Seasonal sea surface temperature contrast between the ... - Wiley

10 downloads 0 Views 313KB Size Report
[1] Annual, summer, and winter sea surface temperatures (SSTs) in the western Arabian Sea were reconstructed through the last 22 kyr using artificial neural ...
PALEOCEANOGRAPHY, VOL. 20, PA1004, doi:10.1029/2004PA001078, 2005

Seasonal sea surface temperature contrast between the Holocene and last glacial period in the western Arabian Sea (Ocean Drilling Project Site 723A): Modulated by monsoon upwelling Pothuri Divakar Naidu National Institute of Oceanography, Dona Paula, Goa, India

Bjo¨rn A. Malmgren Department of Earth Sciences-Marine Geology, Go¨teborg University, Go¨teborg, Sweden Received 2 August 2004; revised 20 October 2004; accepted 2 November 2004; published 26 January 2005.

[1] Annual, summer, and winter sea surface temperatures (SSTs) in the western Arabian Sea were reconstructed through the last 22 kyr using artificial neural networks (ANNs) based on quantitative analyses of planktic foraminifera. Down-core SST estimates reveal that annual, summer, and winter SSTs were 2, 1.2, and 2.6C cooler, respectively, during the last glacial period than in the Holocene. A 2.5C SST increase during Termination 1A (hereinafter referred as glacial to Holocene transition) in the western Arabian Sea. The study reveals a strong seasonal SST contrast between winter and summer from 18 to 14 calendar kyr owing to the combined effect of weak upwelling and strong cold northeasterly winds. Minor or no seasonal SST changes were noticed within the Holocene period, which is attributed to the intense upwelling during the summer monsoon. This causes a lowering of SST to values similar to those of the winter season in analogy with the present day. A 3C rise in winter SSTs during the glacial to Holocene transition coincides with a strengthening of the monsoon, suggesting a link between winter SST and monsoon initiation from the beginning of the Holocene. Strikingly, annual, summer, and winter SSTs show a cooling trend from 8 ka to the present day, implying tropical cooling in the late Holocene. Citation: Naidu, P. D., and B. A. Malmgren (2005), Seasonal sea surface temperature contrast between the Holocene and last glacial period in the western Arabian Sea (Ocean Drilling Project Site 723A): Modulated by monsoon upwelling, Paleoceanography, 20, PA1004, doi:10.1029/2004PA001078.

1. Introduction [2] The interhemispheric heating contrasts between the land and ocean surface are at the heart of the monsoon phenomenon in the Indian Ocean. The upper ocean heat supply feeds the necessary evaporation and atmospheric moisture transports during the summer monsoon season. Therefore a strong link exists between ocean dynamics and the atmospheric heat and moisture transfer that is critical to the strength of the monsoons [Manghnani et al., 2002]. Overall, the Ekman component causes dominant seasonal variations in meridional overturning and heat transport due to monsoon reversal over the northern Indian Ocean and the change of easterlies over the subtropical southern Indian Ocean [Lee and Marotzke, 1998]. A better understanding of the temporal variability and spatial redistribution of the upper ocean heat content is thus valuable to comprehend the monsoon dynamics. Therefore, to understand the heat budget of the Indian Ocean in the past it is essential to gain a better knowledge of past SST changes. [3] Statistically based SST estimates from Imbrie-Kipp transfer functions show that the average Ice Age zonal SST (the difference in SST between a single latitude point and Copyright 2005 by the American Geophysical Union. 0883-8305/05/2004PA001078

the average temperature of the latitude band) for the entire Indian Ocean was 1.4C cooler during February and 1.5C cooler during August compared with the modern SST [Prell and Hutson, 1979]. SST estimations based on alkenone studies reveal that last glacial maximum (LGM) temperatures were about 2C cooler compared to modern values in the Arabian Sea [Rostek et al., 1993, 1997; Emeis et al., 1995]. Furthermore, alkenone-based SST estimations show nearly the same SST shift from LGM to Holocene, resulting from the change from conditions marked by sporadic upwelling in the eastern Arabian Sea during the LGM to strong upwelling in the western Arabian Sea during the Holocene [Rostek et al., 1997]. Attempts have been made to trace the upwelling strength based on the SST differences between glacials and interglacials mainly derived from oxygen isotopes and alkenones [Zahn and Pedersen, 1991; Emeis et al., 1995]. These studies did not find any systematic relationship between SST changes and upwelling strength on glacial and interglacial timescales, because these authors dealt with annual SSTs rather than seasonal ones. Hitherto, the seasonal temperature estimations in the Arabian Sea based on the Imbrie-Kipp transfer function techniques [Prell and Hutson, 1979] have not been verified by any other means. Here, we have used a novel technique, artificial neural networks (ANNs), to estimate the annual, summer and winter SST changes over last 22 kyr at the

PA1004

1 of 9

PA1004

NAIDU AND MALMGREN: SEASONAL SEA SURFACE TEMPERATURE CONTRAST

PA1004

western Arabian Sea gains heat from atmosphere, whereas Eastern Arabian Sea looses the heat to the atmosphere.

