sign in sign up
<<
Home , Atmospheric circulation, Equatorial wave, Rossby wave

ARTICLE IN PRESS

Deep- Research I 51 (2004) 2123–2136 www.elsevier.com/locate/dsr

The Rossby as a key mechanism of Indian climate variability

Mark R. Jurya, Bohua Huangb

aEnvironmental Science Department, Center for Environmental Studies, University of Zululand, KwaDlangezwa 3886, South Africa bCenter for Ocean Land Studies, Maryland, USA

Received 18 November 2003; received in revised form 3 June 2004; accepted 3 June 2004 Available online 7 October 2004

Abstract

We analyze the time-longitude structure of composite cases from model-assimilated ocean data in the period 1958–1998, following on from earlier work by Huang and Kinter (J. Geophys. Res. 107(C11) (2002) 3199) that studied east–west variability in the . Our analysis focuses on the signal along the thermocline ridge in the tropical SW Indian Ocean (101S, 60–801E), where stress curl is important. Anomalous in the equatorial east Indian Ocean force successive Rossby westward at speeds of 0.1 m s1730%. With a wavelength of 7000 km, the period of oscillation is in the range 1.9–5.2 years. The Indian Ocean Rossby wave is partially resonant with the global influence of the El Nino–Southern Oscillation, except during quasi-biennial rhythm. The presence of the Rossby wave offers potential predictability for east–west systems and climate that affect resources in countries surrounding the Indian Ocean. r 2004 Elsevier Ltd. All rights reserved.

1. Introduction Somalia and surface cooling off Madagascar. A zone of maximum occurs over the The west Indian Ocean (WIO) is a climatically eastern warm pool (51N–101S, 70–1001E) with important region, recording on average 10 tropical rates48 mm day1. Zonal winds in each year between December and March. the tropical Indian Ocean are weak within the These storms often bring devastating conse- -guide because of the prevailing quences to the Mascarene Islands, Madagascar meridional flow of the . The equatorial and Mozambique (Naeraa and Jury, 1998). The thermocline lies around 120 m depth and exhibits mean (SST) in the WIO little east–west slope in contrast with the Pacific varies between 26 and 28 1C, and is cooler than the and Atlantic (Xie et al., 2002). The mean eastern Indian Ocean, due to off structure of the thermocline is a ridge–trough pair at 71Sand201S, respectively, with mean tempera- E-mail address: [email protected] (M.R. Jury). tures of 18 and 22 1C for the upper 234 m and

0967-0637/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.dsr.2004.06.005 ARTICLE IN PRESS

2124 M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 thermocline depths of 80 and 180 m, respectively. (Murtugudde et al., 2000; Behera and Yamagata, Due to the lack of a zonal gradient in the 2001; Xie et al., 2002). equatorial zone, it was thought that the Indian In the present study, we further investigate the Ocean may not have the appropriate features to mechanisms of ocean variability using model-assimi- develop its own El Nino–Southern Oscillation lated data, following Huang and Kinter (2002).Of (ENSO)-like interannual variability (Latif and particular interest is the nature of sub-tropical Barnett, 1995). Rossbywavesandtheirimpactontheclimate Yet, east–west gradients in the thermocline can system. Toward this end, we analyze in situ ocean develop at times (Hastenrath et al., 1993; Webster data assimilated by a numerical model (Schneider et al., 1999; Murtugudde et al., 2000; Ueda and et al., 1999). Key issues to be addressed include: the Matsumoto, 2000), with atmospheric circulation time-longitude structure of Rossby waves, the and convection varying in a manner consistent characteristics of phase propagation, preferred loca- with Bjerknes (1969) feedback. At the onset of a tions for wave amplification, and the extent of Pacific El Nino event as coupling with the atmosphere. We conclude that shifts eastward from the warm pool, anomalous much of ocean variability comprising the east–west easterly winds develop near the in the east dipole in the Indian Ocean can be attributed to an Indian Ocean. These generate a gradual warming off-equatorial ocean Rossby wave that propagates to the west (Nigam and Shen, 1993; Klein et al., slowly from the east following a surge of wind at the 1999; Lau and Nath, 2000) that will be further onsetofmostElNinoevents.Thepaperisorganized explored here. into sections on data and methods, results sub- Although surface fluxes explain changes in SST divided into wave character and zonal phase over much of the tropical Indian Ocean, this is less propagation (hovmoller analysis), and discussion. so in the thermocline ridge (5–121S, 55–851E) according to the results of Klein et al. (1999), Lau and Nath (2000). Ocean model results 2. Data and methods (Murtugudde and Busalacchi, 1999; Murtugudde et al., 2000) and analysis of sea surface height Upper ocean heat content fields (mean tempera- measurements (Chambers et al., 1999) suggest that ture in the top 234 m) for the period 1958–1998 dynamic processes contribute to SST variability in have been derived from an ocean data assimilation the tropical Indian Ocean. Huang and Kinter system operated at COLA as outlined in Huang (2000, 2001, 2002) have used upper ocean heat and Kinter (2002), hereafter HK02. The data content from data-assimilated ocean analyses to assimilation uses a variational scheme (Derber and identify the westward propagating Rossby wave as Rosati, 1989) to combine temperature observa- a key feature of the near-continuous interannual tions in an ocean general circulation model oscillation of the upper tropical Indian Ocean (oGCM). The analysis uses all observations since the late 1950s. This result substantiated the available in a moving 10-day assimilation window. finding from the TOPEX/Poseidon The observations are inserted into the first guess measurements during the 1990s. Observational field of the oGCM and discrepancies are iteratively studies of warm and cool events (Reason et al., minimized. The oGCM is a version of the GFDL 2000; Jury et al., 2002) demonstrate that local heat modular ocean model (Huang and Schneider, and radiative fluxes, and vertical entrainment from 1995; Schneider et al., 1999; Huang et al., 2002). curl account for a significant portion of It is forced by monthly averaged surface wind SST variance during onset phase. However, the stress from NCEP reanalysis available at an surface heat fluxes do not fully explain the irregular grid with a variable resolution of around variations of SST there from fall to next spring. 21 near the equator. The solar flux is prescribed The lack of closure in the budget for local SST (Oberhuber, 1988), surface heat and long-wave changes in the WIO may be attributed to fluxes are parameterized (Philander et al., 1987; horizontal via incoming Rossby waves Rosatiand Miyakoda,1988). ARTICLE IN PRESS

