A novel time-varying threshold level method for evaluating non-stationary hydrological drought

Abstract

Affected by climate change and human activities, the evolution of drought has become more complex. Traditional drought assessment methods based on the assumption of stationarity are facing challenges. Currently, research on the non-stationary improvement of standardized drought indices (SDI) is relatively mature, while research on threshold level methods (TL) is still in its infancy. In this study, we propose a “deconstruct-reshape” framework to construct a time-varying threshold level (TL_var) method for assessing non-stationary hydrological droughts. Deconstruction refers to decomposing non-stationary streamflow series into periodic and trend terms. Then, the periodic terms are combined to generate a de-trend streamflow series to extract the threshold value. Reconstruction refers to the linear superposition of the extracted threshold and trend term to obtain the reconstructed TL_var. These two steps are the innovations of this study. Furthermore, the method was performed in 49 human-induced watersheds around the world. Kendall correlation analysis demonstrated that the correlation coefficients between the drought-flood series characterized by the TL_var method and the well-established time-varying parameter standardized streamflow index (SSI_var) were 0.642 to 0.858 on a 3-month cumulative scale and 0.736 to 0.965 on a 12-month cumulative scale. Meanwhile, a strong correlation has been identified between the drought-flood series characterized by SSI_var and TL_var and the Standardized Precipitation Index (SPI) series, particularly during periods of drought, in both small and large watersheds. In addition, the verification results based on historical drought events in the EM-DAT database demonstrated that the TL_var method can accurately capture historical drought events and reasonably assess their severity, avoiding the problems of underestimation or overestimation of drought events in traditional methods. Overall, the TL_var method proposed in this study is an effective method that can serve as an important complement to existing non-stationary drought assessment methods. This method could be applied with SSI_var to facilitate qualitative comparisons of drought severity and quantify streamflow deficit volumes in specific catchments, thereby achieving a comprehensive qualitative to quantitative assessment of non-stationary hydrological drought.