3. Methods

Figure 1. Annual temperature variations at the location of ODP Site 723 from10, 20, 30, 50, 75, and 100 m water depths. The temperature data are from Levitus [1982].

Ocean Drilling Project (ODP) Site 723A, to address (1) how the seasonal SSTs have changed through time in response to SW Monsoon upwelling and (2) how the air-sea interaction process controls the heat budget of the Arabian Sea during the Holocene and LGM.

2. Hydrography [4] Seasonal reversal of the monsoon winds in the Arabian Sea influences surface circulation, productivity, biogenic and lithogenic flux, CO2 uptake and heat budget in the region. The SW Monsoon is accompanied by strong winds (30 knots), cloudy skies and moist air. In addition, the ocean gains an average of 89.5 Wm 2, but SST decreases by 5.5C and the mixed layer deepens to almost 80 m in the central Arabian Sea and shallows in the Oman Margin as compared to the intermonsoon seasons [Schott and Mc Creary, 2001]. The salinity in the upper water column during the SW Monsoon is reasonably uniform, but decreases from about 36.5 to 36.0% during the course of the SW Monsoon season. The SW Monsoon wind field causes upwelling along the Arabian coast, leading to observed persistent jets of cold upwelled water extending from the Oman coast and often advecting laterally about 600 km offshore [Brink et al., 1998; Fischer et al., 2002]. Outside of the regions of coastal upwelling during the SW Monsoon, SSTs are generally high, varying from 27 to 29C. The NE Monsoon is characterized by moderate winds (10 knots), clear skies and dry air. During NE Monsoon ocean looses an average of 19.7 Wm 2 heat to the atmosphere with a 3C SST decrease [Weller et al., 2002]. The two intermonsoon seasons are quite different from either monsoon season. They are characterized by weak winds and strong sea surface heating. Modern SST varies from 23.2 to 28.4C at the location of ODP Site 723A. Minimum SST during July through September (SW monsoon season) and maximum SST during intermonsoon season are noticed (Figure 1). In the Arabian Sea the heat flux from the ocean to the atmosphere is higher during winter (November, December and January) compared to other months [Hastenrath and Lamb, 1979]. During SW Monsoon the

[5] ODP Site 723A is located in the intense upwelling area along the western Arabian Sea (1803N, 5737E) at a water depth of 808 m. The upper 7.4 m of the hole was sampled at 10-cm intervals, which gives an approximate time resolution of 250 years. The time control of the core is based on the AMS C14 dating (Table 1). C14 dates are calibrated to calendar years before present (1950) using the Calib 4.3 marine 98 program [Stuiver and Reimer, 1993]. Details of the planktic foraminifer census were reported earlier from this site [Naidu and Malmgren, 1996]. [6] Past annual, summer and winter season SSTs were estimated using artificial neural networks (ANNs). An ANN is an information processing system inspired by the way the densely interconnected, parallel structure of the mammalian brain processes information. The ANN is composed of a great number of processing elements that are analogous to neurons and are tied together with weighted connections that are analogous to synapses. The most common type of ANNs is the multilayer perceptron, which is most often trained using the back propagation (BP) algorithm. The ANN is trained to reproduce the target variable(s) from the input variables by adaptively updating the synaptic weights that are associated with the strength of the connections. The optimum weights are determined iteratively by optimizing certain ‘‘energy’’ functions as training proceeds. Comprehensive descriptions of multilayer perceptron ANN can be found in the work of Wasserman [1989], Webb [2002, pp. 204 –216] and Malmgren and Nordlund [1996, 1997]. [7] The objective behind the application of ANNs is thus to attempt at reproducing the output variable(s) from the input variables with a minimum error rate through a specific training process. ANNs have the ability to overcome problems of fuzzy and nonlinear relationships between the sets of input and output signals. In paleoceanographic applications the input variables are relative abundances of a number of species of, for example, planktic foraminifers; the output most often consists of one single variable, such as Table 1. AMS Radiocarbon Ages for ODP Site 723Aa Depth, cm 3 16 39 64 96 121 154 186 218 274 316 520 740

