Monsoon droughts over the Indian subcontinent emanate from failures in the seasonal (June–September) monsoon rains. While prolonged dry-spells (“monsoon-breaks”) pervade on sub-seasonal/intra-seasonal time-scales, the underlying causes for these long-lasting anomalies remain elusive. Based on analyses of a suite of observed data sets, we report an ocean-atmosphere dynamical coupling on intra-seasonal time-scales, in the tropical Indian Ocean, which is pivotal in forcing extended monsoon-breaks and causing droughts over the subcontinent. This coupling involves a feedback between the monsoonal flow and thermocline depth in the Equatorial Eastern Indian Ocean (EEIO), in which an anomaly of the summer monsoon circulation induces downwelling and maintains a higher-than-normal heat-content. The near-equatorial anomalies induce strong and sustained suppression of monsoon rainfall over the subcontinent. It is concluded that the intra-seasonal evolution of the ocean-monsoon coupled system is a vital key to unlocking the dynamics of monsoon droughts.
 The tropical Indian Ocean is characterized by a distinct seasonal cycle of monsoon wind reversals, accompanied by coherent changes in the upper ocean circulation [Schott and McCreary, 2001]. Despite the remarkable regularity, the variations of the monsoon circulation from year-to-year are large enough to exert profound impact on the rainfall distribution over the subcontinent thereby affecting the agricultural economy and human lives in one of the most populous regions of the world. The problem of long-range prediction of monsoon rainfall and the occurrence of droughts over the subcontinent remains one of the great challenges in climate research [Sikka, 1999; Gadgil et al., 2004]. The more recent monsoon drought of 2002, that resulted in economic losses of billions of dollars, is a striking example of a large-scale catastrophic event which eluded long-range forecasts models world over [Gadgil et al., 2004]. Much of the seasonal monsoon rainfall shortage over India occurred due to a prolonged dry-spell commonly referred to as a (‘monsoon-break’) during July 2002 (Figures 1a–1b) when the rainfall deficit was about 49% of the long-term normal. While, such intra-seasonal anomalies associated with variations in the large-scale monsoon circulation, can crucially determine the seasonal monsoon precipitation, state-of-the-art general circulation models have limitations in predicting the evolution of break/active phases of the monsoon intra-seasonal variability [Ferranti et al., 1997; Sperber et al., 2000].
 The monsoon drought witnessed in 2002 also was related to near-equatorial precipitation anomalies; however, for the first time, advances in the ocean observing system allow characterization of subsurface properties. The striking contrast in the rainfall distribution between the EEIO and the Indian landmass is evident in Figure 1b. The rainfall over the equatorial Indian Ocean (75°E–100°E; 10°S-equator) was about 11.7 mm day−1; while that over the Indian landmass (70°E–90°E; 10°N–30°N) was about 4.8 mm day−1. The suppressed precipitation over the subcontinent during July 2002 resulted from anomalous subsidence induced by strong upward motions over the equatorial Indian Ocean as evidenced from the vertical motion (Figure S1), indicating a major weakening of the so-called monsoon Hadley circulation - a thermally-driven meridional circulation during the northern summer [Yasunari, 1979; Krishnamurti and Subrahmanyam, 1982]. These observations motivated us to understand the nature of coupled air-sea interactions during extended monsoon breaks. Here, we examine observed data from panoply of sources that include a network of high-quality subsurface temperature observations in the Indian Ocean, high-resolution satellite measurements of sea surface height anomalies (SSHA) and surface winds; and atmospheric circulation from reanalysis data set. Observed temperature profiles from a network of autonomous Argo floats [Roemmich and Owens, 2000], eXpendable Bathy Thermograph (XBT) lines and Conductivity-Temperature-Depth (CTD) measurements in the Indian Ocean (45°E–120°E; 45°S–25°N) during (January 2002–December 2003) have been gridded to produce a 1° × 1° temperature data set for the upper 500 m (Figure S2). The data set construction [Ramesh and Krishnan, 2005] involves blending the in-situ observations during 2002–2003 with the climatological first guess. The Argo float data was obtained from the Coriolis Center, France (http://www.ifremer.fr/coriolis); XBT observations from (http://www.aoml.noaa.gov/phod/trinanes) and climatological first guess from the World Ocean Atlas (WOA2001) (http://www.nodc.noaa.gov/OC5/WOD01).
