USING A NATIONAL HYDROLOGIC MODEL TO OBTAIN REGIONAL WATER SUPPLY ESTIMATES: THE CASE OF THE HEAVILY-STRESSED CENTRAL ARKANSAS RIVER BASIN (CARB)

Streamflow depletion is a global water resource management issue likely to be impacted by climate change. Locally-designed, calibrated hydrologic models are often used to estimate available water resources and develop management strategies, but are not available in many settings and are limited in their ability to evaluate consequences of distant climate impacts such as changes in mountain snowmelt. Currently available national-scale models could address these issues, but often do not represent the groundwater extractions common in irrigated agricultural areas. A way to quantify water availability and streamflow depletion under climate change is needed. This study assesses the utility of the US Geological Survey National Hydrologic Model (NHM) through a case study in the heavily impacted Central Arkansas River Basin (CARB), which spans parts of Kansas, Oklahoma, Texas, New Mexico, and Colorado and is underlain by the High Plains aquifer. Comparison of NHM results to observed streamflow at 27 gages indicated the NHM overpredicts low flows with increasing disagreement lower in the stream network where groundwater pumping volumes are greater. To address these observed biases, we tested multiple bias correction methods based on Flow Duration Curves (FDC) and Auto-Regressive Integrated Moving Average (ARIMA). The FDC involved mapping generated cumulative probability of measured streamflow to cumulative probability of simulated streamflow and the ARIMA method involved generating a residual-based model from simulated streamflows. Results suggest that for this system the FDC method performs better in low flows and the ARIMA performs better in high flows. A hybrid FDC-ARIMA model provides a better match to historic data at both low and high flows and improves NHM’s streamflow predictions in the CARB. The study suggests that national-scale hydrologic models can be useful tools for evaluating future streamflow in heavily-stressed settings where historical data are available to develop appropriate bias-correction methods.