Recent trends in sea level pressure in the Indian Ocean region

Transcription

1 GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L19712, doi: /2006gl027175, 2006 Recent trends in sea level pressure in the Indian Ocean region Dan Copsey, 1 Rowan Sutton, 1 and Jeff R. Knight 2 Received 9 June 2006; revised 9 August 2006; accepted 12 September 2006; published 14 October [1] During the second half of the twentieth century the Indian Ocean exhibited a rapid rise in sea surface temperatures (SST). It has been argued - largely on the basis of experiments with atmospheric GCMs - that this rapid warming was an important cause of remote changes in climate, in particular an increasing trend in the North Atlantic Oscillation Index and decreases in African rainfall. Here however we present evidence that the Indian Ocean warming was associated with local increases in sea level pressure (SLP). These increases are inconsistent with results from experiments in which an atmospheric GCM is forced by historical SST, which show robust decreases in SLP. The clear discrepancy between the observed and simulated trends in SLP suggests that the response of some atmospheric GCMs to the Indian Ocean warming may not provide a reliable guide to the behaviour of the real world. Citation: Copsey, D., R. Sutton, and J. R. Knight (2006), Recent trends in sea level pressure in the Indian Ocean region, Geophys. Res. Lett., 33, L19712, doi: /2006gl Introduction [2] During the second half of the twentieth century the Indian Ocean exhibited a rapid rise in sea surface temperatures (SST) [Lau and Weng, 1999]. At the same time the index of the North Atlantic Oscillation (NAO), a key measure of European winter climate [Hurrell and VanLoon, 1997], showed a notable trend toward more positive values. Experiments in which atmospheric GCMs are forced with a reconstruction of the historical variations in SST suggest that the warming of the Indian Ocean made an important contribution to forcing the trend in the NAO index [Hoerling et al., 2001; Bader and Latif, 2003; Hurrell et al., 2004]. The mechanism involves increases in precipitation over the Indian Ocean followed by an extratropical response to the associated latent heating [Hoerling et al., 2004]. Other work has suggested that important climate impacts were not restricted to the North Atlantic region but included, notably, major changes in precipitation over both northern and southern Africa [Giannini et al., 2003; Lu and Delworth, 2005; Hoerling et al., 2006]. [3] The evidence that the warming of the Indian Ocean contributed to the observed changes in the NAO and in African rainfall appears convincing. However, there are reasons for caution. First, there is a lack of direct observations that precipitation did in fact increase over the Indian Ocean. Secondly, there is evidence from various studies that experiments in which AGCMs are forced with prescribed 1 Department of Meteorology, University of Reading, Reading, UK. 2 Hadley Centre for Climate Prediction and Research, Met Office, Exeter, UK. Copyright 2006 by the American Geophysical Union /06/2006GL SST can sometimes give misleading results [Kumar and Hoerling, 1998; Lau and Nath, 2000; Kitoh and Arakawa, 1999; Wang et al., 2004; Douville, 2005; Wu and Kirtman, 2005; Wu et al., 2006]. The problem arises (at least in part) because fixing SST interferes with air-sea coupling, and the Indian Ocean has been highlighted by these studies as a particular region where fixing SST may cause problems. [4] The above cited studies focus primarily on the role of SST in the intraseasonal or interannual variability of the Indian Ocean region [e.g., Wu and Kirtman, 2005; Wu et al., 2006], or in the simulated response to greenhouse gas forcing [Douville, 2005]. However they also raise the concern that the response of atmospheric GCMs to the twentieth century Indian Ocean warming may not be a reliable guide to the behaviour of the real world. Ideally one would examine this possibility through analysis of observed trends in precipitation. However, the necessary records for ocean locations do not exist. Satellite records, which do include ocean coverage, do not begin until 1979 and therefore do not span adequately the period of interest, which is from In this study we present, as an alternative, an analysis of trends in sea level pressure (SLP). Although not directly related to precipitation, SLP is a