Ecohydrologic model uncertainty and application in an urban environment: The RHESSys model in Mission Creek.

摘要

In water-stressed semi-arid areas, the impact of urbanization on ecohydrologic processes such as vegetation water use and net primary productivity is of concern. This dissertation seeks to increase understanding of ecohydrologic processes within an urban and semi-arid context through the application of an ecohydrological model (the Regional Ecohydrologic Simulation System, RHESSys) to the Mission Creek catchment in Santa Barbara County, CA. The analysis first quantifies the impact of uncertainties from both calibrated soil parameters and inputs (precipitation scaling, temperature, and outdoor water use) on model output. I find that the relative impact of different sources of uncertainty varies throughout the water year and with the model output of interest. Early in the water year (Oct-Nov), soils are relatively dry and soil parameters exert a strong control on streamflow. As the rainy season progresses and soils become more saturated, streamflow becomes more sensitive to uncertainty in the scaling of precipitation with elevation. In the spring and summer months, modeled ET becomes more sensitive to calibrated soil parameters and outdoor water use as plant water demands increase and soil water stores are utilized for ET. I next devised and tested a series of alternative calibration approaches aimed at reducing the potential for observation and input error to introduce bias in the calibration process. I find that valuing stability of model performance across the calibration period results in the selection of soil parameters with more consistent performance between calibration and evaluation periods. I also find that filtering periods of suspected error or model weakness from the calibration period will greatly alter the distribution of acceptable parameter sets. Surprisingly, calibrating across a range of precipitation scaling uncertainty does not change the relative performance of different parameter sets. Finally, I use the calibrated model to conduct an analysis of the role of effective impervious surface area (EIA) on vegetation water use and productivity. I demonstrate that connectivity between impervious areas and the drainage network can influence modeled vegetation water use and productivity: reducing the EIA fraction of total impervious area increases runoff infiltration and can mitigate or even cancel reduction in vegetation water use and NPP directly due to vegetation loss with increased impervious surface coverage. My results indicate that runoff from impervious surfaces into neighboring vegetated areas can increase transpiration and NPP in these vegetated areas well into the dry season. As an extended dry season is common to Mediterranean sites, this finding is of interest with regards to maximizing vegetation productivity in water stressed urban areas while minimizing the need for additional water inputs from irrigation. A first order approximation of the EIA fraction into a catchment scale model indicates that the effect of accounting for the EIA fraction on model ET and NPP estimates will vary between wet and dry years, with an increasing effect during wet years and a minimal effect during dry years.