14

C Age, Years BP 950b 1680c 2010c 2760c 3890c 4670c 6070c 7700c 8110c 9510c 10,500c 15,920b 19,130b

Calendar Years BP 280 1005 1330 2254 3587 4622 6276 7938 8359 9854 10,877 18,200 22,260

a Radiocarbon ages are calibrated to calendar years before present (1950) using calibration program of Stuiver and Reimer [1993]. b From Naidu and Malmgren [1995]. c From Gupta et al. [2003].

2 of 9

PA1004

NAIDU AND MALMGREN: SEASONAL SEA SURFACE TEMPERATURE CONTRAST

PA1004

Figure 2. Location of core tops from which planktic foraminiferal census data was used in the artificial neural network (ANN) computations. SST. In this study the annual, summer and winter SSTs were analysed separately. Once trained on the basis of a reference set of modern microfossil and hydrographical data, the ANN can be employed for prognostication of past sea surface conditions within some limits of precision using the same set of input signals. [8] The ANN training process involves subdivision of the initial data set into two random portions, a training set, which is used for training the ANN, and a test set to which the trained network is applied for estimates of the error rate. The optimum measure of the error rate is the root mean square error of prediction (RMSEP), which is represented by the square root of the squared differences between the observed and estimated values divided by the number of observations in the test set. One such partition of the original data set is not sufficient to estimate the RMSEP in the data set; instead it is necessary to generate several such partitions to obtain an accurate estimate of the error rate. [9] We used the Neuro Genetic Optimizer (NGO, version 2.5) software #BioComp System, Inc) for the training of the ANNs. Planktic foraminiferal census counts of core tops are from the Brown University Foraminiferal Database [Prell et al., 1999]. In this application, the ANNs were trained on the planktic foraminiferal census of 361 core tops from the Indian Ocean (Figure 2). The species listed in Table 2 were used as input variables to the BP neural network. The training set consisted of 271 samples (75%) and the test set of the remaining 90 samples (25%). The training process was repeated ten times to obtain a reliable estimate of the precision of the predictions of the test set observations. Table 3a indicates that annual SSTs can be predicted with an average precision of 0.92C (range 0.72 – 1.14C), and that the corresponding precision for the summer and winter seasons are 1.02 (range 0.82– 1.19C)

and 1.05C (range 0.77 – 1.40C), respectively. Correlation coefficients for the relationship between observed and predicted SSTs are on the average 0.987 (range 0.981– 0.991) for the annual SSTs, 0.980 (range 0.975 –0.985) for

Table 2. Species Used As Input Variables to the Back-Propagation Neural Networks Number

Species

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Orbulina universa Globigerinoides conglobatus Globigerinoides ruber (pink and white morphotypes) Globigerinoides tenellus Globigerinoides sacculifer without sac Globigerinoides sacculifer with sac Sphaeroidinella dehiscens Globigerinella aequilateralis Globigerina calida Globigerina bulloides Globigerina falconensis Beella digitata Globigerina rubescens Globigerina quinqueloba Neogloboquadrina pachyderma sinistral morphtype Neogloboquadrina pachyderma dextral morphotype Neogloboquadrina dutertrei Globoquadrina conglomerata Globoquadrina hexagona Pulleniatina obliquiloculata Globorotalia inflata Globorotalia truncatulinoides sinistral morphotype Globorotalia truncatulinoides dextral morphotype Globorotaliacrassaformis Globorotalia hirsuta Globorotalia scitula Globorotalia menardii Globorotalia tumida Globorotalia menardii flexuosa Globigerinita glutinata

3 of 9

PA1004

NAIDU AND MALMGREN: SEASONAL SEA SURFACE TEMPERATURE CONTRAST

Table 3a. Root-Mean Squares Errors of Prediction (RMSEP) for the Ten Back Propagation Artificial Neural Networks Established for the Predictions of Annual, Summer, and Winter SSTs in Site 723a Partition/Season

Annual

Summer

Winter

1 2 3 4 5 6 7 8 9 10 Mean

0.923 0.718 0.742 0.834 0.941 0.958 1.142 1.058 0.844 1.036 0.920

1.028 1.193 1.081 1.134 0.820 1.137 0.994 1.068 0.831 0.952 1.024

0.932 0.943 1.073 0.773 1.011 1.404 1.093 0.977 0.908 1.396 1.051

a The RMSEPs estimate the error in C between the observed and predicted SSTs in the 90 test set samples constituting the surface-sediment data set.