useful variable because it is relatively easy to observe, has large correlation scales, and good records exist for ocean locations. Furthermore, in the tropics increases in latent (or other diabatic) heating are typically associated with local decreases in sea level pressure. The reason is that the heating causes columns of air to expand. At a fixed height, pressure rises relative to adjacent locations, and the resulting pressure gradients cause divergence out of the column, causing SLP to fall. 2. Data Sets and Model Simulations [5] The observed SST data set used here is the HadISST1 data set of in-situ measurements of sea surface temperature [Rayner et al., 2003]. Two observed SLP data sets are used: HadSLP1 [Basnett and Parker, 1997], and the Kaplan analysis [Kaplan et al., 2000] of the COADS data [Woodruff et al., 1987]. Note that the HadSLP1 analyses also exploit the COADS data but, in addition, incorporate data from the Met Office marine data bank and from land stations. As a guide, the number of observations per 5 grid box in the HadSLP1 analyses was of order 10 3 in the north east Indian Ocean and of order 10 2 in the south west Indian Ocean in the decade By the 1970s, these numbers had increased to observations per grid box per decade, and by the 1990s to 10 4 in almost all grid boxes. [6] The Atmospheric General Circulation Model (AGCM) used in this study is the Hadley Centre s atmospheric model HadAM3 [Pope et al., 2000]. HadAM3 runs at a horizontal resolution of 3.75 longitude by 2.5 latitude, L of5

2 with 19 vertical levels and a 30 minute timestep. The model uses an Arakawa B grid with hybrid vertical coordinates and a Eulerian advection scheme. [7] Two different HadAM3 experiments are analysed. In the first, HadAM3 was forced by observed SSTs from the HadISST1 data set for the period 1870 to There were no changes to external forcings such as aerosols or greenhouse gases. An ensemble of 6 members, each differing only with respect to their initial conditions, was used to sample the internal variability of the atmosphere. In the second experiment, HadAM3 was forced by HadISST1 and also full natural and anthropogenic changes in external forcings ( [Stott et al., 2000; Johns et al., 2003]) for the period The ensemble size for the second experiment was 12. All data was converted to annual means before further analysis. 3. Observed Trends in SST and SLP [8] Figure 1a shows the trend in HadISST1 between 1950 and 1996 (computed from least squares regression). The stippled areas show where this trend is larger than the standard deviation of the interannual variability. This is roughly equivalent to a significance level of 97%. Note that we chose to compute trends up to 1996 in order to avoid possible influence of the record 1997/98 El Niño event. [9] Over the period SSTs increased in most of the tropics with most rapid warming in the Indian Ocean, the central and eastern Pacific, and the southern Atlantic. Some of this warming is likely to have been due to increases in greenhouse gas concentrations [Intergovernmental Panel on Climate Change, 2001]. However internal variability may also have influenced the SST pattern. In the Pacific, the wedge shaped pattern of increasing SSTs in the central and eastern Pacific and decreasing SSTs in the northwest and southwest Pacific resembles the signature of the Interdecadal Pacific Oscillation (IPO [Stephens et al., 2001]). The IPO underwent a large shift during when central Pacific temperatures warmed and subtropical western Pacific temperatures cooled [Zhang et al., 1997]. [10] The warming of SST in the Indian Ocean may have been partly a response to remote forcing from the Pacific. During El Niño events an anomalously warm central and eastern Pacific reduces cloud cover over the Indian Ocean and allows in more solar radiation to warm the sea surface [Klein et al., 1999]. In addition, changes in the winds affect SST through changes in the turbulent surface heat fluxes [Reason et al., 2000]. The SST pattern in Figure 1a differs from that associated with El Niño, but there are some similarities. [11] Figure 1b shows the average trend in SLP from observations (HadSLP1) between 1950 and It shows decreasing SLP in the central Pacific where SSTs increased. As noted in the introduction, decreasing SLP is a typical feature of the tropical response to increased heating, therefore this association is consistent