Drought propagation dynamics and driving factors from a nonstationarity perspective

Drought is one of the most complex natural disasters, which has serious socioeconomic and ecological impacts across the world. With a changing climate, not only drought events have occurred more frequently, but also the characteristics of drought propagation have been changed. Under the joint effects of climate change and human activities, the assumption on the stationarity of hydrological time series has been overturned, which is of great significance in the field of hydrology. However, the current research on drought propagation is generally based on the assumption of sequence stationarity, in which related results may be biased. Therefore, it is necessary to construct a nonstationary standardized drought index to explore the dynamics of drought propagation and its driving factors for further understanding the mechanism of drought propagation. The Generalized Additive Models for Location, Scale, and Shape (GAMLSS) were applied in Luanhe River Basin to construct a time-varying drought index. The seasonal propagation characteristics from meteorological to hydrological drought were examined based on conditional probability, and the moving window was utilized to explore the dynamic change of propagation characteristics. The driving factors were investigated by using the variable importance in projection. The results indicated that using a time-varying drought index was more reasonable than using a stationary assumption; the propagation time showed a significant downward trend; hydrological drought was more likely to be triggered by meteorological drought in autumn and winter; and the precipitation, decreasing runoff, and increasing evaporation were the main factors affecting the seasonal propagation characteristics. These findings are valuable for clarifying the nonstationary characteristics of drought propagation and its seasonal dynamics, providing scientific support for drought early warning systems.