M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 2125

Ocean temperature observations are assimilated more observations along key lines from Perth to from in situ SST measurements from the COADS the NW, from Madagascar to the NE, from archive (Slutz et al., 1985), temperature profile Mauritius to the north, and from Mombassa to (XBT, MBT, CTD, etc.) measurements (Conk- the east (Masumoto and Meyers, 1998), which are right et al., 1998) and, since 1981, with - important for our present study. However, some blended weekly SST fields (Reynolds and Smith, of these, especially in the South Indian Ocean, 1994). The number of temperature profiles is were established with TOGA in 1985. generally 3000–9000 per year within the square Fig. 1a and b shows the spatial distributions of between 301Eand1201E, 301S and 301N during the temperature profiles per degree-square for the the 41 years (see Fig. 1 a and b). In 1967, 1968, and years 1958–1980 and 1981–1998. It is evident that 1979, it surpassed 15000 profiles. In situ SST measurements on major ship tracks were enhanced measurements exceeded 90,000 per year, reaching in the later period in the South Indian Ocean. Even a maximum of 270,000 per year. However, many after this enhancement, the distribution of the sub- of these observations, especially the sub-surface surface temperature observations was still sparse data, are concentrated along the coasts. There are and may be wider than the typical zonal decorr- elation scale in the equatorial ocean (Smith, 1995). Therefore, the assimilated product is inevitably Distribution of Temperature Profiles strongly affected by the ocean model and the Profile Number, 1958-1980 30N surface forcing fields. The model-derived results 25N are considered in this light and we therefore 20N employ analysis techniques that highlight repeti- 15N tious patterns (e.g., EOF, composite). 10N The anomalies of model-analyzed upper ocean 5N 5000 heat content are highly correlated with the EQ 5S 1000 independently derived sea surface height anoma- lies from the TOPEX/Poseidon satellite altimeter 10S 500 15S measurements (HK02) using monthly data. It 20S 300 suggests that the assimilated temperatures are 25S 250 useful for analyzing the interannual variability in 30S 30E 40E 50E 60E 70E 80E 90E 100E 110E this region. One may attribute this high consis- (a) 200 tency to newly established ship data in recent 150 years. However, our EOF analysis of the heat Profile Number, 1981-1998 100 content anomalies (HCa) in different periods 30N 25N 75 shows very similar patterns. We believe that this data set is superior to an oGCM simulation with 20N 50 15N observed surface forcing only, because of insertion 10N 30 of the oceanographic observations. It significantly 5N 10 reduces the systematic errors of the model so that EQ it can respond to surface forcing in a more realistic 5S 10S manner. The inserted observations also nudge and 15S enhance the simulated interannual signals. 20S HK02 analyze the anomaly fields of detrended 25S seasonal data using extended empirical orthogonal 30S 30E 40E 50E 60E 70E 80E 90E 100E 110E function (eEOF) analysis (Weare and Nasstrom, (b) 1982) for lags up to 2 years. This covers the life- Fig. 1. The spatial distribution of profile measurements on span of a typical anomalous ‘dipole’ event in the 11lat. 11long. grid within 301S–301N, 30–1201E during: (a) Indian Ocean. Here we extend the work of 1958–1980, and (b) 1981–1998. HK02 by analyzing the time-longitude structure ARTICLE IN PRESS

