennoscandian rebound of McConnell (1968) compared with the prediction of a model in the VM2 class and compared with the predictions for two additional models that differ from VM2 by the presence of a 70 km thick layer immediately overlying the 660 km discontinuity in which the viscosity is reduced either by one or two orders of magnitude from the value near 0.45 x 1021Pa s that otherwise obtains in this region of VM2.
18
W . R . PELTIER
predicts precisely this relaxation time in southeast Hudson Bay near the centre of uplift. It is therefore clear on this basis that the inferences of viscosity presented by MFSH are untenable. Further evidence of this fact is clear on the basis of analysis of the upper-mantle and transition-zone viscosity structures of the MFSH models. Those models of the shallow structure include both extremely soft transition zones and higher-viscosity upper mantles. Structures of this kind would appear to be ruled out entirely by the McConnell (1968) spectrum for Fennoscandia rebound as demonstrated in Fig. 9, where the spectrum of the Mitrovica & Forte model (denoted MF) is compared with those of the VMX models as well as with McConneU's data. If models of this type are to be entertained it is clearly incumbent upon their advocates to prove that the McConnell (1968) spectrum is very significantly in error. Otherwise the VM2 model must be strongly preferred, a model which is similar to that earlier advocated by Lambeck et al. (1990) although with significantly lower contrast in viscosity across the spinel-post spinel phase transition at 660 km depth. It is very important to realize, however, that the VM2 family of models may in no sense be construed to represent a 'uniquely' preferred solution to the one-dimensional mantle viscosity inverse problem. Although one can argue this point formally, it is probably more useful to demonstrate it by providing a specific example. To this end, Fig. 1 l a shows results for the McConnell spectrum obtained using a perturbed version of the VM2 profile in which a thin low-viscosity layer is inserted into the structure immediately above the 660 km seismic discontinuity, with the viscosity in this layer being fixed to either 102oPa s or 1019 Pa s. The presence of such a structure has been inferred to be required in viscosity models derived by inversion of the non-hydrostatic geoid anomalies that are supported by the mantle convection process (e.g. Forte et al. 1993a, b; Pari & Peltier 1995). Even though models of the perturbed kind shown in Fig. 11 a are essentially identical with those required by these data, insofar as the radial variation of viscosity is concerned, it is clear that such models do not fit the McConnell (1968) relaxation spectrum and they are therefore untenable insofar as the GIA data are concerned. However, it is possible to further perturb the structure so as to recover the good fit to the McConnell spectrum as shown in Fig. 11b. The results shown in this figure demonstrate that the presence of the soft layer may be easily accommodated simply by increasing the viscosity of the rest of the transition zone back towards the value in VM1, namely 1021 Pas. Models of this kind would not be ruled out by the data yet they
differ significantly from VM2. As the transition zone is rich in garnet and as this mineral has a high creep resistance, one may be tempted to argue that such models are actually to be preferred. For present purposes, however, these analyses are presented simply to demonstrate the extreme degree of non-uniqueness in the radial variation of viscosity that the data allow.
Model-data intercomparisons for rsl history As discussed in the last section, very few of the rsl data that are actually available have been employed to tune the radial viscosity profile of the model; in fact, the only data used in this way are those from sites located near the centres of the Laurentian and Fennoscandian ice sheets, and from Barbados, where the coral-based record of Fairbanks (1989), which extends to LGM, has been used to constrain the total ice melt in the deglaciation model. All of the remaining data may therefore be employed to verify the quality of the ICE-4G (VM2) model. In the discussion to follow, the focus of the first subsection will be upon rsl data that actually constrain the time-dependent elevation of the shoreline and that usually derive from 14C dating of mollusc shells or wood specimens whose indicative meaning in the landscape suggests that the sample records a former level of the sea. In the second subsection the focus will shift to the U/Th-dated coral records of rsl history mentioned in the Introduction.
Relative sea-level histories beyond the ice sheet margins The 14C data that will be employed in this subsection to test the quality of the ICE-4G (VM2) model constitute an extremely small subset of the c. 600 individual rsl records that are now contained in the database at the University of Toronto. This database has yet to be published and constitutes a considerable improvement upon the reconnaissance collection of Tushingham & Peltier (1991, 1992). Rather than being based upon sampling the envelope of the set of age-height pairs that are derived from each sample, the new data base consists of the raw data themselves and it is these data that will be employed for comparison purposes herein. When these data are compared with the predictions of the theoretical model we must of course transform from 14C time to sidereal time, and for this purpose we employ the Calib. 3.0 program of Stuiver & Reimer (1993), which links the extensive tree ring database for the
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS Holocene to the coral-based U/Th calibration of Bard et al. (1990) for the late glacial period. One of the most important regions in which the model may be subjected to rigorous test comprises the entire eastern seaboard of the continental USA. This is because even the earliest analyses based upon the VM 1 model (Peltier et al. 1986; Tushingham & Peltier 1992) demonstrated that there were significant misfits between the observations and the predictions such that the rates of sea-level rise predicted to be occurring as a result of the collapse of the proglacial forebulge were much higher than observed in this geographical region. That these misfits are essentially completely eliminated by ICE-4G (VM2) is demonstrated in Fig. 12, which compares observations with theoretical predictions at Montreal, Quebec (which was located just inside the ice sheet margin), Boston, Massachusetts (which was located very close to the ice sheet margin itself), Brigantine, New Jersey (which was located outboard of the ice sheet near the crest of
240.
MONTEEAL, QUE, CANADA I~ !
l
120f BOSTON,MA.USA ,
-50
I
BRIGANTINE. NJ,USA
iif
#
LUIIPUTCREEK.NC. USA
-2o
-',~
-'~0
-o
Time (ka)
Fig. 12. Examples of 14C-dated sea-level curves from four sites on the east coast of the North American continent: Montreal (Quebec) in Canada, and Boston (Massachusetts), Brigantine (New Jersey) and Lilliput Creek (North Carolina) in the USA. The carbon dates for the individual samples have been converted to sidereal years using the U/Th-based calibration of Stuiwer & Reiner (1993) and the raw data, corrected in this way, are compared with the predictions of the ICE-4G (VM2) model.
19
the forebulge) and Lilliput Creek, North Carolina. All of these data are well fitted by the theoretical predictions, demonstrating, as previously documented by Peltier (1996), that the previously evident misfits are eliminated. This is rather important because the modification to VM1 to produce VM2 primarily involves a reduction of the creep resistance in the upper mantle and transition zone that was required to fit the rsl data from Fennoscandia. As this modification to the structure also allows the model to reconcile North American data, this suggests that the upper-mantle and transition-zone viscosity below North America are essentially the same as beneath Northwestern Europe. A global representation of the marked difference in the rsl history predicted by the VM 1 and VM2 viscosity models when the ICE-4G deglaciation history is employed in the calculation is provided in Fig. 13, which shows the present-day predicted rate of rsl rise for both models as well as the difference between them. The difference between these predictions is clearly largest along the US east coast, in precisely the region where the misfits of the VMl-based theory to the observations were largest. Probably the best location in the world from the perspective of the quality of the post-glacial rsl data that are available from it, however, is the British Isles. This region is also especially interesting because it is not only located in the region of forebulge collapse that surrounds the previously glaciated region of Fennoscandia but it was also glaciated in the north, where a significant mass of ice was located over the highlands of Scotland. Furthermore, the coastline of this region experienced significant variation subsequent to LGM, when a vast land bridge connected Britain to France. Figure 14 shows the locations of four sites from which high-quality rsl data are available, superimposed upon the inundation map which illustrates the way in which the coastline is predicted to have evolved based upon the ICE-4G (VM2) model. Figure 15 compares predicted and observed rsl history at Tay Valley and North Solway Firth, Scotland, both sites in the northern region that was once ice covered, and at the Fenlands and Bristol Channel locations in the south, which remained ice free. Inspection of these comparisons clearly demonstrates that the ICE-4G (VM2) model very accurately predicts even the very complex and highly non-monotonic rsl histories that obtained in the northern region, where a complex interplay occurs between the process of post-glacial rebound of the crust that causes sea level to fall and rising sea levels caused by the melting of distant ice sheets. The
20
W.R.
PELTIER
Fig. 13. Present-day rates of sea-level rise predicted using the VM1 and VM2 viscosity models in conjunction with the ICE-4G deglaciation history. The difference between the predictions of these models is maximum along the US east coast, as shown in the final part of the figure.
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS
21
Fig. 14. Inundation map for the British Isles on which are superimposed the locations of the two sites in Scotland and the two sites in England for which the rsl data are described in the text. TV, Tay Valley; SF, Solway Firth; F, Fenlands; BC, Bristol Channel.
fact that the ICE-4G (VM2) model fits these data extremely well is important because it has been previously argued that the data from this region required a rather high value of the viscosity of the lower mantle, in fact near 1022Pas (Lambeck et al. 1996), and therefore considerably in excess of the viscosity over the same range of depths that had been inferred previously on the basis of rsl records from the immediately adjacent Fennoscandia region (Lambeck et al. 1990). Clearly, models with such large radial viscosity contrasts are not, in fact, required by the data from the British Isles. The interested reader will find a far more detailed discussion of the post-glacial sea-level history of the British Isles in the study by Peltier & S h e n n a n (1998).
Moving further still away from the main centres of glaciation into the region that I have previously referred to as the 'far field' of the ice sheets, Fig. 16 compares predicted and observed rsl histories at a sequence of sites that extends from the Caribbean Sea along the east coast of the South American continent to the northern part of Argentina. The first plate in this sequence once more reproduces the fit to the U/Th-dated coral sequence from Barbados, a record that was actually employed to tune the ice load component of the model (Peltier 1994) as mentioned previously. Moving further south to Recife and Santos-Itanhaem, in Brazil, and Bahia Solano, in Argentina, we note that the most apparent characteristic of the records of rsl history, both observed and predicted, is the existence of a
22
W. R. PELTIER TAY VALLEY. SCOTLAND
BARBADOS "/5 L
I
I
/ -75 _
I I NOR'IN SOLWAYFIRTH. SCOTLAND
[
-150 RECIFE, BRAZIL
0
'~
I~NI.ABDS, I~GIAh~D
'
'
t
o
-10 f -20
I
SANTOS-ITANHAEM. BRAZIL
,/1
-10 f I
-20
BRISTOL CHANNEL, BIq'GLAND
4
BAHIA ~OI~NO, ARGEIfI~A
t
2 -20 0 -20
-10 Time
-
0
(ka)
Fig. 15. Examples of 14C-dated rsl curves from four sites in the British Isles: Tay Valley and Solway Firth in the once ice-covered region of Scotland, and the Fenlands and Bristol Channel, which were both beyond the southernmost extent of the Scottish ice sheet. As discussed in the text, the records from Scotland are highly non-monotonic, because of the superimposition at these locations of the influence of rebound of the crust owing to ice removal and the influence of the collapse of the Fennoscandian forebulge and continuing addition of mass to the oceans caused by the melting of both Laurentide and Fennoscandian ice.
mid-Holocene high stand that occurs at c. 5 ka BP. The fit to the observed highstand is excellent at Santos-Itanhaem, acceptable at Bahia Solano but less so at Recife where a significant phase shift appears to exist. As will be discussed in detail elsewhere, the southernmost east coast of Argentina, Patagonia, appears to be experiencing some tectonic uplift, presumably because of the increasingly close proximity to the Chile trench as one moves further southwards along the east coast. In the final sequence of intercomparisons, shown in Fig. 17, the mid-Holocene highstand of sea level continues to be the most prominent feature of the records of rsl history. These records are all from the Pacific Ocean sector, and from north to south correspond to Osaka Bay, Japan, Rota Island in the Marianas, Balding Bay, Australia, and Christchurch, New
-g -20
15
-10 Time
-5 (ka)
Fig. 16. Examples of 14C-dated sea-level curves from four sites along the east coast of the South American continent from Barbados in the north to Recife and Santos-Itanhaem on the coast of Brazil, and to Bahia Solano in Argentina. It is interesting to note that although the data from the northern sector of the coast are fairly well fitted by the ICE-4G (VM2) based theory, as one moves further to the south the data show increasing evidence of the action of the influence of tectonic uplift (a detailed analysis of this tilting effect will be provided elsewhere).
Zealand. The theoretical prediction of the sealevel history at each of these locations is again dominated by the existence of a mid-Holocene highstand with maximum height above present sea-level occurring at c. 5 ka BP and achieving an amplitude of c. 2 m. As previo,usly mentioned, it was the prediction of this feature which is extremely well expressed across the entire Pacific Ocean, that attracted such attention to this work when solutions of the sea-level equation were first reported by Clark et al. (1978) and Peltier et al. (1978). A more recent discussion of the data that constrain this feature from the archive of Tushingham & Peltier (1992) was presented by Mitrovica & Peltier (1991). All of these records, most of which are constrained by a very small number of data points, are reasonably well explained by the theory. In the next subsection, we will consider a series of records which have been construed to pose a considerable challenge to the global theory.
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS OSAKA
BAY,JAPAN'
23
Eustatic Sealevel Rise
T 11P/.8
-45-304 NOrAI5~1[
0G
,~ I
, ,//..///'"/"" """"""..................
/,,"
~ oo.4
~IIIlRI~ItI~C]I I , NEW
'2
f -4o -20
-so
J15 -lo
1~ _ ~ - 7 ~ -5
t
_
o
T i m e ~ka)
Fig. 17. Examples of 14C-dated sea-level curves from four sites in and on the margin of the Pacific Ocean, ranging from Osaka Bay in the north to Christchurch, New Zealand, in the south. At the intermediate sites of Rota Island and Balding Bay, Australia, the mid-Holocene highstand of sea level is especially apparent.
Coral-based records of rsl change." the Pacific Ocean sector At several points in the above discussion, I have drawn attention to the fact that the ICE-4G (VM2) model was tuned in terms of the total ice amount so that it would fit the U/Th-dated coral-based rsl record from Barbados. The reason for focusing upon this record is that it is unique from a number of points of view. First, it is the only accurately dated rsl record which extends back to LGM and which therefore can be used to constrain the amount by which sea level rose from that time to the present. Second, however, is the fact that this record is composed almost entirely of a g e - d e p t h measurements on the coral species Aquapora palrnata. As this Caribbean species is known to live at a depth that is within 5 m of sea level, this record may be assumed to constitute a good recording of changing sea level itself. Of course, it is also well known that Barbados is rising at a rate that is usually assumed to be near 0.35 mm per year, implying that to derive actual sea level from the raw a g e - d e p t h data for the Barbados sequence
Time (kaBP)
%, ,, ,, ,, ~ 9 , 0 Fig. 18. The total eustatic sea-level rise is shown from the model that both excludes (T t) and includes (T) the influence of implicit ice as discussed in the text in connection with equations (14)-(16). Also shown are the individual contributions from North America (N, Nt), Eurasia (E, E'), Antarctica (A) and Greenland (G, G').
one needs to make a tectonic correction. The magnitude of this correction is therefore such as to require sea level to have risen at Barbados since LGM by c. 7 m more than suggested by the raw data. The position that I will adopt for present purposes, as I have done in the past, is that if the theoretically predicted rsl history at Barbados is such that the prediction goes through the raw data points themselves then, given the observed range of living depths of A. palmata, no further correction to the data is required. This is particularly true as the best estimate of the rate of tectonic uplift at Barbados is probably closer to 0.25 mm per year than to 0.35 m m per year. The fit of the ICE-4G (VM2) model to the Barbados data has been shown previously in Figs 4 and 14, where the predicted rise of sea level at this location is seen to be just slightly less than 120 m. It is important to note that this is slightly greater than one might expect based upon the net mass of 'explicit' ice that is melted across the glacial-interglacial transition, though one must remember that determination of the rise of sea level at any given site requires a correction to the effect expected on the basis of mass alone (see equations (4) and (5)). The sea-level rise based upon the consideration of ice-mass alone (both explicit and implicit as discussed previously), for each of the main sectors in which melting occurs, is shown in Fig. 18,
24
W. R. PELTIER
inspection of which shows that one would expect sea level to rise by only 106.7 m if only the explicit ice melted in ICE-4G were to be added uniformly to ocean basins of fixed present-day area. On the other hand, when the influence of implicit ice is properly taken into account, the discrepancy is much reduced, as the predicted total eustatic rise increases to 117.8 m. Figure 19 shows the locations of the additional sites in the Pacific basin that will be of interest for the remainder of this subsection, namely, the Huon Peninsula of Papua New Guinea, Tahiti and Sumba and Morley Islands. At each of these locations coral-based records of rsl change are also available for which the age control is based upon U/Th dating (Edwards 1988). It is clearly interesting to enquire as to whether or not the global theory is also able to reconcile these additional observations, the most important of which are probably those from the
Huon Peninsula (Edwards 1988; Chappell & Pollach 1991; Ota et al. 1993). On the basis of their observations of the uplifted Pleistocene interglacial terraces near the Kwambu-Kilasairo location, Chappell & Polach (1991) inferred a late Pleistocene average rate of tectonic uplift at Huon of 1.9mm per year. In Fig. 20a, I show not only the raw data for Huon but also the corrected data when this rate of tectonic uplift is employed to reduce them, along with the theoretical prediction based upon the ICE-4G (VM2) model. Whereas the raw data lie somewhat above the theoretical prediction at young age, it is clear that when the data are corrected by assuming the conventional rate of tectonic uplift they lie considerably below the theory, so far below as to suggest that perhaps the theory may be in error (see comments by Edwards (1995)). Also shown in Fig. 20, however, are comparisons between observations and theoretical prediction
PACIFIC ISLANDS
it
!i
1 ~' r,
4
I
'
'
II,"
I1
"k
:
t
~
I
t
i
I-
i
!"~
.
,,I . . . . . . "
i8
I
9
I
__, ....
,'"7~'~ ~ " 4.,. Ittt:5 I "IPu5 .P"~
/"-
__~
'
41-43"~""
"t~
_
"~
'~
_
ill
i.,i.~_u
i t............. tAWli~lI,
!
1
I
, 1 , I
I
!
1 ~,,l~
.........
~ . 67 " 't, _lio .06,,'] ..... ~. . . . . . !
I
~7
!
~.
!
9 .. . . . .
;
. . .
1
~I,,
' .
i
z
i ., .
i
i
:
I ,
s,
Is,
_~"~
#'L
i
i
~"~.
. . . . . ,! $IJMBtl ~ _
r
~!
.....
.
._
!
1Z
I~'h/j ',"l
(,,J ~,.,~
'in-i
.i
I
i
I
,i" Lo'd,
I ~
h i
.
fo
i
l ,
- - 7 i .............. t
ti l__i
~)~'~ i
9
i I
'
tit 9, ,
,
.-.~ ~
T ....
'rAttrzl
I
o
i !
4~
i ,
i I
!
~,, 100it
i 12lib
14013
16013
180
i I
i
I
160W
140W
Fig. 19. Location map for the equatorial Pacific Ocean region showing the locations of the islands from which information on rsl history is available. The data from those islands for which names are given are discussed in the text. The numbering scheme for the other locations is that employed in the global recompilation of rsl data that has been continuing at Toronto for the past several years.
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS HUON PEN. PAP&U NG
20,
I,
,I '
I.
- (Vt=l.Omm/yr)
.
.I
,,.
~
'
r
i
Tk,I-II~
20
I
I
~
-40
.40
-00
.60
-OO
.80 I
I
I
L00 I
I
MORLEY IS. I
I
I
I
I
/
~
"~ % -
.20
I, ' /
l
.living depth corrected ! (Bard, pets. comma.)
-20
-100
25
I
I
9
20
1
ki-
Z
,
,
,
I
!
1
SUMBA IS.
]
I
(V,..?)
, I
I
o
(Vt=O.35mm/yr)
0
-
-30
'
,li
~.
--
-20 [.,.
.
.
-40, -BO
-40
f
-80
"
- 8 0 [-100 -24
-80 I -20
-16
-12
-8
I -4
-100 0
-P.AI
I -20
-16
-12
-O
I -4
Time (ka) Fig. 20. The raw and tectonic uplift corrected coral-based records are shown, along with the predicted sea-level histories based upon model ICE-4G (VM2) for (a) the Huon Peninsula, (b) Papeete Harbour, Tahiti, (e) Morley Island and (d) Sumba Island. for the rsl history at Morley Island (Eisenhauer et al. 1993) and Sumba Island (Bard et al. 1996b),
whose locations are also shown in Fig. 19. Inspection of the intercomparisons at these sites demonstrates that at these locations theory and observations agree rather well, although in the case of Sumba Island the data are sparse (essentially only two points). At Sumba there is an estimate, noted in the figure, of the rate of tectonic uplift active at this location, and it will be noted that when this correction is applied to the data the fit of the two data points to the theory is excellent. At Morley Island the observations and theory agree very well if allowance is made for a small living depth correction. Both of these additional datasets are consistent with the existence of a mid-Holocene highstand in the sea-level record but neither data set actually resolves the feature. It is nevertheless clear that when the Huon data are 'corrected' by assuming that the conventional Pleistocene rate of tectonic uplift also
applies during the Holocene period then the fit of the theory to the data is anomalous in that it is so poor. The most apparent anomaly concerns the absence of the mid-Holocene highstand in the rsl record when the 1.9 mm per year tectonic uplift rate is used to make the correction. It seems clear on this basis that the 1.9mm per year average rate of Pleistocene uplift at the Kwambu-Kilasairo site on the H u o n Penninsula is not characteristic of the Holocene period. To determine the rate that has actually been characteristic of this most recent epoch we might best proceed by asking what is the rate that must be assumed to minimize the misfit between observations and theoretical prediction. The results of this analysis are shown in Fig. 21. Depending upon whether one minimizes the misfit over the entire dataset or only over the last 9000 years one infers a best rate of tectonic uplift between 0.3 and 0.65mm per year, although the minimum in variance is rather flat so that the rate is not accurately determined. As suggested by Peltier
26
W. R. PELTIER PEN. P A P A U NO
HUON
4000,
I
I
I
I
I
I
I
I
I
b e s t Vt(all data, solid line)=0.3 m m / y r
/
3000.
0
2000.
v *,-s
1000,
0.0
0.1
0.2
0.3
0.4
0.5
0.8
0.7
0.8
0.9
1.0
Uplift Rate Vt (mm/yr) 35
I
I
I
Vt=O.65 mm/yr o
A
o
E d w a r d s et al---
Chappell & Polach--, Ota et al(Kwambu-Kilasairo)--, -70 -16
,,
-12
I -8
I -4
0
Time (ka) Fig. 21. (a) Variance reduction between the prediction of the rsl history on the Huon Peninsula for the ICE-4G (VM2) model and the observations as a function of the rate of tectonic uplift assumed for the purpose of correcting the observations. Variance reduction is shown based upon the use of all available data and the use of only data younger than 9 ka. With the former assumption the 'best-fitting' rate of tectonic uplift is found to be 0.35 mm per year. (b) Fit of the tectonically corrected data to the ICE-4G (VM2) prediction. (1995), it therefore seems clear that the rate of tectonic uplift at H u o n during the Holocene period has been much less than the Pleistocene average rate. One final piece of analysis that warrants discussion here, concerning the tectonic contribution to Holocene sea-level history, relates to the data for Tahiti (Bard et al. 1996a), which are also shown in Fig. 19. At Tahiti, unlike at Huon, the theoretically predicted rsl history delivered by the ICE-4G (VM2) model lies above the observations rather than below. However, rather than being subject to tectonic uplift, Tahiti is subject to tectonic subsidence. This is perhaps primarily a consequence of the fact that the
island is emplaced in lithosphere that sinks is it cools while moving away from the East Pacific Rise at a rate near 12 cm per year. Now the rate of subsidence that has been suggested to be characteristic of Tahiti is near 0.2 mm per year, but when this correction is applied to the data it does not significantly improve the fit to the observations. When an additional correction is made for the influence of living depth (E. Bard, pers. comm.) the misfit is further reduced (see the figure). However, the data still do not contain the clearly evident mid-Holocene highstand of sea level that is observed to be characteristic of all rsl data from far field locations. Further analysis of the data at Tahiti is clearly warranted.
G L A C I A L ISOSTATIC A D J U S T M E N T A N D COASTAL T E C T O N I C S On the basis of all of the data discussed in this subsection, it is clear that w h e n the model is tuned to fit the history of post-glacial sea-level change at Barbados, as well as those from icecovered sites, then the m o d e l also fits far field rsl histories at locations that were not e m p l o y e d to tune the m o d e l parameters. The model must therefore be considered highly successful.
Conclusions In the previous sections of this paper, I have both reviewed and extended the global theory of post-glacial sea-level change that has been developed over the past two decades. This theory is n o w rather fully articulated and has been shown to reconcile a wide range of geophysical and geological data. One of the most active areas of current application involves the investigation of the extent to which space geodetic data m a y be b r o u g h t to bear to further constrain m o d e l parameters. Three different types of such data have n o w been shown to be useful adjuncts to the geological and astronomical m e a s u r e m e n t s that have been the conventional focus in such work. These consist o f observations of the nontidal acceleration of axial rotation based u p o n the use of L A G E O S satellite laser ranging data (Peltier 1983; Y o d e r et al. 1983), observations o f the rate of radial displacement based u p o n the use of V L B I observations (Argus 1996) and, most recently, observation of tangential motions associated with the G I A process based u p o n global positioning system (GPS) observations ( B I F R O S T 1997). Application of these measurem e n t systems is expected to prove to be especially useful in u n d e r s t a n d i n g the relative contributions of tectonics and glacial isostasy to individual records o f rsl history.
References ARGUS, D. F. 1996. Postglacial rebound from VLBI geodesy: on establishing vertical reference. Geophysical Research Letters, 23, 973-977. BACKUS, G. E. 1988. Bayesian inference in geomagnetism. Geophysical Journal of the Royal Astronomical Society, 92, 125 142. BARD, E., HAMELIN,B., ARNOLD,M., MONAGGIONI,L., CABIOCH, G., FAURE, G. & ROUGERIE, F. 1996a. Deglacial sea level record from Tahiti corals and the timing of global meltwater discharge. Nature, 382, 241 244. , , FAIRBANKS, R. G. & Z1NDLER, A. 1990. Calibration of the 14C timescale over the past 30,000 years using mass spectrometric U-Th ages from Barbados corals. Nature, 345, 405-409.
27
BARD, E., JOHANNIC, C. et al. 1996b. Pleistocene sea levels and tectonic uplift based on dating of corals from Sumba Island Indonesia. Geophysical Research Letters, 23, 1473 1476. BIFROST 1996. GPS measurement to constrain geodynamic processes in Fennoscandia. EOS Transactions of the American Geophysical Union, 35, 377. BILLS, B. G. & JAMES, T. S. 1996. Late Quaternary variations in relative sea level due to glacial cycle polar wander. Geophysical Research Letters, 23, 3023-3026. CHAPPELL, J. • POLACH, H. A. 1991. Post-glacial sea level rise from a coral record at Huon Peninsula, Papua New Guinea. Nature, 276, 602-604. CLARK, J. A., FARRELL,W. E. & PELTIER,W. R. 1978. Global changes in postglacial sea level: a numerical calculation. Quaternary Research, 9, 265-287. DAHLEN, F. A. 1976. The passive influence of the oceans upon the rotation of the Earth. Geophy-
sical Journal of the Royal Astronomical Society, 46, 363-406. DZIEWONSKI, A. M. & ANDERSON, D. L. 1981. Preliminary reference earth model. Physics of the Earth Planetary Interiors, 25, 297-356. EDWARDS, R. L. 1988. High precision Th-230 ages of corals and the timing qf sea levelfluctuations in the late Quaternary. PhD Thesis, California Institute of Technology, Pasadena. - - 1 9 9 5 . Paleotopography of Ice-Age ice sheets. Science, 267, 535-536. E1SENHAUER, A., WASSERBURG, G. J., CHEN, J. H., BONANI, G., COLLINS, L. B., Znu, Z. R. & WYRWOLL, K. H. 1993. Holocene sea level determination relative to the Australian continent: U-Th (TIMS) and a4C (AMS) dating of coral cores from the Abrolhos Islands. Earth Planetary Science Letters, 114, 529-547. FAIRBANKS, R. G. 1989. A 17,000-year glacio-eustatic sea level record: influence of glacial melting rates on the Younger Dryas event and deep-ocean circulation. Nature, 342, 637-641. FANG, M. & HAGER, B. H. 1995. The singularity mystery associated with radially continuous Maxwell viscoelastic structure. Geophysical Journal International, 123, 849 865. FARRELL, W. E. & CLARK, J. A. 1976. On postglacial sea level. Geophysical Journal of the Royal Astronomical Society, 46, 647-667. FORTE, A. M., DZIEWONSKI,A. M. & WOODWARD,R. L. 1993a. Aspherical structure of the mantle, tectonic plate motions, nonhydrostatic geoid and topography of the core-mantle boundary. In: LE MOUEL, J.-L. (ed.) Dynamics of the Earth's Deep Interior and Earth Rotation, Geophysical Monograph Series, American Geophysical Union, 72, 135-166. , PELTIER, W. R., DZIEWONSKI, A. M. & WOODWARD, R. L. 1993b. Earth's dynamics topography: a new interpretation based upon mantle flow models derived from seismic tomography. Geophysical Research Letters, 20, 225-228. HAM, D. & WAHR, J. 1995. The visco-elastic relaxation of a realistically stratified Earth, and a further analysis of postglacial rebound. Geophysical Journal International, 120, 287-311.