with the idea that the warming of SST led to increased heating of the atmosphere. [12] Over the same period SLP increased in the Indian and Atlantic Oceans. The pattern suggests that atmospheric mass shifted out of the Pacific to the east and west. Inspection of higher latitudes also suggests that mass shifted from the extratropics into the tropics (not shown). In the Figure 1. Average trends during the period 1950 to 1996 for: (a) SSTs from observations (HadISST1); (b) SLP from observations (HadSLP1); (c) SLP from HadAM3 forced with just SSTs; (d) SLP from HadAM3 with all the forcings. The colours are the average rate of change in each field and the stippled areas show where this rate of change is significant. All the HadAM3 plots are computed from ensemble means. Indian Ocean the increases in SLP overlie increases in SST. This association appears to be at odds with the idea that warming of the Indian Ocean SST caused local increases of precipitation and latent heating. It might be explained if the SLP and SST changes were part of a remote response to increased heating over the central Pacific, similar to that which occurs during El Niño events. Alternatively, it might be explained as part of the response to the effect of changing radiative forcings on the heating and cooling of the atmosphere. 2of5

3 Figure 2. Precipitation trends during the period 1950 to 1996 from: (a) HadAM3 forced with just SSTs, and (b) HadAM3 with all the forcings. The colours are the average rate of change in precipitation and the stippled areas show where this rate of change is significant. All plots are computed from ensemble means. [13] To check the SLP trends shown in Figure 1b we computed a similar figure using the Kaplan data set (not shown). Over the Indian and Pacific Oceans the pattern of trends is very similar. There is a difference in the southern Atlantic, where the increasing SLP suggested by HadSLP1 is not seen. As noted in section 2, HadSLP1 incorporates additional data not included in the Kaplan analyses. To provide a further check we examined some individual SLP records for land stations in and around the Indian Ocean. The picture from these records is not entirely consistent. So, for example, records from Singapore (104 E, 1 N) and the Seychelles (57 E, 5 S) shows clear positive trends, whereas a record from Sri Lanka (81 E, 6 N) shows no significant trend of either sign. However, land stations in the Indian Ocean are, of course, few and far between. Given the highly coherent pattern of increasing SLP suggested by Figure 1b, and the comparatively large numbers of marine observations (detailed in section 2), we see no reason to doubt the conclusion that the Indian Ocean region saw a general increase in sea level pressure over the period [14] We also examined the seasonal variation of the SLP trend. While there is seasonal variation, the large scale pattern of decreasing SLP over the central Pacific, and increasing SLP over the Indian and Atlantic Oceans is found in all seasons. Note that increases in sea level pressure over the Indian Ocean in the boreal winter season (DJF) were also shown for the period by Gillett et al. [2005] using reanalysis data. 4. Simulated Trends in SLP [15] Figures 1c and 1d show the ensemble mean SLP trends ( ) from the two experiments with HadAM3. Both panels show decreasing SLP over the central Pacific and increasing SLP over the north Atlantic, consistent with observations. These simulations also reproduce a positive trend in the NAO index (not shown). However, in contrast to the observations they exhibit negative SLP trends over the Indian Ocean. These negative SLP trends are significant over most of the western Indian Ocean in the experiment with SST forcing alone, and primarily over the southwest Indian Ocean in the experiment with all forcings. [16] Figure 2 shows the trends in ensemble mean precipitation from the two model experiments. Comparatively rapid increases in precipitation are seen over the central Indian Ocean, overlying the region of most rapidly warming SST. These increases are very similar to those seen in other atmospheric GCMs [see Hurrell et al., 2004, Figure 7]. It is likely that the decreases in SLP seen in the HadAM3 experiments are a response to increased latent heating associated with the increased precipitation. In particular, the strong negative trend seen to the east of Madagascar in Figures 1c and 