2126 M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 of composite HCa according to the phase of 12 tra, and the thermocline ridge (7–121S, 55–851E) events occurring in the 41 year period of analysis. exhibit high SST variance. Wind stress curl over It should be pointed out that, in this empirically the ridge is zonally extensive and lifts the 20 1C derived oscillation, the periods of the events are isotherm to an average depth of 70 m (Schott and variable. Therefore, faster and slower events are McCreary, 2004). Xie et al. (2001) found a co-mingled within the record. Our preliminary correlation4+0.50 between interannual variabil- analysis showed that each cycle went through ity of Z20 and SST in the thermocline ridge, similar stages so that it is useful to construct a highlighting the role of vertical entrainment. One composite picture based on the phases within these question is whether this ridge is more sensitive to events. To perform the composite analysis, we propagating wave systems than the surrounding select data to average at certain phase-intervals region. where the time score amplitude exceeds 0:75s: To create a more consistent composite analysis, the 3.1. Character of WIO Rossby waves HCa data for fast events are ‘compressed’ whilst data for the slow events are ‘stretched’. Similarly, In HK02 the leading mode of seasonally SSTa and wind stress curl anomalies are analyzed averaged HCa represents a basin-wide westward- over the life cycle of the composite event. The propagating pattern consistent with a Rossby composite results are sliced at the of wave (‘o’ shape). It explains only 15% of maximum HCa amplitude (101S) and plotted in interannual variance due to noise possibly con- hovmoller format to study the environmental tributed by synoptic eddies, infrequent data cover- impacts of the transient Rossby wave. We age and the use of raw unfiltered data at monthly investigate relationships with ENSO using 21 resolution. Fig. 2 illustrates the spatial pattern monthly NCEP wind data in the area 51N–101S, when the east–west dipole achieves maximum 70–1001E and Reynolds reconstructed SST in the amplitude. This occurs south of the equator area 51N–51S, 150–901W (Nino3). We first wavelet around 101S, 751E, during interaction with the filter the two indices to remove cycleso1.5 years thermocline ridge. Simultaneously cool anomalies so that ocean influences on the atmosphere can be occupy the region west of Sumatra. The HCa more clearly described. We then calculate their evolution is comparable to that of the annual cross-modulus spectrum and instantaneous phase lag as outlined in Jury et al. (2002). Using the zonal wind index and the thermocline depth at 5–101S, 651E, we isolate years with a deeper layer of warm water when east equatorial winds are anomalous easterly. We then analyze the compo- site structure of the regional circulation and convection response.

3. Results

Using the eEOF analysis of HCa, we illustrate how thermocline oscillations in the WIO region take the form of Rossby waves propagating from the east. To justify and interpret our results, we first consider those of Xie et al. (2002) and HK02. RMS variance of interannual SST variability is not Fig. 2. Spatial pattern at lag 4 for the dominant eEOF mode of at a maximum along the equator. Rather, the HCa (after Huang and Kinter, 2002). Maximum amplitude is coastal upwelling regimes off Somalia and Suma- found along 10S latitude; the east–west dipole is clearly evident. ARTICLE IN PRESS

M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 2127

Rossby wave according to HK02, although much poleward (White, 2000). HK02 showed that the slower and asymmetrical about the equator. The leading EOF mode of the precipitation anomalies interannual fluctuation is analogous to the ENSO over the Indian Ocean is characterized by the cycle in the Pacific, except that there is little fluctuation of precipitation off the Sumatra coast evidence for a northern Rossby wave due to the and a tendency for opposing anomalies in the west. presence of the Asian land mass and associated Moreover, its time series is significantly correlated atmospheric circulations (monsoon) that inhibit its with the time series of the combined EOF mode, westward propagation. with the precipitation leading winds by a . The interannual time scale is prominent in the This suggests that the convection off the Sumatra phase analysis of HK02. The mean period of coast serves as a driving force of the anomalous oscillation is 3.3 years, with a range from 1.9 to 5.2 wind pattern and its distribution is also affected by years. The period of oscillation changes gradually air–sea feedback. from faster quasi-biennial events (2.0–2.4 years in Here we further demonstrate that independent the period 1962–1971), to slower ENSO-like events extended EOF (eEOF) analyses of both the wind (4.8–5.2 years from 1974 to 1993). In the late 1990s stress and precipitation anomalies produce an the period of oscillation was in the range 2.7–3.1 oscillation similar to the HCa mode. Fig. 3 shows years. Since the oscillation apparently has a long- the first eEOF modes of the surface wind stress term modulation, we have also calculated its mean anomalies and the principal components of the period as 2.0 years in the early time (1962–1970) first two modes together. The time series of the two and 4.2 years in the later time (1978–1995). Given modes show a coherent interannual fluctuation of a wavelength of 7000 km (shown later), phase a propagating mode (Fig. 3e). These two leading speeds of 0.11 and 0.05 m s1 are implied for modes account for a relatively small amount of the oceanic Rossby wave propagation in these two total variance because the wind stress is noisier periods, respectively. The free wave phase speed is than the HCa fields. The principal components of given by C ¼Bg0H=f 2: With values assumed to this oscillatory mode are significantly correlated be B=2.25 1011, g0=3.8 102, H=70, and with those of the HC mode. The first principal f2=6.4 1010 at 101S. The phase speed C is components of both wind stress and heat content 0.1 m s1, a finding consistent with White (2000), eEOF obtain a simultaneous correlation of 0.8. Rao et al. (2002). Hence the observed propagation This enhances the statement that this wind is about two-thirds of the free Rossby wave speed oscillation is physically consistent with the HCa (Chelton et al., 1998). Such a slow speed of oscillation. movement is indicative of coupling with the The time-space maps (Fig. 3b, c, and d) depict a atmosphere, likely involving a shift of the Walker gradual strengthening of easterly winds in the overturning circulation cells. central and eastern equatorial Indian Ocean and Using combined EOF analysis of wind stress anomalous northwest winds in the central south- anomalies, SSTa, and HCa, HK02 demonstrate ern ocean, starting from a state of near normal that the HCa variability is strongly coupled with conditions (Fig. 3a). These two the surface wind stress and precipitation. The together form an anomalous wind curl pattern Rossby wave is affected by two mechanisms; the (Fig. 3c). first being reflection of equatorial Kelvin waves at Fig. 4 shows the pattern of the first eEOF modes the Sumatran coast and local reinforcement by for the precipitation as well as the principal winds. The second is the generation of off- components of its two leading modes, as in Fig. equatorial Rossby waves through the combined 3. These two modes each explain 7% of the total effect of equatorial zonal winds and the southeast variance respectively, close to those derived for the that create strong anticyclonic shear HCa even though the precipitation data have a around 101S in the open ocean. A weakening of shorter span. The temporal–spatial evolution the southeast trade winds results as the equatorial (Fig. 4a-d) shows a shift of positive rainfall winds surge westward toward mid-basin and turn anomalies from the eastern ocean off the Sumatra ARTICLE IN PRESS

