The Extreme Positive Indian Ocean Dipole of 2019 and Associated Indian Summer Monsoon Rainfall Response

The positive Indian Ocean Dipole (IOD) event in 2019 was among the strongest on record, while the Indian Summer monsoon (ISM) was anomalously dry in June then very wet by September. We investigated the relationships between the IOD, Pacific sea surface temperature (SST), and ISM rainfall during 2019 with an atmospheric general circulation model forced by observed SST anomalies. The results show that the extremely positive IOD was conducive to a wetter‐than‐normal ISM, especially late in the season when the IOD strengthened and was associated with anomalous low‐level divergence over the eastern equatorial Indian Ocean and convergence over India. However, a warm SST anomaly in the central equatorial Pacific contributed to low‐level divergence and decreased rainfall over India in June. These results help to better understand the influence of the tropical SST anomalies on the seasonal evolution of ISM rainfall during extreme IOD events.

Here we are particularly interested on the relationship between the IOD and the Indian summer monsoon (ISM). Summer monsoon rainfall over India represents the largest source of annual water for the country (Archer et al., 2010;Mall et al., 2006) and is important for the agrarian economy (Gadgil & Gadgil, 2006;Webster et al., 1998). Despite its annual occurrence, the Indian summer monsoon is highly variable in time and space, with the largest portion of its variability modulated by ENSO, as known since the beginning of the 19th century (Kirtman & Shukla, 2000;Sikka, 1980;Rasmusson & Carpenter, 1983;Ratna et al., 2011;Walker, 1924, as few examples). Toward the end of the 20th century a weakening of the ISM-ENSO relationship has been identified Kumar et al., 1999) with the IOD recognized as a potential trigger of ISM rainfall. Several papers reported the individual and combined influences of ENSO and IOD on ISM rainfall and found that both phenomena, individually and combined, affect ISM rainfall performance (Ashok et al., 2004;Hrudya et al., 2020;Krishnaswamy et al., 2015;Li et al., 2017;.
The active and break spells of monsoons are regulated by the boreal summer intraseasonal oscillation (BSISO), which propagates north from the equator into the Indian monsoon region and substantially affects the monsoon rainfall (Sikka & Gadgil, 1980;Sperber et al., 2000). Within the monsoon season, the mean structure of moisture convergence and meridional specific humidity distribution undergoes significant changes in contrasting IOD years, which in turn influences the meridional propagation of BSISO and hence the related precipitation anomalies over India (Ajayamohan et al., 2008;Kikuchi et al., 2012;Konda & Vissa, 2019;Singh & Dasgupta, 2017). At this timescale, the ocean-atmosphere dynamical coupling has been found to be important to the extended Indian summer monsoon break of July 2002 (e.g., Krishnan et al., 2006). Some recent studies have investigated the causes of the strong IOD event in 2019. In particular, it has been found that the occurrence of 2019 extreme positive IOD event features the strongest easterly and southerly wind anomalies on record, leading to the strongest wind speed that facilitated the latent cooling to overcome the increased radiative warming over the eastern equatorial Indian Ocean, leading to the unique thermodynamical forcing . The thermocline warming associated with anomalous ocean downwelling in the southwest tropical Indian Ocean triggered atmospheric convection to induce an easterly wind anomaly along the equator and the positive feedbacks led to an IOD event (Du et al., 2020). Also, the record-breaking interhemispheric pressure gradient over the Indo-Pacific region induced northward cross-equatorial flow over the western Maritime Continent, able to trigger strong wind-evaporation-SST and thermocline feedbacks that contributed to the strong IOD (Lu & Ren, 2020).  described how the consecutive occurrence of positive IOD in 2018 and 2019, along with the evolution of a Central Pacific El Niño, influenced Australian climate. The 2019 IOD event led to unusually warm conditions in many parts of East Asia during 2019-2020 winter (Doi et al., 2020a(Doi et al., , 2020b, though not necessarily linked with the severe drought that occurred during that fall in East China (Ma et al., 2020). In terms of predictability, such an extreme event like the 2019 IOD could be predictable a few seasons in advance (Doi et al., 2020a(Doi et al., , 2020b. In this study, we intend to investigate the dynamical aspects of the relationship between IOD and Indian summer monsoon rainfall with a specific focus on 2019. That year was peculiar in terms of the seasonal evolution of precipitation over India with dry conditions at the beginning of the monsoon season and very wet conditions toward the end (Yadav et al., 2020). In particular, we designed a set of sensitivity experiments to verify the role of anomalous SST in the Indian Ocean, i.e., the developing IOD that year, and the SST anomalies elsewhere. The work is organized as follows: Section 2 describes the data used for the analysis as well as the model and experiments performed. Section 3 is dedicated to the observed characteristics of IOD and ISM during 2019 with specific attention to the evolution within the summer season. Section 4 shows the results from the sensitivity experiments performed, including a discussion of the main results obtained. Finally, Section 5 summarizes the main finding and provides future perspectives from this analysis.