28
W. R. P E L T I E R
HAYS, J. D., IMBRIE, J. & SHACKLETON, N. J. 1976. Variations in the Earth's orbit: pacemaker of the ice ages. Science, 194, 1121-1132. IMBRIE, J., HAYS, J. D. et al. 1984. The orbital theory of Pleistocene climate: support from a revised chronology of the marine 5180 record. In: BERGER, A. L. et al. (eds) Milankovitch and Climate, Part 1, D. Reidel Publishing Company, 269-305. JACKSON, D. D. & MATSU'URA, J. 1985. A Bayesian approach to nonlinear inversion. Journal of Geophysical Research, 90, 581-591. KARATO, S.-I. & WU, P. 1993. Rheology of the upper mantle: a synthesis. Science, 260, 771-778. LAMBECK, K., JOHNSTON, P. & NAKADA, M. 1990. Holocene glacial rebound and sea-level change in NW Europe. Geophysical Journal International, 103, 451-468. , SMITHER, C. & NAKADA, M. 1996. Glacial rebound of the British Isles III: Constraints on mantle viscosity. Geophysical Journal International 125, 340-354. MCCONNELL, R. K. 1968. Viscosity of the mantle from relaxation time spectra of isostatic adjustment. Journal of Geophysical Research, 73, 7089-7105. MITROVICA, J. X. 1996. Haskell [1935] revisited, Journal of Geophysical Research, 101, 555-569. & FORTE, A. M. 1997. Radial profile of mantle viscosity: results from the joint inversion of convection and postglacial rebound observables. Geophysical Journal International, 102, 2751-2769. & PELTIER, W. R. 1989. Pleistocene deglaciation and the global gravity field. Journal of Geophysical Research, 94, 13 651-13 671. & 1991. On postglacial geoid subsidence over the equatorial oceans. Journal of Geophysical Research, 96, 20 053-20 071. & 1995. Constraints on mantle viscosity based upon the inversion of post-glacial uplift data from the Hudson Bay region. Geophysical Journal International, 122, 353-377. OTA, Y., CHAPPELL, J., KELLEY, R., YONEKURA, N., MATSUMOTO, E., NISHIMURA,T. & HEAD, J. 1993. Holocene coral reef terraces and coseismic uplift at Huon Peninsula, Papua New Guinea. Quaternary Research, 40, 177-188. PARI, G. & PELTIER, W. R. 1995. The heat flow constraint on mantle tomography-based convection models: towards a geodynamically selfconsistent inference of mantle viscosity. Journal of Geophysical Research, 100, 12 731-12 751. PELTIER, W. R. 1974. The impulse response of a Maxwell Earth. Reviews of Geophysics and Space Physics, 12, 649-669. - - 1 9 7 6 . Glacial isostatic adjustment II. The inverse problem. Geophysicals Journal of the Royal Astronomical Society, 46, 669-706. 1982. Dynamics of the ice-age Earth. Advances in Geophysics, 24, 1-146. --1983. Constraint on deep mantle viscosity from LAGEOS acceleration data. Nature, 304, 434-436. -
-
-
-
-
-
1985. The LAGEOS constraint on deep mantle viscosity: results from a new normal mode method for the inversion of viscoelastic relaxation spectra. Journal of Geophysical Research, 90, 9411-9421. 1994. Ice-Age paleotopography. Science, 265, 195-201. -1995. Paleotopography of Ice-Age ice sheets. Science, 267, 536-538. --1996. Mantle viscosity and ice-age ice sheet topography. Science, 273, 1359-1364. - - 1 9 9 7 . Correction to the paper 'Glacial isostatic adjustment and earth rotation: refined constraints on the viscosity of the lower mantle', by W. R. Peltier and Xianhua Jiang, Journal of Geophysical Research, 102, 10 101-10 103. 1998a. Global sea level rise and glacial isostatic adjustment. Global and Planetary Change, in press. - - 1 9 9 8 b . The inverse problem for mantle viscosity. Inverse Problems, in press. - - 1 9 9 8 c . 'Implicit Ice' in the global viscoelastic and topographically self-consistent theory of glacial isostatic adjustment: implications for ice-age paleo-topography, Geophysical Research Letters, in press. & ANDREWS, J. T. 1976. Glacial isostatic adjustment I. The forward problem. Geophysical Journal of the Royal Astronomical Society, 46, 605-646. & JIANG, X. 1996a. Glacial isostatic adjustment and earth rotation: refined constrains on the viscosity of the deepest mantle. Journal of Geophysical Research, 101, 3269-3290. & 1996b. Mantle viscosity from the simultaneous inversion of multiple data sets pertaining to postglacial rebound. Geophysical Research Letters, 23, 503-506. & -1997. Mantle viscosity, glacial isostatic adjustment and the eustatic level of the sea. Surveys in Geophysics, 18, 239-277. - - - , FARRELL, W. E. & CLARK, J. A. 1978. Glacial isostasy and relative sea level: a global finite element model. Tectonophysics, 50, 81-110. - - - , FORTE, A. M., MITROVICA,J. X. & DZIEWONSKI, A. M. 1992. Earth's gravitational field: seismic tomography resolves the enigma of the Laurentian anomaly. Geophysical Research Letters, 19, 1555-1558. --, HILLAIRE-MARCEL, C. & PARI, G. 1998. Postglacial rebound at Richmond Gulf (Quebec), mantle viscosity, and the free air gravity anomaly over Hudson Bay, Nature, submitted. PLATZMAN,G. W. 1971. Ocean tides and related waves. In: REID, W. A. (ed.) Mathematical Problems in -
-
-
-
-
-
-
-
the Geophysical Sciences 2. Inverse Problems, Dynamo Theory and Tides. Lectures in Applied Mathematics, American Mathematical Society, 14, pp. 239-291, SHACKLETON, N. J. 1967. Oxygen isotope analyses and Pleistocene temperatures re-addressed. Nature, 215, 15-17. --, BERGER, A. & PELTIER, W. R. 1990. Transac-
tions of the Royal Society of Edinburgh. Earth Sciences, 81,251-261.
GLACIAL ISOSTATIC ADJUSTMENT AND COASTAL TECTONICS SIMONS, M. & HAGER, B. 1997. Localization of the gravity field and the signature of glacial rebound. Nature, 390, 500-504. STUIVER, M. & REIMER, R. J. 1993. Extended 14C data base and revised Calib. 3.0 14C age calibration program. Radiocarbon, 35, 215-230. TARANTOLA,A. & VALETTE,B. 1982. Inverse problems = quest for information. Journal of Geophysics, 50, 159-170. & 1984. Generalized non-linear inverse problems solved using the least squares criterion. Reviews of Geophysics and Space Physics, 20, 219-232. TARASOV, L. & PELTIER, W. R. 1997. Terminating the 100kyr ice-age cycle. Journal of Geophysical Research-Atmospheres, 102, D18, 21 665-21 693. TUSHINGHAM, A. M. & PELTIER, W. R. 1991. ICE-3G: a new global model of late Pleistocene deglaciation based upon geophysical predictions of postglacial relative sea level change. Journal of Geophysical Research, 96, 4497-4523.
--
-
29
& 1992. Validation of the ICE-3G model of Wfirm-Wisconsin deglaciation using a global data base of relative sea level histories. Journal of Geophysical Research, 97, 3285-3304. Wu, P. & PELTIER, W. R. 1983. Glacial isostatic adjustment and the free air gravity anomaly as a constraint on deep mantle viscosity. Geophysical Journal of the Royal Astronomical Society, 74, 377-449. & 1984. Pleistocene deglaciation and the Earth's rotation: a new analysis. Geophysical Journal of the Royal Astronomical Society, 76, 202-242. YODER, C. F., WILLIAMS, J. G., DICKEY, J. 0., SCHUTZ, B. E., EANES, R. J. & TAPLEY, B. D. 1983. Secular variation of the Earth's gravitational harmonic J2 coefficient from LAGEOS and non-tidal acceleration of Earth rotation. Nature, 303, 757-762. -
Decoupling post-glacial tectonism and eustasy at Huon Peninsula, Papua New Guinea JOHN
C H A P P E L L 1, Y O K O
OTA 2 & COLIN
CAMPBELL 3
1Research School of Earth Sciences, Australian National University, Canberra, A.C.T., Australia (e-mail:
[email protected]) 2Department of Geography, Senshu University, Kawasaki, Japan 3 Research School of Pacific and Asian Studies, Australian National University, Canberra, A.C.T., Australia Abstract: Late Quaternary uplift of coral terraces varies along the coast at northeast Huon Peninsula, Papua New Guinea, but the uplift rate has been assumed to be constant at any given locality in previous studies. Measurements indicate that rates for the last 7000 and 120 000 years were similar but this may be coincidence, because uplift at Huon Peninsula is dominated by isolated, metre-scale events with recurrence intervals around 1000 years. Using age-height data from 54 corals from the post-glacial reef, we examine the uplift rate over the last 13 000 years near Kwambu, where facies changes in a drill core indicate several uplift events before 7 ka BP. To separate uplift from sea level, a eustatic curve for Kwambu was generated by the global sea-level model described by Lambeck, recalibrated to new, Late Pleistocene sea-level data. With Barbados as a test case, predictions compare well with observations reported earlier, but predicted sea levels for Kwambu cannot be reconciled with the coral data unless the water depth of coral growth at the site was greater than estimated previously.
Two important questions in regions of plate convergence are whether vertical movements are intermittent, and whether the mean rate is constant or varies on 104-105 year time scales. Intermittent uplift indicates that vertical movement is associated with large earthquakes and probably with fault movement; variation of uplift rate can reveal something about the way in which the geometry of relatively shallow structures changes in response to plate motion. Constancy of uplift over 105 years usually is assessed by comparing uplift rates based on raised mid-Holocene and Last Interglacial shorelines, because eustatic sea levels are believed to have been similar, within a few metres, at those times (Chappell & Veeh 1978). Variations over the last 104 years can be evaluated from postglacial shorelines provided that local eustatic changes during this period are known. In fact, it has been shown that Holocene uplift has been intermittent and dominated by metre-scale events, with recurrence intervals of hundreds to thousands of years, in many Pacific countries including Japan, New Zealand, Papua New Guinea and the US west coast (Ota 1991; Berryman 1993; Ota et al. 1991, 1993; Merritts 1996). Similar coseismic uplifts were identified from Late Pleistocene regressive terraces at H u o n Peninsula, Papua New Guinea (Chappell et al. 1996b). In this paper we attempt to separate postglacial uplift and eustatic changes at Huon
Peninsula, and to identify whether uplift rate has varied significantly in the last 13 000 years. Post-glacial sea-level changes at H u o n Peninsula were derived by Chappell & Polach (1991) and Ota & Chappell (1998), on the basis of radiocarbon age measurements from post-glacial coral reefs including observations from a 52m drill core. These studies assumed that the local uplift rate at each sampling site has been constant since the Last Interglacial (119-126 ka). This assumption may be faulty, because Ota et al. (1993) found that the uplift rate at Kwambu on the central Huon coast was about l m p e r l 0 0 0 years for the last 2000 years but averaged about 2 m p e r 1000 years for the last 6000 years. This may be a statistical artefact, however, because uplift at Huon Peninsula is dominated by metre-scale events with an average recurrence interval around 1000 years (Ota et al. 1993; Chappell et al. 1996b). By extending the record to 13 000 years, we attempt to resolve this question.
Methods and data This study is based on age-height and age-depth data from the raised post-glacial barrier reef near Kwambu, Huon Peninsula (Fig. 1). The barrier reef was formed during Post-glacial sea level rise, which culminated locally at about 6-6.514Ckabp (6.8-7.3 cal. kaBP).
CHAPPELL,J., OTA, Y. & CAMPBELL,C. 1998. Decoupling post-glacial tectonism and eustasy at Huon Peninsula, Papua New Guinea. In: SXEWA~a',I. S. & VrrA-FINz~, C. (eds) Coastal Tectonics, Geological Society, London, Special Publications, 146, 31-40.
32
J. C H A P P E L L E T
AL.
Fig. 1. Site locations on raised Holocene coral reef near Kwambu, Huon Peninsula, Papua New Guinea. Drill site is location of drill-hole data reported by Chappell & Polach (1991); base of Holocene reef on section XY is inferred to be at 70 m depth from more recent drilling at same site (F. Taylor, pers. comm.). Numbers 1-7 are sample sites of Ota et al. (1993); K is Kilasairo cliff site of Chappell & Polach (1976).
The reef has emerged in the last 6000 years, in a series of metre-scale coseismic uplift events (Ota et al. 1993). We use age-height and age-depth measurements of corals in the transgressive barrier reef from a 52 m drill core reported by Chappell & Polach (1991) and from surface exposures reviewed by Ota & Chappell (1998), together with data from post-6 ka BP regressive terraces cut into the raised barrier reef (Ota et al. 1993). The average uplift rate at Kwambu, based on thermal ion mass spectrometry scanning (TIMS) U-series age measurements of the crest of the Last Interglacial reef (reef VII), is 1.9mper 1000 years (Stein et al. 1993).
Table 1 lists age-height data from outcrops and regressive terraces; Table 2 lists age, depth and facies information for the drill core. Conventional radiocarbon ages were measured in the ANU Radiocarbon Laboratory and were adjusted for the local 400 year seawater reservoir correction (Chappell & Polach 1991). Adjusted radiocarbon ages were converted to calibrated ages (Stuiver & Reimer 1993) with the OxCal program (Bronk Ramsey 1994). Sample heights were surveyed to tide levels and normalized to mean sea level using tide tables for Dreger Harbour, 80 km southeast of Kwambu, which has a similar tide. Water depths in which the corals grew are discussed later,
T E C T O N I S M A N D EUSTASY, H U O N
PENINSULA
33
Table 1. Holocene outcrop samples from Kwambu area, in ascending age order Sample code*
Context~
A5686 W5 W1 W20 W18 W29 A8670 A5685 W15 W3 W2 W30 A5687 A6119 W14 A1250 W32 W10 W9 W31 A5689 W11 A5683 W19 A1249 A1253 A1248 A1252 A 1251
reg. ter. reg. ter. reg. ter. reg. ter. reg. ter. reg. ter. reg. ter. reg. ter. reg. reef reg. reef reg. reef reg. reef crest crest transg. transg. transg. transg. transg. transg. transg. transg. transg. transg. transg. transg. transg. transg. transg.
Height (m amsl):~
Adjusted 14C age (ka BP)w
Calibrated age
2a 2a 2a 2a 2b 5a 8.5 9.5 4.6 7.4 6.7 7.1 13.2 13.0 4.6 8.0 6.7 6.9 8.7 4.5 6.1 5.1 4.4 5.0 3.1 0.1 1.9 0.8 1.5
2330 2450 2490 2510 3420 3880 4430 5310 5560 5630 5690 5800 5800 6070 6190 6210 6390 6400 6420 6600 6740 6750 6780 7190 7410 7430 7580 7740 7780
2.1-2.6 2.4-2.7 2.4-2.7 2.4-2.8 3.5-3.8 3.9-4.4 4.9-5.3 5.9-6.3 6.2-6.6 6.3-6.6 6.3-6.7 6.4-6.8 6.4-6.8 6.8-7.2 6.9-7.3 6.8-7.2 7.0-7.5 7.0-7.5 7.0-7.5 7.3-7.6 7.4-7.7 7.4-7.8 7.4-7.8 7.8-8.1 7.9-8.4 8.0-8.4 8.1-8.6 8.2-9.0 8.4-9.0
(ka BP)
* Samples were first reported by Chappell & Polach (1976), Ota et al. (1993) and Ota & Chappell (1998); A-codes dated at Australian National University, W-codes dated at Nagoya University. t Context: reg. ter., regressive terrace described by Ota et al. (1993); reg. reef, regressive reef and transg., transgressive reef (last two as defined by Chappell et al. (1996b)). Heights of samples are adjusted because those marked a are from raised surf benches, which form at or above high tide level, and b is from coral heads which grew below low tide level (see Ota et al. 1993). wConventional radiocarbon age with local seawater correction of 400 years subtracted (see Chappell & Polach 1991). Age errors are given in original sources but most are around +80 years at 1 SD.
based on stratigraphic and faunal data summarized in Table 2. To investigate possible variations of uplift, we required a post-glacial eustatic curve for Huon Peninsula that was not based on Huon Peninsula data. High-precision observations from tectonically stable sites in the region do not extend much beyond 6-7kaBP and we have made use of data from Barbados, reported by Fairbanks (1989). Recognizing that eustatic changes at Barbados will differ from those at Huon Peninsula, owing to global post-glacial isostasy, we placed the two sites on an equivalent footing by generating local eustatic curves from a model of Late Quaternary ice volume changes, using the global glacio- and hydro-isostatic procedure described by Nakada & Lambeck (1987). We followed Lambeck (1996) in deriving ice volume changes from isotopic sea levels of Shackleton (1987) but also used
late Pleistocene sea-level data reported by Chappell et al. (1996a) to calibrate the ice volume changes. This task ranged well beyond the scope of the present paper and full details will be published elsewhere. The predicted post-glacial sea-level curve for Barbados matches the observations well (Fig. 2) and thus we expect that sea-level predictions for Kwambu should match values calculated from the coral data, provided that assumptions about uplift rate and estimates of water depth of coral growth are correct.
Stratigraphy The raised post-glacial reef at K w a m b u has a long, n a r r o w lagoon with a barrier reef to seaward
J. CHAPPELL E T AL.
34
Table 2. Kwambu drill-core facies and age data
Depth below core top (m)
1.2 2.5 3.4 4.4 5.7 6.2 7.7 9.0 10.6 11.0 14.2 15.4 17.4 .
.
.
.
.
.
.
.
.
.
.
.
.
.
Dominant taxa*
Coral taphonomyf
Foram guild index:~
Matrixw
Por., Acrop. Acrop Por., Acrop. Pocil. Por. Por., Acrop. Acrop.
b,b&r, C b&r, M broken, C i.s, bored, M
5 4
broken, S i.s, bored, A b, bored, A
Low-rag mud Ar/mgCa sand CorFrag CorFrag Ar/mgCa sand Ar/mgCa mud mgCa cem/mud mgCa/Ar sand
Por. Hal., mol Por., Acrop Por., A. hyac. Acrop., F. pal.
i.s, S broken, S broken, C i.s, bored, C i.s, bored, A
.
.
.
.
.
.
.
.
.
.
Acrop., Pocil Por., F. pal. Gonio. Monti. Monti.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19.5 20.2 21.3 22.8 23.7 24.5 25.3 26.1 29.2 30.6 32.0 33.1 34.0
Monti. Por., Monti Por., Hal., Monti., Sty. Por., Acrop Por., Acrop Por.
broken, S i.s, bored, M e.s, A i.s, bored, C i.s, bored, C cor. frags, M e.s, bored, A broken, C i.s, bored, C b&b, M broken, M broken, (,M i.s, C
36.8 38.4 39.8 42.5 42.7 45.2 50.0 51.4
Por., F. pal. Por., Pocil Por., Cyph Acrop., F. pal. Monti., Cyph Por., T. mus Pocil. Monti.
e.s, M i.s, C i.s, A broken, M i.s & e.s, M i.s & e.s, C i.s, M e.s, M
2 0.6 1.5 1.5 2.5 1.3
.
.
.
.
.
.
.
1 2 1.5 1.5 1.5 0.6 0.6 0.6 0.8 3 6 0.3 2
1.5 0.7
.
7.6-7.9 8.0-8.3 8.2-8.4 8.4-8.7 8.3-8.6 8.5-9.0
Ar/mgCa sand Ar/mgCa sand mgCa/Ar sand mgCa cem/sand mgCa cem
1 1.4 4 .
Calib. age (ka BP)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Ar/mgCa sand mgCa mud Ar/mgCa mud mgCa mud mgCa cem mgCa/Ar mud mgCa cem mgCa cem Ar/mgCa sand Ar/mgCa mud mgCa cem mgCa cem mgCa cem/sand mgCa mud/sand mud/sand ] / II
8.3-8.7 9.0-9.4 9.4-9.6 9.4-9.8 .
.
.
.
.
.
.
.
.
.
.
.
.
.
9.4-9.8 9.6-10.0 9.8-10.0 9.6-10.0 10.0-10.4 9.9-10.2 10.1-10.6 10.3-10.9 10.3-10.9 10.3-10.9 10.0-10.4 10.5-11.0 10.9-11.7 11.0-11.7 12.5-12.8 12.0-12.5 12.0-12.6 12.4-12 12.8-13.2
* Dominant taxa: Acrop., Acropora; A. hyac., A. hyacinthus; Cyph., Cyphastrea; F. pal., Favia pallida; Monti, Montipora; Pocil., Pocillopora; Por., Porites; Sty., Stylaster; T. mus., Tubipora musica; Hal., Halimeda; mol, molluscs. t Coral taphonomy: b,b&r, bored, broken and rolled; b&r, broken and rolled; i.s, in situ; e.s, ex situ but not obviously broken. Coral abundance scale: A, abundant, >70% of core section; C, common, 50-70%; M, moderate, 20-50%; S, sparse, merely present to 20%; N, none. Foram guild index = ratio of mobile : sedimented species (see text). wMatrix: mgCa, magnesium calcite, Mg typically 10-20%; Ar, aragonite; CorFrag, coral fragments; cem, cement. II Matrix rarely recovered below 39 m owing to change of drilling technique. Horizontal dotted line indicates facies break. and a fringing reef to landward, and laps against Late Pleistocene coral limestone (Fig. 1). The fringing reef platform rises to 13 m above m e a n sea level (amsl) and the barrier crest varies from 6 to 9 m amsl. Drilling in 1988 on the K w a m b u barrier penetrated 5 2 m into post-glacial coral limestone (Chappell & Polach 1991) and fur-
ther drilling at the same site in 1996-1997 showed that the post-glacial reef is about 7 0 m thick (F. Taylor, pers. comm.). Exposures further southeast indicate that the post-glacial structure wedges out against the Pleistocene raised reef sequence and this is expected to be the case at K w a m b u , also.
TECTONISM AND EUSTASY, HUON
PENINSULA
35
Fig. 2. Predicted eustatic curve for Barbados generated by Nakada-Lambeck global glacio-eustatic model (open circles: for details, see text), and uplift-corrected observations from Fairbanks (1989) with time scale in Th-U years from Bard et al. (1990).
Exposures in coastal cliffs, river-cuts and borrow pits show that the post-glacial reef is composed of several facies: coarse coral limestone of the reef crest and upper forereef, coralgal limestone of the reef platform, lagoon deposits of bioclastic grainstone containing dispersed corals, and coral limestone containing arborescent and less robust corals of the lagoon fringing reef. Each facies has its own characteristic fabric and guilds of corals and molluscs (Pandolfi & Chappell 1994). Stacked shallow-water coral deposits exposed in coastal cliffs, particularly at Kilasairo Stream, 2km southeast of the drill-hole site, suggest that reef growth kept pace with rising sea level (Chappell & Polach 1976; Pandolfi & Chappell 1994). Coral limestone throughout the drill core from the Kwambu barier was interpreted by Chappell & Polach (1991) as crest or upper forereef deposits of a shallow water 'keep-up' reef that kept pace with rising post-glacial sea level. However, compared with cliff exposures, the drill core provides only a small sample of reef material and the water depth of reef growth is difficult to gauge. Corals in the core were identified by J. E. N. Veron, who noted that they are common types of the reef crest and upper forereef: dominantly Porites, Acropora, Montipora, Pocillopora and Favids. Transitions occur in the core from assemblages dominated by strong, upward-branching corals, to assemblages with massive and tabular corals surrounded by coral and algal debris,
which suggest changes of habitat and possibly water depth. Variations of foraminiferal assemblages, matrix and cement within the coral limestone also may indicate changes of habitat and water depth. Some 43 taxa of benthic forams were identified in the core (16 Miliolinae, 25 Rotaliinae, 2 Textulariinae). The number of taxa per sample range from six to 19 and the proportions of foraminiferal guilds vary markedly (an ecological guild is a grouping of organisms which employ similar life strategies, regardless of taxonomic relationship). In particular, we found significant variations in the ratio of the guild of mobile species to the guild of species restricted to sediments, which appear to correspond to other changes in the core. Variations of this particular index are listed in Table 2 along with major coral taxa and summaries indicating composition of the coral limestone matrix. Changes of facies at about 10 m, 18 m and 34 m, shown as dotted lines across Table 2 represent simultaneous occurrences of a sharp up-core increase of the foram guild index, an increase of matrix cement and an increased incidence of bored corals with heavy algal rinds.
Reef growth, uplift and sea-level rise On the basis of the drill-core data, Chappell & Polach (1991) estimated that the coral crest was generally 2-5 m below sea level throughout, and considered that the age-depth data define the
36
J. CHAPPELL E T AL.
course of relative sea-level rise at Kwambu. Using seawater-corrected conventional radiocarbon ages, Chappell & Polach (1991) adjusted the age-height data for uniform uplift at 1.9mper 1000 years, and suggested that the results represent the local post-glacial eustatic curve. Ota & Chappell (1998) interpreted post-glacial sea-level changes in the same way but used calibrated radiocarbon ages and a larger dataset, which includes observations from other Holocene sites at Huon Peninsula. These workers included the Chappell & Polach estimates of water depths in deriving local eustatic sea levels, but did not attempt to identify the effects of intermittent coseimic uplift. Age-height data from Tables 1 and 2 are plotted in Fig. 3 using the calibrated radiocarbon time scale, together with eustatic predictions for Kwambu at the same time-points as the coral samples, derived from the global model that generated the Barbados curve shown in Fig. 2. Also plotted are differences between observed heights and predicted sea levels, and a sloping line representing uniform uplift at 1.9m per 1000 years. The height (H) relative to present sea level of a coral depends on sea level (S), the depth of water (W) when it grew, and the uplift history: H=S-
W+*U.d*
(1)
where uplift U is integrated from the present (0) to the calibrated age of the coral (t). Previous workers assumed that U is constant, i.e., *U.d*=Ut and therefore H - S = U t - W . It follows that values (H - S) should lie distance W below the uplift line in Fig. 3. However, most points around 10-12kaBP are offset from the uplift line by over 10 m and range to 18 m, which is much larger than the 2-5 m estimate of W by Chappell & Polach (1991). It is possible that W ranged to 18m, or U was not constant, or the predicted sea levels are wrong. Errors in predicted sea levels cannot be ruled out but we consider it unlikely that they range to 18 m and, to account for the differences between ( H - S) and the uplift line, we examine the effects of variable uplift rate and water depth by holding one or the other constant. Thus, the apparent uplift rate (Uw) at constant water depth, over time interval *t = tl - t2 can be calculated by assuming that the predicted sea levels are correct: Uw = (112 -
$2 -
H=
(2)
+ S1)/*t
To examine variation of apparent uplift over the last 13 000 years we calculated Uw over short intervals by stepping through the age-height
thousand years before present 0
2
4
6
10
8
12
14
o
1~I
9
.
~ "::%.
~
0
-10
-20
E -30
r'l
%un00 i DO
-40
-50
-60
o
-70
Fig. 3. Age-height and age-depth plot showing coral data listed in Tables 1 and 2 (O) compared with predicted eustatic sea levels at Kwambu for the same time points ([~). O, show height differences between corals and predicted sea levels; sloping line represents uniform uplift at 1.9 m per 100 years.
TECTONISM AND EUSTASY, HUON 0
2
4
6
I ....
t ....
PENINSULA
8
37
10
12
14
10
r/)
E
,
~
*1
. . . . .
I.
.;
9
\;-.
....
-5
-10
-15
thousand years B P. Fig. 4. Apparent uplift rates with water depth assumed constant, assessed over short time intervals between paired sample age-height data (smoothed).
data. Samples were ranked by stratigraphic depth and data were smoothed until moving average ages increased with moving average depth. Uw values for adjacent pairs in the smoothed series are plotted in Fig. 4 and show considerable variation, with several negative peaks between 7 and 11 kaBP. Similarly, the
apparent water depth (Wu) with constant uplift rate is calculated: Wu = S -
H+
(3)
Ut
Results plotted in Fig. 5 again show considerable variation, with highest values around 10-11
k a BP.
o
o
14
o1: o
12
;o
o
t0
.g
e
E
o o
o1:
o
F o i
2
.....
~
oo
o
o i
4
o
~o
1"~ 1!
-
i
i
~
6
8
10
.......................
i
12
,,,
14
thousand of years before present
Fig. 5. Apparent palaeo water depths with uplift rate assumed constant, for samples from Tables 1 and 2. Short vertical bars signify episodes when apparent water depth suddenly decreased, perhaps signifying uplift events.