2. Monsoon-Ocean Coupled Response
 The tropical oceans behave in many ways as a two-layer fluid, with thermocline variations reflected in sea-level heights [McPhaden, 2004]. Sea-level thus provides a convenient proxy of upper-ocean heat-content and a measure of the vertically integrated oceanic response to atmospheric forcing. Here, satellite altimeter measurements of SSHA from TOPEX/POSEIDON (T/P) and JASON; along with QuikSCAT scatterometer winds (http://podaac.jpl.nasa.gov) are used to interpret the ocean-atmosphere coupling during the monsoon drought of 2002. Monthly surface winds during May–August 2002 (Figures 2a–2d) show prevalence of westerlies over Southern India and Sri Lanka extending equatorward south of the Bay of Bengal, associated with a prominent southward curvature and an intensified near-equatorial trough. During weak monsoon phases, the cross-equatorial winds are known to acquire a southward curvature and avoid blowing into the subcontinent [Rodwell, 1997]. The EEIO was particularly unusual during 2002 in that the overlying westerly anomalies were co-located with increased SSHA (Figures 2e–2h) and increased oceanic heat-content around (85°E–105°E; 5°S–5°N). Basically, the anomalous sea-level rise is consistent with an anomalous depression of the thermocline and a deep oceanic mixed layer. As will be seen later, the SSHA evolution during 2002 was characterized by downwelling Kelvin waves, forced by westerly winds over the equator, which propagated to the eastern boundary (Sumatra) and were reflected as Rossby waves that increased the heat-content in the Bay of Bengal, as well as the eastern Indian Ocean as far south as Java.
 Subsurface temperature anomalies in the near-equatorial Indian Ocean provide corroborative support for anomalous heat-content changes during 2002 (Figures 3a–3d). The anomalous warming in the EEIO by as much as 0.8°C during July–August 2002 and cold anomalies of about 0.4°C in the western Indian Ocean correspond to an intensification of the east-west temperature gradient along the equator (Figures 3b–3d). The intensified zonal temperature gradient is also seen in observed SST anomalies during 2002 (Figure S3). We have additionally confirmed the occurrence of warm subsurface anomalies in the EEIO during various other past monsoon droughts from model reanalysis products (Figure S4).
 One of the striking features during 2002 was the steady eastward propagation of positive SSHA in the equatorial Indian Ocean from the spring months onward, leading to accumulation of warm waters in the EEIO (Figure 4a). The daily anomalies of zonal winds (Figure 4b) reveal prominent episodes of westerly anomalies during April–May and June–July 2002 respectively. These enhanced westerly wind events, over the Indian Ocean and west Pacific warm pool, have been linked to the strong activity of the Madden Julian Oscillation (MJO) during 2002 [Saith and Slingo, 2006]. This point will be taken up for discussion again. It can be noticed that by May 2002, the rise of SSHA in the EEIO was nearly 20 cm higher than normal (Figure 4a). This anomalous sea-level rise was followed by westward propagation of positive SSHA toward the central Indian Ocean. Off-equatorial sections of SSHA (Figure S5) reveal steady westward propagating positive anomalies from May 2002 onward, which correspond to off-equatorial Rossby waves reflected from the eastern boundary [Schott and McCreary, 2001] and the SSHA gradually rose in the central Indian Ocean during September to December 2002. However, the western Indian Ocean was mostly dominated by decreased SSHA during the second half of 2002 (Figure S5).
