The variations of the regional monsoons and their predictability from monitoring their evolution

This study utilizes the monitoring of the onset dates of the rainy season across India, Southeast Asia, Central America, and West Africa to predict the upcoming season. By employing a straightforward objective method based on daily rainfall data, we pinpoint the onset date of the rainy season at each grid point of the rainfall analysis by identifying the minimum on the corresponding daily cumulative anomaly curve of rainfall. To accurately estimate the onset and retreat dates for each year, we introduce a perturbation technique that generates an ensemble of 100 time series of rainfall at every grid point.

Our research demonstrates that the onset data anomalies of the rainy season are directly linked to the length and seasonal rainfall anomaly of the season in all these regions. Specifically, seasons with an earlier onset tend to be longer and wetter, while those with a later onset are shorter and drier. Furthermore, we explore the relationship between onset, retreat, seasonal length, rainfall, and various large-scale climate drivers, revealing that although these relationships are local and relatively weaker, the intrinsic connections among the variables are robust.

In this study, we leverage the 12-hour latency product of Integrated Multi-Satellite Retrievals for the Global Precipitation Mission version 6 (IMERG) for near real-time monitoring of the season's evolution. The probabilistic skill scores, assessed using the area under the relative operating characteristic curve method, confirm the high predictive skill of anomalous onset dates.