2.4 Digital filtering method for separating base flow
The digital filtering method is based on signal analysis and processing technology, whose principle is to decompose daily flow data into high- and low-frequency signals. It is believed that surface flow is a high-frequency signal while the base flow is a low-frequency signal, thus fast flow and base flow can be separated accordingly (Nathan and McMahon, 1990). In this paper, Lyne Hollick filtering method (Eqs. and ) was adopted to separate.
where is the surface flow (fast flow) at time t , while and is the total flow and base flow at the current time t respectively. is the filtering parameter, with the value of 0.925 (Chapman, 1991).
By applying the baseflow separation methods to daily streamflow, daily streamflow are schematically partitioned into separate fast flow and slow flow time series, as shown in Figure 1 (a). Processes controlling fast flows are surface runoff generation and routing. The variability of fast flows is governed by stochastic characteristics of the sequences of storm events experienced by the catchment and the properties of surface soils and topography. Processes controlling slow flow include subsurface flow and groundwater discharge. The variability of slow flows strongly reflects climate seasonality and the underlying geology of the aquifer system. In this paper, the total flow, fast flow and slow flow series were normalized with the value of average daily total flow, average daily fast flow and average daily slow flow respectively. In addition, a normalized time series (i.e. daily flow divided by long-term average daily flow) was used to construct the empirical duration curve: TFDC, FFDC and SFDC (Yokoo & Sivapalan, 2011), as demonstrated in Figure 1 (b).