Simulating the change of precipitation-runoff relationship during drought years in the eastern monsoon region of China.

Most hydrological simulation and prediction methods assume that the precipitation-runoff relationship was stationary. However, this assumption was found to be questionable during drought years: the annual runoff coefficients (the ratio of annual runoff to annual precipitation) during drought years tend to be smaller than those during non-drought years. However, little is known about the spatial distribution of the magnitude of runoff coefficient change (RCchange) during drought years, and which factors dominate the spatial pattern of RCchange over a large spatial scale. To answer these questions, this study investigated the RCchange in 265 catchments in China that cover a broad range of climate and landscape conditions. We identified the significant factors affecting RCchange from ten catchment characteristics and developed a multivariate generalized additive model to simulate the spatial pattern of RCchange across the eastern monsoon region of China. Results indicated that the RCchange showed an increasing trend from north to south of China, with values ranging from -67.1% to -0.3%, with the average being -26.4%. The lower RCchange (corresponding to more significant runoff reduction) in drought years was more likely to occur in catchments with dryer climate and lower elevation. The simulated RCchange by the multivariate generalized additive model demonstrated a good agreement with observed RCchange, and the values of Nash-Sutcliffe efficiency between observed and simulated RCchange were 0.77 for training catchments and 0.72 for testing catchments. Finally, we applied the model to extrapolate RCchange to the entire eastern monsoon region of China. The result would benefit water resources management during drought years.