1d is located to the southwest of the main centre of increasing precipitation in an arrangement suggestive of a Gill-type Rossby wave response [Gill, 1980]. There is no sign of such a response in the observed SLP trend (Figure 1b). [17] Thus far we have examined ensemble mean trends. Perhaps the variability between ensemble members is large enough to encompass the observed trend? Figure 3 shows SLP trends averaged over the whole of the Indian Ocean for each of the ensemble members (both experiments) and from the two observational data sets. It is immediately apparent that both estimates of the observed trend lie far outside the range spanned by the individual ensemble members. The estimates of the observed trend are hpa year 1 (Kaplan, average over sea points only) and hpa year 1 (HadSLP1, average over land and sea points). By contrast, the HadAM3 experiment with SST forcing only exhibits a trend of ± hpa year 1 while the experiment with all forcings exhibits a trend of ± hpa year 1. It appears that the direct effect of external forcings does act to weaken the trend of decreasing Figure 3. Average trends in SLP in the Indian Ocean (30 E 130 E, 20 N 30 S) for the period 1950 to 1996 using all the ensemble members of HadAM3 (including those forced with (top) just SSTs and (middle) those with all the forcings). Observed trends from HadSLP1 and Kaplan are shown as the bottom bars. 3of5

4 SLP simulated by HadAM3, but far from sufficiently to bring about agreement with the observations. [18] The contrast between the simulated and observed trends is illustrated further in Figure 4, which shows timeseries of SLP in the southwest Indian Ocean (the region east of Madagascar previously discussed). While there are significant offsets in the mean values, both the observed timeseries show increasing SLP, while both the HadAM3 timeseries show decreasing SLP. The largest decreases in the model simulations occur between the mid 1960s and mid 1970s, while the largest increases in the observations occur in the mid-to-late 1970s. The latter change is coincident with the shift in the phase of the IPO. 5. Discussion and Conclusions [19] We have presented strong evidence that during the period SLP rose over the Indian Ocean region. By contrast, simulations in which the HadAM3 model is forced with the historical record of SST show decreases in SLP. This decrease is associated with increases in precipitation directly over the region of most rapidly warming SST, and is likely to be a response to the associated increases in latent heating. The simulated decrease in SLP is robust in that it is common to all ensemble members. The inclusion of changing external forcings weakens the negative SLP trend, but only by 25%. [20] The contrast between the model results and observations suggests a model error. As noted in the introduction, several studies have shown that fixing SST can lead to an incorrect representation of air-sea interactions particularly in the Indian Ocean. It appears that, whereas in the model simulations the SLP trend is determined by a local response to the warming of SST, in reality this trend is controlled by other factors. In particular, the fact that the largest observed increase in Indian Ocean SLP shown in Figure 4 occurred at the same time as the warming of the Pacific Ocean associated with the shift in the IPO suggests the SLP increase may be a remote response to the changes in the Pacific. There is a hint of a similar increase in the model simulations, but it is notably weaker. Thus, it appears that the model error is characterised by an excessive sensitivity of Indian Ocean SLP to local SST warming, and insufficient sensitivity to remote SST warming. This interpretation is further supported by experiments with the CCM3 model (R. Seager, personal communication, 2006). In GOGA runs forced with global SST data CCM3, like HadAM3, simulates a negative SLP trend over the Indian Ocean, whereas in POGA-ML runs (in which SST is prescribed in the tropical Pacific, but computed using a mixed-layer model in the Indian Ocean) CCM3 simulates a positive SLP trend, consistent with observations. [21] We suggested in the introduction that changes in Indian Ocean SLP may provide a proxy for changes in precipitation; we argued this on the basis that, in the tropics, SLP and precipitation are typically anti-correlated. If this anti-correlation holds for the trends considered