the summer season and 0.984 (range 0.976 – 0.992) for the winter season (Table 3b). Subsequently, estimated annual and seasonal paleo-SSTs in Site 723A are represented by the averages of the estimates of the ten ANNs. [10] Trends in the annual, summer and winter SST in the sense of monotonically increasing or decreasing SST were tested using standard linear regression analysis. An approach similar to the one employed here was described by Marchal et al. [2002]. The existence of randomness in the sequence of the residuals was tested using the Runs test of randomness. A second procedure, the Durbin-Watson test [Ryan, 1997, pp. 46 –48], was employed to analyze the independence among residuals. This method is designed to test for correlation between consecutive residuals in the case of cyclical patterns in time series, where the Runs test would not function well. [11] The conformity of the residuals to a normal distribution was tested by means of the nonparametric test, introduced by Royston [1982a, 1982b], which is an extension of the Shapiro-Wilk W test [Shapiro and Wilk, 1965] applied to sample sizes larger than 50. A useful feature of the Royston test is the conversion of the Royston W statistics to an approximate standard normal deviate. [12] The strategy for testing the existence of trends follows standard regression analysis rules. Thus if the residuals are random, the slope of the regression line (b)

PA1004

is allowed to be used as a measure of the magnitude of the temporal change in SST. Furthermore, if the residuals are normally distributed, the confidence interval can be used to assess whether the slope deviates significantly from zero or not; deviation from zero is indicative of a trend. If the residuals are not normally distributed the Kendall test [Kendall and Ord, 1990] was used instead to test for trend.

4. Results and Discussion 4.1. Annual SST Changes [13] The annual SST varies from 23.7 to 28.0C with the coldest values during the last glacial maximum and the warmest values from 11 to 8 calendar kyr (Figure 3a). The last glacial period SST was 2C colder than the Holocene SST (Table 4), which is in sharp contrast to earlier SST estimates in the Arabian Sea mostly based on the Imbrie-Kipp transfer function technique [Prell and Hutson, 1979; CLIMAP Project Members, 1981], which indicates a minor (1.2 for the hypothesis of no correlation between consecutive residuals not to be rejected [Ryan, 1997, pp. 47 – 48]. Test of normality of the residuals: The Royston WR statistics (all series conform to normality). Regression analysis: The 95% confidence interval of the slope of the regression line (b) was used to test for trend since the residuals are random, consecutive residuals are uncorrelated, and the residuals are normally distributed. All of the annual, summer and winter season display a significant decrease in SST through the Holocene.

SSTs in the region. The seasonal SST contrast is highest during the last glacial period for two reasons: (1) weak upwelling during the last glacial period did not alter the summer SSTs and (2) the intense winter cooling extended further westward into the Arabian Sea due to strong cold winds during the NE monsoon in the Arabian Sea. Winter winds bring extremely cold air masses from the continent to the ocean. The larger air-sea temperature difference during the last glacial period thus aids in cooling the sea surface to a larger extent through direct sensible heat losses from the ocean to the atmosphere. This winter surface water cooling extended up to the Arabian coast during the last glacial period, in contrast to the present-day winter cooling which is restricted to the NE region of the Arabian Sea. [22] Between 15 and 9 ka, the Earth-Sun distance decreased during the northern summer and the axial tilt increased; therefore, the seasonality of climate increased in the Northern Hemisphere and decreased in the Southern Hemisphere. As a result, continental ice sheets and sea ice retreated and the oceans warmed. At about 9 ka the average solar radiation over the Northern Hemisphere was 8% higher in July and 8% lower in January than it is today [Berger and Loutre, 1991]. After 9 ka these seasonal radiation extremes decreased toward modern values (Figure 5). However, the maximum seasonal SST contrast is noticed between 18 and 14 calendar kyr, which indicates that solar radiation alone can not account the observed seasonal SST contrast at this site. Thus supporting monsoon strength and associated SST changes lags the solar insolation maxima [Clemens et al., 1991]. However, the winter cooling plays a greater role in regulating the SST changes during the last glacial period. Emeis et al. [1995] have pointed out that late Quaternary SST fluctuations in the western Arabian Sea are a combination of two temperature signals of which one is the sum of all processes unrelated to upwelling, such as solar insolation and cooling by evaporation, and the other is related to the winds of the NE monsoon during the winter season. Here, we report that the seasonal SST difference does record the upwelling-induced changes in SST. This, therefore, opens up the possibility to quantify the upwelling changes in the past. Modern SSTs are warmer in the eastern Arabian Sea compared to the western Arabian Sea [Levitus, 1982]. Seasonal SST contrasts between winter and summer are slightly greater (0.44C) in the eastern Arabian Sea than in the western Arabian Sea (0.15C). In the eastern Arabian Sea no significant seasonal SST difference is noticed between the

