Land managers and scientists studying rangeland watershed scale responses to changes in climate and management need watershed hydrologic models capable of accurately representing these heterogeneous systems. Currently, hydrologic models do not incorporate ecological sites, which are units of land classification, accepted by the majority of management agencies. This limits their [models] use by land management agencies which have accepted the ecological site as an appropriate scale for monitoring, assessing, and altering land management. We hypothesize that by introducing ecological site scale parameters into rangeland watershed models the accuracy and utility of the models would be improved. We used the Automated Geospatial Watershed Assessment tool (AWGA) to parameterize and run the process-based, Kinematic Runoff and Erosion model (KINEROS2). We compare three methods of parameterizing KINEROS2: default AGWA with parameters estimated from pedotransfer functions; calibrated input parameters for default overland flow planes; and ecological site scale parameters with coefficients of variation (CV) from rainfall simulator experiments. Model outputs were compared to multiple observed runoff events from both a shrub dominated and grassland sub-watershed in southeast Arizona. Model results were compared to observed events using the Nash-Sutcliffe model efficiency coefficient.
Results/Conclusions
Our study results found a significant improvement in model performance using model input parameters developed for specific ecological sites compared to default AGWA. The improvement is potentially due to the inclusion of a coefficient of variation (CV) for the effective hydraulic conductivity (Ke) when using the ecological site scale model parameters. Ke varies as a function of vegetation and soil characteristics, making Ke estimates on rangelands highly variable. Incorporating the CV of Ke allows the model to scale Ke up from plot scale to ecological site scale accounting for the inherent variability of the infiltration rate on the sites. A significant issue with process-based models has been their inability to effectively represent highly heterogeneous systems. The results of this study demonstrate the ability to improve rangeland watershed modeling and account for the natural variability of rangeland systems. Additionally, this study provides a basis for continued development of ecological site specific model parameterization techniques.