38
J. CHAPPELL E T AL.
Discussion Uplift v. w a t e r depth
Estimates of uplift rate and palaeo water depth obviously are not independent. However, extreme values of Uw and Wu are not equally compatible with stratigraphic data from the study site. Anomalous values of Uw suggest that the assumption of constant water depth is faulty; anomalous Wu values suggest that uplift rate varied. Negative values of Uw in Fig. 4 are anomalous because they imply subsidence, for which no evidence has been found at northeast Huon Peninsula. Uplift has been recorded after historical earthquakes but not subsidence (Pandolfi & Chappell 1994). We are confident that prehistoric subsidence would be recorded in coastal land-forms and reef deposits. Detailed studies at more than 20 sites revealed excellent evidence for repeated uplifts during the last 6000 years but no trace of subsidence was found, nor were any subsidences recognized in the detailed record of uplift events for the interval 3 0 - 5 5 k a a P (Ota et al. 1993; Chappell et al. 1996b). It seems unlikely that processes should have been very different, 6000-10 000 years ago, and we dismiss the assumption that water depth was constant. Turning to apparent water depth, most values of Wu between 2 and 7 ka BP are 2-4 m (Fig. 5). These clearly are anomalous because the samples are of attached corals on regressive terraces, formed as intertidal platforms or surf benches, where the real water depth of coral growth was very close to zero. The anomalies reflect the fact that uplift was not uniform but proceeded in isolated, metre-scale events, which are reflected by sharp vertical steps in the Wu series: three occur within the last 7000 years and correspond to uplift events at Kwambu identified by Ota et al. (1993), namely, KK1 at 2.5kaBP; KK5 at 6.2kaBP and KK6 at 6.9 ka BP. Three similar steps occur earlier in the Wu series, at 8.5, 9.6 and 10.5 ka BP We suggest that these also represent uplift events and note that significant facies changes occur in the drill core, close to these times (Table 2). Furthermore, the recurrence interval between these events is close to 1000 years, similar to that of the last 6000-7000 years on the northeast Huon coast (see Chappell et al. 1996b). The question of whether uplift was statistically uniform, though proceeding by sudden events, cannot be answered from the apparent water depths, because the high Wu values prior to 9.5 ka may be realistic. Water depths of 10-15 m are not incompatible with the observed coral
facies in the lower half of the core, which, except for an apparent event at 35 m, is dominated from 24 to 40 m by ramose growth forms with a uniform matrix of high-magnesium calcite muddy sand (Table 2).
C a t c h - u p or k e e p - u p reefs?
The reef at Kwambu was established before 13kaBP and ceased growth around 7kaaP, when it emerged owing to uplift. According to Fig. 5, water depth over the reef at the drill-hole site increased after the reef was established, reached a maximum about 11 000 years ago and has decreased, apparently stepwise, since 10.5kaBP Perhaps rising sea level at first outpaced reef growth but later the reef later caught up, assisted by either a decrease of the rate of eustatic sea-level rise or an increase of uplift rate. However, our drill hole does not necessarily trace the highest point of the growing reef; therefore we cannot identify whether reef growth really has accelerated since about 10kaBP and, if so, what caused it to do so. The question would be resolved with a transect of drill holes, landwards of the drill site on the Kwambu barrier. By sampling basal deposits from the transgressive reef, which should climb to landward with rising sea level, there would be less uncertainty about palaeo water depths. Sea-level predictions thus would be tested more sharply.
Conclusions The Late Quaternary uplift rate of coral terraces at northeast Huon Peninsula, Papua New Guinea, has been assumed by previous workers to have been constant at any given locality, although the mean rate increases southeastwards along the terraced coast. At several sites, measurements show that the local mean uplift rate for the last 6000-7000 years is very similar to that for the last 120 000 years, but this may be coincidence. The 7000 year rate is a statistical average, because uplift at Huon Peninsula is dominated by sudden events of up to several metres with an average recurrence interval around 1000 years, which are considered to be coseismic and were identified and dated from small regressive terraces (Ota et al. 1993; Chappell et al. 1996b). We have investigated the uplift rate over the last 13 000 years at the Kwambu-Kilasairo site, where the rate is known to have diminished during the last 6000 years (Ota et al. 1993), with age-height (or age-depth) measurements
T E C T O N I S M A N D EUSTASY, H U O N of 54 corals from the post-glacial reef structure. By examining uplift and palaeo water depth separately, we conclude that there is no evidence that the uplift rate during the last 13 000 years was significantly different from the 7000 year or 120000 year averages. Variations of a p p a r e n t water depth indicate coseismic uplift events before 7 ka BP, similar to those previously identified for the last 7000 years, at 8.3, 9.5 and 10.3 ka ~p The analysis rests on a eustatic curve for K w a m b u generated by the global m o d e l described by L a m b e c k (1996), recalibrated to new, Late Pleistocene sea-level data. The m o d e l performs well against observations from Barbados reported by F a i r b a n k s (1989) but should be verified at other sites. N o n e the less, the estimate by Chappell & Polach (1991) of 2 - 5 m water depth t h r o u g h o u t the K w a m b u drill-hole record is both imprecise and too restrictive, and it is clear that better accuracy is necessary in studies of this kind. So far, this appears to have been better achieved for Caribbean than for Indo-Pacific reefs, but we suggest that better definition at K w a m b u w o u l d be obtained by extending a transect of drill holes landwards of the single hole, drilled previously. Finally, we note that the H u o n Peninsula drill-hole data examined here have been used previously to infer variations of the rate of postglacial sea-level rise, possibly related to the Y o u n g e r D r y a s event (Edwards et al. 1993). Given the uncertainties about water depth and the episodic nature of uplift at K w a m b u , together with the likelihood that the drill hole does not pass continuously t h r o u g h the highest growing point of the reef, it appears to us that variations of the rate of sea-level rise should not be derived from these data. We thank K. Lambeck, K. Smithers and K. Fleming for their collaboration with the sea-level model used here, which will be reported in full elsewhere.
R e f e r e n c e s
BARD, E., HAMELIN, B., FAIRBANKS, R. & ZINDLER, A. 1990. Calibration of ~4C time scale over the past 30,000 years using mass spectrometric Th-U ages from Barbados corals. Nature, 346, 456-458. BERRYMAN, K. R. 1993. Age, height, and deformation of Holocene marine terraces at Mahia Peninsula, Hikurangi Subduction margin, New Zealand. Tectonics, 12, 1347-1364. BRONK RAMSEY, C. 1994. Radiocarbon calibration and analysis of stratigraphy: the OxCal Program. Radiocarbon, 37, 425 430.
PENINSULA
39
CHAPPELL, J. & POLACH, H. A. 1976. Holocene sea level change and coral-reef growth at Huon Peninsula, Papua New Guinea. Geological Society of America Bulletin, 87, 235-240. - & POLACH, H. 1991. Post-glacial sea level rise from a coral record at Huon Peninsula, Papua New Guinea. Nature, 349, 147-149. -& VEEH, H. H. 1978. Late Quaternary tectonic movements and sea-level changes at Timor and Atauro Island. Geological Society of America Bulletin, 89, 356-368. --, OMURA, A., ESAT, T., MCCULLOCH, M., PANDOLFI, J., OTA, Y. & PILLANS, B. 1996a. Reconciliation of late Quaternary sea levels derived from coral terraces at Huon Peninsula with deep sea oxygen isotope records. Earth and Planetary Science Letters, 141,227-236. , OTA, Y. & BERRYMAN, K. R. 1996b. Holocene and late Pleistocene coseismic uplift of Huon Peninsula, Papua New Guinea. Quaternary Science Reviews, 15, 7-22. EDWARDS, R. L., BECK, J. W., BURR et al. 1993. A large drop in atmospheric 14C/12C and reduced melting in the Younger Dryas, documented with 23~ ages of corals. Science, 260, 962-968. FAIRBANKS, R. G. 1989. A 17,000 year glacio-eustatic sea level record: influence of glacial melting rates on the Younger Dryas and deep ocean circulation. Nature, 342, 637 642. LAMBECK, K. 1996. Sea-level change and shore-line evolution in Aegean Greece since Upper Palaeolithic time. Antiquity, 70, 588-611. MERRrrTs, D. 1996. The Mendocino triple junction: active faults, episodic coastal emergence, and rapid uplift. Journal of Geophysical Research, 101, 6051-6070. NAKADA, M. & LAMBECK, K. 1987. Glacial rebound and relative sea-level variations: a new appraisal.
Geophysical Journal of the Royal Astronomical Society, 90, 171-224. O-rA, Y. 1991. Coseismic uplift in coastal zones of the western Pacific rim and its implication for coastal evolution. Zeitschrift fiir Geomorphologie, N.F., Supplementband, 81, 163-179 - & CI4APPELL,J. 1998. Holocene sea-level rise and coral reef growth on a tectonically rising coast, Huon Peninsula, Papua New Guinea. Quaternary International, in press. -- - , KELLEY, R., YONEKURA,N., MATSUMOTO, i~., NISHIMURA, T. & HEAD, J. 1993. Holocene coral terraces and coseismic uplift of Huon Peninsula, Papua New Guinea. Quaternary Research, 40, 177-188. --, HULL, A. G. & BERRYMAN, K. R. 1991. Coseismic uplift of Holocene marine terraces in the Pakarae River area, eastern North Island, New Zealand, Quaternary Research, 35, 331 346. PANDOLFI, J. & CHAPPELL, J. 1994. Stratigraphy and relative sea level changes at the Kanzarua and Bobongara sections, Huon Peninsula, Papua New Guinea. In: OYa, Y. (ed.) Study on Coral
Reef Terraces of the Huon Peninsula, Papua
40
J. C H A P P E L L E T AL.
New Guinea-Establishment of Quaternary Sea Level and Tectonic History. Department of Geography, Senshu University, Kawasaki, 119-139. SRACKLETON, N. J. 1987. Oxygen isotopes, ice volume and sea level. Quaternary Science Reviews, 6 183-190. STE1N, M., WASSERBURG, G. J., AHARON, P., CHEN, J. H., ZHU, Z. R., BLOOM, A. L. & CHAPPELL, J.
1993. TIMS U-series dating and stable isotopes of the last interglacial event in Papua New Guinea. Geochimica et Cosmochimica Acta, 57, 2541-2554. STUIVER, M. & REIMER, P. J. 1993. Extended 14C data base and revised CALIB 3.0 14C age calibration program. In: STUIVER, M., LONG, A. & KRA, R. S. (eds) Calibration 1993. Radiocarbon, 35, 215-230.
Holocene uplift and subsidence of the Helike Delta, Gulf of Corinth, Greece STEVEN
SOTER
Department o f Astrophysics, American M u s e u m o f Natural History, Central P a r k West at 79th Street, N e w York, N Y 10024, U S A (e-mail."
[email protected]) Abstract: The southwestern coast of the Gulf of Corinth, known as Aigialeia, lies in a region
of rapid tectonic uplift and extension. Using age and elevation data from raised relic shorelines, and Lambeck's model for local isostatic sea-level rise, I re-examine the uplift of the coastal footwall block in Aigialeia. The average Holocene uplift rate is 2.4 -4-0.8 m ka -1 , significantly higher than the Quaternary uplift rates associated with the raised terraces near Corinth on the southeastern coast. The footwall movement in Aigialeia consists of coseismic uplift events separating periods of relatively aseismic uplift. A footwall uplift of about 2 m apparently accompanied the earthquake that destroyed and submerged ancient Helike 373 8c. The city was built on a Gilbert-type fan delta adjacent to the area of raised relic shorelines. Using dated samples from bore holes drilled in the delta, I estimate that the delta itself subsided by at least 3 m during the earthquake. The opposition between gradual uplift and coseismic subsidence events apparently resulted in a relatively small absolute net displacement of the delta during Holocene time.
The Gulf of Corinth is a marine basin about 105 km long and up to 30km wide, in central Greece (Fig. 1). It occupies an active asymmetrical rift zone (or half-graben), which undergoes north-south extension in connection with uplift of the northern Peloponnesos (Ori 1989; Armijo et al. 1996). The present rate of extension is about 13 mm/year near the western end of the Gulf, and decreases toward the east (Clarke et al. 1997). The southern margin of the rift zone is marked by a series of W N W trending normal extension faults dipping to the north (Doutsos & Poulimenos 1992; Rigo et al. 1996; Roberts & Koukouvelas 1996). A major onshore element of this tectonic system is the Helike Fault, which has been traced for about 40 km (Mouyaris et al. 1992; Poulimenos 1993; Dart et al. 1994; Stewart 1996; Stewart & Vita-Finzi 1996). During the evolution of the rift zone, seismic activity has progressively moved northward (Ori 1989; Dart et al. 1994) and is now concentrated in the Helike and Aigion Faults, and perhaps in other faults located to the north under the Gulf itself (Bernard et al. 1998). This fault system controls the almost linear southern shore of the Gulf of Corinth. In the western sector, modern footwall-derived Gilberttype fan deltas occur wherever rivers have incised northward through the uplifting mountains of the Peloponnesos (Seger & Alexander 1993; Collier & Gawthorpe 1995). The marine delta slopes are eroded by fissures and rotational slumping (Ferentinos et al. 1988). The coastline between the deltas is dominated by limestone conglomerate footwalls. The modern deltas in-
crease in size along this coast from east to west (the direction of increasing tectonic activity) and consist mainly of coarse-grained deposits. The largest of these deltas, here called the Helike Delta, extends for about 13 km between Aigion and Diakopto in the region of Aigialeia (Fig. 1). It is actually a coalesced fan delta, fed by the three rivers Selinous, Kerynites and Vouraikos. The Helike Delta lies on the hanging-wall block of the Helike Fault, which separates it from the footwall mountains to the south. These in turn consist of older (Plio-Pleistocene) uplifted fan deltas of the same kind (Ori 1989). The most prominent of these exhumed structures is the ancient Kerynites fan delta, whose uppermost topset beds have been elevated as much as 1200 m above sea level (Dart et al. 1994). The active Helike Delta has evolved under the influence of long-term tectonic uplift and earthquake-related subsidence. There is good historical evidence for the latter in the delta. In 373BC, a catastrophic earthquake and seismic sea wave destroyed and submerged Helike, then the principal city of ancient Achaea (Marinatos 1960; Soter & Katsonopoulou 1998). According to Pausanias, who visited the site on the Helike Delta around AD 174, the city was located about 7 km southeast of Aigion. In his description of the disaster, Pausanias (7.24.12) wrote that 'the sea flooded a great part of the land and encircled the whole of Helike. Moreover, the flood from the sea so covered the sacred grove of Poseidon that only the tops of the trees remained visible.' Pausanias is generally regarded as a reliable source. Assuming that the grove of
SOTER, S. 1998. Holocene uplift and subsidence of the Helike Delta, Gulf of Corinth, Greece. In: STEWART,I. S. & VITA-FINZI,C. (eds) Coastal Tectonics. Geological Society, London, Special Publications, 146, 41-56.
42
S. SOTER
Fig. l. Aigialeia, on the southwestern shore of the Gulf of Corinth. The Helike Delta, between Aigion and Diakopto, is part of the hanging-wall block of the Helike Fault. The Holocene rate of uplift of this block is calculated based on raised shorelines at Diakopto, Trapeza, Paralia Platanou and Aigeira (Table 1). Uplift is also found for similar sites on the Perachora Peninsula at the eastern end of the Gulf. The inset at lower left shows the central part of the Helike Delta with the locations of bore holes B1 B5 and B18. Seismic fault locations adapted from Koukouvelas & Doutsos (1996).
Poseidon was situated at least a metre above sea level and that the trees were at least 3 m high, this account suggests earthquake-related subsidence by at least 3 m. Pausanias goes on to say that in his time the ruined walls of Helike were still visible in the sea. Other ancient writers (Seneca, Strabo, Ovid) also mentioned the submerged ruins. Aigialeia lies in one of the most seismically active areas in the Mediterranean. In the last 300 years, 12 earthquakes with magnitudes estimated in the range 6-7 have occurred within 25 km of Aigion (Ambraseys & Jackson 1997). Macroscopic anomalies have been observed before earthquakes in this area (Soter 1998). On 26 December 1861, an earthquake of estimated magnitude 6.6 struck the Helike Delta (Ambraseys & Jackson 1997). Schmidt (1862, 1875) reported that the earthquake submerged a coastal fringe 13 km long and up to 200 m wide and left a 2 m scarp and fissure of the same length along the base of the foothills. He also found evidence of extensive soil liquefaction, particularly near the mouth of the Vouraikos River. Water and sand erupted from numerous fissures and 'sand volcanoes' up to 20 m across. One of the
eruptions was violent enough to kill a man working in the fields. Schmidt conjectured that the 1861 earthquake destabilized the entire coastal plain, causing it to slip seaward along the steeply inclined basement rock. He suggested that an event of the same nature but greater magnitude had destroyed ancient Helike in the same location. The mechanism he proposed for these earthquakes has some resemblance to the current view of normal faulting. Schmidt's account of what is now called the Helike Fault is one of the earliest scientific descriptions of a seismic fault. Other faults also contribute to the seismicity of the region. On 15 June 1995 an earthquake of magnitude 6.2 seriously damaged Aigion. Its epicentre was near Eratini on the northern shore of the Gulf (Bernard et al. 1998), but the shock caused widespread sediment failures on the Helike Delta, including submarine landslides, shoreline subsidence as a result of shallow rotational sliding, and soil liquefaction with sand blows through fissures and craters (Lekkas et al. 1996; Papatheodorou & Ferentinos 1996). Papatheodorou & Ferentinos (1996) suggested
UPLIFT AND SUBSIDENCE OF THE HELIKE DELTA that the earthquakes of 373 Bc, 1861 and 1995 all induced the liquefaction of a subsurface horizon, setting in motion the translation and subsidence of the overlying sediments. Koukouvelas & Doutsos (1996) identified the east-west trending Aigion Fault as the source of the 1995 earthquake. They traced its surface break on land for 7 km eastward across Aigion until it disappeared about 1.5kin before the shore. Our 1988 sonar survey (Soter 1998) shows that the Aigion Fault continues offshore for at least another 1 km. However, Bernard et al. (1998) suggested that the 1995 earthquake involved a fault located about 10kin ENE of the Aigion Fault. They concluded that the surface breaks on the Aigion Fault in 1995 were secondary features of the earthquake. In view of the historical evidence for subsidence of the Helike Delta, Soter & Katsonopoulou (1998) began a search for the site of ancient Helike in the sea southeast of Aigion. An extensive sub-bottom and sidescan sonar survey showed no unambiguous signs of a city on or under the sea floor. Accordingly, they shifted the search to the subaerial delta, where they drilled 60 bore holes to obtain sediment core profiles. In almost all of the bore holes located in the upper part of the delta between the Selinous and Kerynites Rivers, they found ceramic fragments in occupation horizons dating from Early Bronze Age through Classical and Roman to Byzantine times (Katsonopoulou & Soter 1997). One surprising result was that virtually all the occupation horizons, including the oldest ones, were located above present sea level. If this was in fact the site of ancient Helike, submerged in 373 Rc, then it appears that the delta was subsequently uplifted tectonically. The geological evidence for tectonic uplift along the northern coast of the Peloponnesos is compelling. From Corinth to Xylokastro, the erosion of an uplifting footwall modulated by Quaternary sea-level oscillations has produced an impressive flight of raised coastal terraces. These dated features show that uplift has proceeded at an average rate of order 1.4 m per 1000 years (m ka l) during the last 350 000 years (Kerauden et al. 1995; Armijo et al. 1996). In Aigialeia, on the southwestern shore of the Gulf, elevated erosion notches and emergent marine fauna provide evidence of Holocene elevation of the footwall behind the Helike Fault. Stewart & Vita-Finzi (1996) used the radiocarbon ages of these features to estimate the average rate of footwall uplift there at about 1.5 m ka -l . Here we re-examine this estimate, using Lambeck's (1995) model for local isostatic sea level and radiocarbon dates calibrated for the
43
reservoir effect of the Gulf of Corinth. The analysis invokes an absolute frame of reference to measure the changing elevation of local sea level and of the footwall block. This allows one to visualize the relationship between the present elevation of a relic shoreline and its absolute elevation when created. The data can then be used graphically to reconstruct a Holocene trajectory for the footwall. It then becomes possible, on the basis of the age and depth of selected core samples from bore holes drilled in the Helike Delta, to obtain a very rough estimate for the subsidence of the delta related to the earthquake of 373 BC.
Modelling absolute sea level and coastal uplift The local sea level is the sum of contributions representing the global eustatic sea level (determined by the amount of glacial ice), the local isostatic adjustment (related to the redistribution of mass between glacial ice and ocean water), and the local vertical tectonic movement of the shore. In the absence of vertical tectonic motion, the 'corrected' local sea level ~-c can be written as ~c = ~e + ~i
(1)
where Ce is the global eustatic sea level and ~i is the local glacio-hydro-isostatic correction, both of which are time dependent (see Lambeck 1995). We adopt the model eustatic sea-level curve of Peltier (1994, fig. 3B), shown by r in Fig. 2 for the last 12000 years. The curve shows the deceleration of sea-level rise around 7kaBP, and the cessation of net glacial melting at 5kaBp. The actual eustatic curve may have additional fine structure (see Blanchon & Shaw 1995) but as this has not been firmly established, I will retain the model of Peltier. Lambeck (1995, 1996) modelled the isostatic corrections for local sea level in Greece and Aegean Turkey during the last 20000 years. According to his results, the corrections appropriate for the Gulf of Corinth are almost identical to those for Kavalla in Thrace. As Lambeck provided a time series of r data for Kavalla (1995, figs. 4d and 6d), one can apply this correction factor to the Gulf of Corinth. Adding Ci to the eustatic curve G, one obtains the corrected local sea-level curve ~c shown in Fig. 2. The absolute frame of reference for these curves is fixed relative to the centre of the Earth, with zero elevation corresponding to the present local sea level.
44
S. SOTER
Fig. 2. Global eustatic sea level Ce from Peltier (1994) and the corrected local absolute sea level ~c for the Gulf of Corinth based on the glacio-hydro-isostatic model of Lambeck (1995). The corrected sea-level curve rises steadily during the last 5000 years, despite the flatness of the corresponding eustatic curve. This is due mainly to local isostatic subsidence of the mantle in response to the post-glacial uplift of distant Fennoscandia. Lambeck (1995) showed that in the absence of vertical tectonic movement, all the coastlines in Greece and Aegean Turkey would experience rising sea levels (i.e., coastal submergence) through post-glacial times. To illustrate the effect of vertical tectonic movement on local relative sea level, we can use a simplified schematic model. Let a vertical sea cliff move steadily upward parallel to itself at rate r. Its surface defines a relative frame of reference that moves vertically with respect to the absolute frame. Imagine a baseline scratched on the cliff face at the position that now coincides with sea level. At any time t the absolute elevation ~b of that baseline mark is ~b = - r t
(2)
where t is positive in the past. In Fig. 3 the line CBA represents the trajectory Cb of the cliff for the case r = 2 m k a -1. Point A locates the baseline at the present time t = 0. In reality, of course, the rate of tectonic uplift will not be constant over 10000 yeaers, particularly where abrupt seismic offsets occur, but the long-term average rate of uplift is still a useful quantity. Later, I will consider discontinuities in the rate of uplift. The curve CEA in Fig. 3 is the corrected sealevel trajectory r for the last l0 000 years, taken
Fig. 3. Model of the corrected local sea level s and the footwall baseline elevation Cb (assuming a constant uplift rate of 2 m per 1000 years). The present elevation h of a relic shoreline above sea level equals the difference between the absolute sea level and the footwall baseline at the time the shoreline was created. The curve HKA is thus the difference between the curves for r and Cb.
from Fig. 2 but smoothed with a ninth-order polynomial. Following the sea level and tectonic trajectories in time, we see that the baseline was submerged from about 9 ka BP (point C) to the present (A). Suppose that wave action and biological activity eroded a notch in the cliff at D when sea level was 5 m above the baseline, at B. Then with the steady ascent of the entire cliff, the notch follows the trajectory D F G , parallel to the baseline track CBA. For thousands of years the notch remains below sea level. But with the continued deceleration in sea-level rise, it eventually reappears above water (at F) and today will be 5 m above sea level (at G). The present absolute elevation A G of the notch thus equals the separation BD between baseline and sea level when the notch was created. One can express the present absolute elevation h of any relic shoreline feature on the cliff face (or footwall) as h -- r - Cb
(3)
the difference between the absolute sea level and the baseline at the time the shoreline feature was created. In the upper part of Fig. 3 the changing interval between sea level and baseline (r - ~ b ) is replotted relative to the present sea level (e.g. IJ = BD). The curve H K A represents the function h(t) for the case of constant tectonic
UPLIFT AND SUBSIDENCE OF THE HELIKE DELTA uplift with r = 2 m k a -1. It traces the rise and fall of sea level with respect to the reference frame of the sea cliff. At about 9 ka BP, the rapidly rising sea level overtakes the baseline (at C). Any relic shoreline produced at that time would now have zero elevation (H), and any older relic shoreline features would still be submerged. In this model, the relic shoreline with the highest possible present elevation (at K) was produced when sea level (at E) reached its maximum distance above the baseline. In the example shown in Fig. 3 this occurred about 6000 years ago. It follows that any relic shoreline found today at a given elevation above sea-level (at G for example) can have two possible ages (corresponding to J and L), depending on whether it was produced on the rising or falling leg of the relative sea-level curve (at D or F respectively). This bimodal distribution of age v. height has been observed for dated Holocene shoreline sequences in Japan (Shimazaki & Nakata 1980) and Papua New Guinea (Ota et al. 1993), as well as for the Gulf of Corinth. To interpret the relic shoreline data (elevation v. age) for the Gulf of Corinth, I have plotted a set of h(t) curves for different values of r
Fig. 4. The age and elevation above sea level of raised relic shoreline samples from Aigialeia (filled symbols) and the western Perachora Peninsula (open circles), listed in Table 1. Changing the linear uplift rate r of the footwalt produces a family of curves representing the expected present elevation of relic shorelines as a function of their age. These curves are labelled with the corresponding values of r in m ka -1 . The curve (c represents the corrected local sea level through time and is thus equivalent to setting r = 0.
45
(the model tectonic uplift rate) taken to be time invariable. Figure 4 shows these curves, calculated from equations (2) and (3) and labelled with the corresponding values of r, ranging from 1.0 to 3 . 0 m k a -1. The curve 2.0 is thus equivalent to H K A in Fig. 3. These curves all have bimodal ages for any positive elevation. By plotting the present elevation and age of any relic footwall shoreline for the Gulf of Corinth in Fig. 4 one can estimate its average rate of tectonic uplift from the corresponding background curve of h(t). To do so, however, one must first apply an appropriate calibration to the radiocarbon ages of marine samples from the Gulf of Corinth.
The reservoir effect in the Gulf of Corinth Carbon samples derived from modern ocean surface water have lower 14C/12C ratios than atmospheric CO2, because of vertical mixing with 'older' (14C-depleted) deep water. This 'reservoir effect' increases the radiocarbon age of modern ocean surface water to about 400 years, on average. However, because the magnitude of this effect varies with location, a local correction factor AR is often added to the average (400 year) reservoir effect. Because the Gulf of Corinth is a restricted marine basin, this correction is important here. The Gulf is about 850 m deep, but its connection to the Ionian Sea through the silled Rion Strait is only about 2 km wide and 65 m deep. Nielsen (1912) noted that temperature and salinity below 100m in the Gulf of Corinth were both lower than in the Ionian Sea, and concluded that the deep Gulf water is formed locally by convection during winter. The relatively high and nearly uniform concentration of dissolved oxygen at all depths in the Gulf provides further evidence for efficient vertical mixing (Poulos et al. 1996). Heezen et al. (1966) measured radiocarbon ages in seawater samples taken from the Gulf of Corinth in July 1956. They found the 14C/12C ratio in the surface water to be 4.5% less than in North Atlantic surface water; the comparable figure for water sampled at 800m depth in the Gulf was 4.1%. The equivalent values of AR would be 380 and 350 years, respectively. That is, the waters of the Gulf of Corinth, both deep and shallow, appeared to be nearly twice as 'old' as normal ocean surface water. To Heezen et al. the nearly identical results from both depths suggested 'vigorous and thorough vertical mixing' of the Gulf water. They also noted that radioactively 'dead' carbon
46
S. SOTER
in the surrounding limestone mountains is continuously washed into the Gulf and that the limited exchange of waters through the Rion Straits 'may help to preserve this great apparent age by isolating the Gulf of Corinth waters from the Ionian Sea'. Indeed, it appears likely that carbon dissolved in river and ground water discharging into the Gulf of Corinth is strongly depleted in lac. This is suggested by the anomalous radiocarbon ages we obtained for organic carbon samples from bore holes drilled on the Helike Delta (Maniatis et al. 1996). Samples of wood and marine sediments yielded radiocarbon ages in reasonable agreement with an age-depth profile based on luminescence and archaeological dating of ceramic fragments from the bore holes. However, most of the samples of freshwater organic sediment gave radiocarbon ages 3000-5000 years older than the ceramic age-depth profile (Soter & Katsonopoulou 1998). These anomalously old carbon dates are probably due to the 'hard water effect', involving the dissolution of carbonates by ground water carrying CO2 from the atmosphere. Isotopic reequilibration introduces 'dead' carbon from the carbonates into the CO2. The ground water then feeds lakes where aquatic organisms take up the 'old' CO2 and deposit it as 14C-depleted organic sediments. The hard water effect can increase the apparent radiocarbon age of freshwater organic sediments by as much as the half-life of 14C, nearly 6000 years (MacDonald et al. 1991). Organisms that obtain their CO2 directly from the atmosphere remain unaffected. The case for a hard water effect in the Helike Delta is supported by the presence of carbonate concretion nodules in many of our bore holes. The reservoir effect for the Gulf of Corinth has undoubtedly varied during Holocene time, as a result of changes in climate and sea level. However, the value AR = 380 years measured for the surface water in 1956 (before the global contamination by bomb carbon) is the best available. I will adopt it in preference to the value AR = - 8 0 years estimated for other parts of the Mediterranean (Stiros et al. 1992) and previously applied to the Gulf of Corinth (Pirazzoli et al. 1994; Stewart 1996). Consequently, the calibrated ages of radiocarbon-dated relic shorelines adopted here will be significantly younger than those found previously.
