Local‐scale rainy season onset detection: A new approach based on principal component analysis and its application to Vietnam

Abstract

In rain‐fed farming systems, the rainy season onset date (RSOD) is of utmost importance. RSODs are often defined by applying fixed thresholds to station data or by quantifying changes in the large‐scale atmospheric circulation between the dry and wet seasons. In this study, a new improved approach is proposed and applied to determine RSODs for Vietnam based on daily rainfall station data for the period 1979–2019. In this approach, the inter‐annual similarity between station‐based RSODs and regional‐scale RSOD signal, obtained by applying principal component analysis to multiple fields, was examined. The method allows to quantify the sensitivity of RSODs to three thresholds chosen; the rainfall amount of an initial wet spell (P), the number of wet days (N), and duration of control period (D). It is demonstrated that the sensitivity to the choice of the thresholds varies across stations and sub‐regions. In particular, P is a principal parameter to determine RSODs with the significant control at 66.4% stations over entire Vietnam. While D plays an important role at 55.7% stations, and N seems to be less important, which influences only 31.3% stations. The results also reflected that the selected optimal criteria for each station captured very well the rainy seasonal cycle of Vietnam. Trends towards earlier onset dates are found for large parts of central Vietnam, whereas several northern stations show a trend towards later onsets. The new method can be applied to other monsoon regions to define more spatially coherent thresholds to define local, agriculturally relevant rainy season onsets.