2128 M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136

(a) (b)

(c) (d)

(e)

Fig. 3. The temporal–spatial evolution of the first eEOF mode of the surface wind stress anomalies (a–d) and the first and second principal component scores (e). Spatial patterns are displayed every other season over 2 years (8 lags) in (a, b, c, d, 0, +2, +4, +6 , respectively). For (a)–(d), the unit is dynes cm2 with regions of stress magnitude40.1 outlined. The thick and thin curves in (e) are the first and second principal components, respectively. The time series are normalized by their maxima. coast toward the west, while the negative anoma- nents of HCa and precipitation eEOF reaches 0.9, lies do the reverse. The time series (Fig. 4e) exhibit while the precipitation leads the HCa by one an interannual fluctuation that is consistent with season. those derived from the HCa in the same period, Similar oscillatory modes can also be found in even mimicking the same long-term modulation. SSTa and the 850 hPa wind anomalies. However, For their common period, the correlation coeffi- the relationship between the leading modes of the cient between the original first principal compo- SSTa and HCa is more complicated. Although the ARTICLE IN PRESS

M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 2129

(a) (b)

(c) (d)

(e)

Fig. 4. The temporal–spatial evolution of the first eEOF mode of the anomalies of precipitation (a–d) and the first and second principal components (e). The spatial patterns are displayed as in Fig 3. For (a)–(d), the contour interval is 1 mm day1 and regions greater than 1 mm day1 and those lower than 1 mm day1 are darkly and lightly shaded, respectively. The thick and thin curves in (e) are the first and second principal components, respectively, as in Fig 3. time series of the SSTa modes bear some correlation coefficients with wind stress and resemblance to those of the HCa modes, there precipitation. This is because of influences from are differences in the long-term modulations the sub-tropical SST fluctuations, which are not between the two sets of principal components. directly related to thermocline fluctuations. In- The correlation is 0.6 between the first modes of stead, it is more strongly related to evaporative the SSTa and HCa after the long-term trend is heat fluxes, as pointed out by Behera and removed. It is significant though not as high as the Yamagata (2001). ARTICLE IN PRESS

2130 M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136

Hence equatorial rainfall is modulated west of depending on the disposition of the thermocline the zonal wind anomaly in agreement with ridge and trough associated with the Rossby wave. Reverdin et al. (1986). Poleward flow links the One question is the extent to which the equatorial convection associated with the Rossby wave to the wind is related to Pacific ENSO phase. To test this atmospheric circulation in sub-tropical , we analyze east Indian zonal wind anomalies and giving rise to climate impacts in Australia and Nino3 SSTa over the same period 1958–1998 using southern Africa (Nicholls, 1989; Reason and a wavelet analysis technique described in Jury et Mulenga, 1999). When the ocean thermocline is al. (2002). The result is shown in Fig. 5. deep in the WIO, the adjacent southern hemi- The two time series are closely associated with sphere continents tend to experience increased r=–0.56 at zero lag, significant at po:01 account- southwesterly flow and subsident motions that ing for auto-correlation imposed by the filter. In contribute to drought. the period 1958–1998 about 30% of interannual Our analysis suggests a dynamical coupling variance of the zonal wind anomaly in the east between the sub-surface Rossby wave, the atmo- Indian Ocean is attributable to Pacific ENSO spheric convection and the overlying circulation: phase, as indicated by Nino3 SSTa. The instanta- both the zonal tropical system and the sub-tropical neous phase lag shows that Nino3 leads by a few trough. Centers of action amplify and decay months except during the early 1980s. Inspection

