Threshold behavior in a fissured granitic catchment in southern China: 2. Modeling and uncertainty analysis

Abstract

Key Points Influence of model uncertainty on threshold In general, preferential flow occurring in forest catchments cannot be reasonably described using classical partial differential equations. As a result, linear or nonlinear reservoir models are widely used in hillslope and catchment hydrology. Currently, few studies have simulated the hydrological threshold behavior that has been observed in many experimental catchments. In this study, five models with different structures were constructed using linear reservoir method to explore the inherent mechanisms of threshold behavior and to analyze the uncertainty of model structure in threshold simulations. According to the model results, the average bedrock depression storage over the study catchment (0.99 km2), which can be represented using the height of the lowest lateral outlet from the reservoir bottom (h1), was 1.5-5.1 mm. Substantial movable water percolated into the fissure bedrock and was discharged into the streamflow as base flow after storm events, illustrating why the slope of the linear relationships between the total event precipitation plus the antecedent soil moisture index (P + ASI) and the event quick flow depth above the rising threshold was less than one. Streamflow was simulated effectively by all five models with (h1 > 0) or without (h1 = 0) bedrock depression storage; however, different ratios of annual infiltration into bedrock to annual interflow discharged to the stream were obtained by models with h1 > 0 (2.5-2.8) and h1 = 0 (1.8-1.9). Namely, the calculated infiltration into bedrock was larger by models with h1 > 0 than that by models with h1 = 0. At the storm event scale, the simulated total bedrock flow was larger than the interflow for models with h1 > 0 by a ratio of 1:0.7, whereas for models with h1 = 0 the ratio was ∼1:1.5. ©2013. American Geophysical Union. All Rights Reserved.