Tectonic uplift bordering the Gulf of Corinth Table 1 lists the dated Holocene samples from relic shorelines on the Gulf of Corinth. The
samples consist of faunal species that lived just below sea level or, in the case of the boring mollusc Lithophaga lithophaga, that produced erosional notches with a distinct upper limit at sea level (Laborel & Laborel-Deguen 1994). One group of samples was collected on the Perachora Peninsula at the eastern end of the Gulf by Vita-Finzi (1993) and Pirazzoli et al. (1994). The other samples are all from Aigialeia, on the southwestern shore, collected at Aigeira, Paralia Platanou, Trapeza and Diakofto (Fig. 1). S. Stiros kindly provided samples of the vermetid gastropod Dendropoma petraeum collected in 1996 from an uplifted shore below the old Trapeza railroad station. The other samples from Aigialeia were collected by Mouyaris et al. (1992), Papageorgiou et al. (1993), VitaFinzi (1993) and Stewart & Vita-Finzi (1996). One sample from Aigeira was dated by the uranium-series method (Vita-Finzi 1993). All other samples in Table 1 were dated by radiocarbon and are calibrated here using the marine calibration curves of Stuiver & Braziunas (1993) with a reservoir correction factor AR = 380 years. Because of unknown variations in the reservoir effect during Holocene time, the actual uncertainties of the calibrated ages must be larger than the errors listed in Table 2. The relic shoreline elevations are plotted against calibrated age in Fig. 4. The ageelevation data for the shorelines of the Perachora Peninsula (open circles) would fall among the curves representing constant tectonic uplift rates of r = 1.35 + 0.25 m ka -1, and their distribution suggests a fairly steady rate of uplift over 7000 years. The data points for the shorelines of Aigialeia (filled symbols) would fall among the curves with rates of r = 2.4 4- 0.8 m km -1 ka -1. The average Holocene uplift rates for both Aigialeia and Perachora are significantly larger than those found by Stewart & Vita-Finzi (1996). The Holocene uplift rates for Aigialeia are also significantly larger than the average Quaternary uplift rates derived from the raised terraces near Corinth (c.l.4mka-1). Noting that footwall topography increases westward along the southern margin of the Gulf of Corinth, Dart et al. (1994) had suggested that 'rates of uplift may increase toward Aigion'. In fact, this appears to be the case. The present extension rate of the Gulf of Corinth also increases to the west (Clarke et al. 1997), which is consistent with the uplift rates and the topography. The distribution of the data for Aigialeia suggests that the rate of uplift there has not been steady but rather has been affected by major discontinuities. Even so, the smooth curves labelled with constant r values are still useful in
U P L I F T A N D S U B S I D E N C E OF T H E H E L I K E D E L T A
47
Table 1. Holocene shoreline data h (m) Aigeira
Fauna
14C age (a BP)
Cal. age (a BP)
@(m)
Ref.
1.0 6.0 6.0 6.5 7.5
D CC L L L
1420 + 60 2965 + 50 U-series 4880 4- 270 8040 4- 85
600 + 50 2245 4- 75 6400 4- 200 4665 4- 335 8055 4- 85
- 1.2 -7.2 -11.5 -8.9 -19.6
2 3 4 2
P. Platanou
2.3 3.7 4.0 6.2 6.5
L L L L L
2785 4- 50 2420 4- 130 8730 4- 340 3285 4- 65 8050 4- 60
2020 4- 70 1590 4- 150 8875 4- 475 2695 4- 45 8065 4- 65
-3.4 -4.7 -20.4 -7.5 -18.6
2 2 4 2 2
Trapeza
3.5
D
6920 4- 50
7030 4- 60
-11.2
5
Diakopto
1.5 3.5
L L
1210 + 100 2190 4- 60
435 4- 85 1340 4- 50
-1.6 -4.3
2 2
Perachora
0.8 1.4 1.7 1.7 1.7 2.2 3.0 3.1
V C N N M L L L
1865 4- 55 1990• 100 6890 4- 90 7100 4- 1300 7200 4- 350 4120 4- 60 4705 4- 50 58204-60
1010 4- 60 11604- 110 6985 4- 115 7095 + 1325 7230 4- 290 3640 4- 70 4450 4- 60 5810+80
- 1.4 -2.1 -9.0 -9.6 - 10.2 -4.0 -5.2 -7.1
6 6 3 3 3 6 6 6
1
h(m) is sample elevation in metres above sea level. Dated shoreline fauna: C, Chthamalus; CC, Cladocora caespitosa; D, Dendropoma petraeum; L, Lithophaga lithophaga; M, Mytilus galloprovincialis; N, Notirus irus; V, Vermetus triqueter. All radiocarbon calibrations are from the marine series of Stuiver & Braziunas (1993) with a reservoir correction factor AR = 380 years. Cb (In), footwall baseline elevation (in metres) corresponding to the calibrated age of each relic shoreline. References: 1, Papageorgiou et al. (1993); 2, Stewart & Vita-Finzi (1996); 3, Vita-Finzi (1993); 4, Mouyaris et al. (1992); 5, this paper; 6, Pirazzoli et al. (1994). For the third Aigeira sample, reference 3 gives an incorrect elevation (C. Vita-Finzi, pers. comm., 1996). The Trapeza sample was collected in 1996 by S. Stiros and dated by AMS at Woods Hole Oceanographic Institution (OS-10187). providing the integrated long-term average rate of tectonic uplift for any raised relic shoreline feature since the time of its formation. To investigate the discontinuities in the tectonic uplift in Aigialeia, we refer again to Fig. 3. In reality, the curve Cb representing footwall elevation is not linear in time, as expressed by equation (2), but has a complex form, with earthquake-related discontinuities and periods of relatively steady aseismic motion. The difference between the smoothly varying sea-level curve ~c and the discontinuous baseline curve ~b will therefore p r o d u c e a discontinuous curve for the present elevation h(t) of relic shorelines as a function o f their age. Conversely, for each relic shoreline plotted in Fig. 4 the c o r r e s p o n d i n g absolute elevation of the footwall baseline w h e n the shoreline was formed is given by Cb = r -- h
(4)
The quantity Cb for each relic shoreline is listed in Table 1 and plotted, for the Aigialeia group, by open symbols in Fig. 5. (The upper part of Fig. 5 repeats the raised shoreline data for
Aigialeia from Fig. 4.) F o r each relic shoreline, the depth below sea level of the footwall baseline (open symbols) at the time of origin equals the present elevation of the feature (solid symbols) above sea level. Figure 5 suggests that the footwall baseline (and hence the entire footwall block) ascended steadily between 7 and 2.2 ka BP. A least-squares linear fit to that segment of the data has a slope of a b o u t 0 . 9 m k a -1. There are p r o n o u n c e d discontinuities at both ends of this segment. Around 2.1kaBp the footwall shoreline ascended rapidly by a b o u t 2 m. This was probably due to the great e a r t h q u a k e of 373 Bc. The discrepancy between the dates is less than 200 years, which is small, considering the uncertainties in the r a d i o c a r b o n ages due to u n k n o w n variations in the ancient reservoir effect for the G u l f of Corinth. Stewart (1996) previously noted this discontinuity in the uplift data for Aigialeia and suggested that it was due to the Helike earthq u a k e of 373Bc, but the present analysis makes the case m u c h stronger. F o r r a d i o c a r b o n
48
S. S O T E R
0
o
_o
~.
~
.~
.~
"~
~
~
~
~
_
_
~i
~
~
o
~
k;
~
i:~
~ o~.~ -~.~
.~
~
~.~ ~
.o.~ .c ~" II o
~=
~,= ~
.~o_.
~o
~
ANALYSIS,
SOUTHERN
"~=~
.o ~ =
o~"~ 9 o ~.~ o o . O~. ~ ~ ~ o
=-~= ~ ~
-~ 'C = ~
9
0a .,..~ ...~
~_~ eO
-~
;~
a 0
.,
>.>-,
~
.~.~
~~ m
9~
~_~
0
~-~ .g=>~
~,-~I
~
Oa
~ ~
t--:. ,.--ZOO
~
~ ~o
~.~.~
,a::l
"~cs~
0a
~
~o=
~a,.--
- >,
~- =~
~ = ~
~,o~
~,>
~
= o
=~ 3 2 k m wavelength), low-amplitude structure (c. 45m) reflected in the outcrop pattern of a wave-cut surface (isotope stage 7 or 5) and outwash (isotope stage 4), as well as in observable dips of older strata. The stratigraphic setting of the syncline
209
permits detailed analysis of the character of the deformation and of corresponding geodetic data. Interseismic strain accumulation appears to dominate geodetically measured uplift. Geological uplift rates determined from the buried wave-cut surface are low (31 ka40 ka (~C) !" ~ .... MSL +5.5 m: 4.7 ka (~4C) "7 +I m
EASTERN QESHM ISLAND N ofTula (Lot.3)
Tourgan (Loc.4)
Fig. 3. Schematic profile of the marine terraces along a cross-section between Locality 3 and Locality 4, eastern Qeshm Island. The continuous line represents Quaternary marine deposits capping Mio-Pliocene deposits. Arrows correspond to locations where the altitude was measured. Dating results are also summarized.
1.5 m near Bushehr to a maximum of 5 m in the narrow strait between Qeshm Is. and the mainland, falling to between 1.8 and 2.7m along the coasts of the O m a n - M a k r a n Sea. Indian Ocean cyclones and atmospheric depressions during the southwesterly monsoons may cause strong swells on the Makran coast.
Methods In the field, geographical coordinates were determined using a pocket global positioning system (GPS) device, with an estimated accuracy of the
order of 50m. Altitudes are with reference to mean sea level (MSL), if not otherwise stated. Elevations were generally measured with reference to the sea surface for Holocene samples (and then corrected according to tide predictions), or to high-water marks (and then corrected for the local tidal range), by means of a spirit level and a folding rule. For higher terraces, elevations were estimated by averaging readings on two pocket altimeters. Because interpolated corrections for barometric changes were often not possible, altitudes of inland higher terraces are probably accurate to no better than 4-5 m.
246 m
+210 m 00m
@,-~.
m
+16o m
+14o
m.-?,~".~" x~10
m
' ~
....
Fig. 4. As for Fig. 3 but near Chah Bahar.
+36 m '/ +28 m z +20m ~ t ~ / + 4 m: 3.7ka (NC) -- ~ - -
MSL
QUATERNARY UPLIFT, SOUTH COAST OF IRAN To reconstruct former MSL positions, an additional vertical uncertainty equalling half the local tidal range was ascribed to sampled beach deposits. More precise estimations were possible when fossilized organisms collected in growth position (e.g. barnacles) could be measured with reference to the position of their modern counterparts. Corals are known to develop below the low-tide level, reaching depths that vary according to the species considered, though generally not exceeding 15-20 m in the area considered because of limited water transparency. Accordingly, it has been assumed that coral samples collected in growth position generally indicate minimum former low-tide level elevations, and are thus usually associated with the next slightly higher terrace level. Well-preserved horizontal reef flats are thought to correspond closely to former low-tide positions. Reconstructions of former sea levels are more difficult on sloping terraces where a polycyclic origin seems likely. In these cases, the stratigraphical position and altitude of the sample to be dated remain the most objective altitudinal criterion, whereas the elevation of the sloping terrace surface must be interpreted on a case-bycase basis. Samples selected for absolute age determination consisted of shells and corals which had already been investigated by X-ray diffraction. In spite of the aridity of these regions, most of the corals were recrystallized, suggesting more humid conditions in the past. Fourteen 14C dates were estimated by 13-counting on shell samples from the lowest terrace. Six additional dates obtained from a previous survey were also considered (Table 1). Conventional ages were expressed as recommended by Stuiver & Polach (1977) and were calibrated using the marine calibration curve proposed by Stuiver & Braziunas (1993), with an apparent age of 190 years for the Persian Gulf (J. Southon & M. Fontugne, unpubl, data). Except for Gif-8622 and -8624 the ~13C values are in the range of variation observed for modern marine carbonates and do not indicate recrystallization. Shell samples consist of aragonite or calcite. A content of 3.3 m. One of the uncertainties in these estimates concerns the undeformed surface level of the Bund before the earthquake, because it is to this level that estimates of coseismic deformation must be referred. As discussed above, there is little doubt that significant surface morphology was formed entirely by coseismic deformation associated with the 1819 sequence of earthquakes. Oldham (1926, p. 23) argued that if the observations are referred to a datum just north of the Allah Bund taken from Baker's map, absolute uplift may have been 1 m less, and absolute subsidence 1 m more, than the maximum values estimated from local datum levels, e.g. the bed of the Puran or the level of Lake Sindri, respectively. This, however, does not take into account the preearthquake seaward gradient of the land surface. The smooth surface of the bank of the Puran, mapped in Baker's profile, provides a surface whose extrapolation permits the absolute amplitude of uplift beneath the Allah Bund to be estimated. Baker's measurement datum was the level of water dammed behind the Mora Bund, but no precise estimate of sea-level elevation is provided. The absolute level of Baker's datum above sea level can be estimated to c. +0.3 m because it was possible to navigate to the base of the Lallan Puttun Dam before the earthquake, and because the level of Lake Sindri was replenished by high tides after the earthquake. With this assumption, two approximations to the pre-seismic land surface beneath the Allah Bund are possible: in one, a smooth curve is fitted to the bank of the Puran between 17 and 50 km north of the Allah Bund and extrapolated beneath the Bund, and in the other the curve is, in addition, constrained to fit to lowest estimated sea level at a distance of 50 km south of the Bund (Fig. 4). Using the first approximation the peak elevation c. 1 km north of the Bund increases from 6.2 to 6.6 m, and using the second it reduces to 6.1 m. The goodness of the fit to Baker's river bank data is superior (see residuals in Fig. 4) if the sea level constraint is ignored, and because it is by no means certain that the bank surface should asymptotically approach sea level, the higher estimate is thus considered more reliable. Data used in subsequent models are summarized in Table 1. The recorded region of maximum subsidence occurs along the Puran river and in the Sindri region along a well-travelled trade route. Numerical data for the Allah Bund are obtained only
303
where navigation was impeded and where a clear uplift profile was visibly manifest after the 1826 flood. Few roads exist to the east and yet fewer roads to the west that might have been explored after the event, and our knowledge of deformation is biased by this historical circumstance. Neither Baker nor Burnes travelled the length of the Bund to establish its lateral extent. In 1827 Burnes estimated its width as +16 miles (c. 50 km), but in 1828 revised it on the basis of travellers' accounts of newly necessary, circuitous routes around Lake Sindri, to 18 miles west to Ghari, and 24 miles east to Pacham Island (c. 80 km). Baker indicated its length would be too difficult to survey because of the absence of drinking water. The Survey of India maps later in the century were used by Oldham to confirm that the Bund was at least 80 km long, and that morphological features suggest faulting for more than 150 km. The southerly facing scarp width is likely to have been underestimated in 1826, as part of it was submerged, and in 1844, because it may then have been covered partly by sediments. Thus the south facing scarp could have been greater than the 600 m width estimated by Burnes, but because it determines only the closest approach of the subsurface fault to the surface, and has minor influence on deep slip parameters, its true width is of little consequence in the following analysis. The southern extent of subsidence is perhaps the most clearly defined because this formed a freshwater lake that eventually became saline and finally dried up. Unfortunately, because a deep channel existed through the lake, some of the depths in subsequent descriptions relate the channel depth and lake depth in ways that do not permit true bathymetry to be evaluated precisely. When Burnes visited this after the 1926 flood the main river channel was fresh, as was the surrounding water in Lake Sindri. In following years the Rann shallowed, and although much of this may have been due to sedimentation, it is possible that post-seismic deformation occurred (Oldham 1926). In 1827 the width of the channel though the Bund was 40 m wide but by 1828 the flow had ceased and the waters of the Rann were saline. Deformation tapers to low values near the northern and southern limits of rupture, but no deformation is apparent >24 km from the Allah Bund (Table 1). Slip on the fault is modelled as uniform slip in an elastic half-space using the formulation of Okada (1985). The procedure adopted is to compare the surface deformation arising from one or more subsurface dislocations with the observed surface deformation, and to reject those whose theoretical geometries do not result in satisfactory
304
R. B I L H A M
Fig. 4. Baker's 1844 profile, projected on a north-south line, with exponential approximations to the slope for the river bank (above), and residual elevations when these are subtracted from the observed data (below). A better fit to Baker's river-bank data is obtained if assumptions concerning morphological relations to inferred sea level are ignored. agreement. By assuming two-dimensional (2D) u n i f o r m slip in a northerly direction, for example, five u n k n o w n s remain to be determined: fault dip, slip, latitude, and the depth to the top and b o t t o m of the rupture. In practice, the latitude and depth of the rupture are determined to first order by the width and location of the section t h r o u g h the
Allah Bund. That is, the a p p r o x i m a t e ant• metry of the vertical d e f o r m a t i o n field requires an almost vertical fault whose surface extension must cut the Earth's surface at the s o u t h e r n m o s t expression of the Allah Bund. The half-width o f the s o u t h w a r d slope of the Allah Bund is approximately equal to the depth to the u p p e r
Table 1. Deformation estimates for north-south section through the Allah Bund Parameter
Maximum
Minimum
Model input
Northerly extent of uplift -0.1 m
+24 km +6.6 m + 1 km 0 - 1 km -4.5 m -3.5m - 6 km -40 km
+6 km +6.1 m +0.6 km 0 -200 m -2.5 m -1.5m - 5 km -24 km
6• 6.34-0.3 m 14- 0.2 kill -3.5i0.3m -1.5• -6• -24•
Distances are measured relative to the inferred deformation-null separating uplift from subsidence. Uncertainties indicated as model input are used to estimate confidence levels of solutions.
THE 1819 RANN OF KACHCHH EARTHQUAKE
305
in Fig. 6. Assuming the values listed in Table 1 the model misfits in Fig. 6 are consistently lower for down-dip widths of less than 10 km than for larger fault widths. Acceptable combinations of dip and coseismic slip for a down-dip width of 10 km are 11.5 4- 1 m and 68 ~ + 7~ respectively (lo-, Fig. 7a).
6 5 .4 4 0
"" 3
Planar dislocations constrained by the Allah Bund profile
E 2
..~ 1 .o
~0
I
50
I
I
I
I
I
I
60 70 80 dip of fault in degrees
I
90
Fig. 5. The ratio of maximum uplift to maximum subsidence determines the dip of the fault in an elastic half-space. A 1.8 ratio of uplift to subsidence favours a dip close to 70~ largely independent of down-dip parameters (top and bottom of down-dip width indicated for two different solutions) surface of the dislocation. In addition, simple numerical tests (Fig. 5) show that a ratio of uplift to subsidence of 6:3.3 requires the dip on the causal rupture to be to the north at 65-70 ~ requiring reverse slip on the fault.
Planar dislocations constrained by maximum and minimum vertical deformation In principle, having assumed the location and depth of the dislocation by inspection, only three observations are required to constrain the dip, slip and down-dip width of the Kachchh rupture. Five data are available in Table 1 in addition to the continuous but incomplete spatial coverage from Baker's levelling data. Forward models were developed to estimate the sensitivity of the interpretation to each of the available data. The first suite of models (Fig. 6) emulates the subsidence at Sindri (+0.3m), maximum uplift north of the rupture (+0.3 m), maximum subsidence south of the rupture (+0.3 m), and far-field constraints of uplift and subsidence less than 2 0 + 10cm at distances + 2 4 k m from the Bund. These models ignore the location of maximum uplift and subsidence, and instead use only their amplitudes as constraints. Least-squares misfits between observations and model results for a range of possible slips and down-dip depths are estimated and contoured in terms of l - 3 a confidence intervals
In models illustrated in Figs 6 and 7 Baker's profile has not been used to constrain the 1819 rupture parameters. Baker's map view of the Bund (reproduced from a tracing of the original by Oldham (1898)) shows the section (lower expanded view in Fig. 3) to have been taken at N30E approximately normal to the strike of the Bund. Although Baker's numerical data are presumably more precise than the estimates for subsidence and uplift listed in Table 1 the profile has some puzzling characteristics. The transition between the almost linear northern slope and the undeformed surface of the desert (6 km from the southern edge of the Allah Bund) is too abrupt to be caused by elastic deformation. If the shallowing of the bed of the Puran is used to estimate the northern limit of deformation, the width of the uplifted Allah Bund might be estimated to extend perhaps 4 km further north. Baker suggested that the channel may have filled by bank collapse at the mouth of the incised cut through the Bund but not to the north. A possible reason for the abrupt transition between the Bund and the apparently undisturbed Rann north of the Bund may be erosion of the bank caused by drainage of the impounded waters in 1926. A further problem with the data concerns the surprisingly linear northerly dip to the deformation field, again atypical of elastic deformation. Notwithstanding these perceived problems with the levelling data, slip parameters were estimated from Baker's sectional profile of the Allah Bund by comparing the slope at 0.5km intervals with the theoretical slope estimated from a dislocation model. An observational uncertainty of 0.2 m per 0.5 km was assigned to these slope data, limited mainly by digitizing errors from Oldham (1898). Appropriate models require a down-dip width of 5 + 1 km, dipping at 504-5" with 1 2 4 - l m of slip (Fig. 7b). The shallower dips for the Bund profile result from the model's attempt to fit the steep northerly slope, in addition to the subsidence evident between 6 and 8 km north of the southern edge of the Bund (Fig. 4). A suite of models in which data were examined only from the southern
306
R. BILHAM
Fig. 6. Misfit contours for solutions for slip and down-dip width for four alternative northerly dips to the inferred rupture zone using constraints listed in Table 1. A down-dip width of c. 10km is favoured by the data, with a dip of at least 60~
5.5 km of the uplifted Bund favoured similar slip parameters. The models favoured by the Bund data evidently give shallower down-dip widths than the maximum-minimum deformation field used in Figs 6 and 7a, and are inconsistent with the reported subsidence in Lake Sindri. Subsidence at Fort Sindri is required to be less than 1 m, and maximum subsidence is preferred to be less than 2.2m values lower than those listed in Table 1.
Listric fault models The above models use planar dislocations with uniform slip. More realistic slip distributions and more complex geometries can be proposed that also fit the data, but their exploration would be conjectural in the absence of additional constraints. Because > 1 0 m of slip is large for a fault with < 1 0 k m down-dip width, a search for faults with longer down-dip width but similar surface deformation is of utility. A class of listric faults was examined with a 2D subsurface geometry of the form d = a + b e -Cx, where d is depth and x is distance from the southern edge of the Bund, and a, b and c are constants
chosen to best fit the observed surface deformation field. In the examples shown the listric rupture surface was approximated by ten short planar fault segments. These subsurface faults have steep near-surface dip and shallow dip at depth. Two geometries of several that approximate the data (but which predict a broader Bund profile) are shown in Fig. 7c. The total down-dip length to the faults increases by a factor of 2-3 in these models to 15-23 km. A detailed examination of listric fault models was considered unwarranted in the absence of constraints from horizontal deformation.
Location and Magnitude of the 1819 earthquake No accurate estimates for the epicentre for the 1819 earthquake have hitherto been proposed, although Chung & Gao (1995) attributed inappropriate locations south of the Allah Bund to Quittmeyer & Jacob (1979) and Chandra (1977), on the basis of those researchers' approximated coordinates. The above analytical solutions suggest that the 1819 rupture occurred 5-15 km north
THE 1819 RANN OF K A C H C H H EARTHQUAKE
307
Fig. 7. (a) Contours showing a range of slip and dip solutions that fit point data from Table 1, assuming a 10 km deep dipping planar dislocation. Best fitting solution shown left. (b) Best fitting solutions for a 5 km wide planar dislocation using only Baker's levelling profile (enlarged right) as a constraint. (c) Surface deformation arising from listric faulting. Listric faulting results in similar subsidence, much increased down-dip fault width, but broader uplift than that associated with planar faults.
or northeast of the Allah Bund. The longitude is not determined by the available data to better than 1~ The along-strike length of the Allah Bund was estimated by Oldham (1926) as 80-150 km, from which a geometric seismic moment can be estimated for the rupture. Because the eastward extension of the fault and its sense of slip is unconstrained by the historical record, it is ignored in the following estimates for magnitude. Using the relation M o = # x slip x L W, where M w = 2 / 3 ( l o g M o ) + 1 0 . 6 , the range of parameters
determined above corresponds to a local magnitude of MI, = 7.7 + 0.2 using typical values for the rigidity modulus #, and a range of down-dip widths of 10-20 km. An alternative method to estimate the magnitude is to use the intensity data reported for the event. An empirical relation between intensity and area of shaking constrained by six Indian events including the Anjar (1956) event was discussed by Johnston (1996), who offered a magnitude of M = 7.5-8 for the 1819 event (7.8 by Johnston & Kanter (1992)). A slightly lower
308
R. BILHAM
Table 2. Isoseismal areas and estimated moment magnitudes for the 1819 earthquake
Intensity
a b c Radius (km) Mo F94C (logl0(dyncm)) M
Felt
VIII
17.3 0.959 0.00126 1600 27.23 7.4
24.1 0.44 0.00586 140 27.66 7.7
Constants a, b, and c are preferred values from Johnston (1996). intensity magnitude can be estimated from the data of Fig. 1, where significant attenuation of intensity northward is evident, causing the isoseismals to be non-circular, especially for lower intensities. Moment magnitude, Mo, of an earthquake in the F94 model is related to the area, S enclosed within a specified isoseismal intensity contour, by an expression of the form logMo=a+blogS+cv/S
(1)
where the constants a, b and c are determined empirically for each isoseismal area. Intensity magnitudes are shown in Table 2, although the intensity data from which they are derived are sparse and of uncertain quality. A mean magnitude estimated from the intensity data is M = 7.5 4- 0.2 in reasonable agreement with the deformation data. Combining the intensity and deformation data, a preferred magnitude of M = 7.7 + 0 . 2 is assumed for the 1819 event. A more careful evaluation of authentic felt reports is needed to improve significantly on the mapped isoseismal estimation of magnitude. Discussion
Although several forms of subsurface slip geometry are admitted by the data, they share several features. Maximum and minimum vertical deformation values estimated 6 years after the earthquake yield preferred solutions for a 67 4-5 ~ dipping fault, with a down-dip width of 6-10 km, and the short profile across the Bund measured 25 years after the earthquake favours a dislocation with 50 + 5 ~ northerly dip, 12 + 1 m of slip and a down-dip width of 5-6 km. Listric faulting with dip shallowing to the NE is an alternative subsurface geometry that requires c. 11 m of slip on a steeply NE dipping nearsurface fault. As noted above, it is assumed that slip vector was at N30E, normal to the Bund, and
consistent with both the NE directed IndoAsian plate convergence vector (Paul et al. 1995) and the p-axes of regional earthquakes (Chung 1993; Chung & Gao 1995). A N45E slip vector would require steeper dips, and shorter downdip widths. Hence, the estimated dips are lower bounds, and a steep fault plane is a necessary, common feature of any interpretation of the data near the Allah Bund. At dips of 50-59 ~ 'Byerlee' friction causes a fault to 'lock' in response to horizontal compression, unless significant fluid overpressuring is available to reduce friction on the fault (Sibson & Xie 1998). The geometry of the Kachchh rupture is thus severely misoriented for reverse slip, and would have required fluid overpressuring to promote rupture. Fluid overpressuring is believed to be widespread in the lower crust (Sibson 1992). Reservoir-induced seismicity throughout India suggests that fluid pressures play an important role in triggering shallow seismicity, and it is possible that this may be a common slip mechanism for the Indian subcontinent. The thick sediments in the Indus fan and the Rann of Kachchh, moreover, have favourable conditions for overpressuring. A consequence of probable northeasterly directed slip is that Oldham's inferred fault east of the Allah Bund, if it indeed exists, could absorb a large left-lateral strike-slip component with insignificant convergence, and thus minor vertical deformation (Fig. 8). Oldham (1926) suggested that post-seismic surface changes occurred in the Sindri region that were not entirely the result of silting or precipitation of evaporites. The various sketches of Fort Sindri show it to have been initially surrounded by water close to the high-tide or monsoon-surge level, and a few decades later to have been surrounded by dry land. Oldham suggested that this change was caused by a relaxation of coseismic subsidence. The wavelength of the vertical changes is considered too short to result from viscoelastic adjustment of the elastic crust. However, J. N. Brune (pers. comm., 1997) has demonstrated in computer simulations and foam rubber models that dynamic effects associated with propagating wrinkles along the fault plane (Brune et al. 1993; Andrews & Ben Zion 1997) can cause over-shoot or undershoot during rupture, which may differ substantially from the static-frictionless deformation of models examined in this paper. Presumably, aftershocks, afterslip or post-seismic creep would bring surface deformation closer to the static deformation field. However, if this were the case in the Sindri region, and initial subsidence at Fort Sindri were an artefact of dynamic rupture, we would expect
THE 1819 RANN OF KACHCHH EARTHQUAKE
reverseslip/
local
I
50km Fig. 8. Simplified sketch of inferred NE-directed convergence during the 1819 earthquake. Steep reverse slip causes uplift NE of the Allah Bund (barbs). The existence and style of deformation east of the Bund is speculative, because contemporary field observations are absent.
