Dynamically downscaled climate outputs for estimating hydrological responses for a Wyoming watershed

Abstract:

Potential impacts of climate on surface hydrology in western Wyoming were assessed using the Weather Research and Forecasting (WRF) model in conjunction with spatially explicit hydrological models. The study focused on Crow Creek watershed, which is one of the main watersheds providing water to the city of Cheyenne, Wyoming. Pronounced water shortages were occurred between 2011 and 2013, leaving no water in the streams by the end of July each year. We developed climate scenarios by downscaling the predictions from General Circulation Models (GCM’s) and Regional Climate Models (RCM’s). Therefore, WRF was employed downscale the existing GCM’s and RCM’s in to local climate conditions and to obtain a higher spatial resolution. The data assimilation system and software architecture enables parallel processing developed by the National Center for Atmospheric Research (NCAR). The Automated Geospatial Water Assessment tool (AGWA) interface was used to parameterize and execute two hydrologic models: the Soil and Water Assessment Tool (SWAT) and the KINEmatic Runoff and EROSion model (KINEROS2). We used freely available data including SSURGO soils, Multi-Resolution Landscape Consortium (MRLC) land cover, and 10m resolution terrain data to derive suitable initial parameters for the models. Observed daily rainfall and temperature inputs as a function of elevation were used for model validation. Cellular Automation used in predicting future land cover. Future scenarios were developed for different global emissions scenarios proposed by the Special Report on Emissions Scenarios (SRES). Daily rainfall and surface temperature series were simulated for Crow Creek watershed for the year 2050 and used as an input to AGWA model. Results were used to find the impacts of the climate on water resources and the flow regimes of the watershed. The results from different data sources were compared for percentage of explained variance, mean bias for temperature and rainfall to produce spatially explicit maps. Most of the GCM’s and RCM’s underpredict the stream flow in the area and in different scenarios. Nash-Sutcliffe for the coefficient of efficiency (CE) was best with the simulations performed with station data and above 0.8 or higher.