The Asian summer monsoon's interannual variability profoundly affects the lives and economies of the densely populated Asian region. Variations in seasonal precipitation, surface temperatures, and the frequency of floods, droughts, and tropical cyclones (TCs) are all significantly influenced by this variability. Summer rainfall and river discharges are critical for water security and food production, making accurate and long-lead-time predictions of the Asian summer monsoon extremely valuable for various sectors. However, achieving long-range prediction accuracy for the Asian summer monsoon has been challenging due to complex atmosphere-land-ocean interactions, interbasin interactions, and unpredictable atmospheric internal variability. While improvements in climate models and initialization techniques have extended seasonal predictions, long-range prediction beyond six months has remained elusive. Previous research has hinted at the potential for longer-lead predictions, suggesting that the Indo-western Pacific Ocean (IWPO) summer monsoon may be predictable longer than previously thought. This study aims to demonstrate the feasibility of predicting the Indo-WNP summer monsoon significantly longer than current seasonal predictions, investigating the underlying mechanisms that drive this extended predictability.

Existing literature highlights the significant challenges in long-range prediction of the Asian summer monsoon. Complex interactions between the atmosphere, land, and ocean, as well as global interbasin interactions, limit the ability of models to accurately represent and predict monsoon behavior. Studies have shown that anomalous precipitation over the western North Pacific (WNP) and South China Sea, crucial components of Indo-WNP monsoon variability, are negatively correlated with local sea surface temperatures (SSTs). This suggests a dominance of atmospheric forcing on the ocean and limits predictability from local SSTs. Despite these challenges, improvements in climate models and prediction techniques have led to some progress in seasonal predictions, reaching lead times of approximately six months. However, the full potential for longer lead times remains largely unexplored. Recent advances in understanding the mechanisms driving Indo-WNP monsoon variability suggest the possibility of extending these lead times significantly.

A 16-month ensemble prediction experiment with 52 members was conducted using the Japan Meteorological Agency/Meteorological Research Institute-Climate Prediction System version 2 (JMA/MRI-CPS2). The experiment spanned 37 years (1980-2016), with hindcasts initiated every April. The prediction skill for the second-year boreal summer (June-August, with a 13-month lead) was evaluated against historical observations and reanalysis data. The JMA/MRI-CPS2 model, an atmosphere-ocean-sea ice-land coupled model with a high spatial resolution, was employed. The ensemble was created by slightly perturbing initial conditions and using a stochastic physics scheme. The initial conditions for the atmosphere and ocean were derived from an ocean analysis system (MOVE/MRI.COM-G2) and an atmospheric analysis system (JRA-55), respectively. JRA-55 land analysis data provided the initial land conditions. Various climate indices were calculated to represent the interannual variability of SST, atmospheric circulations, and TC activity. These included the NINO3.4 index for ENSO, the Indian Ocean Basin (IOB) index for Indian Ocean SST variability, the WNP monsoon index based on 850 hPa zonal winds, rainfall indices for the WNP and Ganges regions, and the Indochina land surface temperature. Prediction skill was assessed using the Pearson correlation coefficient between the ensemble mean predictions and the observations. The theoretical skill dependence on the ensemble size was also estimated based on the perfect model assumption. Tropical cyclone (TC) activity was evaluated by calculating TC density using a previously validated objective detection and tracking method. The monsoon trough index was also utilized to understand TC activity.

The JMA/MRI-CPS2 model demonstrated skillful prediction of key indices representing the interannual variability of the Indo-WNP monsoon and ENSO one year ahead. The WNP monsoon index, representing the dominant variability of the WNP monsoon, showed a significant correlation skill (r = 0.50, p < 0.005) for the second summer. The predictability of the Indo-WNP monsoon was found to originate from slowly evolving IOB SST, which was predicted with a much higher correlation skill (r = 0.70, p < 0.001) than NINO3.4 SST (r = 0.41, p = 0.012). The model also showed statistically significant skill in predicting WNP rainfall (r = 0.52, p < 0.001), Ganges River Basin precipitation (r = 0.48, p < 0.005), and Indochina land surface temperature (r = 0.75, p < 0.001). Spatial correlation maps confirmed the prediction skills, showing high correlations for 850 hPa zonal wind and precipitation matching the dominant patterns of Indo-WNP monsoon variability. The model retained meaningful skill even after linear detrending, indicating its ability to predict interannual variability. The study also found a strong correlation between the model's prediction skills and the inherent predictability of the climate system. Analysis of the underlying mechanisms revealed that the delayed influence of ENSO, mediated by Indian Ocean anomalies (the Indo-western Pacific Ocean capacitor, or IPOC, mode), was the primary driver of the skillful one-year-ahead prediction. The model successfully reproduced IPOC-related climate anomalies, which are consistent with previous research. The ensemble size dependency analysis revealed that a large ensemble size (>20 members) was necessary to achieve statistically significant prediction skill for the second summer. Finally, the model showed significant skill in predicting WNP TC density for the first summer (r = 0.67, p < 0.001) and moderate skill for the second summer (r = 0.39, p = 0.017). This was linked to the model's ability to capture the variability of the monsoon trough, a key modulator of WNP TC activity.

This study demonstrates the potential of state-of-the-art climate modeling to overcome long-standing challenges in predicting the Asian summer monsoon. The skillful one-year-ahead predictions of the Indo-WNP monsoon, including precipitation, surface temperature, circulation, and WNP TC activity, highlight the significant progress in climate modeling capabilities. The findings are consistent with the inherent predictability of the climate system, indicating that the model accurately captures the underlying dynamics. The key mechanism driving this extended predictability is the delayed influence of ENSO mediated by the IPOC mode, emphasizing the importance of accurately representing the coupled atmosphere-ocean interactions in the Indian Ocean-WNP system. The successful reproduction of IPOC-related climate anomalies underscores the model's fidelity. The results also emphasize the necessity of large ensembles for achieving robust prediction skills in systems with low signal-to-noise ratios. While the current study shows considerable promise, further improvements in prediction skill could be achieved by increasing the ensemble size and refining the model's ability to predict ENSO and its delayed impacts on the WNP. These findings have significant implications for various sectors in Asia.

This study showcases a significant advancement in the prediction of the Asian summer monsoon, achieving skillful predictions with a lead time exceeding one year. The success relies on accurate simulation of ENSO and the subsequent IPOC mode, highlighting the importance of coupled atmosphere-ocean dynamics in the Indian Ocean-WNP system. A large ensemble size is crucial for achieving this skill. Further improvements could be achieved by increasing ensemble size and enhancing model representation of ENSO and IPOC. The results hold considerable promise for improved seasonal prediction, with wide-ranging implications for risk management and resource allocation in Asia.

The study's findings are based on a single climate model (JMA/MRI-CPS2), limiting the generalizability of the results. While the model's performance is encouraging, future research should examine the extent to which these findings hold across other state-of-the-art climate models. The prediction skill may vary in different regions of Asia. Although the ensemble size was large (52 members), the study notes that there is a potential for further improvements in prediction skill by increasing the ensemble size. Additionally, although the model successfully reproduced the IPOC mode in many cases, there are instances where the predictions fall short, potentially because of limitations in simulating the transition from El Niño to La Niña.