Predictability of the Onset of the Rainy Season Across Thailand: Mechanisms and Non‐Stationary Behaviour

ABSTRACT The onset of the rainy season across 36 stations of Thailand is analysed from 1951 to 2022. The country‐scale mean onset occurs in late April with an interannual standard deviation of 11.5 days. It tends to be anomalously late (respectively, early) during decaying warm (respectively, cold) El Niño Southern Oscillation (ENSO) events. Warm ENSO events delay the onset through a deep anticyclonic anomaly over the western North Pacific, intensification of the southern margins of the subtropical westerly jet, sustained easterly low‐level anomalies north of the equator and an anomalous anticyclonic circulation combined with in situ cold sea surface temperature (SST) anomalies over the tropical NW Pacific. The predictability of the country‐scale onset from February or March SST is large and statistically significant from ~1980, suggesting a high level of confidence for operational forecasts. However, the analysis of 1950–1970s reveals a lack of stationarity to the extent that the SST–onset relationships were then hardly present. Consistently, the spatial coherence of the local‐scale onset as well as its predictability and its interannual variability all increase from the start to the end of the period. This non‐stationarity seems fundamentally due to a change in the seasonal structure of the decay of warm ENSO events themselves, that is, from ~1980 onwards, they last longer retaining warm equatorial Eastern Pacific SST anomalies through the boreal spring, while the decaying cold ENSO events have undergone no clear change in seasonality. In consequence, in recent decades, the warm ENSO events significantly impact the April–May atmospheric structure over South East Asia, thus delaying the rainy season onset across Thailand. A long‐term significant warming SST has occurred over the tropical North West Pacific, which may potentially reduce the impact of warm ENSO events by reducing the in situ anticyclonic anomaly.