Abstract: Baseflow is the major recharge source of river runoff. Analyzing changes in time series of baseflow characteristics using a stable and reliable baseflow separation method is of great importance to water resources planning and ecological environment protection in river basins. Based on measured daily runoff data from 1957 to 2020 at Zhimenda Hydrological station in the headwaters of the Yangtze River, we selected nine methods including filtering method, BFI method and HYSEP method to investigate the characteristic values of baseflow index in the headwaters of the Yangtze River. On this basis, we can determine a baseflow separation method suitable for the Yangtze River, and analyze the time series of baseflow index including annual maximum, annual minimum and annual average values. The results demonstrate that there is a great difference among the results of various methods, with the maximum baseflow index being 0.899 and the minimum being 0.502. The filtering method, especially F4, generates small variance and small inter-annual and intra-annual variations. Thus, we determine that F4 is the method suitable for Zhimenda station. Calculated by F4 method, all the characteristic values of baseflow index at Zhimenda show an increasing trend and will continue to increase in the future. Statistics of the characteristic values using the Bayesian method indicates that baseflow index BFI, annual maximum baseflow, annual minimum baseflow and annual average baseflow of Zhimenda station varied significantly in 2002, 2004, 2017, and 2004, respectively. Warming and wetting of the headwaters is one of the causes of baseflow change. Meanwhile, the continuous rising of BFI suggest that the proportion of baseflow in annual runoff also increases, which impies that the water conservation capacity in the region has been enhancing.

Key words: baseflow separation, baseflow variation, time series analysis, Bayesian change point test, headwaters of the Yangtze River