that relaxation of the footwall would be evident also in relaxation of the hanging wall measured by Baker. The data of both Baker and Burnes were obtained from isolated samples of a feature whose along-strike length and surface geometry renders approximate any simple elastic deformation model. The amplitude of slip required in each solution is considerable for along-strike dimensions of 80 km, hence Oldham's somewhat weak evidence for 150 km of along-strike slip is consistent with this aspect of the data. In addition, a down-dip width of 6kin is unexpectedly small to permit > 10 m of surface rupture. Had this occurred, mean dilatational extension along each side of the fault plane would have exceeded 1000# strain, with a correspondingly high stress drop. The modelling presented here is insensitive to along-strike slip, and to variations of slip along strike. It is curious that the impressive dip-slip component resulted in no surface fault scarp, as it appears to have reached at least to within a few hundred metres of the surface. Presumably, for this to occur, the surface alluvium would have to have been draped over the rupture in the near surface. It is possible also that a fault scarp, or several scarps and fissures, did occur along parts of the Bund, the details of which were not related in second-hand accounts of the event. A curious feature of the region is the absence of a pronounced physiographical feature along the Bund (the Bund is a mound, not a mountain). This suggests that recurrence intervals are low, or that reverse slip is a relatively recent process for the Kachchh fault, which, like nearby faults associated with the Kachchh rift system, is currently being reactivated in a reverse sense. The recurrence of earthquakes in the Kachchh region would appear to be accessible to palaeoseismic techniques, and several issues associated with the 1819 event are worthy of field investigation. For example, the 1826 flood
309
will have deposited fresh-water sediments above the salt deposits on the floor of Lake Sindri, providing a measure of the current form of the subsidence basin. Investigations of ponding south of the Bund, along the Rann to the west and far east of the mapped expression of the 1819 Bund, would clarify the along-strike length of faulting, and its potential sinistral component. Investigations of ponding north of the Bund might also reveal the transient strand line of the 1826 flood. Surface studies of the eastern expression of the fault might reveal evidence for left lateral slip, although many of the drainages across the scarp would have been initiated only after the earthquake. The rate of secular strain contraction of India is not known in the Kachchh region but is immeasurably small (
UPPER PLATE DEFORMATION AND CASCADIA EARTHQUAKES fault is up to the north and may have a strikeslip component, but Quaternary deformation is unconfirmed. Deformation opposite Tillamook Bay is more complex than at Nehalem Bay, but a broad low between the Happy Camp fault at Cape Meares and an uplifted region at Twin Rocks can be seen in Fig. 7. This low is interrupted by smaller fold axes. The uplifted region north of the bay may be related to the Tillamook Bay fault (Figs 6 and 7). Owing to the thin Quaternary section on the innermost shelf, offshore Quaternary deformation cannot be confirmed, but the onshore Tillamook Bay fault suggests Quaternary activity. Rapid marsh burial has been identified at both bays (Atwater et al. 1995; Barnett 1997). Netarts Bay. Netarts Bay, on the northern Oregon coast (Fig. 1), is bounded to the north by WNW-trending, NE-dipping high-angle reverse faults which deform coastal sediments, similar in style to Tillamook Bay to the north (Wells et al. 1992, 1994). Faults thrust middle Miocene Columbia River Basalt Group (15 Ma) over late Pleistocene gravels, and may also offset the youngest Pleistocene marine terrace surface (Wells et al. 1994). The fault is known as the Happy Camp fault onshore (Parker 1990; Wells et al. 1992, 1994) and is the southernmost extension of the prominent Nehalem Bank fault zone, which deforms Miocene to Holocene sediments offshore (Fig. 6; Niem et al. 1990; Goldfinger 1994). This complex zone of deformation trends roughly N-S on the outer shelf and upper slope off the northern Oregon coast (25 km west of the coastline), but changes to the southeast at its southern end to project onshore just north of Netarts Bay (Fig. 6). No evidence of major strike-slip offset has been found along the northern segment of the fault, but its orientation (suggesting margin-parallel rightlateral offset where oblique convergence is partitioned into a compressional and strike-slip component), the presence of minor strike-slip faulting, and the linearity of the fault, which truncates bedding planes in AMS 150 kHz sidescan images, support a strike-slip component. However, the fault also shows significant vertical offset (both along its N-S and SE-trending segments) and we interpret it as a reverse fault system downthrown to the west and south. As the fault trends southeasterly, the zone of deformation becomes less complex, being characterized by a N-dipping (possibly blind) reverse fault, with Miocene sediments uplifted and exposed at the sea floor, and an asymmetrical syncline to the south (south end of Fig. 7), which lies immediately opposite Netarts Bay. Sidescan
329
images show bedrock within the hanging-wall anticline exposed on the sea floor and offset by minor N- to NNE-trending right-lateral faults. The vertical motion on the northern segment of the fault may be a flower structure or transpressional deformation. The nearshore Nehalem Bank fault clearly deforms and offsets the middle Miocene Columbia River Basalt Group (highly reflective in seismic reflection data, south end of Fig. 7) and overlying sediments. Absence of structural growth of strata within the syncline in Fig. 7 suggests this fault and associated fold post-date late Miocene sedimentation. Investigation of other seismic reflection data across both the southern and northern sections of the fault shows minimal thinning in late Miocene sediments and some thinning of Pliocene sediments across the fault-controlled anticline. This indicates that the fault was active as early as the late Miocene, but the bulk of deformation has taken place during the Pliocene and Quaternary. Vertical seafloor offset of 10-20m across the
Fig. 8. Location map of Siletz Bay on the central Oregon coast showing structures mapped offshore from single-channel and multichannel seismic reflection profiles and onshore fi'om Pleistocene marine terrace deformation (structures in Fig. 9 crosssection). Bold line indicates position of single-channel line in Fig. 10. Onshore and offshore deformation suggests that Siletz Bay is structurally downwarped within a syncline or on the downthrown side of a fault. Generalized locations of subsided marshes after Darienzo et al. (1994).
330
L . C . McNEILL ET AL.
Fig. 9. Cross-section of beach exposure of Pleistocene sediments underlying the youngest marine terrace at Siletz Bay. Marker horizons, including a clay horizon, gravel beds, and the wavecut platform were used to determine deformation of Pleistocene sediments and the location of Quaternary structures. The laterally continuous clay horizon is apparently offset across the bay (up to the north). Faults also offset the wavecut platform at Fishing Rock and Fogarty Creek. The trends of these structures are poorly defined.
fault zone is estimated from sidescan and seismic records, presumed to post-date late Pleistocene lowstand erosion on the shelf (Goldfinger 1994). Cooper (1981) and Parker (1990) also suggested that a west-plunging Miocene syncline is centred about Netarts Bay. Siletz Bay. Structures deforming the underlying sediments and wavecut platform of Pleistocene marine terraces (presumed 80 ka BP Whiskey Run terrace, West & McCrumb 1988; Kelsey 1990) have been identified in the Siletz Bay region of the central Oregon coast (Figs 1, 8 and 9). The wavecut platform and a locally continuous carbonaceous clay horizon, interpreted as a lagoonal deposit or Palaeosols and assumed to be initially sub-horizontal, were shown to be deformed. Variations in elevation of the wavecut platform and the clay horizon/ palaeosol may alternatively be controlled by existing topography at the time of formation, and not by deformation. Variations in altitude of these marker horizons indicate faulting, with vertical offsets of 5 - 3 0 m and broad folding,
with a wavelength of 8 - 1 2 k m (Figs 8 and 9), assuming the terrace is the same age throughout. Beach exposures alone indicate trends between N N W and SSW, but offshore data (see below) provide more precise trends. The clay horizon of the youngest terrace (80 ka BP) dips gently north between 5 km and 3 km south of the Siletz River mouth, where it is below beach level and projected below sea level (Fig. 9). This clay horizon is exposed again c. 10m above beach level just north of the river mouth, where the wavecut platform is at c. 2 m elevation (Fig. 9). The platform is presumed below sea level south of the river, inferred from the elevation and dip of the clay horizon. Projection of the clay horizon below beach level suggests maximum fault offset across the Siletz River mouth of c. 30m up to the north (offset could be a combination of folding and faulting). If this horizon is assumed to be the same age as the terrace (80 ka BP), this produces a late Quaternary vertical slip or subsidence rate relative to terrace levels across the river mouth of 0.4mm/ year. This is a maximum slip rate as sediments
Fig. 10. Line drawing of N-S trending OSU single-channel sparker profile, 6 km offshore Siletz Bay (bold line in Fig. 8). The MP (late Miocene-early Pliocene) unconformity is projected to the sea floor at the southern end of the profile. Profile shows synclinal deformation of presumed late Miocene strata off Siletz Bay, and sea floor offset by possible flexural-slip faults within an active synclinal fold west of Fishing Rock and Fogarty Creek. Fault dips are poorly constrained by seismic data.
UPPER PLATE DEFORMATION AND CASCADIA EARTHQUAKES underlying the terrace are somewhat older than 80kaBp. Siletz Bay may lie in a Quaternary syncline controlled by a fault at the northern end of the bay, similar to the structure observed at Netarts Bay. Fault offset (downthrown to the north) of the wavecut platform and marine terrace was also documented at Fishing Rock and Fogarty Creek (Fig. 8; Priest et al. 1994). Orientation of these two faults is poorly defined, but previously mapped faults onshore are oriented NW-SE and NE-SW. Poor exposure prevents the determination of any strike-slip component on onshore faults. The possible correlative of a syncline at Siletz Bay is traced on N-S trending seismic reflection profiles 4-17km offshore. Figure 10 is a line drawing of a N-S single-channel seismic profile 6kin west of Siletz Bay, which clearly shows synclinal deformation opposite the bay. The late Miocene-early Pliocene (MP) unconformity is truncated at the sea floor and therefore the age of the youngest strata is late Miocene. The sea floor indicates no synclinal deformation, therefore offshore Quaternary deformation at this scale is unconfirmed. The trend of the syncline across several profiles is between E-W and ESE-WNW. Deformed synclinaI sediments on middle- to outer-shelf profiles are truncated by the MP unconformity, indicating little or no activity since the late Miocene or early Pliocene time to the west; however, this unconformity is deformed by the southern bounding anticline which may project into the Gleneden Beach area. Onshore deformation of late Pleistocene marine terraces by the syncline points to the recent activity of these structures. Possible flexural-stip faults north of the river mouth (Fig. 8) were poorly imaged in the single-channel sparker profile and therefore have uncertain offset or dip. Faults with similar offset to those at Fishing Rock and Fogarty Creek (Fig. 9), identified in single-channel sparker lines between 2 and 6km offshore, may also be flexural-slip faults (Fig. 10). A flight of uplifted Pleistocene marine terraces is preserved north and south of Yaquina Bay (Figs 1 and 11) on the central Oregon coast. These terraces have been differentiated by age using amino acid enantiomeric (D:L) ratios in conjunction with the palaeoecology of fossil shells (Kennedy 1978; Kennedy et al. 1982) and a soil chronosequence (Ticknor et al. 1992; Ticknor 1993; Kelsey et al. 1996). These techniques indicate offset of marine terraces on the inferred Yaquina Bay fault of 75 m down to the south (Ticknor 1993; Kelsey et al. 1996). The fault juxtaposes the 80 ka BP terrace Yaquina and Alsea Bays.
331
(Qn) north of the bay against the 125 ka BP (Qy) terrace south of the bay (Fig. 11; Kelsey et al. 1996). This offset yields a slip rate of 0.6 ram/year. The continuation of the Yaquina Bay fault to the east was mapped by Snavely (1976), giving an ENE fault orientation (Kelsey et al. 1996). All core locations of buried marshes identified by Peterson & Priest (1995) are located on the south or downthrown side of the Yaquina Bay fault. Similar studies at Alsea Bay (Ticknor et al. 1992; Ticknor 1993; Kelsey et al. 1996) show that Quaternary faults strike generally N-S. The N-S striking Waldport fault zone vertically
Fig. 11. Central coastal Oregon between Yaquina Head and Alsea Bay, showing late Pleistocene marine terrace backedges and Quaternary faulting (after Kelsey et al. 1996). The onshore Yaquina Bay fault trends NE and downdrops marine terraces to the south. The Waldport fault zone (including the Lint Slough fault) downdrops terraces to the east, with greatest offset at Alsea Bay. Marine terraces in order of increasing age based on soil development chronology (Kelsey et al. 1996): Qn, Newport; Qw, Waconda; Qy, Yaquina; Qc, Crestview; Qfr, Fern Ridge; Qag, Alder Grove. Qh, Holocene beach and dune sands. 9 General subsided marsh locations (after Darienzo et al. 1994; Peterson & Priest 1995; Peterson & Darienzo 1996). Reproduced by kind permission of GSA.
332
L . C . McNEILL E T AL.
displaces terrace platforms down to the east (Fig. 11), with cumulative offset apparently greatest at Alsea Bay, suggesting a structural origin for this embayment (Kelsey et al. 1996). All rapid subsidence sites are on the downthrown side of the Waldport fault zone. Kelsey et al. (1996) concluded, from the evidence of Pleistocene terrace deformation, that both Yaquina and Alsea Bays are downwarped and structurally controlled by faults. Offshore data neither support nor refute the onshore terrace evidence. South Slough. Many active structures with N - S trends on the southern Oregon coast and shelf, where the deformation front is closer to the coastline, are interpreted to be part of accretionary prism-related deformation (Fig. 12). One example is the South Slough syncline which deforms Quaternary sediments southwest of Coos Bay on the southern Oregon coast (Fig. 12; Nelson 1987; Peterson & Darienzo 1989; Kelsey 1990; McInelly & Kelsey 1990) and may have produced multiple buried peats as an independent local structure (e.g. Nelson & Personius 1996). The syncline has been traced onto the shelf on seismic reflection profiles (Fig. 12; Goldfinger et al. 1992a; Goldfinger 1994). Both offshore and onshore deformation suggests that many faults are flexural-slip faults bounding active folds, such as the South Slough syncline, with fault slip parallel to bedding planes (McInelly & Kelsey 1990; Goldfinger 1994).
Coquille River. The Coquille fault (Fig. 12; Clarke et al. 1985; Goldfinger 1994) comes onshore just south of the Coquille River mouth, where it deforms Pleistocene marine terraces (McInelly & Kelsey 1990). The Whisky Run (80 ka BP) platform descends from an altitude of 35 m at Cape Arago to sea level just north of the Coquille River, deformed by the Pioneer anticline (Fig. 12; McInelly & Kelsey 1990). The terrace abruptly gains altitude to 18 m above sea level just south of the Coquille River at Coquille Point (Fig. 12). This altitude change is accompanied by a change in dip of platforms from southwest north of the river to west or seaward south of the Coquille River (Fig. 12; McInelly & Kelsey 1990). The Whisky Run platform is tilted slightly landward at Coquille Point, possibly resulting from deformation by the Coquille fault. The terrace elevation descends once again south of Coquille Point to reach sea level c. 10 km to the south (McInelly & Kelsey 1990). Fold trends in Tertiary and Mesozoic formations underlying the Whisky Run wavecut plat-
Fig. 12. Location map of the south-central Oregon coast and shelf showing Pliocene and Quaternary structures. Structures onshore based on deformed late Pleistocene terraces mapped by Kelsey (1990) and McInelly & Kelsey (1990). Offshore structures adapted from Goldfinger (1994), from seismic reflection and sidescan sonar data. Typical fractures and shears (with strike-slip offset) within Miocene and older sediments uplifted in the complex fault zone are also shown. Elevation contours of the Whisky Run wavecut platform (Qwr) shown in 10 m intervals (from McInelly & Kelsey 1990) indicate deformation by the Pioneer anticline and the Coquille fault. Coastal terrace altitudes are highest at Cape Arago (CA) and at Coquille Point (CP). Reproduced by kind permission of GSA. form also vary from north to south across the Coquille fault. North to south, fold trends are generally consistent with the variations in dip of the Pleistocene terraces and offshore Pleistocene fold and fault trends (Fig. 12). No strike-slip offset or recent deformation could be determined on onshore faults in Jurassic to Eocene strata. In seismic reflection profiles offshore, the fault zone appears as a ridge flanked by synclinal folding deforming the sea floor. Clarke
UPPER PLATE DEFORMATION AND CASCADIA EARTHQUAKES
et al. (1985) suggested the fault downdrops Pleistocene sediments to the northeast on the innermost shelf as observed in the deformed onshore marine terraces, but other seismic data across the fault indicate a fairly symmetrical ridge. Humboldt Bay. Two large thrust fault systems, the Little Salmon fault and the Mad River fault zone, deform Holocene sediments in the Humboldt Bay region of northern California (Fig. 1), and are interpreted to be part of the onshore expression of the southern Cascadia accretionary prism (Clarke & Carver 1992). The Freshwater syncline lies between these two fault systems within Humboldt Bay and Mad River Slough, and has produced Holocene subsidence resulting in stacked buried marsh and forest horizons (Clarke & Carver 1992). The most recent subsidence event is dated at 250-300 radiocarbon a BP, contemporaneous with the most recent event recorded throughout much of the subduction zone. Locations with inconclusive evidence for Quaternary subsidence Other abrupt subsidence sites are characterized by inconclusive evidence or no evidence of Quaternary deformation, in the form of crustal downwarping. Marsh burials have been found in Neah Bay (Waatch River) and the Pysht River on the northernmost Olympic Peninsula (Fig. 1; Atwater 1992). No active faults have been mapped and directly linked to evidence of rapid subsidence in Neah Bay. Buried marshes at Pysht River (Atwater 1992) lie on the downthrown side of a high-angle fault through Tertiary formations (Gower 1960; Tabor & Cady 1978). This fault is mapped parallel to the river (NE), downthrown to the west, and projects offshore to a similar fault in the Strait of Juan de Fuca which deforms an acoustic unit of Holocene age (Wagner & Tomson 1987). A second fault trends parallel to the coast (WNW) and projects into the bay. Buried marshes may lie on the upthrown side of this fault, but there is no evidence that it is recently active. High-resolution seismic profiles along the lower reaches of the Columbia River (Ryan & Stevenson 1995) indicate possible evidence of Quaternary faulting, including the NE-trending Fern Hill fault, which offsets Miocene Astoria Formation onshore (Niem
333
& Niem 1985). These faults have not yet been linked directly to locations of rapid marsh burial or liquefaction in the Columbia River (Atwater 1992, 1994; Obermeier 1995). Niem & Niem (1985) also mapped a WNW-trending syncline through Youngs Bay (Fig. 1), south of Astoria, where a drowned forest and rapidly buried marshes have been identified (Peterson et al. 1997, C. D. Peterson, pers. comm., 1997). Quaternary deformation across this structure remains unproven. An older syncline (deforming late Miocene and possibly Pliocene strata but with no conclusive evidence of Quaternary deformation) has been mapped on the outer shelf opposite Nestucca Bay (Fig. 1). No evidence of Pleistocene deformation has been reported onshore, but it has been suggested that Cape Kiwanda, the headland to the north of the bay which is composed of Miocene Astoria Formation and Smugglers Cove Formation, may be a structural high (Parker 1990). Evidence of Quaternary downwarping at Vancouver Island sites, the Necanicure River, and Salmon River also remains inconclusive, to judge from the available data.
Marsh burial located near structural uplifts Two possible exceptions to the hypothesis that buried marshes lie within tectonic downwarps are the Sixes River, southern Oregon, and the Copalis River, central Washington (Fig. 1). Kelsey (1990) mapped an E-W trending anticline, which deforms Pleistocene terraces, just north of Cape Blanco and coincident with the lower reaches of the Sixes River (Fig. 1). Buried marshes and tsunami sands have been identified on the southern limb of this anticline (Kelsey et al. 1993, 1998) in a cutoff meander of the Sixes River. Further examination of marshes in a N-S transect across the southern flank of the anticline may reveal differential subsidence (H. M. Kelsey, pets. comm., 1997). This apparent anomaly of local uplift and regional subsidence could be explained by subsidence being the net result of local uplift and regional subsidence, by the anticline not being triggered by every subduction zone earthquake, or by the fact that the anticline was active in Pleistocene but not during Holocene time. A second possible exception is an ENE-trending ridge (Langley ridge) located 5 km south of the Copalis River. Deformation is interpreted as anticlinal folding and diffuse faulting above a possibly N-dipping blind thrust fault (McCrory 1996), with buried marshes at the Copalis River on the upthrown or north side of this fault. In addition to buried marshes, Atwater (1992)
334
L . C . McNEILL ET AL.
identified liquefaction evidence from 900-1300 years ago at Copalis River with no indication of accompanying tectonic subsidence. This event could be attributed to movement on a local crustal thrust fault such as that underlying Langley Ridge. The dip of a blind fault is often difficult to determine from geomorphology and surface faulting, and this thrust fault may, in fact, dip to the south, with the Copalis River marshes lying on the downthrown side of a fault. Alternatively, a syncline to the north of this ridge may coincide with the Copalis River. This is supported by N-S single-channel and multichannel seismic profiles which indicate a series of closely spaced approximately E-W trending ridges and intervening synclines on the continental shelf. The Copalis River buried marshes may in fact lie within one of these synclines rather than being associated with the Langley Ridge structure 5 km to the south.
Discussion Implications f o r the Cascadia subduction zone earthquake record
The record of prehistoric subduction earthquakes on the Cascadia subduction zone, in the form of rapidly buried marshes, documents sudden submergence, inundation of coastal lowlands, and burial of the former land surface. Correlation of coseismic events between coastal bays, based on radiocarbon ages and dendrochronology, has allowed rupture lengths, magnitudes, and recurrence intervals of prehistoric Cascadia earthquakes to be proposed. In addition, estimates of amounts of coastal subsidence can be used to approximate the position of the rupture zone and earthquake magnitude using elastic dislocation modelling. The possible nontectonic origin of some submergence events should, however, be considered when assessing potential earthquake hazards. In this study, it has been demonstrated that many abruptly buried marsh locations can be linked to Quaternary structures (synclines and downdropped side of faults) which produce downwarping. The influence of upper-plate crustal deformation on the prehistoric earthquake record may lead to inaccuracies in calculations of magnitudes and recurrence intervals if based on the Holocene stratigraphy of coastal bays. Evidence of Quaternary deformation offshore is equivocal in some cases, but the use of offshore datasets and coastal exposures together has increased the probability of identifying recent activity. Only two possible examples of
rapid subsidence coincident with crustal uplift were identified. However, these apparent anomalies could be explained by active upper-plate structures not deforming during every subduction event. Localised upper-plate deformation at other subduction zones. Localized upper-plate deformation has been documented at subduction margins world-wide, with deformation both synchronous with and independent of subduction zone events. Examples include the Hikurangi margin of New Zealand (Berryman et al. 1989; Cashman & Kelsey 1990; Berryman 1993a, b), the Alaskan margin (Plafker 1972), the Nankai forearc of SW Japan (Maemoku 1988a, b; Maemoku & Tsubono 1990; Sugiyama 1994), and the Huon peninsula of Papua New Guinea (Pandolfi et al. 1994). Holocene terraces along the coastal Hikurangi margin off eastern North Island, New Zealand, are uplifted by movement on steep reverse faults of the onshore accretionary prism (Berryman et al. 1989), with clustering of terrace ages along the coast. Stratigraphic and ecological studies of Holocene terrace sediments on the Mahia Peninsula reveal that sedimentation was progradational between events, implying a lack of interseismic subsidence that would be expected with a subduction earthquake cycle and supporting the formation of Holocene coseismically uplifted terraces by local crustal structures (Berryman et al. 1997). Other earthquakes within the accretionary prism include the 1931 Hawkes Bay earthquake (Ms 7.8), caused by a fault cutting up from the megathrust (Hull 1990), and the 1855 Wairarapa earthquake, which may have originated on the megathrust and propagated into the upper plate along a blind thrust fault (Darby & Beanland 1992). In Alaska, significant deviations from the regional subsidence or uplift patterns during the 1964 earthquake (up to 12 m uplift across the Patton Bay fault on Montague Island, relative to a regional 2-4 m of uplift) were associated with movement on crustal faults contemporaneous with the subduction zone earthquake (Plafker 1969, 1972). Along the Nankai margin of Japan, two types of subduction earthquake have been inferred from coseismically uplifted terraces (Fig. 13; Sugiyama 1994). Subduction events where no permanent crustal deformation and therefore no uplifted terrace preservation occurred are known as Taisho type events (T, Fig. 13). Preserved uplifted terraces resulting from the triggering of crustal deformation are known as Genroku type events (G, Fig. 13).
UPPER PLATE D E F O R M A T I O N AND CASCADIA EARTHQUAKES
~
G I bench
levell T1
~,~
G 2 bench / s ....
T
~
'
"
~'
'
T2
G1
T3
T4
G2
T5
time
Fig. 13. Illustration of the resulting record of Taisho and Genroku subduction earthquakes on the Nankai margin (after Sugiyama 1994; reproduced by kind permission of Geofisica International). No permanent inelastic crustal deformation occurs during Taisho events (Tl-5); coseismic deformation is recovered in the interseismic period leaving no permanent record of the earthquake. Genroku events (G1, G2) involve local faulting or other inelastic crustal deformation leading to preservation of an uplifted bench. It should be noted that the coseismic and interseismic vertical motions are opposite to those expected on much of the Cascadia coastline.
Effects o f local crustal deformation on subsidence records. Recorded coseismic subsidence is the net result of regional elastic strain release from a subduction zone earthquake and local crustal deformation (permanent and/or elastic), assuming a tectonic origin for subsidence. For each event, subsidence could result from strain release on the plate boundary or on local structures, or a combination of the two. The contributions of each cannot be determined for prehistoric events, although the amount and pattern of subsidence at each location may suggest a particular mechanism. The apparent rupture length (and hence magnitude), amount of subsidence, and timing of coseismic events can be specifically affected in the following ways. (1) Triggering of local crustal faults beyond (along strike) the subduction rupture zone, as the release of the elastic load on the upper plate in one area causes loading in other areas, thereby increasing the apparent rupture length and magnitude (e.g. LS2 and SZE1 in Fig. 14). Similar patterns occurred during the Landers earthquake, where a sequence of delayed ruptures occurred within the fault zone (Sieh et al. 1993; Wald & Heaton 1994; Spotila & Sieh 1995), and more distant earthquakes were also triggered following the mainshock (Hill et al. 1993), although these events lacked surface rupture and geodetic change. (2) The amount of subsidence per event at each location is dependent not only on the magnitude of the subduction zone earthquake and position of the rupture zone, but also on the amount of localized upper-plate deformation
335
accompanying the earthquake. The amount of subsidence during a given earthquake cannot be used for elastic dislocation modelling of the locked zone. (3) Local crustal faults may move independently of subduction zone earthquakes and produce anomalous local coseismic subsidence and marsh burial (LS1 of Fig. 14). Correlation of subsidence events from site to site is dependent on age control with sufficient precision to distinguish such events. The majority of radiocarbon ages from marsh burials have large error bars, of the order of +50 years to hundreds of years (e.g. Atwater 1992; Atwater et al. 1995), with errors often larger than the suggested recurrence intervals. More recently, AMS and high-precision radiocarbon and dendrochronology ages in some locations have significantly reduced errors to 4-10-20 years or even to within a year or season (Nelson et al. 1995; Jacoby et al. 1997; Yamaguchi et al. 1997), but suitable material for such precise dating techniques is often unavailable (Nelson et al. 1996b). The most abundant high-precision data are available for the most recent subsidence event, which is dated within a few decades of AD 1700, and is consistent with evidence for a remote tsunami in Japan at that time (Satake et al. 1996). Older
Fig. 14. Coastal marsh stratigraphy in hypothetical cores showing subsided marsh and soil deposits produced by a variety of coseismic events. LS, localized subsidence; SZE, subduction zone earthquake. SZE1 represents a small subduction earthquake which triggers localized subsidence to the north, increasing the apparent rupture length. SZE2 (Taisho type) does not trigger inelastic deformation and does not preserve coseismic subsidence. SZE3 is a widespread subduction event with subsidence at all four locations. This earthquake may also trigger local structures which may contribute to subsidence and help preserve the buried soils (Genroku type).