last glacial period and Holocene [Cayre and Bard, 1999]. Several studies suggest a greater influence of the NE monsoon during the LGM along the eastern Arabian Sea [Duplessy, 1982; Cayre and Bard, 1999]. We argue that if the NE Monsoon wind effect is more severe in the SE Arabian Sea one would expect cold winter SSTs and normal summer SSTs resulting in more seasonal contrast. However, such seasonal temperature changes were not documented from core MD77194 during the last glacial period [Rostek et al., 1997]. This reveals that NE monsoon winds were stronger during the glaciation in the NW Arabian Sea than in the SE Arabian Sea. 4.5. Tests of Trend [23] Tests of trend in the Holocene annual, summer and winter SSTs records (Figure 3) indicate the existence of statistically significant decreases in SST for all records through the last 7.9 calendar kyr (Table 5). The decrease is of a magnitude of 2.4C for the winter season SST, and 1.6 and 1.5C for the annual and summer season SSTs, respectively (Figure 3). Tests of differences in the slopes of the regression lines (Table 5) show that those of the summer and winter season differ statistically significantly (t = 2.04, which is statistically significant at the 5% level for 32 degrees of freedom). Hence the cooling trend observed through the last 7.9 calendar kyr appears to be more pronounced during the winter season (0.4C/kyr) than during the summer season (0.2C/kyr). No statistically significant differences were found when comparing the slopes of the summer and annual, and winter and annual regression lines. [24] Similar Holocene cooling trends have been noticed in the northeast Atlantic, Mediterranean [Marchal et al., 2002] and western pacific [Stott et al., 2004]. Several continental records such as bore hole temperature measurements in the Green Land Ice Sheet [Dahl-Jensen et al., 1998] glacier advances and retreats in the north east Atlantic area [Lubinski et al., 1999] and pollen sequences documenting vegetational changes in Europe [Cheddadi et al., 1998] also provide an evidence of a cooling during late Holocene. Climate model simulations also support the hypothesis of a long-term cooling during the Holocene [Pollard et al., 2000]. The rate of late Holocene cooling varies from place to place, for example northeast Atlantic and Mediterranean show cooling rate of 0.27 to 0.15C kyr 1 [Marchal et al., 2002], western pacific 0.05C kyr 1 [Stott et al., 2004], and in the Green Land 0.6C kyr 1 [Dahl-Jensen et al., 1998].

7 of 9

PA1004

NAIDU AND MALMGREN: SEASONAL SEA SURFACE TEMPERATURE CONTRAST

Although, our estimate of the annual cooling rate (1.6C) comprises only the last 7.9 calendar kyr of the Holocene, where the cooling trend starts in the Site 723A record, it is well within the range of the estimates for the northeast Atlantic, Mediterranean and western Pacific. This represents evidence for a general tendency for climate cooling in many parts of the world ocean through the Holocene or parts thereof. This cooling trend is most likely controlled by the decrease in solar insolation from 10 to about 2 kyr (Figure 3a) along with changes in cloud albedo, evaporation and ocean dynamics arising from the precessional forcing. [25] However, warming trend during late Holocene is noticed in the eastern subtropical Pacific [Yamamoto et al., 2004], western subtropical Atlantic and eastern Mediterranean [Rimbu et al., 2003]. A comprehensive picture of sea surface cooling and warming during late Holocene show a see saw type of trend between the eastern and western Pacific and Atlantic oceans and Mediterranean Sea. Good high-resolution records are not available from the eastern Arabian Sea to compare the Holocene SST changes between western and eastern Arabian Sea. However, the present study infers a long term cooling in the western Arabian Sea during the Holocene, this cooling pattern is similar to that of western Pacific, eastern Atlantic and western Mediterranean.