For rainfall we used the Global Precipitation Climatology Project (GPCP) data (Adler et al., 2003) available at 2.5° resolution. We also used the Homogeneous Indian Monthly Rainfall Data Sets (Kothawale & Rajeevan, 2017) from https://tropmet.res.in/static_pages.php?page_id=53. Other atmospheric variables and the global SST field are taken from National Center for Environmental Prediction-Department of Energy (NCEP-DOE) Reanalysis 2 (Kanamitsu et al., 2002) available at 2.5° resolution. All anomalies are calculated with respect to the 1981-2010 climatology.

The IGCM4 Model and Sensitivity Experiments
The Intermediate General Circulation Model version 4 (IGCM4; Joshi et al., 2015) is a global spectral primitive equation atmospheric model with a spectral truncation at T42 (corresponding to 128 × 64 grid points in the horizontal) and 20 layers in the vertical, with the top at 50 hPa. This configuration, i.e., T42L20, is the standard for studies of the troposphere and climate (Joshi et al., 2015). IGCM4 has been extensively used in climate research, process modeling and atmospheric dynamics (O'Callaghan et al., 2014;Ratna et al., 2020;van der Wiel et al., 2016). The IGCM4 gives a good representation of the mean climate state (Joshi et al., 2015), in particular the simulated climatology and annual cycle over Asia is in reasonable agreement with the reanalysis for temperature and precipitation (Ratna et al., 2020). The physical parameterization schemes used here are the same as in Joshi et al. (2015) and Ratna et al. (2020).
The set of experiments performed with the IGCM4 consist of a control simulation (CTRL) with prescribed SST obtained from a climatology  of the skin temperature in the NCEP-DOE Reanalysis 2 (Kanamitsu et al., 2002) and two sensitivity experiments where the 2019 SST anomaly is added to the CTRL climatology globally (IODglob) and only over the Indian Ocean (IODreg). All other boundaries conditions are the same as in CTRL. The surface albedo has been adjusted to indicate the presence or absence of sea ice according to whether the new surface temperature was below freezing. We used the greenhouse gas concentration in the model which is close to the 1995 value, the midpoint of the 1981-2010 climatology. For each simulation, the model is integrated for 55 years and the mean of the last 50 years is analyzed, excluding the first five years as model spin up. These simulations are long enough to allow a clear separation of the response to the SST anomalies from the internally generated variability, especially for "noisy" variables such as precipitation.

