Divergent nonlinear trends of global drought and its multivariate characteristics

Drought has become a critical scientific issue in the context of global climate change. It is very important to monitor the long-term trends of multivariate characteristics of drought to mitigate and address drought risk. However, the nonlinear trend of drought has not been well understood, which needs more attentions. Hence, this study used the 1-month global standardized precipitation evapotranspiration index (SPEI) from 1901 to 2018 to extract the multivariate characteristics of drought using a 30-year sliding window, and explored their nonlinear trends with Ensemble Empirical Mode Decomposition (EEMD) method. The results showed that: (1) During the period 1901-2018, although SPEI mainly showed a significant decreasing trend globally (34.38%), but only 15.51% was monotonic decreasing and 18.87% was shift from increasing to decreasing. From hyper-arid zone to hyper-humid zone, monotonic decreasing and shift from increasing to decreasing decreased, while shift from decreasing to increasing increased. (2) Based on linear trend analysis, drought events were increasing, with an enhancement in the drought severity (DS) and a decrease in the drought intensity (DI), while the drought duration (DD) exhibited complex changes. However, when considering trend transitions, in the future, drought events will decrease in frequency. Nonetheless, their intensity, severity, and duration will increase, resulting in more severe drought risks. (3) In dryer regions, DD and DS are becoming longer and stronger, while in wetter regions, they are becoming shorter and weaker. As for DF and DI, different regions mainly exhibited trends of increasing to decreasing or decreasing to increasing. (4) The four drought features exhibited divergent trends not only from SPEI, but also from themselves, except for the high consistency between the trends of DD and DS. Ignoring the trends of multivariable characteristics of drought events will underestimate the drought risk. Our results highlight the importance of using multivariate drought characteristics for monitoring drought risk with nonlinear analyses.