5. Summary and Conclusions [26] We have quantified annual, summer and winter sea surface temperature (SST) by using artificial neural network (ANN) technique based on the census counts of planktic foraminifera in the Arabian Sea. [27] 1. Annual, summer and winter were 2C, 1.1C and 2.6C cooler, respectively, during the last glacial period than in the Holocene. A 2.5C SST shift is noticed during the glacial to Holocene transition and a 0.8C cooling during the Younger Dryas in the western Arabian Sea. About 2C change is documented in the annual, winter and summer

PA1004

SST within the Holocene. This indicates that Holocene temperatures are not stable as believed earlier. [28] 2. An observed annual SST difference of 2C between the last glacial period and Holocene is not related to the strength of upwelling. However, the temperature difference between summer and winter seasons is directly related to the upwelling intensity in the western Arabian Sea. Therefore we conclude that the temperature difference between winter and summer seasons can be used to quantify upwelling changes in the past. [29] 3. Both summer and winter SSTs show a greater magnitude of change during the last glacial period compared to the Holocene. However, the winter SST shows a wider range of variation compared to the summer SST during the last glacial period. This suggests that prevailing strong cold northeasterly winds extend further westward during the last glacial period. Furthermore, a 3C rise in winter SST during the glacial to Holocene transition coincides with a strengthening of the monsoon, which suggests a possible link between winter SST and monsoon initiation from the beginning of the Holocene. [30] 4. Annual, summer and winter SSTs show statistically significant cooling trends from 7.9 calendar kyr. The cooling rate over the interval from 7.9 calendar kyr to the present day is 2.4C, 1.6C and 1.5C for the winter season, annual and summer season, respectively. This observation, together with similar cooling trends in the northeast Atlantic, Mediterranean, and western Pacific suggest a widespread cooling in the late Holocene world ocean are found to be forced mainly by the precession-driven changes in mean annual insolation.

[31] Acknowledgments. PDN would like to thank Satish Shetye, Director, National Institute of Oceanography, Goa, India, for his support and encouragement. We thank Satish Shenoi for his help in retriving the sea surface temperature data from the Levitus data set. We thank anonymous reviewers for offering constructive comments, which improved the quality of interpretations. The samples were provided by Ocean Drilling Program, which is sponsored by the U.S. National Science Foundation and participating countries under management of Joint Oceanographic Institutions. This is National Institute of Oceanography Contribution 3941.

References Anderson, D. M., and R. C. Thunell (1993), The oxygen-isotope composition of tropical ocean surface water during the last deglaciation, Quat. Sci. Rev., 12, 465 – 473. Anderson, D. M., T. Overpeck, and A. K. Gupta (2002), Increase in the Asian southwest monsoon during the past four centuries, Science, 297, 596 – 599. Berger, A., and M. F. Loutre (1991), Insolation values of the climate of the last 10 million years, Quat. Sci. Rev., 10, 297 – 317. Berger, W. H. (1990), The Younger Dryas cold spell—A quest for causes, Palaeogeogr. Palaeoclimatol. Palaeoecol., 89, 219 – 237. Brink, K., R. Arnone, P. Coble, C. Flagg, B. Jones, J. Kindle, C. Lee, D. Phinney, M. Wood, C. Yentsch, and D. Young (1998), Monsoons boost biological productivity in Arabian Sea, Eos Trans. AGU, 79, 168 – 169.

Cayre, O., and E. Bard (1999), Planktonic foraminiferal and alkenone records of the last deglaciation from the eastern Arabian Sea, Quat. Res., 52, 337 – 342. Cheddadi, R., H. F. Lamb, J. Guiot, and S. van der Kaars (1998), Holocene climate change in Morocco: A quantitative reconstruction from pollen data, Clim. Dyn., 14, 883 – 890. Clemens, S., W. L. Prell, D. Murray, G. Shimmield, and G. Weedon (1991), Forcing mechanisms of the Indian Ocean Monsoon, Nature, 353, 720 – 725. CLIMAP Project Members (1981), Geological Society of America Map and Chart Series, MC 36, Geol. Soc. of Am., Boulder, Colo. Dahl-Jensen, D., K. Mosegaard, N. Gundestrup, G. D. Clow, S. J. Johnsen, A. W. Hansen, and N. Balling (1998), Past temperatures directly

