Spectral analysis of base flow separation with digital filters

Base flow separation has often been portrayed as the process of removing a high‐frequency event (runoff) from a streamflow time series to determine the low‐frequency component (base flow). Fourier decomposition of several models of streamflow components suggests that this view is inaccurate. Base flow is a predominantly low‐frequency phenomenon, but runoff has a broad bandwidth with a significant low‐frequency “signal.” Base flow separation with digital filters is the attempt to isolate these two signals. Perfect separation is not possible because of the overlapping frequency content. Removal of low‐frequency runoff signal will also partially attenuate base flow. Some principles of filter theory are used to suggest optimal filtering procedures for one commonly used recursive filter. To maximize runoff removal while minimizing base flow attenuation and phase distortion, it is usually best to use two high‐attenuation filter passes as opposed to several low‐attenuation passes.