336
L . C . McNEILL E T AL.
events are less accurately dated. Where error bars are large, it is impossible to distinguish between regional subsidence events and locally anomalous events which may have resulted from independent deformation on crustal structures. These anomalous events may be wrongly correlated with regional coseismic subsidence events, producing an inaccurate picture of the earthquake rupture zone. In addition, subduction earthquake recurrence intervals may be underestimated if independent local subsidence events contribute to the marsh stratigraphy. Synchronous v. independent movement of local structures and subduction earthquakes. It is unknown, in the absence of historic subduction zone earthquakes and little upper-plate seismicity, whether local crustal structures in Cascadia are triggered by subduction zone earthquakes or operate independently, or both. If movement on crustal structures is always synchronous with and triggered by subduction events, estimates of the subduction earthquake recurrence interval will be unaffected but magnitude calculations may be inaccurate. If these structures operate independently, deforming both during and between subduction events, both the magnitude and recurrence interval of subduction zone events will be affected. If patterns of strain release are similar to those of the Alaskan and Nankai subduction zones, we might expect crustal structures to be triggered by slip on the megathrust (PlaNer 1969, 1972; Sugiyama 1994). Minimal historic seismicity in the coastal and shelf region supports the hypothesis that these structures are predominantly triggered by subduction zone earthquakes, which also lack seismicity. Buried marshes similar in age to regional subsidence events have been attributed to upper-plate structures in South Slough, southern Oregon coast, and Humboldt Bay, northern California coast, with little visible evidence of significant rapid subsidence in other bays in this region, such as the Siuslaw River (Clarke & Carver 1992; Nelson 1992; Nelson & Personius 1996; Nelson et al. 1996a). If rapid subsidence is not regionally extensive, these are examples of crustal structures that were triggered by subduction zone events. Regional subsidence in this area may be small and only detectable by biostratigraphic investigations (compare with Mathewes & Clague (1994)), and pronounced sudden subsidence may only be recorded where local structures were triggered. Upper-plate structures are likely to have longer recurrence intervals than subduction zone earthquakes and may not be triggered by every subduction event.
Local structures may also produce tectonic subsidence independently of subduction events. Independent coseismic subsidence has been suggested as a likely cause of marsh burials on the southern Oregon and northern California coasts (Nelson 1992; Nelson & Personius 1996).
Preservation o f buried marshes. The sequence of Cascadia buried marshes indicates net submergence of the land or net relative sea level rise of 2-5m in the last 2000-4000 years. If the earthquake strain cycle were completely elastic and no other factors were involved, coseismic subsidence and interseismic uplift would cancel out and no buried marshes would be preserved. This argument is used for the Nankai subduction zone, where coseismically uplifted terraces are only preserved permanently when synchronous upper-plate uplift occurs (Sugiyama 1994). If similar patterns of deformation to those at Nankai occur along the Cascadia subduction zone, permanent deformation by local structures may help to preserve marsh burial. Not all subduction zone earthquakes would be recorded (e.g. SZE2 in Fig. 14) and subduction zone earthquakes would appear to be less frequent with longer recurrence intervals. One major difference between the Nankai and Cascadia coastlines is the sense of coseismic motion: Nankai experiences uplift whereas Cascadia experiences subsidence. Therefore, preservation of buried marshes in Cascadia may also be influenced by the following nontectonic factors, producing relative sea-level rise; (1) late Holocene eustatic sea-level rise; (2) isostatic forebulge collapse following the last glacial maximum; (3) compaction; or (4) changes in the geometry of coastal estuaries (this could cause relative sea-level rise or fall). These factors would not contribute to the preservation of uplifted terraces in Nankai, unless sea level was falling in late Holocene time, or this region experienced some form of isostatic uplift. Buried marsh preservation in Cascadia could also be attributed to regional tectonic subsidence resulting from, for example, subduction erosion. The rates and effects of these factors in late Holocene time are poorly known and therefore their contributions to the preservation of buried marshes can only be approximated. The rate of late Holocene global eustatic sea-level rise is hotly debated, with sorne estimates pointing to negligible rise during the last 5000 years (P. Clark, W. R. Peltier, pers. comm., 1997), very low rates (Clark & Lingle 1979; Bard et al. 1996), or a value which is currently very difficult to separate from local and regional factors,
UPPER PLATE DEFORMATION AND CASCADIA EARTHQUAKES including isostatic and tectonic factors, which dominate relative sea-level rise (Bloom & Yonekura 1990; Nelson et al. 1996b). Estimates of forebulge collapse on much of the Cascadia margin associated with isostatic re-equilibration following the last glacial maximum (LGM) are given by the models of Peltier (1996). Subsidence rates (or relative sea-level rise) for much of the Washington and Oregon coastline are estimated as 0-1 ram/year, with a maximum on the northern Oregon coast (M2 model of Peltier (1996)). However, the northern Olympic Peninsula and Vancouver Island, which underlay the Cordilleran ice sheet during the LGM, should be experiencing isostatic rebound. Tectonic subsidence may indeed dominate the preservation of buried marshes, but until other variables are better resolved, this hypothesis remains untested. Quantitative subsidence calculations. In general, subsidence patterns measured along the Cascadia subduction zone appear to be fairly consistent, with subsidence of 0.5-2 m for each burial event. Very large differences in the amount of subsidence per event, which might indicate localized subsidence contributions, have not been observed, as pointed out by Clague (1997). The large deviation in uplift magnitude recorded across the Patton Bay fault in Alaska, is not observed in subsidence on the Cascadia margin. However, the Patton Bay fault, within the coseismically uplifted zone, is within the accretionary prism and close to the deformation front, and therefore might be expected to experience more pronounced deformation. Measurements of Cascadia subsidence are invariably imprecise because biostratigraphic markers such as diatoms and plant assemblages have large vertical water depth ranges, but fairly large differences in subsidence can be detected (Atwater & Hemphill-Haley 1996). Small variations in measured subsidence through careful lithostratigraphic and biostratigraphic studies, such as those of Long &Shennan (1994), Nelson et al. (1996b), and Shennan et al. (1996), may eventually allow the contribution of local structures to the subsidence record to be determined. The estuarine stratigraphy at many Cascadia sites is strikingly similar to that observed at passive margin sites (Long & Sherman 1994), where a coseismic origin is unlikely. Some Cascadia subsidence events may therefore have non-seismic origins such as natural succession of intertidal environments from local changes in sea level, sedimentation rates, and ocean currents (Long &Shennan 1994; Nelson et al. 1996a, b).
N-S
337
compression
Most active structures on the inner shelf and coast in the Cascadia subduction zone are characterized by roughly E-W trends, in contrast to the predominantly N-S to NW-SE trends which result from plate convergence on the continental slope. We agree with Snavely (1987) and Goldfinger et al. (1992b) that this landward region of the forearc is under N-S compression, which is in agreement with regional N-S compression throughout the continental northwestern USA derived from late Tertiary upper-plate fault orientations, earthquake focal mechanisms, and borehole breakouts (Werner et al. 1990; Zoback & Zoback 1989). The regional N-S compressional stress field extends onto the middle to outer shelf in Washington and much of Oregon. In contrast, the southern Oregon and northern California shelf and coastal region are within the active accretionary prism, and deformation is in response to plate convergence leading to structures with N-S to NW-SE trends. Wang et al. (1995) suggested that the NE-directed strain accumulation caused by plate convergence can be considered a time-dependent local perturbation superimposed on the regional N-S compressive stress field, and thus the regional stress field and cyclic loading may coexist. The transition from regional N-S compression to predominantly plate convergence driven compression represents a significant structural domain boundary. This transition may act as a backstop and may be related to the long-term average position of the downdip end of the seismogenic locked zone. Despite the apparent independence of upper-plate structures from subduction zone deformation, it seems likely that these structures could be triggered by rupture of the subduction zone. Independent fault movement in response to regional compression is also possible and therefore these structures pose independent seismic hazards.
Conclusions Evidence of Quaternary deformation on the Cascadia coast and inner shelf is widespread, with crustal downwarping or fault offset coincident with many coastal lowlands. Rapidly buried marshes at these locations may be due to elastic strain release on the subduction megathrust, downwarping or fault displacement on upper-plate crustal structures, or both. Calculations of Cascadia subduction zone earthquake recurrence intervals, rupture zones, and magnitudes based on correlations of marsh burial
338
L. C. McNEILL E T AL.
events between sites may be complicated by the possibility of localized crustal fault movements and fold growth, in addition to non-seismic origins of the observed stratigraphy. The earthquake record is likely to be more difficult to resolve with the interaction of these multiple factors (Fig. 14). Prehistoric subduction zone earthquakes may have been of lower magnitude than previously estimated. Recurrence intervals for such earthquakes may be overestimated, if some events are not preserved as a result of little permanent deformation, or underestimated, if anomalous local subsidence events are wrongly linked to similar-age regional events along the margin. Higher-resolution records of marsh chronology and estimates of subsidence may eventually lead to the separation of regional and local factors. Meanwhile, the use of these records for modelling of the subduction earthquake cycle and prediction of prehistoric earthquake rupture zones should be undertaken with caution. The inner shelf and coastal structures are consistent with the regional N - S compressional stress field and inconsistent with subduction-driven compression. Despite low seismicity, these crustal faults may be seismic and pose significant shaking and ground deformation hazards to the coastal communities. We acknowledge the Minerals Management Service Pacific OCS Region at Camarillo, California, for supplying data used in this study. We thank the crews of the Jolly Roger and Cavalier, DELTA submersible pilots, Williamson and Associates of Seattle, Washington for sidescan sonar operations, and the scientific crews of the 1992-1995 research cruises. L.C.M. wishes to thank B. Atwater, P. Clark, E. Clifton, H. Kelsey, P. McCrory, A. Nien, S. Obemeier, R. Petier, C. Peterson, G. Priest and I. Shennan for helpful discussions. However, the interpretations and conclusions of the paper are entirely the responsibility of the author. We also thank two anonymous reviewers for their helpful comments and suggestions. This study was supported by NOAA Undersea Research Program at the West Coast National Undersea Research Center, University of Alaska Grants UAF-92-0061 and UAF-93-0035, National Science Foundation Grants OCE-8812731 and OCE-9216880, and US Geological Survey National Earthquake Hazards Reduction Program awards 14-08-0001-G1800, 1434-93-G-2319, 1434-93-G2489, and 1434-95-G-2635.
References ADAMS, J. 1990. Paleoseismicity of the Cascadia subduction zone: evidence from turbidites off the Oregon-Washington margin. Tectonics, 9, 569-584.
ANDO,M. & BALAZS,E. I. 1979. Geodetic evidence for aseismic subduction of the Juan de Fuca plate. Journal of Geophysical Research, 84, 3023-3027. ATWATER, B. F. 1987. Evidence for great Holocene earthquakes along the outer coast of Washington State. Science, 236, 942-944. - - 1 9 9 2 . Geologic evidence for earthquakes during the past 2000 years along the Copalis River, southern coastal Washington. Journal of Geophysical Research, 97, 1901-1919. - - (compiler) 1994. Geology of Liquefaction Features about 300 Years Old along the Lower Columbia River at Marsh, Brush, Price, Hunting, and Wallace Islands, Oregon and Washington. US Geological Survey Open-File Report, 94-209. & HEMPHILL-HALEY, E. 1996. Preliminary Estimates of Recurrence Intervals for Great Earthquakes of the Past 3500 Years at Northeastern Willapa Bay', Washington. US Geological Survey Open-File Report, 96-001. -& YAMAGUCHI,D. K. 1991. Sudden, probably coseismic submergence of Holocene trees and grass in coastal Washington state. Geology, 19, 706-709. , NELSON, A. R., CLAGUE, J. J. et al. 1995. Summary of coastal geologic evidence for past great earthquakes at the Cascadia subduction zone. Earthquake Spectra, 11, 1-18. BARD, E., HAMELIN, B., ARNOLD, M., MOTAGGIONI, L., CABIOCH, G., FAURE, G., & ROUGERIE, F. 1996. Deglacial sea-level record from Tahiti corals and the timing of global meltwater discharge. Nature, 382, 241-244. BARNETT,E. T. 1997. Potentialjor coastalflooding due to coseismic subsidence in the central Cascadia margin. MS thesis, Portland State University, Portland. BERRYMAN,K. R. 1993a. Age, height, and deformation of Holocene marine terraces at Mahia Peninsula, Hikurangi subduction margin, New Zealand. Tectonics, 12, 1347-1364. - - 1 9 9 3 b . Distribution, age, and deformation of late Pleistocene marine terraces at Mahia Peninsula, Hikurangi subduction margin, New Zealand. Tectonics, 12, 1365-1379. , BEU, A. & IRWIN, S. 1997. A high resolution model for the formation of a coseismic Holocene marine terrace sequence. INQUA Commission on Neotectonics, Late Quaternary Coastal Tectonics Abstracts. , OTA, Y., & HULL, A. G. 1989. Holocene paleoseismicity in the fold and thrust belt of the Hikurangi subduction zone, eastern North Island, New Zealand. Tectonophysics, 163, 185-195. BLOOM, A. L. & YONEKURA, N. 1990. Graphic analysis of dislocated shorelines. In: NATIONAL RESEARCH COUNCIL (ed.) Studies in Geophysics." Sea-Level Change. National Academy Press, Washington, DC, 104-115. CASHMAN, S. M., & KELSEY, H. M. 1990. Forearc uplift and extension, southern Hawke's Bay, New Zealand: Mid-Pleistocene to present. Tectonics, 9, 23-44.
UPPER PLATE DEFORMATION CLAGUE, J. J. 1997. Evidence for large earthquakes at the Cascadia subduction zone. Reviews of Geophysics, 35, 439-460. CLARK, J. A., & LINGLE, C. S. 1979. Predicted relative sea level changes (18,000 BP to present) caused by late-glacial retreat of the Antarctic ice sheet. Quaternary Research, 11, 279-298. CLARKE, S. H. JR 1990. Map Showing Geologic
Structures of the Northern CaliJbrnia Continental Margin. Scale 1:250000. US Geological Survey, Map MF-2130. - - 1 9 9 2 . Geology of the Eel River Basin and adjacent region: implications for late Cenozoic tectonics of the southern Cascadia subduction zone and Mendocino triple junction. American Association of Petroleum Geologists Bulletin, 76, 199-224. - & CARVER, G. A. 1992. Late Holocene tectonics and paleoseismicity, southern Cascadia subduction zone. Science, 255, 188-192. , FIELD, M. E. & HIROZAWA, C. A. 1985. Re-
connaissance Geology and Geologic Hazards of the Offshore Coos Bay Basin, Oregon. US Geological Survey Bulletin, 1645. CLIFTON, H. E. 1983. Discrimination between subtidal and intertidal facies in Pleistocene deposits, Willapa Bay, Washington. Journal of Sedimentary Petrology, 53, 353-369. - - 1 9 9 4 . Transgressive and Early Highstand System
Tracts: Pleistocene terrace deposits, Willapa Bay, Washington. Field guide prepared in conjunction with a Society of Economic and Petroleum Geologists Research Conference, Clastic Deposits of the Transgressive Systems Tract: Facies, Stratigraphy and Reservoir Character. COOPER, D. M. 1981. Sedimentation, stratigraphy and
facies variations of the lower to middle Miocene Astoria Formation in Oregon. PhD thesis, Oregon State University, Corvallis. CROSSON, R. S. & OWENS, T. J. 1987. Slab geometry of the Cascadia subduction zone beneath Washington from earthquake hypocenters and teleseismic converted waves. Geophysical Research Letters, 14, 824-827. DARBY, D. & BEANLAND, S. 1992. Possible source models for the 1855 Wairarapa earthquake, New Zealand. Journal of Geophysical Research, 97, 12375-12389. DARIENZO, M. E. & PETERSON, C. D. 1990. Episodic tectonic subsidence of late Holocene salt marshes, northern Oregon central Cascadia margin. Tectonics, 9, 1-22. & CLOUGH, C. 1994. Stratigraphic evidence for great subduction-zone earthquakes at four estuaries in northern Oregon, USA. Journal of Coastal Research, 10, 850-876. DEMETS, C., GORDON, R. G., ARGUS, D. F. & STEIN, S. 1990. Current plate motions. Geophysical Journal International, 101, 425-478. GOLDFINGER, C. 1994. Active deformation of the
Cascadia forearc: Implications for great earthquake potential in Oregon and Washington. PhD thesis, Oregon State University, Corvallis. , KULM, L. D. & YEATS, R. S. 1992a. Neotectonic
Map of the Oregon Continental Margin and
AND CASCADIA EARTHQUAKES
339
Adjacent Abyssal Plain. Scale 1:500 000. Oregon Department of Geology and Mineral Industries, Open-File Report, 0-92-4. , APPELGATE, B., MACKAY, M. & COCHRANE, G. R. 1996. Active strike-slip faulting and folding of the Cascadia plate boundary and forearc in central and northern Oregon. In: ROGERS, A. M., WALSH, T. J., KOCKELMAN, W. J. & PRIEST, G. (eds) Assessing and Reducing
Earthquake Hazards in the Pacific Northwest. US Geological Survey Professional Paper, 1560, 223-256. & MOORE, G. F. 1992b. Transverse structural trends along the Oregon convergent margin: implications for Cascadia earthquake potential. Geology, 20, 141-144. , MCNEILL, L. C., & HUMMON, C. ]997a.' Oblique strike-slip faulting of the central Cascadia submarine forearc. Journal of Geophysical Research, 102, 8217-8243. --, MCNEILL, L. C. & HUMMON, C. 1997b. Case study of GIS data integration and visualization in submarine tectonic investigations: Cascadia subduction zone. Marine Geodesy, 20, 267-289. GOWER, H. D. 1960. Geologic Map of the Pysht Quadrangle, Washington, scale 1:62 500. US Geological Survey Geologic Quadrangle Map, GQ-129.
GRIM, M. S. & BENNETT, L. R. JR 1969. Shallow seismic profiling of the continental shelf off Grays Harbor, Washington. University of Washington Department of Oceanography Special Reports, 41, 72-92. HEATON, T. H. & KANAMORI, H. 1984. Seismic potential associated with subduction in the northwestern United states. Bulletin of the Seismological Society of America, 74, 993-941. HILL, D. P. et al. 1993. Seismicity remotely triggered by the magnitude 7.3 Landers, California earthquake. Science, 260, 1617-1623. HULL, A. G. 1990. Tectonics of the 1931 Hawke's Bay earthquake. New Zealand Journal of Geology and Geophysics, 33, 309-330. JACOBY, G. C., BUNKER, D. E. & BENSON, B. E. 1997. Tree-ring evidence for an AD 1700 Cascadia earthquake in Washington and northern Oregon. Nature, 25, 999-1002. KELSEY, H. M. 1990. Late Quaternary deformation of marine terraces on the Cascadia subduction zone near Cape Blanco, Oregon. Tectonics, 9, 983-1014. , TICKNOR, R. L., BOCKHEIM, J. G. & MITCHELL, C. E. 1996. Quaternary upper plate deformation in coastal Oregon. Geological Society of America Bulletin, 108, 843-860. , WITTER, R. C. & POLENZ,*M. 1993. Cascadia paleoseismic record derived from late Holocene fluvial and lake sediments, Sixes River Valley, Cape Blanco, south coastal Oregon (abstract).
Eos Transactions, American Geophysical Union, 74, 199. , & HEMPHILL-HALEY, E. 1998. Response of a small Oregon estuary to coseismic subsidence and postseismic uplift in the past 300 years. Geology, 26, 231-234.
340
L.C.
M c N E I L L E T AL.
KENNEDY, G. L. 1978. Pleistocene paleoecology, zoogeography and geochronology of marine invertebrate faunas of the Pacific Northwest coast (San Francisco Bay to Puget Sound). PhD thesis, University of California, Davis. , LAJOIE, K. R. & WEHMILLER, J. F. 1982. Aminostratigraphy and faunal correlations of late Quaternary marine terraces, pacific Coast, USA. Nature, 299, 545-547. KULM, L. D. & SUESS, E. 1990. Relation of carbonate deposits and fluid venting: Oregon accretionary prism. Journal of Geophysical Research, 95, 8899-8915. KVENVOLDEN, K. A., BLUNT, D. J. & CLIFTON, H. E. 1979. Amino-acid racemization in Quaternary shell deposits at Willapa Bay, Washington. Geochimica et Cosmochimica Acta, 43, 1505-1520. LONG, A. J. & SHENNAN, I. 1994. Sea-level changes in Washington and Oregon and the 'earthquake deformation cycle'. Journal of Coastal Research, 10, 825-838. LUDWIN, R. S., WEAVER, C. S. & CROSSON, R. S. 1991. Seismicity of Washington and Oregon. In: SLEMMONS, D. B., ENGDAHL, E. R., BLACKWELL,D. & SCHWARTZ, D. (eds) Neotectonics of North America. Geological Society of America, DNAG CSMV-1, Boulder, CO, 77-98. MAEMOKU, H. 1988a. Holocene crustal movement in Muroto Peninsula, southwest Japan. Geographical Review of Japan (Series A), 61(10), 747-769. - - 1 9 8 8 b . Holocene crustal movement around Cape Ashizuri, southwest Japan. Geographical Sciences, 43, 231 240. - & TSUBONO, K. 1990. Holocene crustal movements in the southern part of Kii Peninsula, outer zone of southwest Japan. Journal of Geography, 99, 349 369. MATHEWES, R. W. & CLAGUE, J. J. 1994. Detection of large prehistoric earthquakes in the Pacific Northwest by microfossil analysis. Science, 264, 688-691. MCCAFFREY, R. & GOLDFINGER, C. 1995. Forearc deformation and great subduction earthquakes: Implications for Cascadia earthquake potential. Science, 267, 856-859. MCCRORY, P. A. 1996. Tectonic model explaining divergent contraction directions along the Cascadia subduction margin, Washington. Geology, 24, 929 932. MCINELLY, G. W. & KELSEY, H. M. 1990. Late Quaternary tectonic deformation in the Cape Arago-Bandon region of coastal Oregon as deduced from wave cut platforms. Journal of Geophysical Research, 95, 6699-6713. MCNEILL, L. C., PIPER, K. A., GOLDFINGER, C., KULM, L. D. & YEATS, R. S. 1997. Listric normal faulting on the Cascadia continental shelf. Journal of Geophysical Research, 102, 12 123-12 138. NELSON, A. R. 1987. Apparent gradual rise in relative sea level on the south-central Oregon coast during the late H o l o c e n e - implications for the great Cascadia earthquake hypothesis (abstract). Los Transactions, American Geophysical Union, 68, 1240.
-1992. Holocene tidal-marsh stratigraphy in SouthCentral O r e g o n - evidence for localized sudden submergence in the Cascadia subduction zone. In: FLETCHER, C. H. & WEHMILLER, J. F. (eds) Quaternary Coasts of the United States." Marine and Lacustrine Systems. Society for Sedimentary Geology Special Publication, 48, 287-301. --, & PERSONIUS, S. F. 1996. Great-earthquake potential in Oregon and Washington an overview of recent coastal geologic studies and their bearing on segmentation of Holocene ruptures, central Cascadia subduction zone. In: ROGERS, A. M., WALSH,T. J., KOCKELMAN,W. J. & PRIEST, G. (eds) Assessing and Reducing Earthquake hazards' in the Pacific Northwest. US Geological Survey Professional Paper, 1560, 91-114. --, ATWATER, B. F., BROBOWSKI, P. T. et al. 1995. Radiocarbon evidence for extensive plate-boundary rupture about 300 years ago at the Cascadia subduction zone. Nature, 378, 371-374. --, JENNINGS, A. E. & KASHIMA, K. 1996a. An earthquake history derived from stratigraphic and microfossil evidence of relative sea-level change at Coos Bay, southern coastal Oregon. Geological Society of America Bulletin, 108, 141-154. --, SHENNAN, I. • LONG, A. J. 1996b. Identifying coseismic subsidence in tidal-wetland stratigraphic sequences at the Cascadia subduction zone of western North America. Journal of Geophysical Research, 101, 6115-6135. NIEM, A. R. & NIEM, W. A. 1985. Oil and Gas Investigations of the Astoria Basin, Clatsop and Northernmost Tillamook Counties, Northwest Oregon. Scale 1:250000 Oregon Department of Geology and Mineral Industries Oil and Gas Investigation, 14. , SNAVELY, P. D., JR & NIEM, W. A. 1990. Onshore-Offshore Geologic Cross Section from the Mist Gas Field, Northern Oregon Coast Range, to the Northwest Oregon Continental Shelf Oregon Department of Geology and Mineral Industries Oil and Gas Investigation, 17. OBERMEIER, S. F. 1995. Preliminary Estimates of the Strength of Prehistoric Shaking in the Columbia River Valley and the Southern Hall" of Coastal Washington, with Emphasis for a Cascadia Subduction Zone Earthquake about 300 Years ago. US Geological Survey Open-File Report, 94-589. OPPENHEIMER, D., BEROZA, G., CARVER, G. et al. 1993. The Cape Mendocino, California, earthquakes of April 1992: subduction at the triple junction. Science, 261, 433-438. ORANGE, D. L., GEDDES, D. S., & MOORE, J. C. 1993. Structural and fluid evolution of a young accretionary complex: the Hoh rock assemblage of the western Olympic Peninsula, Washington. Geological Society' of America Bulletin, 105, 1053-1075. PALMER, S. P. & LINGLEY, W. S. 1989. An Assessment o/ the Oil and Gas Potential of the Washington Outer Continental Shelf. Washington State Division of Geology and Earth Resources Report, WSG 8 9 - 2 .
UPPER PLATE DEFORMATION AND CASCADIA EARTHQUAKES PANDOLFI, J. M., BEST, M. M. R. & MURRAY, S. P. 1994. Coseismic event of May 15, 1992, Huon Peninsula, Papua New Guinea: comparison with Quaternary tectonic history. Geology, 22, 239-242. PARKER, M. J. 1990. The Oligocene and Miocene geology of the Tillamook embayment, Tillamook County, northwest Oregon. MS thesis, Oregon State University, Corvallis. PELTIER, W. R. 1996. Mantle viscosity and ice-age topography. Science, 273, 1359-1364. PETERSON, C. D. & DARIENZO, M. E. 1989. Episodic, abrupt tectonic subsidence recorded in late Holocene deposits of the South Slough syncline: an onland expression of shelf fold belt deformation from the southern Cascadia margin. Geological Society of America Abstracts with Programs, 21, 129. & 1996. Discrimination of climatic, oceanic, and tectonic mechanisms of cyclic marsh burial, Alsea Bay, Oregon. In: ROGERS, A. M., WALSH, T. J., KOCKELMAN, W. J. & PRIEST, G. (eds) Assessing and Reducing Earthquake Hazards in the Pacific Northwest. US Geological Survey Professional Paper, 1560, 115-146. & PRIEST, G. R. 1995. Preliminary reconnaissance survey of Cascadia paleotsunami deposits in Yaquina Bay, Oregon. Oregon Geology, 57, 33 40. --, BARNETT, E., BRIGGS, G. G., CARVER, G. A., CLAGUE, J. J. & DARIENZO, M. E. 1997. Subsidence from Great Earthquakes in the Cascadia Subduction Zone, Vancouver Island, B.C., Washington, Oregon, and Northernmost California. Department of Oregon Geology and Mineral Industries, Open-File Report, 0-97-5. PLAFKER, G. 1969. Tectonics of the March 27 Alaskan earthquake. US Geological Survey Professional Paper, 543-I. - - 1 9 7 2 . Alaskan earthquake of 1964 and Chilean earthquake of 1960: implications for arc tectonics. Journal of Geophysical Research, 77, 901-925. PR1EST, G. R., SAUL, I. & DIEBENOW,J. 1994. Chronic Geologic Hazard Maps and Erosion Rate Database, Coastal Lincoln Count)', Oregon: Salmon River to Seal Rocks, 1:4800 scale maps. Oregon Department of Geology and Mineral Industries Open-File Report, O-94-11. RAU, W. W. 1973. Geology of the Washington Coast between Point Grenville and the Hoh River. Washington Division of Geology and Earth Resources Bulletin, 66. RYAN, H. F. & STEVENSON,A. J. 1995. Cruise Reportfor C1-94-0W: Reconnaissance High Resolution Geopulse Data Acquired for Seismic Hazard Studies along the Columbia River from July 18 22, 1994. US Geological Survey Open-File Report, 95-668. SATAKE, K., SHEMAZAKI,K., YOSHINOBU,T. & UEDA, K. 1996. Time and size of a giant earthquake in Cascadia inferred from Japanese tsunami records of January 1700. Nature, 379, 246-249. SHENNAN, I., LONG, A. J., RUTHERFORD, M. M e t al. 1996. Tidal marsh stratigraphy, sea-level change and large earthquakes, I: a 5000 year record in Washington, USA. Quaternary Science Reviews, 15, 1023-1059. -
-
-
-
341
SIEH, K., JONES, L., HAUKSSON, E. et al. 1993. Nearfield investigation of the Landers earthquake sequence, April to July 1992. Science, 260, 171-176. SNAVELY, P. D., JR 1976. Geologic map of the Yaquina and Toledo Quadrangles, Lincoln County, Oregon, scale 1:62 500. US Geological Survey Miscellaneous Investigations Series Map, 1-867. 1987. Tertiary geologic framework, neotectonics, and petroleum potential of the Oregon-Washington continental margin. In. SCHOLL, D. W., GRANTZ, A. & VEDDER, J. G. (eds) Geology and Resource Potential of the Continental Margin oJ Western North America and Adjacent Ocean Basins - Beau[ort Sea to Baja California. CircumPacific Council for Energy and Mineral Resources, Houston, TX, 305-335. -& WELLS, R. E. 1996. Cenozoic evolution of the continental margin of Oregon and Washington. In." ROGERS, A. M., WALSH, T. J.; KOCKELMAN, W. J., & PR1EST, G. (eds) Assessing and Reducing Earthquake Hazards in the Pacific Northwest., US Geological Survey Professional Paper, 1560, 161 182. SPOTILA, J. c~zSIEH, K. 1995. Geologic investigations of the 'slip gap' in the surficial ruptures of the 1992 Landers earthquake, southern California. Journal of Geophysical Research, 100, 543-559. SUGIYAMA, Y. 1994. Neotectonics of Southwest Japan due to the right-oblique subduction of the Philippine Sea plate. Geofisica lnternacional, 33, 53 76. TABOR, R. W. & CADY, W. M. 1978. The Structure of the Olympic Mountains, Washington-Analysis of a Subduction Zone. US Geological Survey Professional Paper, 1033. TICKNOR, R. L. 1993. Late Quaternary crustal deJormation on the central Oregon coast as deduced from uplifted wave-cut platforms. MS thesis, Western Washington University, Bellingham. --, KELSEV, H. M. & BOCKHEIM, J. G. 1992. Late Quaternary tectonic deformation along the central Oregon portion of the Cascadia margin as deduced from deformation of wave-cut platforms. Geological Society of America Abstracts with Programs, 24, 86. WAGNER, H. C. & TOMSON, J. H. 1987. Offshore Geolog)' of the Strait of Juan de Fuca, State of Washington and British Columbia, Canada. Washington Division of Geology and Earth Resources Open File Report, 87-1. WALD, D. & HEATON, T. 1994. Spatial and temporal distribution of slip for the 1992 Landers, California, earthquake. Bulletin of the Seismological Society of America, 84, 668-691. WALSH, T. J., KOROSEC, M. A., PHILLIPS, W. M., LOGAN, R. L. & SCHASSE, H. W. 1987. Geologic Map of Washington- Southwest Quadrant. Washington Division of Geology and Earth Resources, Geologic Map, GM-34. WANG, K., MULDER, T., ROGERS, G. C. & HYNDMAN, R. D. 1995. Case for very low coupling stress on the Cascadia subduction fault. Journal of Geophysical Research, 100, 12 907-12 918.
