Modeling hydrological non-stationarity to analyze environmental impacts on drought propagation.

Climate variation and anthropic activities are two key driving forces that impact the hydrologic cycle as well as the relationships between different drought types. Thus, it is essential to evaluate the impacts of environmental variations on the relationship between meteorological and hydrological droughts. In this study, abrupt changes in the yearly hydrological time series (streamflow) of the Han River Basin (HRB) were detected using a non-linearity-based empirical segmentation approach. The Standardized Precipitation Evapotranspiration Index (SPEI) was employed to model meteorological drought, while the Generalized Additive Model for Location, Scale and Shape (GAMLSS) algorithm was adopted to model the non-linear hydrological time series to obtain the non-stationarity based Standardized Runoff Index (SRI NS). Correlation analyses were conducted on meteorological droughts (as presented by SPEI) and the hydrological drought data (as presented by the SRI NS). A Bayesian network model (BNM) was employed to calculate the propagation likelihood of different categories of meteorological droughts resulting in hydrological droughts. Change points in the hydrological regime were identified based on the empirical segmentation analysis after the 1990s. Significant increasing trends in urbanization, gross domestic product, and population were observed after the change points. The correlation analysis showed that the seasonal (3-month) timescale of SPEI corresponded best to the three-month SRI NS. The BNM revealed that the average propagation likelihoods of severe and extreme categories of meteorological drought resulting in severe and extreme categories of hydrological drought were 23.6% and 18.2%, respectively, due to the influence of climate change. These probabilities were increased by 53.9% and 70.8%, respectively, in the human impacted era due to high pressure on water resources caused by increased population, industrialization, water extraction, etc. In conclusion, the likelihood of extreme conditions of meteorological drought resulting in extreme hydrological drought was increased significantly after the change points. • Novel framework to analyze the impact of environmental forces on drought propagation. • A non-stationary approach was proposed for change point detection in streamflow series. • Bayesian Network model to investigated propagation probability between SPEI and SRI NS. • Increase in drought propagation probability was observed in human induced period. [ABSTRACT FROM AUTHOR]