Fig. 5. Wavelet analysis of east Indian Ocean zonal wind anomaly (bold) and Nino3 SSTa (inverted), illustrating variability in the filtered time series (upper), the cross-spectral modulus and the instantaneous phase lag (lower), where negative refers to Nino3 leading. ARTICLE IN PRESS

M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 2131 of the filtered time series suggests that the faster lowering its frequency prior to 1977. Hastenrath Indian Ocean oscillations in the 1960s are only (2000) points out that this connection between partially replicated in the Pacific SST, hence a these two oceans primarily occurs in SON season. lower degree of resonance between the two ocean basins then. In the period 1978–1982, there is a 3.2. Transient Rossby wave structure hiatus in oscillations followed by repeated ENSO events. The cross-modulus spectrum illustrates a Considering that the propagating signal in HCa gradual shift in period from quasi-biennial to is most strongly represented along 101S, we slice interannual (2-4 years) from the early 1960s to the oGCM-assimilated data there and composite the mid-1980s, in agreement with the Rossby successive events (12) based on their amplitude wave-modulated phase. The slight delay of the and phase, compressing/expanding different events Indian Ocean circulation is consistent with earlier to yield a hovmoller plot of consistent phase, as comparisons with Pacific ENSO (Nigam and Shen, discussed earlier. We do this for the HCa, SSTa 1993; Klein et al., 1999). and wind stress curl and plot the data in hovmoller These results are consistent with Huang and format in Fig. 6a–c. From these patterns we can Shukla’s (2002) recent study on the decadal derive a wavelength of 7000 km7500 km along modulation of the tropical Indian Ocean oscilla- 101S. Within each cycle, the Rossby waves are tion. They noted that the variability in the Indo- estimated to propagate at a variable phase speed, Pacific domain was dominated by ENSO and ranging from 0.05 to 0.10 m s1.The composite associated Walker overturning during the period Rossby wave amplifies around two centers: the of 1977–1998. On the other hand, ENSO forcing Sumatra upwelling region (1051E) and the thermo- over the Indian Ocean appears weaker during the cline ridge (701E). Propagation of the Rossby wave period 1958–1976 and more influenced by the in terms of HCa is disrupted around 901E but Indian monsoon. The shift of the convection is appears enhanced across the mid-ocean ridge from controlled by SST anomalies in the Indian and 601Eto721E. The presence of the African Pacific Oceans, so that there is a feedback among continent diminishes the signal west of 551E, them. However, it is not yet clear whether the within 1500 km of the coast. Xie et al. (2001) tropical Indian Ocean affected the ENSO by found a similarly distinct westward propagation

Fig. 6. Composite hovmoller analysis of HCa, SSTa and wind stress curl (left to right) sliced at 101S from 40–1201E. The y-axis represents time as phase lag from seasons one to eight, covering the life cycle of a Rossby wave-modulated warm event (72 years). ARTICLE IN PRESS

2132 M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 by considering the lag-correlation of Z20 varia- SSTa, and wind stress curl are not significantly bility along 101S from a point at 701E. They argue different in these two periods. However, this does that the forcing of the Rossby wave is seasonally not necessarily mean that the Rossby waves were dependent and governed by linear wave dynamics. ‘slower’ during the latter period. It is more likely Composite hovmoller plots for SSTa and wind that, because of the more persistent atmospheric stress curl are shown in Fig. 6b and c respectively. forcing, there is a stronger stationary response. The pattern of propagation for both variables is For instance, the HCa center in the Southwest not as consistent as HCa. Presumably localized Indian Ocean simply remains longer west of 701E. forcing by the surface fluxes reduces the coherency Thus 0.1 m s1 seems a reasonable estimate of the of SSTa propagation. The western center of action propagating Rossby wave. varies between warm and cool phases, being To study the impact of the seasonally phase located at 701E in the former and 551E in the locked component of the WIO Rossby wave on the latter. The Sumatra upwelling region exhibits a , we study composite plots of NCEP strong signal in cool phase (0.7 1C at phase 6), meridional upper wind and out-going long-wave but a weak signal in warm phase, as expected. radiation (OLR) selected for cases when the Such variability means that a static dipole index thermocline is deep in the west (phase 7 in Fig. cannot completely capture the basin-wide oscilla- 6a). Here we analyze November–February seaso- tion. SSTa associated with the Rossby wave in the nal differences from the long-term mean. The WIO thermocline ridge decay more slowly (3 meridional upper wind (Fig. 7a) reveals a wave phases, equivalent to 3 seasons in a biennial cycle) train pattern in conjunction with warm conditions than in the eastern center of action. Although in the WIO. The wave train extends well upstream, these changes appear small, they occur near the and is related to the ENSO-modulated atmo- convective threshold (28 1C), and significantly spheric circulation over the Atlantic (Jury et al., impact on atmospheric convection and circulation 2002). Significantly, the wavelength of this sub- as shown here. tropical atmospheric Rossby wave (77000 km), The hovmoller plot for wind stress curl exhibits matches that of the ocean Rossby wave. Some of little propagation, with strong anomalies confined the impacts of this coupled system are evident in to the region 70–951E. The curl alternates symme- Fig. 7b: a convective dipole is observed with trically between anticyclonic () and opposing values over southern Africa and the cyclonic (upwelling) events, each lasting 3 phases WIO. Hence dry conditions occur over land when (seasons). The coastal regions are dissimilar, rainfall is increased over the ocean. Similarly, Xie suggesting a semi-closed basin-scale circulation et al. (2001) show that WIO tropical system. Considering the alternation of sign in the frequency is related to the Rossby wave-induced context of SSTa, we surmise that horizontal wind dipole. Their comparison of shear drives vertical entrainment processes that occurrences for deep and shallow thermocline govern changes in SST. Since the anomalous wind conditions demonstrates a 66% change in tropical stress curl does not ‘follow’ the westward-traveling cyclone days around 151S, 601E (Mauritius). The Rossby wave, the forcing may be considered to be increase in tropical cyclone activity is consistent partially impulsive and wind-induced Ekman with warmer SST and an anomalous cyclonic transport is unlikely to drive changes in SST circulation west of the poleward atmospheric flow further west. The structure of composite winds associated with the Rossby wave. suggests a possible ‘standing’ component for Rossby wave forcing. We consider this further in the discussion section. 4. Discussion We have made separate composite life cycles for the events occurring before and after 1976, We have studied the nature of climate varia- considering the significant period shift around this bility in the Indian Ocean, using ocean model- time. We found that that the patterns of HCa, assimilated data in the period 1958–1998. A ARTICLE IN PRESS

