ObjectiveXinshui River is a representative river in the key watershed of the Loess Plateau in northern China and is one of the most serious soil erosion areas in this area. Selecting a suitable baseflow separation method is very important in Xinshui River Basin in the Loess Plateau.

MethodIn this paper, based on the actual daily runoff data from 1956 to 2014, with digital filtering methods (Lyne-Hollick method (DF one) and Eckhardt method (DF two), baseflow index method (BFI), time step method (HYSEP method), and PART method, the Nash-Sutcliffe coefficient and the mean relative error method (MRE) were used to compare and evaluate the data. The suitability of different baseflow separationmethodsin Xinshui River Basin was analyzed from the perspective of the multi-year baseflow index, the stability of the baseflow index and baseflow process line.

ResultThe results indicated that baseflow was evaluated using the Nash-Sutcliffe coefficient and MRE value, and the accuracy of DF one method (0.925, 3 times) was higher (E=0.76, MRE=1.24%). The digital filtering method was the most stable method to estimate baseflow index.

ConclusionThe application of DF one method (0.925, 3 times) was a relatively stable and reliable method for the baseflow separation in Xinshui River Basin in the Loess Plateau. The average annual baseflow rate of Xinshui River Basin was 0.045 billion m³ and perennial average baseflow index was 0.366.