A dynamic multiple regression approach for quantifying the relative impact of precipitation variations and streamflow generation conditions on runoff
Based on the observed data of monthly precipitation and runoff from 1960 to 2011 in the catchment controlled by the Sandaohezi hydrological station, the double mass curve of precipitation-runoff and Kolmogorov–Smirnov dual sample test were applied to divide the years 1960–2011 into four periods: 1960–1979, 1980–1991, 1992–1999 and 2000–2011. And then, through several trial calculations, a water year of the catchment was divided into wet season and dry season, and the precipitation-runoff dynamic multiple regression (DMR) models were constructed for the four periods. The computed annual runoff values of the DMR models all passed the Fisher test. Besides, the four models' average residuals were all less than 14.15%, and the determination coefficients were all greater than 77.96%. Using the established models, we quantified to what extent the runoff was affected by precipitation variations and streamflow yield capability changes. The result showed that compared with the period of 1960–1979, the annual runoff in 1980–2011 decreased by 14.42% due to precipitation variations and by 23.11% due to streamflow generation condition changes. Furthermore, we found that the runoff of the dry season was dominantly related to the precipitation during the flood season in the previous year.