M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 2133

30N NCAA-CIRES/Climate Diognastles Centar

20N

10N

EQ

10S

20S

30S

40S

50S 60W 50W 40W 30W 20W 10W 0 10E 20E 30E 40E 50E 60E 70E 80E 90E 250 mb V-wind Component (m/a) Composite Anomaly

EQ NCAA-CIRES/Climate Diognastine Center 5S

10S

15S

20S

25S

30S

35S

40S 10W D 10E 20E 30E 40E 50E 60E 70E 80E 90E Outgoing Longwave Radiation (W/m-2) Composite Anomaly

Fig. 7. (a) Composite 250 hPa meridional winds for the October–February period based on NCEP reanalysis data fields for 10 years when the thermocline is deep in the WIO. The historical mean is subtracted to produce anomalies. (b) Composite OLR field formulated as in (a). A pronounced dipole in atmospheric convection is found with centers of action over the WIO and southern Africa. Contour interval is 0.4 m s1 and 3 W m2, where blue/purple tones are southward winds and increased rainfall, respectively. comprehensive analysis has been performed by the mid-ocean thermocline ridge and the area of Huang and Kinter (2002), here we focus on the Sumatra upwelling. The Rossby wave shifts characteristics of the propagating ocean signal. We westward at a mean rate of 0.08–0.1 m s1, find a slow westward propagation of sub-tropical yielding cycles in the range 1.9–5.2 years. Zonal Rossby waves most clearly in sea temperatures in winds in the east Indian Ocean oscillate simulta- the upper 234 m. The southern band of HCa neously with Nino3 SST, and may be considered intensifies in the latitude band 5–151S and persists an active component of the global ENSO. up to 3 seasons. Two centers of amplification Although the transient nature of Indian Ocean occur when the trough and ridge of the 77000 km variability is highlighted in the eEOF analysis, it Rossby wave are located at 701 and 1051E, being accounts for about 15% of total variance. A larger ARTICLE IN PRESS