2019 Indian Ocean Dipole and Indian Summer Monsoon
The Indian Ocean Dipole (IOD) was unusually strong in 2019 (Figure 1a). The positive IOD event was the strongest of the last two decades, and possibly the strongest of the last 38 years. The September-November 2019 DMI was four standard deviations above the 1981-2010 climatology in the ERSST data. This exceeded the previous strong event of 1997 in the ERSST and NOAA-OI-SST datasets, while 1997 remained the strongest in HadISST ( Figure S1). The 2019 positive IOD phase arose from both negative SST anomalies over the eastern equatorial Indian Ocean (EEIO) and warm SST anomalies over the western equatorial Indian Ocean (WEIO) from June to October (Figures 1c-1h). However, the evolution of the event was strongly determined by the EEIO, which largely cooled from climatological conditions in May to almost 1 K cooler than normal by October. On the other hand, the WEIO stayed more constant (i.e., less than 1 K warmer than normal) throughout the period (Figure 1b).
The total seasonal (June-September) rainfall over India was 110% with respect to its long period average, with the June rainfall quite low (67%) while the September one quite excessive (152%) (Yadav et al., 2020).   The annual evolution of the IOD index is compared with ENSO associated indices for the year 2019 (Figure 1b). The IOD is strong compared to the rest of indices during 2019 so it is interesting to consider the role of IOD on the seasonal evolution of ISM rainfall. The IOD index intensified from July and reached its peak during October-November (Figure 1b), due to the strengthening of the SST anomaly in the EEIO, as noted above. Nino3.4 SST indicates that ENSO condition was slightly positive in June, before decreasing in strength to reach zero anomaly in September. El Niño-Modoki index, which is indicator of a central Pacific SST anomaly, remained slightly above normal throughout the year (Figures 1c-1h).
We have compared ( Figure S2) the seasonal evolution of the IOD, Pacific indices and ISM rainfall (Table S1) with three other strong IOD events (1994,1997,2006) to consider if they support our finding that a strengthening positive IOD may be associated with a wetter ISM when not overwhelmed by ENSO influences. In 1994, a positive IOD strengthened further from June to August. Although the central Pacific was warmer than normal, El Niño state were not reached, perhaps allowing the IOD to dominate and contribute to above-average ISM rainfall in most months and in the seasonal total (Behera et al., 1999;Ashok et al., 2004;. By contrast, 1997 was dominated by a very strong El Niño, though the expected RATNA ET AL. 10.1029/2020GL091497 5 of 11 Figure 2. (a, d, and g) Observed GPCP rainfall anomaly (mm/day, shaded) and NCEP2 850 hPa wind anomaly (m/s, vectors) for June-September mean, June and September, respectively. (b, e, and h) and (c, f, and l) are the same as (a, d, and g) but for IODglob and IODreg experiments, respectively. Shaded precipitation anomalies are significant at 90% level using a Student's t-test. GPCP, Global Precipitation Climatology Project.

(a) (b) (c) (f) (e) (d) (g) (h) (i)
ENSO-induced anomalous subsidence may have been neutralized/reduced by anomalous IOD-induced convergence over the Bay of Bengal (Ashok et al., 2001) and contributed to a near-normal ISM season. During 2006, the onset of positive IOD was late compared to the other years considered, perhaps contributing to above-normal rainfall in the final months of the ISM (the Modoki index was close to normal and Nino3.4 only warmed to an El Niño state later in the year). Overall, out of these 4 years, the two with the strongest positive IOD and relatively weak Nino3.4 anomalies (1994 and 2019) had excess ISM rainfall (+15% and +16% with respect to the 1981-2010 climatology, Table S1). There was a smaller increase in ISM rainfall in 2006 (+9%) when the IOD event developed later, while 1997 had a strong El Niño and a normal ISM season (+2%).

