Model diagnostic analysis of seasonal switching of runoff generation mechanisms in the Blue River basin, Oklahoma

This paper analyses the observed intra-annual variability of the hydrological response in the Blue River basin (Oklahoma, USA) and explores the underlying process controls with the use of the Tsinghua Representative Elementary Watershed (THREW) model, which is a semi-distributed, physically based model. Analysis of rainfall and runoff data at the event scale revealed a seasonal switching pattern of runoff coefficients. Event runoff coefficients in the moderately wet and wet periods (October-April) are consistently higher than those in the dry period (May-September). The switching pattern can be attributed to the higher potential evaporation in the summer season (relative to precipitation), which contributes to a reduction of soil moisture and thus higher initial losses during storm events. This is partly confirmed by the simulation results with the use of the THREW model. However, the results also show that soil drying alone cannot fully explain the sharp drop of the estimated runoff coefficients between spring and summer. This may be indicative that additional abstraction processes may be operative. Three additional abstraction processes are hypothesized as potential explanations, namely, surface depressions, canopy interception and soil cracking. Each of these processes is introduced into the THREW model sequentially, and evaluated in terms of their ability to reproduce various signatures of intra-annual runoff variability, including the switching behavior of event runoff coefficients. Finally, we present simulation results, where all three abstraction processes are included in the THREW model, and the model are recalibrated against the observed data, resulting in improved reproduction of observed streamflows over the baseline results.