The South Asian Summer Monsoon (SASM) significantly impacts over a billion people and influences global atmospheric circulation. Despite its importance, accurate seasonal prediction of SASM interannual variability remains a challenge. Existing indices, like the All India Rainfall Index (AIRI), primarily focus on rainfall over the Indian subcontinent, neglecting the substantial rainfall over the Bay of Bengal (BOB), a crucial global atmospheric heat source. Other indices consider rainfall and wind shear but often overlook the crucial role of the Indian Monsoon Trough (IMT) over the BOB. The IMT, extending from the northern BOB to western India, plays a critical role in rainfall, particularly extreme rainfall events originating over the warm northern BOB waters and moving along the IMT. The interannual variability of the SASM is influenced by sea surface temperature (SST) anomalies associated with El Niño-Southern Oscillation (ENSO), with their relationship exhibiting interdecadal changes. The AIRI largely represents the first leading mode of Indian summer monsoon rainfall, linked to equatorial Pacific SST anomalies. However, the second mode, associated with the IMT over northern India and exhibiting distinct rainfall anomalies, is a lagged response to ENSO, hindering accurate prediction using AIRI due to the spring predictability barrier related to ENSO. This study aims to investigate SASM interannual variability by analyzing the coupled variability of rainfall and atmospheric circulation across South Asia, focusing on convection and the IMT's role over the northern BOB. A new index is proposed to represent the first mode, capturing the coupled features of SASM rainfall and low-level wind, differing from existing indices. This index is linked to tropical SST anomalies, enabling one-season-ahead prediction and showcasing a predictable SASM aspect.

Extensive research has been conducted on the South Asian Summer Monsoon (SASM), exploring its mechanisms and predictability. Studies have highlighted the significant variations in rainfall and atmospheric circulation across various timescales. The interannual variability of SASM rainfall profoundly affects the lives and livelihoods of more than a billion people and regulates both local and remote atmospheric circulations. The All India Rainfall Index (AIRI) and other indices have been proposed to quantify interannual variability, but their predictive skill remains low. Dynamic indices utilize convection-circulation relationships and key SASM features like jet streams and Hadley circulation, but often lack consideration of the crucial IMT variability over the BOB. The IMT's mesoscale convective systems contribute significantly to South Asian rainfall, and the dynamic and thermodynamic conditions differ across its eastern and western flanks. Previous research has established the impact of ENSO on SASM, but the relationship has shown interdecadal changes. The predictability of the SASM has been a persistent challenge, partly due to the spring predictability barrier linked to ENSO. Therefore, this study aims to improve the prediction by focusing on a more predictable mode of SASM variability linked to the IMT and its interaction with SST anomalies.

This study employed singular value decomposition (SVD) analysis on normalized and detrended 850-hPa wind and rainfall data (June-August, 1979-2020) over South Asia to identify dominant modes of coupled rainfall and low-level circulation. The SVD analysis was performed on specified domains for wind and rainfall. The two leading modes, accounting for 75.89% of the total squared covariance, were analyzed. The first mode (SVD1), explaining 51.91%, showed a low-level cyclonic circulation over the northern BOB and south of the Tibetan Plateau, coupled with a rainfall pattern of positive anomalies along central India-northeastern BOB and negative anomalies on its sides. This mode indicated a strengthened and eastward-extended IMT. The second mode (SVD2), accounting for 23.98%, displayed a cyclonic circulation over the Indian subcontinent with rainfall anomalies primarily over the Indian subcontinent. The study examined the relationship between the SVD modes and SSTs in the preceding winter and spring seasons. It was found that SVD1 correlated significantly with SSTs in the central-eastern tropical Pacific and Indian Ocean in the preceding winter and spring, indicating a lagged response to ENSO. SVD2 showed a weaker, simultaneous connection with SSTs in the central-eastern tropical Pacific. To improve real-time monitoring and prediction, a new index, the IMT index (IMTI), was constructed using area-averaged 850-hPa zonal wind over two domains reflecting the southern and northern flanks of the low-level cyclonic anomalies in SVD1. The IMTI was compared with existing indices (WYI, AIRI, EIMRI, MHI, CI, IMI, SASMI, SASMII, SASM11) for their correlation with SVD modes and their relationships with rainfall and circulation anomalies. Finally, a statistical prediction model for IMTI was developed using SST data (eastern tropical Indian Ocean and tropical Atlantic) from the preceding spring and the preceding winter Niño-3 index. The least absolute shrinkage and selection operator (LASSO) was employed to construct the regression function.