Mechanisms Contributing to the Anomalous 2019 Indian Summer Monsoon Rainfall
To understand the contribution that SST forcing may make to the 2019 rainfall variability over the Indian landmass, we compared the model simulated anomaly (IODglob and IODreg as explained in Section 2) with the observed anomaly. Following the design of the experiments, the comparison is focused in the identification of the rainfall pattern anomalies in the different cases. Of course, we do not expect perfect agreement, even were the model perfect, because of internal atmospheric variability unrelated to the 2019 SST anomalies. Nevertheless, both sensitivity experiments reproduce a dipole precipitation anomaly over the south equatorial Indian Ocean (dry in the east, wet in the west; Figures 2a-2c) during the whole monsoon season (June-September) that closely resembles the observed pattern. Observed Jun-Sep precipitation is above average over the Indian land mass and over the Bay of Bengal, and both experiments simulate a qualitatively similar pattern. Instead, the intensity of the anomaly is larger when the model is forced with only Indian Ocean SST anomalies (IODreg; Figure 2c) compared to the global SST (IODglob) anomaly (Figure 2b). This indicates the importance of the 2019 Indian Ocean SST anomaly in contributing to wet conditions over India, though it is modulated by SST anomalies elsewhere.
The comparison of the sensitivity experiments also illuminates on the possible mechanisms behind the two contrasting months of the season (i.e., dry June and wet September). In June, the model response to Indian Ocean SST forcing produces a stronger south-westerly monsoon flow and wet anomalies over western India (IODreg; Figure 2f), whereas including SST anomalies from other ocean basins (IODglob; Figure 2e) suppresses the wet anomaly and brings the simulated response closer to the observations (with the exception of the western Indian Ocean). The negative rainfall anomaly over EEIO is also stronger in IODglob compared to IODreg and more similar to the observations. On the other hand, both IODglob and IODreg experiments have a wet anomaly over India in September, as is also seen in the observations (though the observed anomaly is stronger and more extensive). These results indicate that the 2019 Indian Ocean SST anomalies suppress rainfall in the EEIO and favor a wetter than normal Indian monsoon, but that in June the latter is more than offset by a response to the SST anomaly outside the Indian Ocean, resulting in the dry anomaly, as it is observed.
Considering the whole 2019 season, stronger low-level southerly wind anomalies dominated over the Bay of Bengal due to low-level divergence over EEIO associated with the very positive IOD (Figures 2a-2c). The low-level winds are similar to Behera and Ratnam (2018) where they show low-level westerlies and southerlies toward India originated from the EEIO but they do not show any significant cross-equatorial flow in their positive IOD events composite. Over the Arabian Sea, the IODreg simulation has stronger south-westerly anomaly compared to IODglob and hence simulates excess rainfall (Figures 2a-2c). In June, the dry anomaly observed over India is related to low-level anomalous anticyclonic circulation over central-east India and adjacent Bay of Bengal and to anomalous easterlies prevailing in the peninsular India (Figure 2d). Both circulation features reduced the monsoon flow toward India and hence contributed to the negative rainfall anomaly over India. IODglob realistically simulated both these anomalous circulation features (Figure 2e), whereas IODreg did not and it shows strong south-westerly flow reaching the Indian landmass (Figure 2f). In September 2019, observations show that there was a strong anomalous south-westerly flow toward Indian landmass and associated cyclonic circulation over central west India, contributing to the excess rainfall (Figure 2g). Both sensitivity experiments (Figures 2h and 2i) simulated anomalously strong south-westerly flow and anomalous cyclonic circulation over India, though they are not as strong as observed.
Consistent with precipitation and low-level wind patterns, there is convergence in the upper troposphere over the Maritime Continent and EEIO in September when the IOD is at its peak (Figure 3b), but such convergence does not appear in June (Figure 3a) when the IOD is developing and there are still warm SST anomalies over the equatorial Pacific (Figure 1). In the IODglob experiment (Figure 3c) we see that the model responds strongly to these equatorial Pacific SST anomalies in June, causing strong upper-level divergence over east equatorial Pacific and convergence over the Maritime Continent. The opposite circulation is seen at lower levels (see Figure S2 for the 850 hPa velocity potential and divergent winds) which causes low-level divergence extending from the Maritime Continent to the Bay of Bengal and Indian landmass, contributing to negative rainfall anomaly in June. In IODreg, where the model is forced with the 2019 SST anomaly only over the Indian Ocean, the model responds with upper-level (lower level) divergence (convergence) over the Indian Ocean and over India ( experiments simulate upper-level divergence over EEIO region in September, although in IODglob it is stronger than in IODreg, and this explains the link between the Indian Ocean SST anomaly and the circulation and rainfall anomalies.

