Comparison of baseflow separation methods in the conterminous united States

Baseflow is critical in maintaining freshwater quality and quantity on a global scale. The estimation of the baseflow index is required for activities such as managing water resources, determining available irrigation supply and water quality, allocating water for cooling, recreation, and navigation, calibrating hydrological and climate models, and assessing ecosystem productivity. Since quantitative estimates of baseflow are required but no direct observations are available, hydrograph filtering appears to be a viable solution for quantifying the baseflow process. The study aimed to assess BFI using different graphical and digital separation algorithms. In this study, 60 catchments have been selected within the Continental United States (CONUS) based on three criteria: (1) location of the stream gauge sites; (2) length of time series; (3) the least number of gaps in time series. Moreover, the study used data processing techniques, statistical analysis, and baseflow separation methods to assess the behavior of BFI values in well-drained and poorly-drained hydrological systems at a long-term scale. To separate baseflow from streamflow time series, we used graphical and digital filters, including the Smoothed Minima technique (SMT), Lyne&Hollick (LH), and Exponential Weighted Moving Average (EWMA) methods. These methods require one parameter, and the results of baseflow separation can vary depending on the method and filter parameters used. We applied all methods to the daily data of 60 catchments in the database, and after separating the baseflow, we calculated BFI as the ratio of baseflow volume to total streamflow volume. To assess the ability of the applied methods, we also used the tracer-based mass balance method (MBF), which has a clear physical basis and is considered one of the most objective baseflow separation methods, as a reference method. Statistical calculations were performed for both the well-drained and poorly-drained groups. The results indicate that in the poorly-drained group, the determination coefficients for all applied methods showed a weaker relationship between the variation of BFI relative to BFI calculated by MBF, as compared to the well-drained group.