8 of 9

from the Greenland ice sheet, Science, 282, 268 – 271. Duplessy, J. C. (1982), Glacial to interglacial contrasts in the northern Indian Ocean, Nature, 295, 494 – 497. Emeis, K.-C., D. M. Anderson, D. Kroon, and D. Schulz-Bull (1995), Sea-surface temperatures and history of monsoon upwelling in the northwestern Arabian Sea during the last 500,000 years, Quat. Res., 43, 355 – 361. Fischer, A. F., R. A. Weller, D. L. Rudnick, C. Eriksen, C. M. Lee, K. Brink, C. A. Fox, and R. R. Leben (2002), Mesoscale eddies, coastal upwelling, and the upper-ocean heat budget in the Arabian Sea, Deep Sea Res. Part II, 49, 2231 – 2264. Gupta, A. K., D. M. Anderson, and J. T. Overpeck (2003), Abrupt changes in the Asian southwest monsoon during the Holocene and their links to

PA1004

NAIDU AND MALMGREN: SEASONAL SEA SURFACE TEMPERATURE CONTRAST

the North Atlantic Ocean, Nature, 421, 354 – 356. Hastenrath, S., and P. J. Lamb (1979), Climatic Atlas of the Indian Ocean, Part 1, Surface Circulation and Climate, 109 pp., Univ. of Wisc. Press, Madison, Wisc. Kendall, M., and J. K. Ord (1990), Time Series, 3rd ed., 296 pp., Edward Arnold, London. Lee, T., and J. Marotzke (1998), Seasonal cycles of Meridional overturning and heat transport of the Indian Ocean, J. Phys. Oceanogr., 28, 923 – 943. Levitus, S. (1982), Climatological atlas of the world ocean, Prof. Pap. 13, 173 pp., NOAA, Silver Spring, Md. Lubinski, D. J., S. L. Forman, and G. H. Miller (1999), Holocene glacier and climate fluctuations on Franz Jozef Land, Artic Russia, 80N, Quat. Sci. Rev., 18, 85 – 108. Madhupratap, M., S. P. Kumar, P. M. A. Bhattathiri, M. D. Kumar, S. Raghukumar, K. K. C. Nair, and N. Ramaiah (1996), Mechanism of the biological response to winter cooling in the Arabian Sea, Nature, 384, 549 – 551. Malmgren, B. A., and U. Nordlund (1996), Application of artificial neural networks to chemostratigraphy, Paleoceanography, 11, 505 – 512. Malmgren, B. A., and U. Nordlund (1997), Application of artificial neural networks to paleoceanographic data, Palaeogeogr. Palaeoclimatol. Palaeoecolecol., 136, 359 – 373. Manghnani, V., J. M. Morrison, L. Xie, and B. Subrahmanyam (2002), Heat transports in the Indian Ocean estimated from TOPEX/ POSEIDON altimetry and model simulations, Deep Sea Res. Part. II, 49, 1459 – 1480. Marchal, O., et al. (2002), Apparent long-term cooling of the sea surface in the northeast Atlantic and Mediterranean during the Holocene, Quat. Sci. Rev., 21, 455 – 483. Martin, P. H., J. W. Hurrell, and T. Xu (2001), Tropical origins for recent north Atlantic climate changes, Science, 292, 90 – 92. Naidu, P. D. (2004), Isotopic evidences of past upwelling intensity in the Arabian Sea, Global Planet. Change, 40, 285 – 293. Naidu, P. D., and B. A. Malmgren (1995), A 2200 years periodicity in the Asian monsoon system, Geophys. Res. Lett., 22, 2361 – 2364. Naidu, P. D., and B. A. Malmgren (1996), A high-resolution record of late Quaternary upwelling along the Oman Margin, Arabian Sea based on planktonic foraminifera, Paleoceanography, 11, 129 – 140. Pollard, D., J. C. Bergengren, L. M. StillwellSoller, B. S. Felzer, and B. S. Thomson