342
L.C.
M c N E I L L E T AL.
WEAVER, C. S. & BAKER, G. E. 1988. Geometry of the Juan de Fuca plate beneath Washington and northern Oregon from seismicity. Bulletin of the Seismological Society of America, 78, 264-275. WELLS, R. E., N~M, A. R. & SNAVELu P. D., JR 1994. Young deformation at Netarts Bay and elsewhere, northern Oregon coast (abstract). Proceedings of the Oregon Academy of Sciences, XXX, 39. , SNAVELY, P. D., JR & NIEM, A. R. 1992. Quaternary thrust faulting at Netarts Bay, northern Oregon coast. Geological Society of America Abstracts with Programs, 24, 89. WERNER, K. S., GRAVEN, E. P., BERKMAN, T. A. & PARKER, M. J. 1990. Direction of maximum horizontal compression in northwestern Oregon determined by borehole breakouts. Tectonics, 10, 948-958.
WEST, D. O. 1986. WNP-3 Geologic Support Services Coastal Terrace Study. Washington Public Power Supply System, Richland, 853-1010. -& MCCRUMB, D. R. 1988. Coastline uplift in Oregon and Washington and the nature of Cascadia subduction zone tectonics. Geology, 16, 169-172.
YAMAGUCHI, D. K., ATWATER, B. F., BUNKER, D. E., BENSON, B. E. & REID, M. S. 1997. Tree-ring dating the 1700 Cascadia earthquake. Nature, 389, 922-923. ZOBACK, M. L. & ZOBACK, M. D. 1989. Tectonic stress field of the continental United States. In. PAKISER, L. & MOONEY, W. (eds) Geophysical Framework of the Continental United States. Geological Society of America, Memoir, 172, 523-540
Late Holocene coastal tectonics at Falasarna, western Crete: a sedimentary study DALE
DOMINEY-HOWES
1"2, A L A S T A I R
DAWSON 2 & DAVID
SMITH 2
1 Coventry Centre Jor Disaster Management, School of the Built Environment, Coventry University, Coventry CV1 5FB, UK (e-mail:
[email protected]) 2 Centre for Quaternary Science, Division of Geography, Coventry University, Coventry CV1 5FB, UK Abstract: The late Holocene sedimentary record of Falasarna Harbour, western Crete, includes detailed evidence of tsunamis and serves as an independent dataset to evaluate the magnitude and timing of coastal tectonic movements in an area affected by contrasting tectonic regimes. Analysis of a foraminiferal assemblage makes it possible to identify suites of tsunami-deposited sediments within normal sedimentary sequences. The palaeo-environmental record is then complemented with a sequence of raised fossil marine notches. The transitional boundary between marine and terrestrial sedimentation indicates tectonic uplift at C.AD63-75+90 radiocarbon years BP, which is in conflict with previously published interpretations. No sedimentary evidence can be found for a tsunami believed to be associated with a large uplift event during AD365.
This paper presents the results of an investigation at Falasarna Harbour, western Crete, of sediments deposited by tsunamis reported to have occurred in the Aegean Sea region of Greece. Falasarna is located within an active extensional domain inboard of the compressional front associated with subduction of the Mediterranean plate, and, is therefore affected by different tectonic regimes. Any sedimentary evidence for tsunamis may help to shed light on the nature of coastal tectonic activity in this area. Biostratigraphic (Foraminifera) and lithostratigraphic evidence was used to determine the palaeoenvironmental history of Falasarna. We show that, following the construction of the harbout, sedimentation progressed until the site was affected by tectonic movement. The sedimentological data indicate that the harbour of Falasarna was raised above sea level, as reflected by a change from marine to terrestrial conditions. The data do not preserve evidence for a large vertical coseismic displacement reported to have occurred in AD 365 (Pirazzoli 1986).
Tectonic setting On a regional scale, the present form of the Aegean is the result of a set of complex interactions between phases of compressional and extensional tectonic normal faulting which result from the southward stretching and subsidence of the Aegean plate (Le Pichon & Angelier 1981; Mercier 1981). The Aegean region is composed of the Inner Hellenic Volcanic Arc and the Outer
Hellenic Arc (with subducting trench system) (Fig. 1). The Outer Hellenic Arc forms part of a rigid body thrusting over the Mediterranean basins and is associated with the development of the accretionary complex and a compressional front located south of Crete at the junction between the African and European plates (Lallemant et al. 1994). The Falasarna study site lies within an active extensional domain inboard of the compressional front and is found at the western end of the island of Crete, the most prominent feature of the Outer Hellenic Arc and which has been affected by uplift during the Holocene. The uplift is related to the development of the compressional front and the accretionary wedge by underplating of sediments from the downgoing plate beneath the continental upper plate (Le Pichon & Angelier 1981; Lallemant et al. 1994). Figure 1 provides a schematic representation of the main structural and tectonic components for the region in which Falasarna is located. It should be noted that the area to the south of western Crete is characterized by a series of E - W trending faults, whereas to the west of Crete, the main offshore faults trend N-S. To the NW of Crete, the orientation of submarine faults changes to NW-SE. Such radically different tectonic regimes in closely adjoining areas have prevailed since early Quaternary time (Angelier 1978); Jackson (1994) noted that the faults associated with the Outer Hellenic Arc are affected by normal, reverse and thrust movement. The area considered in this study is affected by both relatively shallow earthquakes associated with extensional faulting of the overriding
DOMINEY-HOWES, D., DAWSON, A. & SMITH, D. 1998. Late Holocene coastal tectonics at Falasarna, western Crete: a sedimentary study. In: STEWART, I. S. & VrrA-FINZI, C. (eds) Coastal Tectonics. Geological Society, London, Special Publications, 146, 343-352.
344
D. DOMINEY-HOWES E T AL.
SEA OF CRETE
%
I
km
0
t
1 O0
Fig. 1. Schematic representation of the main structural components of the Outer Hellenic Arc, subduction system and compressional front (Mediterranean Ridge and deformation front). Location of the study site is shown by the square.
upper-crustal plate and by deeper earthquakes associated with thrust movement of the subducting plate (Taymaz et al. 1990; Lallemant et al. 1994; Stewart, pers. comm. 1997).
Conceptual framework and methods It has been reported that the harbour of Falasarna contains a record of late Holocene sedimentation and that the harbour sediments include
deposits laid down by tsunamis which flooded the area in AD 66 and AD 365 (Pirazzoli et al. 1992). To elucidate the palaeoenvironmental history of this site, the lithostratigraphy of sedimentary sequences was investigated in five trenches excavated by the Greek Regional Archaeological Service at Falasarna. The elevation of all units was determined by instrumental levelling using a Zeiss Autoset Level. All levelling traverses were closed with no error greater than +0.01 m. In the
COASTAL TECTONICS IN CRETE absence of a Greek Datum, present sea level was assumed to be true mean sea level, as tidal variations rarely exceed 10 cm (P. A. Pirazzoli, pers. comm., 1994). It is recognized, however, that sea-level variations as a result of tidal cycles, atmospheric conditions, the nature of the geoid, storminess and seasonality may result in variations of mean sea level approaching 50cm (IAPSO 1985; Emery et al. 1988; Flemming & Woodworth 1988; Flemming 1992). Variations in the stratigraphy based upon changes in clast content, colour, lithology, matrix, shell content, stone content, clast roundness, structure and texture were recorded. Contiguous 0.05 m bulk samples (125 cm 3) of sediments were collected successively through the sequence from the base to the top of each trench for laboratory investigation including detailed biostratigraphic analysis based on Foraminifera. Foraminifera counting was carried out using a VMT 12 microscope at x 1 and x4 magnification. Reference was made to type collections at the Department of Micropalaeontology at the Natural History Museum, London, and the accounts of Sidebottom (1904-1909) and Cimerman & Langer (1991). Wherever possible, 300 individuals were counted in each sample to ensure statistical confidence. Particle size analysis was unsuccessful because detailed variations were obscured by the sample size adopted and since over 80% of the sediment is composed of marine biogenic CaCO3 (Pirazzoli et al. 1992).
Falasarna Harbour, site description Falasarna was a pirate port which operated between approximately the middle of the fourth century BC and the late first century BC (Hadjidaki 1988). The date at which the harbour became functional and the date at which it was finally abandoned are not known exactly (Hadjidaki 1988; Frost 1997), although the harbour was in existence by the time of Scylax of Caryanda in the middle of the fourth century BC (Hadjidaki 1988). The harbour is situated next to the Bay of Livadi in an enclosed position behind the eastern side of Cape Kutri at the southern end of the Grammvousa peninsula in western Crete. It is an quadrangular artificial harbour cut into the surrounding Mesozoic limestone and Scylax referred to Falasarna's status as a closed harbour (Pirazzoli et al. 1992, p. 375). The main harbour is 1 0 0 m • in size and is divided from a secondary basin 50m x 35m lying immediately to the east by a complex of walls and buildings. The main harbour is connected to the sea at its western side by a
345
channel which stretches 100m to the present shoreline (Fig. 2). The ground surface within the main harbour is at an elevation of 6.6 m above sea level (m a.s.1.). Hadjidaki (1988) suggested that Falasarna was probably destroyed by the Romans in 67 BC. She stated that the Romans sent Caecillius Metellus as a praetor to Crete to destroy a number of pirate strongholds. According to Pirazzoli et al. (1992), after its destruction the harbour rapidly filled with marine and terrestrial sediments and was inundated by tsunamis in AD66 and in AD 365, when it was uplifted to 6.6 m a.s.1, by an earthquake on 21 July. There is no evidence at Falasarna of continued occupation after the destruction of 67 BC.
Sedimentology In this paper only the results of the sedimentological analysis of Trench A are presented because this trench displays the clearest sedimentary record (Dominey-Howes 1996). Trench A is located within the main harbour basin (Fig. 2). The surface is at an elevation of 6.2m a.s.1, and the base is at 4.6 m a.s.1. (Fig. 3). Five lithostratigraphic units occur within the sedimentary sequence. The basal unit is a Foraminifera- and molluscrich, well-sorted fine to medium sand. The unit contains coarse rounded to angular grit. There are many small to medium-sized well-rounded to subangular limestone, sandstone and quartz clasts. The matrix contains many whole and comminuted marine molluscs. Many small subrounded pottery sherds are visible. Some crude horizontal bedding is apparent. The upper boundary of this unit is defined by an undulating unconformity. The basal unit is overlain by a mollusc- and Foraminifera-rich marine sand which contains rounded fine grit. The unit is further characterized by large numbers of whole and broken marine molluscs with no obvious orientation. Medium to large subrounded to subangular limestone and sandstone clasts and blocks (a-axis up to 10 cm) are present. There is no obvious bedding or structure, although clast a-axis tends towards the horizontal. The upper boundary of this unit is defined by an undulating unconformity. The middle unit is composed of a fine to medium-sized sand and grit and is further characterized by abundant well-rounded to subangular limestone and sandstone clasts. The smaller clasts are matrix supported whereas the larger blocks (a-axis from 20cm) are clast supported. The a-axis of the clasts tends towards the horizontal although there is no other
346
D. DOMINEY-HOWES E T A L .
Fig. 2. Detail of Falasarna Harbour and the position of the trenches. This paper presents the results of only Trench A. Cape Kutri is located immediately to the north of the harbour. The main channel connects with the Bay of Livadi.
obvious bedding or structures. Both whole and comminuted molluscs are present, as are small rounded pottery sherds. This unit is conformably overlain by a red silty clay sand which extends to the surface of the trench. There are many small to large subrounded to very angular limestone, sandstone and quartz clasts which are matrix supported. A few broken marine molluscs are found towards the base of this unit and there is no apparent bedding or structure. This unit is interrupted by numerous subrounded limestone blocks at 5.70-5.90ma.s.1. (Fig. 3). This unit represents the sediments reported to
have been deposited by a tsunami in AD365 (Pirazzoli et al. 1992). The a-axis of these blocks is up to 18 cm in length, and is approximately horizontal, and the blocks are matrix supported. More than 5200 Foraminifera were extracted and identified from the Trench A sediments and 28 species were identified (Table 1). The number of individuals (300+) per sample is high from the bottom of the trench upwards as far as sample 14 (Table 1); between samples 15 and 20 it declines rapidly and remains low (nine individuals in sample 21 to four in sample 23). None is recorded from sample 24 upwards to the surface
COASTAL TECTONICS IN CRETE
347
Fig. 3. Stratigraphy of Trench A according to the present authors and Pirazzoli et al. (1992). It can be clearly seen that the elevations of the base and top of Trench A vary between the two studies. (For a description of the lithostratigraphic units, refer to the text.) of the trench. The assemblage of Foraminifera present in samples 1-17 is dominated by Amm-
onia ber Ammonia parkinsoniana, Ammonia tepida, Elphidium advenum, Elphidium crispum, Globigerina ruber, Quinqueloculina aspera, Quinqueloculina bicornis and Quinqueloculina vulgaris (Table 1). These species make up 64% of the Foraminifera present within the Trench A samples. The rarest species of Foraminifera preserved are Quinqueloculina jugosa, Lachlanella variolata and Triloculina tricarinata, which constitute 0.4%, 1.6% and 1.8% of the total count, respectively. The number of individuals per sample of Cibicides advenum increases from an average of six specimens per sample in samples 1-14 to an average of 11 specimens per sample in samples 15-20. Similarly, Eponides repanda increases from an average of 2.2 specimens per sample in samples 1-14 to seven specimens per sample in samples 15-20. It is also worth noting that there is an increase in the number of broken tests for all species from an average of 21% of tests in samples 1-14 and 21-23 to an average of 55% of tests in samples 15-20. Pennate forms of Foraminifera total 555 specimens and the average percentage of broken tests is 26% in samples 1-14 and 21-23, compared with 83% in samples 15-20. The average percentage of broken centric forms in samples 1-14 and 21-
23 is 21%, and in samples 15-20 is 18%. Therefore, a higher percentage of pennate to centric Foraminifera are broken in samples 15-20. Table 2 gives the results of radiocarbon dating. The AMS (accelerator mass spectrometry) technique was used as no sample weighed more than 2.0 g. For the purpose of consistency all samples submitted for dating comprised specimens of the marine mollusc Hydrobia acuta. Calibration to calendar years was made by reference to the data of Stuiver & Braziunas (1993). The 13C/12C ratios are compared with those given as - 5 + 40%0 for the eastern Mediterranean by Stuiver & Braziunas (1993). As they lie well within the - 4 5 to +35%o range (Table 2) the shell ages reported here are taken to be reliable.
Interpretation of palaeoenvironmental history The most striking aspect of the lithostratigraphy of Trench A is that there is a clear change in the pattern of the sedimentation at c. 5.70 m a.s.l. and this change dates from AD 63 to AD 75 + 90 radiocarbon years BP. The abrupt change of sedimentation is from marine conditions characterized by the deposition of Foraminiferamollusc-rich sands to terrestrial silts and clay
348
D. D O M I N E Y - H O W E S
/-.,
'~
~
E T AL.
~
~
rT1
,.--..
~6
b
,~.~
~
~.'t.
:~ ~
~
~
~
~....~
.~
,..,,-.-.~
~
~
~
~
.-,.
~
,~
COASTAL TECTONICS IN CRETE
349
Table 2. 14C AMS dates of samplesfrom Trench A Sample
Measured
Conventional
13C/nC(%0)
Calibrated
Beat-81412 Beta-81413 Beta-81414 Beta-81415
2460 + 70 1900 + 60 1890 + 90 21104-60
2870 + 70 2270 + 60 2290 + 90 2530+60
0.0 -2.4 -0.8 +0.7
770-555 Bc AD25-I 50 35 B~AD160 335-155BC
The dates are plotted on Fig. 3.
(with some sand). This change is sudden, as indicated by the boundary unconformity, and is thought to represent relative sea- and land-level changes associated with vertical coseismic deformation. The conditions in which the deposition of large, well-rounded to subangular limestone and sandstone blocks between 5.20 and 5.60m a.s.1. were deposited are thought to be different from those that had operated during marine sedimentation. It is believed by the present authors that deposition of the larger blocks at this level probably relates to a high-energy low-frequency event such as a tsunami. The evidence to support this interpretation is described below. Samples 15-20 in Table 1 correspond lithostratigraphically to the sediments ascribed by Pirazzoli et al. (1992) to a tsunami in AD66 (Fig. 3). Sample 15 lies unconformably on top of the underlying sediments. The evidence provided by the Foraminifera shows clearly that the last episode of marine sedimentation within the harbour was probably during or just after the AD 66 tsunami event reported by Pirazzoli et al. (1992). There is no evidence for marine sedimentation taking place anywhere within the harbour after this event. Of the Foraminifera recovered from Trench A, 64% belong to the Ammonia, Elphidium and Quinqueloculina genera (Table 1). According to Murray (1991), these three genera are all found together in the shallow inner-shelf region. The Ammonia and Elphidium genera are also representative of brackish conditions, but the presence of Quinqueloculina is representative of open marine-lagoonal conditions. Parker (1958) identified a 'typical' shallow-depth assemblage between 0 and 25 m, which he referred to as the 'bay-open marine' assemblage. The presence of the high numbers of Ammonia, Elphidium, Miliolidae and Peneroplidae indicate a clear shallow, fully marine (lagoonal) assemblage at Falasarna similar to that identified by Parker (1958). Murray (1991) noted that Cibicides species range from the inner shelf (0-100 m) through the outer shelf (100-200 m) and in to the upper slope (200-2000m or upper bathyal). E. repanda,
however, was noted as characteristic of depth ranges from the outer shelf (100-200m) to the abyssal plain (4000 m). Its presence may be the result of post-mortem transport processes operating from the outer shelf-bathyal-abyssal depths into the shallow marine-lagoonal sediments of Falasarna. However, the processes that resulted in the deposition of the higher numbers of E. repanda and C. advenum in samples 15-20, were markedly different from those normally operating and are equated with a tsunami because tsunamis occur less frequently than other high-energy phenomenon such as storm surges. A tsunami may also explain the rise in the percentage of broken pennate Foraminiferal tests from 21% to 83% in samples 15-20. In laboratory experiments on the relative resistance of Foraminiferal tests to crushing, globular forms were more resistant than pennate forms (Wetmore 1987). Wetmore recognized that test strength is likely to be related to a complex set of factors such as general shape, character of partitions between chambers, the arrangement of chambers and test wall thickness. However, the strongest foraminiferal tests are those which belong to species that have biconvex to globular shapes and sutures which are only slightly depressed. Such a morphology, Wetmore believes, would allow a more uniform dispersion of compressive stresses associated with sediment impact in the marine environment. If pennate Foraminiferal forms are more susceptible to crushing (and breakage) associated with impact stresses, higher percentages of broken pennate forms in sediments thought to have been deposited during highenergy tsunami inundation would be expected. The data presented here support this hypothesis. Furthermore, such dramatic increases in the percentage of broken pennate forms of diatoms have been reported for tsunami-deposited sediments associated with the Storegga tsunami in Scotland (S. Dawson, pers. comm. 1995). Radiometric dating of marine shells of the species H. acuta from the base of the Trench A (Fig. 3) gave a conventional radiocarbon date of 2870+70 radiocarbon a BP (calibrated age 662 BC 4-70). This dates the onset of
350
D. DOMINEY-HOWES E T A L .
sedimentation within the harbour. A conventional radiocarbon date of 2270 + 60 radiocarbon abe (calibrated age AD75 4-60) has been obtained for sample 11 which comes from below the inferred AO 66 tsunami layer, and a conventional radiocarbon age of 2290 4- 90 radiocarbon a aP (calibrated age AD634-90) has been obtained for sample unit 16, which comes from within the proposed tsunami unit. These dates imply that the high-energy event which led to the deposition of the high percentages of broken (pennate) Foraminifera in samples 15-20 occurred between 35 Bc and AD 160 (but probably between AD 63 and 75 4- 90). The preceding interpretation strongly suggests that deposition of sediments associated with a tsunami of c. AD66 appears to be preserved within the Trench A stratigraphy and it is also noted that no Foraminifera are recorded above sample 23. Most significantly, no Foraminifera are recorded in samples 24-28 which correlate with the sedimentary unit that according to Pirazzoli et al. (1992), had been deposited by a tsunami during AD 365. The results of the present investigation broadly support the findings of a previous study by Pirazzoli et al. (1992) which sought to understand the palaeoenvironmental history of Falasarna. However, there are some discrepancies between the findings of the two investigations. Pirazzoli et al. (1992) reported that the base of Trench A is at an elevation of 5.0 m a.s.1, and the surface is at 7.0 m a.s.l., whereas the present study gave elevations of 4.6 m a.s.1, and 6.2 m a.s.l., respectively (Fig. 3). There are four possible explanations for these variations. First, tectonic activity resulting in relative sea- and land-level changes could have occurred between the two successive phases of investigations, but such tectonic activity is not known to the present authors. Second, the elevations of the present authors could be erroneous. However, the closing error for the Trench A traverse was 0.01 m. Third, the elevations of Pirazzoli et al. (1992) could be incorrect. P. A. Pirazzoli (pets. comm., 1997) stated that as the trench elevations reported by him and his coworkers were calculated by the Regional Archaeological Service, significant error may have been introduced. Fourth, errors associated with the assumption that present sea level is true mean sea level may result in variations between successive phases of investigation of up to 50 cm. The lithostratigraphy described in this paper is similar to that previously published, and the foraminiferal assemblages identified in the present investigation are similar to those with those identified by Pirazzoli et al. (1992), although
those workers reported that Foraminifera are only present within the sediments from the base of Trench A only as high as the layer they ascribed to the AD66 tsunami (sample 14 in Table 1). They stated that Foraminifera are not present in the AD66 tsunami layer, although they reappear above this unit (e.g. from sample 21 onwards). This conflicts with the findings of the present investigation, perhaps because the earlier interpretation was based on just 15 samples taken at regular intervals between 5.2 and 6.6ma.s.1. Finally, the calibrated age of 662 Bc 4-70 for the base of Trench A contrasts with a date of 522-340 BC at 4-1 SD for a sample 20cm farther up the stratigraphic column reported by Pirazzoli et al. (Fig. 3), and suggests that the harbour may actually have been in existence two centuries earlier than proposed by Pirazzoli et al. (1992). Discussion
In many parts of Crete, by mapping and dating sequences of uplifted, superimposed raised fossil marine notches, it is possible to identify those areas believed to have been affected by coseismic deformation associated with earthquakes (Spratt 1865; Flemming 1978; Pirazzoli 1986; Kelletat 1991). At Falasarna, a sequence of uplifted palaeo-shorelines suggests that a series of small, uniform episodes of subsidence occurred during the 2000-3000 years before c. 1530 4-40 radiocarbon aBl~ (Pirazzoli et al. 1981, 1982, 1992; Thommeret et al. 1981; Pirazzoli 1986) (Fig. 4). Radiometric dates on these raised shorelines show ages which decrease with increasing altitude ~"
Destru~ive
r
ea~hquake
11o
~kTsunami +7 267
66
.~.
"6
g
368
,_ . . . . . . . . . . . . . . . . . ,
,I .
. . . . .
255 t--
I
r--',
II I ....
,v ; .......
i-
I I I
I
__,
,,
$FA9
,,
'L,
I
i
I
I
'
7FA4 i
l
@ A
............
o
] III
, Iv'
,
365
__, ..... ---
,,
,, ,,
g
=
'v'
e~ []
Drop to present MSL
+5
1000
.
.
.
.
500
'
'BC'
'
0
'
'AD'
'
500
Year
Fig. 4. Band of relative sea-level changes at Falasarna between 1000 Bc and AD500 compared with historically destructive earthquakes and tsunamis in western Crete. Radiocarbon dates for the shoreline displacements are shown in Table 3. (After Pirazzoli et al. (1992).)
COASTAL TECTONICS IN CRETE
351
Table 3. Relative sea-level changes at Falasarna during historical times deduced from radiocarbon-dated shorelines. Shoreline
0 I II III IIIa IV IVl
Elevation (m a.s.1.)
+0 +6.6+0.1 +6.5+0.1 +6.35 4- 0.1 +6.25 + 0.1 +6.1 4-0.1 +5.9 :k:0.1
Displacement age
-,
aBP
Calibrated age range*
Inferred historical event
1530 + 40 i 600-1710 1780-1800 1880-1900 1950-1980 2250-2300 2500-2610
AD341-439 AD89-404 1613C~D 169 141 BC--AD69 235-18 Bc 728-378 Bc 991-759 Bc
AD 365 (?) AD 66 (?)
The shoreline numbers correspond to those shown in Fig. 4. (Adapted from Pirazzoli et al. (1992)). * Calibration according to Stuiver et al. (1986).
(Pirazzoli et al. 1992) (Table 3). The sedimentological analyses presented in this paper provide an independent record of late Holocene coastal tectonic movements. From Fig. 4 and Table 3 the palaeo-shoreline data ofPirazzoli et al. (1992) imply that an earthquake subsidence occurred c.AD 66, which displaced the contemporary shoreline from +6.35 4-0.1 to +6.5 +0.1 m. However, the sedimentology indicates a sudden change from marine to terrestrial conditions, which is believed to be associated with uplift of the harbour rather than subsidence. The sedimentology is thus in direct conflict with the geomorphological evidence. The last major tectonic displacement determined from the shoreline data in Fig. 4 relates to the uppermost of the emerged notches at c. 6.5 m a.s.1, radiometrically dated at 1530 4- 40 radiocarbon a ~31~(Table 3). The stratigraphic record preserves no evidence of this event in the form of tsunami-deposited sediments. It is difficult to understand why the stratigraphy at Falasarna records no evidence of such a large displacement. The implications of these findings are that, in general, it is difficult to correlate the very precise stratigraphic record with the raised palaeoshorelines at Falasarna and in particular with the +6.5 m shoreline associated with the inferred AD 365 tectonic uplift. Furthermore, the sedimentology does not reflect the suggested pattern of Holocene coseismic tectonic movements deduced from uplifted shoreline features.