3. Discussions and Conclusion
 The above results, when juxtaposed together, lend credence to the occurrence of a coupled feedback between the monsoon circulation and the tropical Indian Ocean dynamics. It is important to recognize that the antecedents of the anomalous warming in the EEIO appeared in the SSHA evolution as early as the spring preceding the summer monsoon drought of 2002. It is known from ship drift climatology that eastward surface currents referred to as “Wyrtki Jets” prevail during the spring and fall seasons in the equatorial Indian Ocean [Wyrtki, 1973; Schott and McCreary, 2001]. The eastward propagating positive SSHA in the equatorial region (Figure 4a) indicate enhancement of the spring Wyrtki Jet during 2002. This is well-supported by current measurements in the EEIO (90°E), available for the period (2001–2003), that provide evidence for intensified eastward flow of nearly 150–160 cm s−1 at a depth of about 40 m during May–June 2002; with high-resolution ocean model simulations demonstrating the influence of wind forcing on the upper-ocean currents (http://iprc.soest.hawaii.edu/meetings/workshops/IOM2004/Day2/Masumoto.pdf). Further, Saith and Slingo  have pointed out the occurrence of strong westerly wind bursts along the equator and enhancement of the eastward component of the MJO during 2002; although the northward propagation of convective anomalies over the Indian longitudes was rather weak and contributed substantially to the intensity and duration of the prolonged monsoon-break. From the results described above, it is hypothesized that the warm water accumulation in the EEIO and the development of zonal temperature gradient (warmer in the east compared to the west) was a key element that contributed to the strengthening of westerly anomalies over the EEIO; in turn the westerly anomalies reinforced the eastward currents and accumulation of warm waters in the east.
 The present findings indicate that continued accumulation of warm water in the EEIO can lead to a positive feedback between the Southwest monsoon circulation and the Indian Ocean dynamics on intra-seasonal time-scales. In this feedback process, the warm anomalies in the EEIO along with the intensified zonal gradient of SST promote sustained low-level moisture convergence and destabilization of the overlying atmosphere [Neelin and Held, 1987; Graham and Barnett, 1987; Lindzen and Nigam, 1987; Emanuel, 1987] required to maintain the convective activity locally over the EEIO. The enhanced convective activity over the EEIO weakens the monsoon circulation by inducing subsidence over the subcontinent. Conversely, the restricted northward penetration and southward curvature of the weak monsoon flow gives rise to an intensified equatorial trough and favors downwelling in the EEIO thereby maintaining the depressed thermocline and warm subsurface anomalies in the region. The rainfall, wind and subsurface temperature anomalies during several instances of past monsoon droughts are found to be rather similar to those of 2002, indicating the robustness of the coupled system response in the near-equatorial Indian Ocean (Figure S4).
 A question arising at this stage is what determines the termination of the positive feedback. Our understanding indicates that the termination can be influenced by processes that counter the warming tendency of the oceanic anomalies such as cloud-radiative effects associated with the enhanced convection over the EEIO; heat loss from the ocean due to increase of latent and sensible heat fluxes in the near-equatorial ocean. For a typical increase of the net cooling flux, a simple calculation shows that a SST change of 1°C occurs at a slower rate (time-scale ∼1.5 months) for a deep thermocline with mixed layer depth of about 55 m; as opposed to a faster SST cooling rate (time-scale ∼2 weeks) for a shallow mixed layer of about 20 m (see auxiliary material). Therefore, the deep temperature anomalies in the EEIO are important in that they provide persistence to the SST anomalies and allow sustained atmospheric convection to occur over the region.
 The findings take on added significance in view of the antecedents that are initiated during the spring months preceding the monsoon drought, when enhanced eastward currents along the equator cause accumulation of warm waters in the EEIO. If the coupled interactions among the Indian Ocean dynamics, monsoon circulation and the MJO activity hold the key to the evolution of the tropical intra-seasonal variability, the implications of these coupled interactions on the regional and global climate warrant careful consideration. In recent years, there has been renewed interest to understand the role of the Indian Ocean in the regional climate variability [Annamalai and Murtugudde, 2005], generated by anomalous events in the region during 1994 and 1997 and the identification of the Indian Ocean Dipole/Zonal Mode [Saji et al., 1999; Webster et al., 1999]. Essentially, a major challenge at the present time is to understand and predict the oceanic structure and its variability on intra-seasonal to seasonal time-scales, which requires continuous and sustained surveillance of the state of the ocean and the atmosphere. Recognizing the fact that there is an overall lack of observations in the Indian Ocean region, efforts are underway for the design and implementation of a basin scale mooring array and an Integrated In-situ Observing System in the region (http://www.clivar.org/organization/indian/index.htm) which is required to further quantify the details of monsoon-ocean coupling on intra-seasonal time-scales and track the evolution of coupled variabilities, leading to major advances in our ability to predict the monsoonal rains on time-scales of days-to-weeks.
 This research was funded by the DOD/INDOMOD-SATCORE project ISP 1.5, Govt. of India.