in this paper, then our evidence of a positive trend in Indian Ocean SLP calls into question the reality of the significant increases in Indian Ocean precipitation that have been postulated as a key driver of changes in the NAO index and in African precipitation [Hurrell et al., 2004; Hoerling et al., 2004; Figure 4. Timeseries of SLP in the southwest Indian Ocean (45 E 80 E, 10 S 30 S) from the ensemble means of HadAM3 forced with just SSTs (dashed), HadAM3 with all the forcings (dotted), HadSLP1 (black) and Kaplan (grey). The annual means are shown plus those computed using a low pass filter (19 point Henderson filter). Giannini et al., 2003; Lu and Delworth, 2005; Hoerling et al., 2006]. Doubt about these increases has also been expressed in another recent study [Deser and Phillips, 2006]. These authors argue on the basis of an analysis of surface marine cloudiness and wind divergence that substantial increases in precipitation are unlikely to have occurred. In addition, Norris [2005] shows evidence that between 1952 and 1997 in the equatorial Indian Ocean there was a negative trend in upper level cloud cover, suggestive of decreases in precipitation. [22] A further interesting feature of the Deser and Phillips [2006] study is that the authors compare results from GOGA experiments with two different atmospheric GCMs: CCM3 and CAM3. In CCM3 they find a substantial increase in Indian Ocean precipitation similar to that simulated by HadAM3, but in CAM3 the increase is much smaller. They argue that the CAM3 result is more consistent with observations; if true, this finding could suggest that the failure of some atmospheric GCMs to simulate correctly the response to Indian Ocean warming may not be due simply, or even mainly, to the use of a prescribed SST boundary condition. Rather, it may reflect specific failings of the model, such as a convection scheme that is overly sensitive to near surface temperature and humidity, or errors in the simulation of teleconnections that may act to suppress convection over the Indian Ocean region even in the presence of significant surface warming. To the extent that such errors are present in atmospheric models, they could also compromise results from coupled climate models [e.g., Selten et al., 2004]. [23] The discrepancy we have shown between observed and simulated trends in Indian Ocean SLP suggests that the response of some atmospheric GCMs to the Indian Ocean warming may not provide a reliable guide to the behaviour of the real world. The most important area for further work is the need to investigate fully the relationship between trends in tropical SLP and trends in precipitation. If this further work confirms that the observed positive trend in Indian Ocean SLP is indeed inconsistent with significant 4of5

5 increases in Indian Ocean precipitation, there may also be a need to seek alternative explanations for the observed trend in the NAO index. [24] Acknowledgments. This work was funded by the Natural Environment Research Council with some extra support from the University of Reading. Rowan Sutton is supported by a Royal Society University Research Fellowship. We would also like to acknowledge help from Adam Scaife and Jim Hurrell. References Bader, J., and M. Latif (2003), The impact of decadal-scale Indian Ocean sea surface temperature anomalies on Sahelian rainfall and the North Atlantic Oscillation, Geophys.Res.Lett., 30(22), 2169, doi: / 2003GL Basnett, T. A., and D. E. Parker (1997), Development of the Global Mean Sea Level Pressure Data Set GMSLP2, CRTN79, Hadley Cent. for Clim. Predict. and Res., Met Office, Exteter, U. K. Deser, C., and A. Phillips (2006), Simulation of the 1976/1977 climate transition over the North Pacific: Sensitivity to tropical forcing, J. Clim., in press. Douville, H. (2005), Limitations of time-slice experiments for predicting regional climate change over South Asia, Clim. Dyn., 24(4), Giannini, A., R. Saravanan, and P. Chang (2003), Oceanic forcing of Sahel rainfall on interannual to interdecadal time scales, Science, 302(5647), Gill, A. E. (1980), Some simple solutions for heat-induced tropical circulation, Q. J. R. Meteorol. Soc., 106, Gillett, N. P., R. J. Allan, and T. J. Ansell (2005), Detection of external influence on sea level pressure with a multi-model ensemble, Geophys. Res. Lett., 32, L19714, doi: /2005gl Hoerling, M. P., J. W. Hurrell, and T. Y. Xu (2001), Tropical