(2000), Climate simulations for 10,000 and 6000 years BP using the GENESIS global climate model, Paleoclim. Data Modell., 2, 183 – 218. Prell, W. L., and W. H. Hutson (1979), Zonal temperature-anomaly maps of Indian Ocean surface waters: Modern and ice-age patterns, Science, 206, 454 – 456. Prell, W. L., A. Martin, J. Cullen, and M. Trend (1999), The Brown University foraminiferal database, Data Contrib. Ser. 1999027, http://www.ngdc.noaa.gov/paleo/ paleo.html, Natl. Clim. Data Cent., Asheville, N. C. Reichart, G. J., L. J. Lourens, and W. J. Zachariasse (1998), Temporal variability in the northern Arabian Sea Oxygen Minimum Zone (OMZ) during the last 225,000 years, Paleoceanography, 13, 607 – 621. Reichert, B. K., R. Schnur, and L. Bengtsson (2002), Global ocean warming tied to anthropogenic forcing, Geophys. Res. Lett., 29(11), 1525, doi:10.1029/2001GL013954. Rimbu, N., G. Lohmann, J.-H. Kim, H. W. Arz, and R. Schneider (2003), Arctic/North Atlantic Oscillation signature in Holocene sea surface temperature trends as obtained from alkenone data, Geophys. Res. Lett., 30(6), 1280, doi:10.1029/2002GL016570. Rostek, F., G. Ruhland, F. C. Bassinot, P. J. Muller, L. D. Labeyrie, Y. Lancelot, and E. Bard (1993), Reconstructing sea surface temperature and salinity using d18O and alkenone records, Nature, 364, 319 – 321. Rostek, F., E. Bard, L. Beaufort, C. Sonzogni, and G. Ganssen (1997), Sea surface temperature and productivity records for the past 240 kyr in the Arabian Sea, Deep Sea Res. Part II, 44, 1461 – 1480. Royston, J. P. (1982a), Algorithm AS 177: Expected normal order statistics (exact and approximate), Appl. Statist., 31, 161 – 165. Royston, J. P. (1982b), An extension of Shapiro and Wilk’s W test for normality to large samples, Appl. Statist., 31, 115 – 124. Ryan, T. P. (1997), Modern Regression Methods, 515 pp., John Wiley, Hoboken, N. J. Schott, F. A., and J. P. Mc Creary Jr. (2001), The monsoon circulation of the Indian Ocean, Prog. Oceanogr., 51, 1 – 123. Schulz, H., U. von Rad, and H. Erlenkeuser (1998), Correlation between Arabian Sea and Greenland climate oscillations of the past 110,000 years, Nature, 393, 54 – 57. Shapiro, S. S., and M. B. Wilk (1965), An analysis of variance test for normality (complete samples), Biometrika, 52, 591 – 611.

9 of 9

PA1004

Sirocko, F., M. Sarnthein, H. Erlenkeuser, H. Lange, M. Arnold, and J. C. Duplessy (1993), Century-scale events in monsoonal climate over the past 24,000 years, Nature, 364, 322 – 324. Sonzogni, C., E. Bard, and F. Rostek (1998), Tropical sea surface temperature during the last glacial period: A view based on alkenones in Indian Ocean sediments, Quat. Sci. Rev., 17, 1185 – 1201. Stott, L., K. Cannariato, R. Thunell, G. H. Haug, A. Koutavas, and S. Lund (2004), Decline of surface temperature and salinity in the western tropical Pacific Ocean in the Holocene epoch, Nature, 431, 56 – 59. Stuiver, M., and P. J. Reimer (1993), Extended 14 C data base and revised CALIB 3.0 14C age calibration program, Radiocarbon, 35, 215 – 230. Visser, K., R. Thunell, and L. Stott (2003), Magnitude and timing of temperature change in the Indo-Pacific warm pool during deglaciation, Nature, 421, 152 – 155. Wasserman, P. D. (1989), Neural Computing— Theory and Practice, 230 pp., Van Nostrand Reinhold, Hoboken, N. J. Webb, A. (2002), Statistical Pattern Recognition, 2nd ed., 496 pp., John Wiley, Hoboken, N. J. Weller, R. A., A. S. Fischer, D. L. Rudnick, C. C. Eriksen, T. D. Dickey, J. Marra, C. Fox, and R. Leben (2002), Moored observations of upper-ocean response to the monsoons in the Arabian Sea during 1994 – 1995, Deep Sea Res. Part II, 49, 2195 – 2230. Yamamoto, M., T. Oba, J. Shimamune, and T. Ueshima (2004), Orbital-scale anti-phase variation of sea surface temperature in midlatitude North Pacific margins during the last 145,000 years, Geophys. Res. Lett., 31, L16311, doi:10.1029/2004GL020138. Zahn, R., and T. F. Pedersen (1991), Late Pleistocene evolution of surface and middepth hydrography at the Oman Margin: Planktonic and benthic isotope records at Site 724, Proc. Ocean Drill. Program Sci. Results, 117, 291 – 308.

B. A. Malmgren, Department of Earth Sciences-Marine Geology, Go¨teborg University, Box 460, SE-40530 Go¨teborg, Sweden. P. D. Naidu, National Institute of Oceanography, Dona Paula, 403 004 Goa, India. (divakar@ darya.nio.org)