A voxel-based three-dimensional framework for flash drought identification in space and time

Abstract

Flash drought (FD) is an extreme event causing severe impacts on ecosystems and agriculture due to its fast velocity of evolution and high frequency. Here, we develop a voxel-based three-dimensional FD (V3DFD) method to track the spatio-temporal dynamics of FD based on the soil moisture (SM) decline method proposed by Yuan et al. (2019). The key idea of V3DFD is that FD is identified in terms of the connectivity of drought voxels (i.e., cubes with width, length and height representing latitude, longitude and time, respectively) in 3D space instead of directly forming the drought events by merging drought clusters. The V3DFD is systematically evaluated and compared with the commonly used 3D severity-area-duration (SAD) method in four provinces in China: Guangdong (GD), Guangxi (GX), Guizhou (GZ), and Yunnan (YN). Results show that V3DFD can better reflect the dynamic development and recovery of FD events compared with SAD. Regional analyses indicate that FD occurs more frequent with a larger intensity and longer duration in GD and GX than YN and GZ. The onset (terminal) of FD in YN occurs mostly in April-May (April-June). The onset of FD in GX and GD is scattered at the intra-annual scale, but the terminal occurs mostly in November. In addition, the first FD event in a year is mostly located during April-May in YN and GZ, while the last FD event of a year occurs mostly during October-November in GX and GD.