Conclusions
One of the strongest positive IOD events in the historical period occurred in 2019. The evolution of the 2019 IOD was characterized by a cold anomaly over the EEIO which started strengthening from June and reached its peak in October, remaining strong until November. In the same year, the Indian summer monsoon season experienced peculiar behavior with weak rainfall during June (despite the IOD index being already in its positive phase). Then the monsoon gained its strength from July, ending with an anomalous wet September and contributing to above-normal seasonal rainfall.
With a suite of atmospheric GCM experiments we have been able to evidence the role of the IOD and of the SST anomalies elsewhere in the seasonal evolution of rainfall and circulation anomalies during the 2019 summer monsoon. The anomalous SST gradient between the west and east equatorial Indian Ocean drives a dipole in equatorial precipitation anomalies and anomalous low-level circulation that would, in isolation, lead to a wetter than normal Indian summer monsoon across the monsoon season including June and September. However, when forcing the IGCM4 model with the global pattern of SST anomalies observed in 2019, the response changes, particularly in June. Although not considered to be an El Niño, the first half of 2019 did exhibit anomalously warm conditions in the central Pacific (visible in the Nino3.4 index) that dissipated by September. The model responds to this equatorial Pacific warmth with upper-level divergence over the equatorial Pacific and convergence over the Maritime Continent. This causes low-level divergence extending from the Maritime Continent to the Bay of Bengal and the Indian landmass, contributing to a negative rainfall anomaly there in June. By September, this response to remote forcing from the Pacific weakens (likely linked in part to the weakening of the Nino3.4 SST anomaly there), leaving the response to the Indian Ocean SST anomalies (linked to the very strong IOD) to dominate. This response arises from strong IOD-related low-level divergence over EEIO and convergence over the Indian landmass, contributing to excessive rainfall.
The similarity between the model simulations and observed/reanalysis data provides evidence that these mechanisms occurred in the real world in 2019, i.e., that there was a contrasting contribution from the Pacific and Indian Ocean SST anomalies to ISM rainfall. The tropical Pacific SST contributed to a drying tendency over India while the IOD contributed to anomalous wet conditions over India. The Pacific effect dominated in June, contributing to the dry anomalies observed, but the weakening Pacific SST anomalies and especially the dramatic strengthening of the IOD led to the latter dominating by September and having a significant contribution to the very wet September observed.
The observed June and September rainfall anomalies were more extreme than those simulated in these SSTforced experiments, reinforcing the role that internal atmospheric variability plays in any particular month or season. Nevertheless, the results from this study help to understand the role of SST anomalies within and outside the Indian Ocean in affecting ISM rainfall intensity and seasonal evolution during extreme IOD events. This is important for improving seasonal predictions of Indian summer monsoon, and our results also highlight that, to predict the seasonal evolution of ISM rainfall, Pacific SST anomalies must be considered even when there is an extremely strong IOD. For example, Li et al. (2017) show that the majority of CMIP5 models simulate an unrealistic present-day IOD-ISMR correlation due to an overly strong control by ENSO and hence a positive IOD is associated with a reduction of ISM rainfall in the simulated present-day climate. Hence, coupled climate models need to improve their simulation of these type of linkages.

Acknowledgments
This study is supported by the Belmont Forum and JPI-Climate project INTE-GRATE (An Integrated data-model study of interactions between tropical monsoons and extratropical climate variability and extremes) with funding by UK NERC Grant NE/P006809/1. The model experiments were conducted on the High-Performance Compute Cluster at the University of East Anglia. The authors thank the two anonymous reviewers for their constructive comments.