2134 M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 part of interannual variance may be contributed can be expected to vary. Such is the chaotic nature by ‘stationary oscillations’, but these are less of the climate system (Wang et al., 1999). One predictable and not well underpinned by theory. conclusion is that the coupled Rossby wave Indeed, the composite hovmoller results for both anticipates the phase of the dipole and provides SSTa and wind stress curl reveal a quasi-stationary considerable potential for predicting climate im- dipole that is about twice the amplitude of the pacts around the Indian Ocean. moving wave. The existence of such a dipole was Our results on the westward propagation of the suggested by Jury (1992) in the context of atmo- oceanic Rossby wave are consistent with other spheric wave-trains and further confirmed in the recent studies (e.g., HK02, Xie et al., 2002; Rao et analysis of Sajiet al. (1999) as an intermittently al., 2002). Here more emphasis is placed on the coupled oceanic feature. role played by the off-equatorial wind curl At the onset of an El Nino event, a surge of associated with coherent wind surges that enhance equatorial easterly winds creates twin anticyclonic the southern branch of the Rossby wave as gyres in the east Indian Ocean. The surge described by Masumoto and Meyers (1998). This progresses westward at 70.2 m s1 in the compo- process is not simply an air–sea interaction on the site months August to November (Matitu, 2003). equator associated with the east–west gradient. The northern gyre fades out whilst the southern The slower evolution and persistence associated one intensifies, creating the downwelling (warm- with oceanic signals from higher southern latitudes ing) effect about 151 south and west of the surging has significant implications on the climate varia- wind system. Xie et al. (2002) suggest that tions around the region. thermocline adjustments preferentially impact This off-equatorial process is consistent with the SSTa and the overlying atmosphere in the ridge most recent results of Yamagata et al. (2003), along 8–101S. In our analysis of precipitation in which demonstrated using data from observations the period 1979–1998 a high-amplitude signal is and a coupled ocean–atmosphere general circula- found in the Sumatra upwelling zone where SSTs tion model the connections between the wind stress are warmer and more variable. This region is the curl and the westward propagating oceanic dis- ‘gateway’ for interaction between the Indian and turbance. Yamagata et al. (2004) also pointed out Pacific Walker Cells during austral spring season. that the wind curl anomalies are related to the Although further studies are needed to resolve the ENSO cycle. Furthermore, they argued that a ocean–atmosphere coupling processes, we specu- major distinction between the ENSO-related di- late that changes in amplitude of the zonal see-saw pole events and those triggered by other mechan- depend on how Rossby wave trains resonating isms is in that the former are equatorially trapped within the Indian Ocean interact with atmospheric while the latter are stronger to the south of 101S. zonal injected from the western Pacific The implication seems to be that, apart from during austral spring. perturbing the over the Jury et al. (2002) have demonstrated that local Indian Ocean, ENSO affects the southeast trades heat fluxes account for about half of the variance in a broader way. This issue deserves further in SSTa in the WIO, but tend to lag by a few examination. months. Vertical motions estimated from local wind stress curl provide a significant influence at lead times of a few months, yet much of this effect Acknowledgments is attributable to incoming Rossby waves (Murtu- gudde et al., 2000; Xie et al., 2001). east–west M. Jury was supported by grants from the oscillations in the thermocline evolve slowly and NRF, WRC and DACST in South Africa. B. predictably, varying in period730% over the Huang was supported by the grants from the NSF, years 1958–1998. The equatorial wind forcing is NOAA, and NASA in the USA. The SST and resonant with the global ENSO much of the time, oceanic heat content data are derived from ocean yet uptake of the signal through the Rossby wave data assimilation carried out at COLA. Part of the ARTICLE IN PRESS

M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136 2135 material is adapted from COLA Technical Report and links with continental climate. Journal of Geophysical 94 (2001). Research C 10 (15), 1–17. Klein, S.A., Soden, B.J., Lau, N.C., 1999. Remote sea surface temperature variations during ENSO: evidence for a tropical atmospheric bridge. Journal of Climate 12, References 917–932. Latif, M., Barnett, T.P., 1995. Interactions of the tropical Behera, S.K., Yamagata, T., 2001. Subtropical SST dipole oceans. Journal of Climate 8, 952–964. events in the southern Indian Ocean. Geophysical Research Lau, N.-C., Nath, M.J., 2000. Impact of ENSO on the Letters 28, 327–330. variability of the Asian–Australian as simulated Bjerknes, J., 1969. Atmospheric from the in GCM experiments. Journal of Climate 13, 4287–4309. equatorial Pacific. Monthly Review 97, 163–172. Masumoto, Y., Meyers, G., 1998. Forced Rossby waves in the Chambers, D.P., Tapley, B.D., Stewart, R.H., 1999. Anom- southern tropical Indian Ocean. Journal of Geophysical alous warming in the Indian Ocean coincident with El Nin˜ o. Research 103, 27589–27602. Journal of Geophysical Research 104, 3035–3047. Matitu, M., 2003. The response of tropical African rainfall to Chelton, D.B., de Szoeke, R.A., Schlax, M.G., El Naggar, K., zonal circulations over adjacent oceans. M.Sc. Thesis, Siwertz, N., 1998. Geographic variability of the first University of Zululand, South Africa, 178pp. baroclinic Rossby radius of deformation. Journal of Murtugudde, R., Busalacchi, A.J., 1999. Interannual variability Physical 28, 433–460. of the dynamics and thermodynamics, and Conkright, et al., 1998. World Ocean Database 1998: doc- processes in the Indian Ocean. Journal of Climate 12, umentation and quality control. NODC Internal Report 14, 2300–2326. NOAA/NESDIS, Maryland, 43pp. Murtugudde, R., McCreary, J.P., Busalacchi, A.J., 2000. Derber, J., Rosati, A., 1989. A global oceanic data assimila- Oceanic processes associated with anomalous events in the tion system. Journal of 19, Indian Ocean with relevance to 1997–1998. Journal of 1333–1347. Geophysical Research 105, 3295–3306. Hastenrath, S., 2000. Zonal circulations over the equatorial Naeraa, M., Jury, M.R., 1998. Tropical cyclone composite Indian Ocean. Journal of Climate 21, 2746–2756. structure and impacts over eastern Madagascar during Hastenrath, S., Nicklis, A., Greischar, L., 1993. Atmospheric– January–March 1994. and Atmospheric Phy- hydrospheric mechanisms of climate anomalies in the western sics 65, 43–53. equatorial Indian Ocean. Journal of Geophysical Research Nicholls, N., 1989. Sea surface temperature and Australian 98, 219–235. winter rainfall. Journal of Climate 2, 965–973. Huang, B., Kinter III, J.L., 2000. The interannual variability in Nigam, S., Shen, H.S., 1993. Structure of oceanic and atmo- the tropical Indian Ocean. Preprints of 10th Conference on spheric low-frequency variability over the tropical Pacific Interaction of the Sea and Atmosphere, pp. 10–12. and Indian Oceans. Part I: COADS observations. Journal of Huang, B., Kinter III, J.L., 2001. The interannual variability in Climate 6, 657–676. the tropical Indian Ocean and its Relation to El Nin˜ o/ Oberhuber, J.M., 1988. An atlas based on the COADS data set: Southern Oscillation. COLA Technical Report 94, 48pp. budgets of heat and turbulent kinetic energy at Huang, B., Kinter III, J.L., 2002. Interannual variability in the the surface of the global ocean. Report 15, Max-Planck tropical Indian Ocean. Journal of Geophysical Research 107 Institut fur Meteorology, Hamburg, Germany, 199pp. (C11), 3199–3229. Philander, S.G.H., Hurlin, W.J., Seigel, A.D., 1987. A model of Huang, B., Schneider, E.K., 1995. The response of an ocean the seasonal cycle in the tropical Pacific Ocean. Journal of general circulation model to surface wind stress produced by Physical Oceanography 117, 1986–2002. an atmospheric general circulation model. Monthly Weath- Rao, S.A., Behera, S.K., Masumoto, Y., Yamagata, T., 2002. er Review 123, 3059–3085. Interannual subsurface variability in the tropical Indian Huang, B., Shukla, J., 2002. The interannual variability in the Ocean with special emphasis on the . tropical Indian Ocean and its decadal modulation. Preprints Research 49, 1549–1572. for 13th Symposium on and Climate Reason, C.J.C., Mulenga, H.M., 1999. Relationships between Variations, 225–227. South African rainfall and SST anomalies in the SW Indian Huang, B., Kinter, J.L., Schopf, P.S., 2002. Ocean data Ocean. International Journal of 19, 1651–1673. assimilation with intermittent analyses and continuous Reason, C.J., Allan, R.J., Lindesay, J.A., Ansell, T.J., 2000. model error correction. Advances in Atmospheric Sciences ENSO and climatic signals across the Indian Ocean basin in 19, 965–992. the global context: Part 1, interannual composite patterns. Jury, M.R., 1992. A climatic dipole governing the interannual International Journal of Climatology 20, 1285–1327. variability of convection over the SW Indian Ocean. Trends Reverdin, G., Cadet, D., Gutzler, D., 1986. Interannual in Geophysical Research 1, 165–172. displacements of convection and surface circulation over Jury, M.R., Enfield, D.B., Melice, J.L., 2002. Tropical the equatorial Indian Ocean. Quarterly Journal of the Royal monsoons around Africa, stability of ENSO associations Meteorological Society 112, 43–67. ARTICLE IN PRESS

