Multi-temporal analysis for drought classifying based on SPEI gridded data and hybrid maximal overlap discrete wavelet transform

Drought is one of the complex phenomena among all extreme climate events. Temporal–spatial variations of drought in an area can have numerous effects on the engineering, management and planning of water resources. In this study, a hybrid temporal-preprocessing and spatial-classification-based method was used to take the advantage of multiscale properties of the standard precipitation evapotranspiration index (SPEI) series. The SPEI gridded data from 1950 to 2019 for 60 points in the northwest of Iran were used for this aim. The maximal overlap discrete wavelet transform (MODWT) was applied to obtain the time-series time–frequency attributes, and multiscale regionalization was done via K-means clustering method. For determining the input dataset in spatial clustering, different combinations were considered based on the wavelet and scaling coefficients. Based on the results, it was found that the SPEI series had an inverse relationship with the energy values of stations, in which the SPEI values increased with decreasing energy values. It was observed that the MODWT-K-means method performed more successfully than the classical K-means method. The obtained results showed that the clusters obtained via MODWT-K-means method recognized homogenous drought districts very well. The silhouette coefficient of clustering based on historical data was obtained as 0.45, while for the proposed method it was obtained as 0.86.