The SVD analysis revealed two dominant modes of SASM variability. The first mode (SVD1), accounting for 51.91% of the total squared covariance, is characterized by a low-level cyclonic circulation over the northern BOB and south of the Tibetan Plateau, strongly coupled with rainfall anomalies. Positive rainfall anomalies were observed over central India and the northeastern BOB, while negative anomalies appeared on the western coast of India and the southeastern edge of the Tibetan Plateau. This mode reflects a strengthened and eastward-extended IMT. The second mode (SVD2), explaining 23.98%, showed a cyclonic circulation over the Indian subcontinent with rainfall anomalies primarily concentrated over the Indian subcontinent. SVD1 exhibited a lagged response to ENSO, while SVD2 showed a weaker simultaneous relationship. The new IMT index (IMTI), a physical proxy for SVD1, demonstrated high correlation with the PCs of 850-hPa wind and rainfall (0.896 and 0.938 respectively). Most existing indices showed stronger correlation with SVD2 than SVD1, except for SASMI and SASMII which exhibited reverse relationships. The IMTI showed significant correlation with preceding winter Niño-3 index (-0.472) and moderate correlation with simultaneous summer Niño-3 index. Statistical prediction of IMTI using preceding spring SSTs (eastern tropical Indian Ocean and tropical Atlantic) and preceding winter Niño-3 index achieved a skill of R = 0.698 (1979-2020), suggesting a high predictability of the first mode. The LASSO model, trained on 1979-2010 data, exhibited high predictive skill (R=0.719) and only two false hits (2013,2016) during the validation period (2011-2020). Even using only the preceding winter Niño-3 index resulted in a significant correlation (R=0.472).

This study's findings challenge the conventional view of SASM unpredictability by highlighting the highly predictable first mode linked to the IMT. The superior predictability of this mode stems from its strong connection to persistent SST anomalies during the ENSO decaying phase, unlike the second mode which lacks significant previous-season oceanic signals. The study's new IMTI effectively captures this predictable mode, exhibiting superior representativeness and objectivity compared to existing indices. The high predictability of IMTI (R=0.698) suggests that a substantial portion (48.7%) of SASM variance is predictable. The statistical prediction model, based on the physical link between IMTI and preceding SST anomalies, demonstrates the potential for improved SASM forecasting. Further research could explore the commonalities between IMTI and SASMI2, which also exhibits high correlation with the first mode and preceding ENSO. The study also suggests exploring multi-scale monitoring and prediction using IMTI.

This research reveals a significant predictable aspect of the SASM by focusing on its first dominant mode, strongly linked to the IMT. The new IMTI effectively captures this mode, showing high predictability (R=0.698) through a statistically derived model using preceding SST anomalies. This contrasts with existing indices, which often correlate more with a less predictable second mode. The results advocate for utilizing the IMTI for real-time monitoring and prediction and for further investigation into multi-scale analysis using IMTI. Future work should focus on refining the prediction model to enhance accuracy and expand the model's applicability.

While the study demonstrates the high predictability of the first mode of SASM variability using the IMTI, several limitations exist. The statistical model's performance might be affected by the limited sample size (42 years). The study primarily focuses on interannual variability and may not capture the full spectrum of SASM variability. The chosen predictors (SSTs in the eastern tropical Indian Ocean and tropical Atlantic) might not be exhaustive, and other predictors could improve predictive skill. Future research should explore the inclusion of other variables in the model and investigate the extent to which this predictability extends to different regions within the SASM domain.