A virtual experiment on the effects of evaporation and intensity smoothing by canopy interception on subsurface stormflow generation

Abstract

The effects of canopy evaporation and intensity smoothing during rain events on hillslope subsurface stormflow are poorly understood. While watershed manipulation experiments have suggested that these processes are important at long timescales, such processes may also be important in storm-timescale responses. Notwithstanding, there are few hillslopes for which both internal subsurface stormflow generation processes and canopy processes are known, so canopy interception effects on subsurface stormflow have not been tested mechanistically. Furthermore, it has not yet been possible to separate the effects of canopy evaporation from intensity smoothing in terms of which component of interception most affects hillslope response. We report a series of virtual experiments (numerical experiments driven by collective field intelligence) using HYDRUS-2D to model flow in a well-studied and characterized research hillslope in Georgia, USA. Previous work has shown that HYDRUS-2D approximates well both measured subsurface stormflow and internal pore pressures at this site. Our virtual experiments compared modeled hillslope response to rainfall and throughfall characteristic of known forest canopy processes in Washington, USA. The experiments generated subsurface stormflow using measured rainfall and throughfall data from three sites within the forest, and using synthetic, simplified throughfall signals containing either evaporation alone or intensity smoothing alone. As expected, results of our virtual experiments driven by field-measured throughfall data showed that evaporative loss delayed the onset of subsurface stormflow, lowered and delayed stormflow peaks, and decreased total flow and the runoff ratio. Virtual experiments based on simplified modeled throughfall (where we separated evaporation from intensity smoothing) showed that canopy evaporation was responsible for most of these effects, while intensity smoothing showed measurable differences only in peak subsurface stormflow rate. Overall, this work has implications for the calibration of watershed models. Not only would ignoring interception miss a major effect of vegetation on subsurface stormflow generation, our work also shows that simply applying some fractional reduction as a scaled input signal (as is customary in watershed modeling studies) may mask important effects on peak flow response in some situations.