Incorporating channel network information in hydrologic response modelling: model development and validation using ecologically relevant indicators

Abstract:

Many studies in the past have revealed that hydrologic response of a basin carries imprints of its channel network. However, accurate representation of channel networks in hydrologic models has been a challenge. In addition, dominating flow processes during high flow periods are not the same as those during recession periods, and there is a need for models that can represent these varying behaviors. In this study, we develop two model structures that aim to address the challenges above. The first model assumes that flow processes can be classified into two main categories: i) pure surface flow (PSF) and ii) mixed surface-subsurface flow (MSSF). The second model is a special case of the first model which neglects PSF. Using channel networks extracted from digital elevation models, we develop instantaneous unit hydrographs (IUHs) separately for PSF (PSFIUHs) and MSSF (MSSFIUHs). PSFIUH is descried by the channel ‘network width function’, whereas MSSFIUH is obtained by modifying a recently developed channel network morphology based recession flow model. To obtain the simulated streamflow time series for a basin, we convolute the PSFIUH and the MSSFIUH with the respective effective rainfall time series. The effective rainfall time series is obtained by using the probability distributed model (PDM). For comparison purposes, we also use a dual linear-bucket model for routing flow. Comparing model performance across 78 watersheds in the United States using the Nash Sutcliffe efficiency (NSE), we find that the two model structures that incorporate channel network information outperform the linear-bucket model in 56 watersheds. Further testing model performance using indicators that capture frequency and duration of low and high flows shows that the two developed models outperform the linear-bucket model in four out of five indicators.