2136 M.R. Jury, B. Huang / Deep-Sea Research I 51 (2004) 2123–2136

Reynolds, R.W., Smith, T.M., 1994. Improved global sea Wang, C., Weisberg, R.H., Virmani, J.I., 1999. Western Pacific surface temperature analyses using optimal interpolation. interannual variability associated with the El Nin˜ o–- Journal of Climate 7, 929–948. Southern Oscillation. Journal of Geophysical Research Saji, N.H., Goswami, B.N., Vinayachandran, P.N., Yamagata, 104, 5131–5149. T., 1999. A dipole mode in the tropical Indian Ocean. Weare, B.C., Nasstrom, J.S., 1982. Examples of extended Nature 401, 360–363. empirical orthogonal function analysis. Monthly Weather Schneider, E.K., Huang, B., Zhu, Z., DeWitt, D.J., Kinter, J.L., Review 110, 481–485. Kirtnman, B.P., Shukla, J., 1999. Ocean data assimilation, Webster, P.J., Moore, A.M., Loschnigg, J.P., Leben, R.R., initialization, and predictions of ENSO with a coupled 1999. Coupled oceanic–atmospheric dynamics in the Indian GCM. Monthly Weather Review 127, 1187–1207. Ocean during 1997–98. Nature 401, 356–360. Schott, F.A., McCreary, J.P., 2004. The monsoon circulation of White, W.B., 2000. Coupled Rossby waves in the Indian Ocean the Indian Ocean. Progress in Oceanography, in press. on interannual timescales. Journal of Physical Oceanogra- Smith, N.R., 1995. An improved system for tropical ocean phy 30, 2972–2988. subsurface temperature analyses. Journal of Atmospheric Xie, S-P., Annamalai, H., Schott, F.A., McCreary, J.P., 2002. and Oceanic Technology 12, 850–870. Structure and mechanisms of south Indian Ocean climate Ueda, H., Matsumoto, J., 2000. A possible triggering process of variability. Journal of Climate 15, 864–878. east–west asymmetric anomalies over the Indian Ocean in Yamagata, T., Behera, S.K., Guo, J.J., Masson, S., Jury, M.R., relation to 1997/98 El Nino. Journal of the Meteorological Rao, S.A., 2004. Coupled ocean–atmosphere variability in Society of Japan 78, 803–818. the tropical Indian Ocean. AGU monograph series, in press.