origins for recent North Atlantic climate change, Science, 292(5514), Hoerling, M. P., J. W. Hurrell, T. Xu, G. T. Bates, and A. S. Phillips (2004), Twentieth century North Atlantic climate change: Part II. Understanding the effect of Indian Ocean warming, Clim. Dyn., 23(3 4), Hoerling, M. P., J. W. Hurrell, and J. Eischeid (2006), Detection and attribution of 20th century northern and southern African monsoon change, J. Clim., 19(16), Hurrell, J. W., and H. VanLoon (1997), Decadal variations in climate associated with the North Atlantic oscillation, Clim. Change, 36(3 4), Hurrell, J. W., M. P. Hoerling, A. Phillips, and T. Xu (2004), Twentieth century North Atlantic climate change: Part I. Assessing determination, Clim. Dyn., 23(3 4), Intergovernmental Panel on Climate Change (2001), Climate Change 2001: The Scientific Basis. Contribution of Working Group 1 to the Third Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge Univ. Press, New York. Johns, T. C., et al. (2003), Anthropogenic climate change for 1860 to 2100 simulated with the HadCM3 model under updated emissions scenarios, Clim. Dyn., 20(6), Kaplan, A., Y. Kushnir, and M. A. Cane (2000), Reduced space optimal interpolation of historical marine sea level pressure: , J. Clim., 13(16), Kitoh, A., and O. Arakawa (1999), On overestimation of tropical precipitation by an atmospheric GCM with prescribed SST, Geophys. Res. Lett., 26(19), Klein, S. A., B. J. Soden, and N. C. Lau (1999), Remote sea surface temperature variations during ENSO: Evidence for a tropical atmospheric bridge, J. Clim., 12(4), Kumar, A., and M. P. Hoerling (1998), Specification of regional sea surface temperatures in atmospheric general circulation model simulations, J. Geophys. Res., 103(D8), Lau, K. M., and H. Y. Weng (1999), Interannual, decadal-interdecadal, and global warming signals in sea surface temperature during , J. Clim., 12(5), Lau, N. C., and M. J. Nath (2000), Impact of ENSO on the variability of the Asian-Australian monsoons as simulated in GCM experiments, J. Clim., 13(24), Lu, J., and T. L. Delworth (2005), Oceanic forcing of the late 20th century Sahel drought, Geophys. Res. Lett., 32, L22706, doi: / 2005GL Norris, J. R. (2005), Trends in upper-level cloud cover and surface divergence over the tropical Indo-Pacific Ocean between 1952 and 1997, J. Geophys. Res., 110, D21110, doi: /2005jd Pope, V. D., M. L. Gallani, P. R. Rowntree, and R. A. Stratton (2000), The impact of new physical parametrizations in the Hadley Centre climate model: HadAM3, Clim. Dyn., 16(2 3), Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan (2003), Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century, J. Geophys. Res., 108(D14), 4407, doi: / 2002JD Reason, C. J. C., R. J. Allan, J. A. Lindesay, and T. J. Ansell (2000), ENSO and climatic signals across the Indian Ocean Basin in the global context: Part 1. Interannual composite patterns, Int. J. Climatol., 20(11), Selten, F. M., G. W. Branstator, H. A. Dijkstra, and M. Kliphuis (2004), Tropical origins for recent and future Northern Hemisphere climate change, Geophys. Res. Lett., 31, L21205, doi: /2004gl Stephens, C., S. Levitus, J. Antonov, and T. P. Boyer (2001), On the Pacific Ocean regime shift, Geophys. Res. Lett., 28(19), Stott, P. A., S. F. B. Tett, G. S. Jones, M. R. Allen, J. F. B. Mitchell, and G. J. Jenkins (2000), External control of 20th century temperature by natural and anthropogenic forcings, Science, 290(5499), Wang, B., I. S. Kang, and J. Y. Lee (2004), Ensemble simulations of Asian- Australian monsoon variability by 11 AGCMs, J. Clim., 17(4), Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer (1987), A Comprehensive Ocean-Atmosphere Data Set, Bull. Am. Meteorol. Soc., 68(10), Wu, R. G., and B. P. Kirtman (2005), Roles of Indian and Pacific Ocean airsea coupling in tropical atmospheric variability, Clim. Dyn., 25(2 3), Wu, R. G., B. P. Kirtman, and K. Pegion (2006), Local air-sea relationship in observations and model simulations, J. Clim., in press. Zhang, Y., J. M. Wallace, and D. S. Battisti (1997), ENSO-like interdecadal variability: , J. Clim., 10(5), D. Copsey, 86 Scrub Rise, Billericay, Essex CM12 9PE, UK. (dan.copsey@googl .com) J. R. Knight, Hadley Centre for Climate Prediction and Research, Met Office, Fitzroy Road, Exeter EX1 3PB, UK. R. Sutton, Department of Meteorology, University of Reading, PO Box 243, Reading RG6 6BB, UK. 5of5