Conclusions Data presented from Falasarna Harbour records a late Holocene palaeoenvironmental history which also contains detailed evidence of earthquake-related tsunamis. The results provide a useful independent dataset for evaluating the
magnitude and timing of coastal tectonic movements in an area affected by contrasting tectonic regimes. Sedimentation within the harbour began c. 662 Bc + 70, which is approximately two centuries earlier than previously reported. The results suggest that, on the basis of foraminiferal assemblage, assemblage variation and individual test preservation, it is possible to identify suites of tsunami-deposited sediments within normal sedimentary sequences. The evidence provided by the foraminifera shows clearly that the last episode of marine sedimentation within the harbour was during or just after the AD66 earthquake-tsunami event. Consequently, the sedimentology strongly suggests tectonic uplift of the harbour c. A D 66, rather than subsidence, as inferred from the raised palaeo-shoreline data. Significantly, there is no bio- or lithostratigraphic evidence to infer sedimentary deposition associated with a tsunami reported to have been generated by a large vertical tectonic displacement c. AD 365. It is not possible to interrelate distinct sedimentary horizons with raised shoreline features at Falasarna which have previously been used to describe late Holocene coastal tectonic movements in this area. Funding for this research was provided by the European Union under Contract EV5V-CT92-0175: Genesis and Impact of Tsunamis on the European Coasts (GITEC), administered by Directorate General XII (Science, Research and Development), Climatology and Natural Hazards Unit. We would also like to acknowledge the generous permission of E. Hadjidaki to undertake fieldwork at Falasarna Harbour.
References ANGELIER, J. 1978. Tectonic evolution of the Hellenic Arc since the Late Miocene. Tectonophysics, 49, 23-36.
352
D. D O M I N E Y - H O W E S E T AL.
CIMERMAN, F. & LANGER, M. R. 1991. Mediterranean Foraminifera. Ceneter Sazu Paleotoski Institut Ivana Takovca, Ljubljana, 2. DOM1NEY-HOWES, D. T. M. 1996. The geomorphology
MERCIER, J. L. 1981. Extensional-compressional tectonics associated with the Aegean Arc: comparison with the Andean Cordillera of south Peru-north Bolivia. Philosophical Transactions
and sedimentology of five tsunamis in the Aegean Sea region, Greece. Unpublished PhD thesis,
of the Royal Society of London, Series A, 300,
Coventry University. EMERY, K. O., AUBREY, D. G. & GOLDSMITH,V. 1988. Coastal neotectonics of the Mediterranean from tide-gauge records. Marine Geology, 81, 41-52. FLEMMING, N. C. 1978. Holocene eustatic changes and coastal tectonics in the northeast Mediterranean: implications for models of crustal consumption.
Philosophical Transactions of the Royal Society of London, Series A, 289(1362), 405-445. - - 1 9 9 2 . Predictions of relative coastal sea level change in the Mediterranean based on archaeological, historical and tide-gauge data. In: JEFTIC, L., MILLIMAN, J. D. & SESTtNI, G. (eds)
Climatic Change and the Level of the Seas. Vol. I The Mediterranean, 247-281. --
& WOODWORTH, P. L. 1988. Monthly mean sea levels in Greece during 1969-1983 compared to relative vertical land movements measured over different timescales. Tectonophysics, 148, 59-72. FROST, F. J. 1997. Tectonics and history at Phalasarna. In: SWINY, S., HOHLFELDER, R. L. & WYLDE SWlNY, H. (eds) Res Maritimae: Cyprus and the
Eastern Mediterranean from Prehistory to Late Antiquity. Cyprus American Archaeological Research Institute, Monograph Series, l, 107115. HADJIDAKI, E. 1988. Preliminary report of excavations at the harbour of Phalassarna in western Crete. American Journal of Archaeology, 92, 463-479. IAPSO 1985. Changes in relative mean sea level. Working Party Report of the International Association for the Physical Sciences of the Ocean Advisory Committee on Tides and Mean Sea Level. Eos Transactions, American Geophysical Union, 66, 754-756. JACKSON, J. 1994. Active tectonics of the Aegean region. Annual Review of Earth and Planetary Sciences, 22, 239-271. KELLETAT, D. 1991. The 1550BP tectonic event in the eastern Mediterranean as a basis for assessing the intensity of shore processes. Zeitschrift fu'r GeomotThologie , Supplementband, 81, 181-194. LALLEMANT,S., TRUFFERT, C., JOLIVET,L., HENRY, P., CHAMOT-ROOKE, N. & DE VOOGD, B. 1994. Spatial transition from compression to transition in the western Mediterranean Ridge accretionary complex. Tectonophysics, 234, 33-52. LE PICHON, X. & ANGELIER, J. 1981. The Aegean Sea.
Philosophical Transactions of the Royal Society' of London, Series A, 300, 357-372.
337-355. MURRAY, J. W. 1991. Ecology and Palaeoecology of Benthic Foraminifera. Longman, Harlow, UK. PARKER, F. L. 1958. Eastern Mediterranean Foraminifera. Swedish Deep Sea Expedition, 8, 219-279. PIRAZZOLI, P. A. 1986. The Early Byzantine tectonic paroxysm. Zeitschrift J~ir Geomorphologie, Supplementband, 62, 31-49. --, AUSSEIL-BADIE, J., GIRESSE, P., HADJIDAKI, E. & ARNOLD, M. 1992. Historical environmental changes at Phalasarna harbor, West Crete. Geoarchaeology, 7, 371-392. - - , THOMMERET, J., THOMMERET,Y., LABOREL,J. & MONTAGGIONI, L. F. 1981. Les rivages 6merg+s d'Antikithira (Cerigotto): corr61ations avec la Crete occidentale et implications cin6matiques et geodynamiques. Actes du Colloque Niveaux
Marins et Tectonique Quaternaires dans l'Aire Mdditerran~ene. University of Paris I, 49-65. & 1982. Crustal block movements from Holocene shorelines: Crete and Antikythira (Greece). Tectonophysics, 86, 2 2 4 3 . SIDEBOTTOM, H. (1904-1909). Report on the Recent Foraminifera from the coast of the Island of Delos (Grecian Archipelago). Manchester Literary and
Philosophical Society Memoirs and Proceedings. 48, 1-26; 49, 1-22; 50, 1-18; 51, 1-28; 52, 1-28; 53, 1-32. SPRATT, T. S. 1865. Travels and Researches in Crete. Vol. 2. J. van Voorst, London. STUIVER, M. & BRAZIUNAS, T. F. 1993. Modelling atmospheric 14C influences and ~4C ages of marine samples to 10,000B.C. Radiocarbon, 35, 137-189. --, PEARSON, G. W. & BRAZIUNAS,T. 1986. Radiocarbon age calibration of marine samples back to 9000 cal yr BP. Radiocarbon, 28, 980-1021. TAYMAZ, Y., JACKSON, J. & WESTAWAY, R. 1990. Earthquake Mechanisms in the Hellenic Trench near Crete. Geophysical Journal International, 102, 695-731. THOMMERET, Y., THOMMERET, J., LABOREL, J., MONTAGGIONI, L. F. & PIRAZZOLI, P. A. 1981. Late Holocene shoreline changes and seismotectonic displacements in western Crete (Greece). Zeitschrift jh'r Geomorphologie, Supplementband, 40, 127 149. WETMORE, K. L. 1987. Correlations between test strength, morphology and habitat in some benthic Foraminifera from the coast of Washington. Journal of Foraminiferal Research, 17, 1-13.
Possible tsunami deposits from the 1855 earthquake, North Island, New Zealand JAMES
R. G O F F 1'3, M I C H A E L URSULA
C R O Z I E R 1, V E N U S
COCHRAN 1 &
SUTHERLAND
l,
PHIL SHANE 2
1 School o f Earth Sciences, Victoria University of Wellington, PO Box 600, Wellington, New Zealand 2 Department of Geology, School o f Environmental and Marine Sciences, University of Auckland, Tamaki Campus, Private Bag 92019, Auckland, New Zealand 3 IGNS, PO Box 30-368, Lower Hutt, New Zealand (e-mail.
[email protected]) Abstract: A series of three fining-upward sequences from deposits in the Okourewa Stream
bank on the south coast of the North Island, New Zealand, investigated by grain-size, diatom, radiocarbon, geochemical and macrofaunal analyses have been tentatively interpreted as the products of a tsunami. The proposed event consisted of three separate waves (the second being the largest) generated by a surface rupture of a local fault. Changes in diatom assemblages and the presence of marine shells, pumice, and beach pebbles may represent a tsunami advancing inshore over beach, freshwater channel, and coastal wetland enviromnents. Deposition occurred between AD500 and 1890. The event in question may have currently the AI~1855 rupture of the West Wairarapa fault. Internationally, there has been a recent increase in the amount of literature concerning tsunami research on both contemporary and palaeotsunami events (e.g. Dawson et al. 1996). Previous research concentrated on palaeo-tsunami in two main regions, the Pacific (Australia (Bryant et al. 1992), Hawaii (e.g. Moore et al. 1994), Japan (e.g. Minoura et al. 1994), Pacific Northwest (e.g. Atwater 1987; Clague & Bobrowsky 1994)) and Northwest Europe (e.g. Dawson et al. 1988). This work has served to improve our understanding of tsunami signatures preserved in the sedimentary record, and a summary of diagnostic criteria used to identify palaeotsunami is given in Table 1. In New Zealand, tsunami research is generally limited to either modelling for coastal hazard planning purposes (e.g. Gilmour 1960; Victory et al. 1989; Van Dissen et al. 1994) or establishing a record of palaeo-tsunami events (Eiby 1982; DeLange & Healy 1986). DeLange & Healy (1986) have produced a valuable, yet limited, record of at least 32 documented tsunami dating back to AD 1840 which indicates that locally derived tsunami produce far larger waves than their exogenic counterparts. The lack of tsunami research in New Zealand is somewhat surprising bearing in mind that the country sits astride the boundary of the Pacific and Australian plates, and thus is subject to considerable locally generated tectonic activity. Moreover, the country is also exposed to several exogenic tsunami sources. The east coast of New
Zealand is exposed to tsunami generated by earthquakes occurring almost anywhere around the Pacific Ocean, but particularly South America (Heath 1976, 1977). The west coast is exposed to the Tasman Sea and Australia, where tsunami may be refracted around New Zealand or reflected from the continental shelf of Australia (Braddock 1969). There is evidence that the tectonically inactive coast of New South Wales, Australia, has been inundated by several large tsunami events (Bryant et al. 1992; Young et al. 1993, 1996), the most recent about 800 years ago. These records point to the existence of other tsunamigenic sources such as subaquatic landslides off the continental shelf or meteor impacts (Bryant et al. 1996). The Wellington region is located at the southern end of the North Island (Fig. l a) and is separated from the South Island by Cook Strait, a 22km wide, 1500m deep channel in which the edge of the continental shelf is within 0.5-2.0 km of the present-day coastline (Mitchell & Lewis 1980; Carter et al. 1988). There are four class-one active faults in the Wellington region (that is those that are known to have moved at least once in the last 5000 years): the Wairarapa, Wellington, Ohariu and Shepherds Gully faults (Van Dissen & Berryman 1996), all of which are believed to extend across Cook Strait (Carter et al. 1988). Berryman (1990) considered that movement on the Wellington, Wairarapa and Ohariu faults and the offshore extension of the Wairau fault (South Island)
GOFF, J. R., CROZmR, M., SUTHERLAND,V., COCHRAN,U. & SHANE, P. 1998. Possible tsunami deposits from 1855 in North Island, New Zealand. In: STEWART, I. S. & VITA-FINZI, C. (eds) Coastal Tectonics. Geological Society, London, Special Publications, 133, 353 374.
354
J. R. G O F F E T A L .
Table 1. S u m m a r y o f diagnostic criteria used to identify tsunami deposits
(1) (2) (3)
(4) (5) (6) (7)
(8)
(9) (10)
(11) (12) (13) (14)
Diagnostic characteristics
References (e.g.)
Generally fines inland and upwards Each wave can form a distinct sedimentary unit, although this is not often recognized in the sedimentary sequence Distinct upper and lower sub-units representing run-up and backwash can be identified, B U T investigation of recent tsunami deposits indicates that there is still considerable uncertainty about when most deposition occurs (during run-up or backwash) and so these sub-units may be related to other processes Lower contact is unconformable or erosional Can contain intraclasts of reworked material, but these are not often reported Often associated with loading structures at base of deposit Particle and grain sizes range from boulder layers (up to 750 m3), to coarse sand to fine mud. However, most deposits are usually recognized as anomalous sand units in peat sequences Generally associated with an increase in abundance of marine to brackish-water diatoms, but reworking of estuarine sediments may simply produce the same assemblage; preservation of frustules can be excellent, although many are often broken Marked changes in Foraminifera (and other microfossils) assemblages. Deeper-water species are introduced with catastrophic saltwater inundation Increases in the concentrations of sodium, sulphate, chlorine, calcium and magnesium occur in tsunami deposits relative to under- and overlying sediments; indicates saltwater inundation Individual shells and shell-rich units are often present Often associated with buried vascular plant material and/or buried soil Shell, wood and less dense debris often found 'rafted' near top of sequence Dating of tsunami sediments is problematic. Best results for dating are from units above and below to 'bracket' the event. Radiocarbon ages often equivocal because of older reworked carbon; dating of introduced marine macrobiota is preferred (and successful). Optical dating (OSL) is the best method available assuming the sediments were exposed to daylight during reworking by the tsunami
Foster et al. 1991; Dawson 1994 Ota et aI. 1985; Moore & Moore 1988; Clague & Bobrowsky 1994
could account for 80-100% of the strain rate measured in Wellington. The most recent movements of faults in the C o o k Strait region occurred on the southern portion of the Wellington fault c. AD 1350, c. AD 1250 and c. AD 1450 (Best 1923; Stirling 1992; Van Dissen et al. 1992; Ian R. Brown Associates Ltd 1995; Van Dissen & Berryman 1996); the AD1855 earthquake occurred on the West Wairarapa fault 2 0 k i n
Moore & Moore 1988; Dawson et al. 1996
Dawson et al. 1988; Moore & Moore 1988 Dawson 1994; Moore et al. 1994 Foster et al. 1991; Minoura & Nakaya 1991 Ota et al. 1985; Moore & Moore 1988; Minoura & Nakaya 1991; Dawson 1994; Minoura et al. 1994; Young et al. 1996 Dawson et al. 1988; Minoura et al. 1994; Hemphi11Haley 1996
Patterson et al. 1996; Sherman et al. 1996 Minoura et al. 1994; Goff & Chagu&Goff 1998
Moore & Moore 1988; Bryant et al. 1992 Clague & Bobrowsky 1994; Dawson 1994 Albertho & Martins 1996; Imamura et al. 1997 Dawson et al. 1988; Hansom & Briggs 1991; Clague & Bobrowsky 1994; Dawson 1994; Huntley & Clague 1996
east of Wellington (e.g. D a r b y & Beanland 1992; Stirling 1992). The AD 1855 earthquake represented a surface rupture of the West Wairarapa fault (Fig. l a) along a 200 k m segment from southern Hawke's Bay to the middle of C o o k Strait (Robinson 1986; Carter et al. 1988; Barnett et al. 1991). The offshore fault trace is sub-parallel to the coastline and marks the edge of the continental shelf
POSSIBLE TSUNAMI DEPOSITS (1855?), NEW ZEALAND
355
Fig. 1. Okourewa Stream: (a) regional location map; (b) study area (showing transect A-B), (e) Surveyed transect from storm ridge A (modern), to Okourewa valley floor, B. (Begg & Mazengarb 1996). Hull & McSaveney (1996) have produced a model to explain the complex pattern of surface faulting and ground deformation at the southern end of the West Wairarapa fault, which indicates that the centre of maximum uplift (>5 m) was near Turakirae Head as opposed to being several kilometres to the NE. The uplift rate is believed to have been about 2.9m per 1000 years over the last 7200 years (Hull & McSaveney 1996). Although the West Wairarapa fault is to the west of the study site (Fig. la), it gives an indication of the complex tectonic history of the region. The study site is in an area where the tectonic history is poorly understood, although the Okourewa valley is believed to follow the Pirinoa fault, c. 2 km west
of the Moanatahi Syncline (Ghani 1974, 1978). The projected average uplift rates for synclines in the region is between 0 and 2.2m per 1000 years for the last c. 125 000 years (Ghani 1974, 1978; Pillans 1986). There are few contemporary accounts of tsunami along the south coast of the Wellington region and all refer to the AD 1855 earthquake. This event generated a 9 - 1 0 m high tsunami in Cook Strait and a 3 - 4 m high wave at the entrance to Wellington Harbour, and caused severe seiching inside (Fox 1855; Barnett et al. 1991). In Palliser Bay, three waves inundated the coast, the first being larger than the following two (Mason 1855). Computer modelling of tsunami inside Wellington Harbour concurs with
356
J. R. GOFF ET AL.
contemporary accounts, but separates the signal of external (to the harbour) 'seismic forcing' from that of internal effects (Barnett et al. 1991). This effectively explains the unique response of Wellington Harbour to the AD 1855 earthquake, but places the events in Palliser Bay outside the parameters of the study by Barnett et al. However, it is believed that contemporary accounts underestimated tsunami height outside Wellington Harbour (A. Barnett, pers. comm., 1997).
Physical setting Okourewa valley (41~ 175~ is located in southern Wairarapa, c. 30 km east of Wellington, on the south coast of the North Island, New Zealand (Fig. l a). Okourewa Stream is oriented NE-SW, draining a catchment of about 10km 2 and emptying into a lagoon formed on the landward side of the AD 1855 storm ridge (Figs lb and c, and 2). The lagoon appears to drain by seepage through the storm ridge (about 5m above sea level (asl)), maintaining a maximum lagoon depth of about 1.0 m. There is little existing geological information relating specifically to Okourewa valley, although reconnaissance surveys of Tertiary- and Pleistocene-age geological features in the region were carried out by McKay (1878, 1879), King (1930, 1933) and Cotton (1942). The local geology comprises the Hautotara Formation of bluegrey mudstone (Pliocene) laid down in a marine inner-shelf environment (Begg & Mazengarb 1996), and unnamed Pleistocene marine terraces cut during several stages of the last interglacial (Palmer & Vucetich 1989). Pre-European settlement in the area is indicated by a Maori Pa (fort), which is situated on the west side of the valley, and the remnants of several shell middens occur on the valley slopes (Adkin 1959). The length of occupation is unknown, but Maori occupied the valley at the time of European settlement (c. AD 1840), providing the initial ferry service across the mouth of Lake Onoke (McIlraith, pers. comm., 1996). No field evidence exists for Maori occupation of the valley floor, although early European documents indicate the existence of a Pa adjacent to the east bank of the stream (Adkin 1959). At the time of European settlement, the lake drained into the sea east of the mouth of Okourewa valley. Subsequent river flooding in the early 1950s redirected the mouth to the west, to the position that it occupies today. The valley floor is currently under farmland, although access to a 1 km length of stream is restricted by forestry plantations.
Fig. 2. Schematic diagram of Okourewa valley showing sections described in the text: (a) Seaward end of valley; (b) landward end, 800 m north of Fig. (a).
Study methods Surveying A topographical profile of the beach, lagoon and Okourewa valley floor as far as site LF2 (Figs lb and c, and 2) was measured in relation to benchmark locations at Lake Ferry and the landward end of the valley (Stewart trig point). Mean sea level (msl) was estimated based upon the Cape Palliser tide levels (which are offset from the Wellington tide tables by +10min). The profile was repeated twice and averaged, with errors of 4-0.2 m caused mostly by natural variations in storm ridge height.
Sampling and sediment analyses Sections of exposed stream bank were cleaned by removing weathered sediment with a spade to expose a fresh surface and brushing loose material from the exposure. The sections were photographed and the exposed stratigraphy was described. Sediment samples were generally taken at O.1m intervals, and at smaller intervals where distinct units were not covered by the
POSSIBLE T S U N A M I D E P O S I T S (1855?), N E W Z E A L A N D standard sampling interval. Sediment samples were taken for grain-size and diatom analyses, wood and charcoal for radiocarbon analysis, and pumice and shells for identification. Grain-size analysis followed procedures described by Barrett & Brooker (1989). Organic material was removed with hydrogen peroxide treatment for 7 days. Salts and acids were removed with distilled water and then centrifuged. After the supernatant liquid was poured off, sediment samples were washed with sodium hexa-metaphosphate and wet sieved at 60 #m to separate the sample into coarse and fine fractions. Coarse fractions were dry sieved at 89 intervals (0.63-4.00mm) using a Frisch shaker. Fine fractions were dried for 24h at 100~ and a 1.5-2 g sub-sample was used in SediGraph analysis. Data were entered into a PC software package (SIZE) to produce grainsize distribution indices. Bulk density was determined by sampling a known volume of sediment (10cm3). Samples were dried for 24h at 100~ then reweighed, and dry bulk density was calculated. Loss on ignition (LOI) was determined by ashing at 550~ for 16h. Diatoms were concentrated by digesting samples in 27% hydrogen peroxide to remove organic matter, adding 32% hydrochloric acid to remove carbonate, and heating at 100~ for 30min. Sand was extracted by decanting and clay was removed by washing in sodium hexa-metaphosphate. Drops of the liquid were dried on a slide, mounted in naphrax and examined under x1600 magnification. Two hundred frustules were counted from each sample. Changes in relative concentrations of diatoms are based on the number of frustules counted along one transect of the cover slip (all samples were prepared with the same weight of sediment and diluted by the same amount, so that relative concentrations can be compared between samples). Fifteen pollen samples were taken from three sites (LF1, LF2, and LF5). Samples were taken at 5cm intervals down-section from the surface. Samples were collected from a 5 - 1 0 m m surface layer using a metal spatula and were prepared for palynomorph analysis using the technique outlined by Moore & Webb (1978), with at least 300 grains studied on each slide. As discussed below, the primary aim was to identify the first appearance of Pinus pollen, and detailed pollen spectra were not produced. Clast roundness and sphericity were measured at nine sites using the criteria laid down by Folk (1980). A total of 1150 greywacke clasts were measured, with the mean roundness and sphericity of 50 clasts being calculated for each sub-site, and a similar number for each of the additional sites. Fabric measurements were taken from prolate clasts at site LF2 to determine flow direction. Two subsamples of 51 clasts each were measured and Schmidt equal-area, lower hemisphere projections of stone a-axis fabrics were produced. In both cases, contours were drawn at every one, three, five, etc., points per (100/n) % of the projected area, where n is the sample size. The technique and presentation of data are well established for glacial and fluvial sediments (e.g. Mark 1973; Hicock et al. 1996; Maizels 1996).
357
Radiocarbon analysis Charcoal and wood fragments were taken from two sections for radiocarbon analysis. The largest available samples were analysed, to avoid possible contamination caused by the incorporation of organic material in sediments of a different age. Conventional ages are reported according to Stuiver & Polach (1977). Calibrated ages were obtained using the bidecadal curve developed by Stuiver & Pearson (1993), with 27 ~4C years subtracted for the Southern Hemisphere offset (T. Higham, pets. comm., 1997).
Identification of pumice and shells Glass shards from pumice samples and tephra were analysed by electron microprobe. Results were compared with the glass chemistry (CaO wt % v. FeO wt %) of sea-rafted pumice and tephra from four major volcanic events, the Kawakawa tephra event at 22 ka BP (Taupo), the Macauley Island rhyolite at 6kaBp, Loisels sea-rafted pumice at c. 0.5kaBP (P. Froggatt, pers. comm., 1997), and the Taupo pumice (tephra) at 1.85 ka BP. Shells were picked from stream bank exposures for subsequent identification. There was concern that reworking of old shell midden sites would produce considerable shell material of diverse origin and age. However, as all shells found were juvenile, with articulated and intact bivalves, it seems unlikely that they are derived from old shell midden sites.
Results Results o f stratigraphic, d i a t o m , r a d i o c a r b o n , geochemical a n d m a c r o f a u n a l analyses carried o u t at sites LF1, L F 2 , L F 5 a n d L F 6 are detailed in Figs 3 a n d 4. Sites L F 3 a n d L F 4 (Fig. 5) are discussed below.
Site LF1 (Figs 3a and 4a) Stratigraphically, three erosional c o n t a c t s can be recognized within the preserved sequence. Sedim e n t a r y units overlying each c o n t a c t represent f i n i n g - u p w a r d s sequences t e r m i n a t e d by a f u r t h e r erosional contact. T h e e x c e p t i o n is the u p p e r m o s t f i n i n g - u p w a r d s sequence, w h i c h a p p e a r s to t e r m i n a t e at a g r a d a t i o n a l c o n t a c t with an overlying s a n d y l o a m unit. T h e lowest erosional c o n t a c t overlies a basal unit (below 1.06 m) o f organic-rich silt. G r a i n sizes in each sequence generally fine t h r o u g h a series o f g r a d a t i o n a l c o n t a c t s f r o m pebbles to silt or sand. E a c h f i n i n g - u p w a r d s sequence has different characteristics. T h e lowest sequence has a distinct a n d u n i f o r m clasts u p p o r t e d unit (with a lens (intraclast?) o f
358
J. R. G O F F E T AL.
Fig. 3. Stratigraphy of sections (including, where applicable, mud and organic percentages, summary diatom assemblage, fabric data and chronological information): (a) section LF1, (b) section LF2, (c) section LF5, (d) section LF6.
POSSIBLE TSUNAMI DEPOSITS (1855?), NEW ZEALAND
359
Fig. 3. (continued)
deformed sand), overlain by a massive sand with organic content decreasing upwards. Decreases in both grain size and organic content continue through a gradational contact into a massive silt. An in situ marine shell, Dosinia lambata, was found at 0.73 m below ground level. The next sequence is a unit composed of a lower clast-supported section and an upper, matrix-supported section. This is overlain by the final sequence, which is predominantly a sand unit, with sand deformed around rare pebblesized clasts, fining upwards into a massive sand. Deformation appears to be associated with loading structures such as water escape features (e.g. flames). Changes in mud and organic content match the stratigraphic record, with the most significant changes occurring between 0.92 and 1.11 m. Both mud and organic contents decrease upsection across the lowest erosional contact. Diatom samples were analysed from 0.53 to 1.18m across two erosional contacts (0.60 and 1.08 m). Frustules for non-marine species were well preserved (marine ones were broken), and relative concentrations decreased above the lower erosional contact. In general, the diatom assemblages remain unchanged, although there are two exceptions. At 0.94 m above the lowest erosional contact, there is a slight increase in the
relative percentage of polyhalobous (marine) and a large increase in halophilous (brackishfresh) assemblages recorded in a sand lens sample. Corresponding decreases in mesohalobous (brackish) and oligohalobous (fresh) assemblages are evident. At 0.53-0.58 m above the next erosional contact, there is a marked increase in oligohalobous (fresh) assemblages, and a corresponding decrease in the others. Pollen samples taken in the upper section of the exposure record a distinct change in pollen assemblage at about 0.10m with the first appearance of Pinus pollen.
Site LF2 (Figs 3b and 4b) The stratigraphic record is similar to that at site LF1, with three erosional contacts. All are associated with overlying fining-upwards sequences. Underlying organic-rich silts are deformed by loading structures (e.g. water escape features or flames). The lowest sequence fines upwards from coarse to fine sand with one marine shell, D. lambata, found at the top (1.23 m) of the unit. The central sequence fines from clast-supported cobbles (a maximum a-axis of 0.20 m) and pebbles to a massive sand incorporating several marine shells near the upper contact (D. lambata). Measurements of 51 clasts each in the lower
Fig. 4. Diatom species, salinity and habitat: for section LF1 (0.53-1.118m) and section LF2 (1.27-1.82m).
(e)
Diatom Salinity, Section LF1
/
/
Depth~ om
/
/ -I
-1 / \ I
.J
J
I
-4 I
J
o
(d)
,,,,,,,,, 10 20 30 4 0
,
,,,,,,,,,, 10 20 30 40
,,,,,,,,, ,,,, ,,,,, 10 20 30 40 SO 60 70 0 10
_
,
Diatom Habitat, Section LF1
-f
J
cn,t
I
J
/
/ /
7
I
6O
7O
t
t-i I 80
"-'1 100
i
,n,n ,i, i, 10 20 30 40
Fig. 4. (continued)
i , i , i , i i , i , L , i , Li I r I , I , L , I , I , I , I , I , 0 10 20 30 O 10 20 30 40 0 10 20 30 40 GO 0 10 20 L~t
I
LiI 10
110
Fig. 4. (continued)
POSSIBLE TSUNAMI DEPOSITS (1855?), NEW ZEALAND
363
Fig. 4. (continued) (1.10-1.20 m) and upper (0.85-1.05 m) sections indicate that clasts possess a weak fabric, with flow directions from 241 ~ and 82 ~, respectively. The upper sequence fines upwards from clast-
supported pebbles and gravels to a massive sand unit, the latter containing wood with a 14C age of 1350+ 190aBP (Table 2). This is overlain by a unit of sandy loam.
Fig. 5. Section LF4: photograph showing stratigraphy of uppermost units.
364
J. R .
GOFF
E T AL.
t~
~ r,~ r.t3
,.._.., ..
,,~