Validating Dynamically Downscaled Climate Projections for Mountainous Watersheds Using Historical Runoff Data Coupled with the Distributed Hydrologic Soil Vegetation Model (DHSVM)

Abstract

The performance of dynamically downscaled climate fields with respect to observed historical stream runoff has been assessed at basin scale using a physically distributed hydrologic model (DHSVM). The dynamically downscaled climate fields were generated by running the Weather Research & Forecasting (WRF) model at 4-km horizontal resolution with boundary conditions derived from the Climate Forecast System Reanalysis. Six hydrologic models were developed using DHSVM for six mountainous tributary watersheds of the Jordan River basin at hourly time steps and 30-m spatial resolution. The size of the watersheds varies from 19 km2 to 130 km2. The models were calibrated for a 6-year period from water year (WY) 1999–2004, using the observed meteorological data from the nearby Snow Telemetry (SNOTEL) sites of the Natural Resources Conservation Services (NRCS). Calibration results showed a very good fit between simulated and observed streamflow with an average Nash-Sutcliffe Efficiency (NSE) greater than 0.77, and good to very good fits in terms of other statistical parameters like percent bias (PBIAS) and coefficient of determination (R2). A 9-year period (WY 2001–2009) was selected as the historical baseline, and stream discharges for this period were simulated using dynamically downscaled climate fields as input to the calibrated hydrologic models. Historical baseline results showed a satisfactory fit of simulated and observed streamflow with an average NSE greater than 0.45 and a coefficient of determination above 0.50. Using volumetric analysis, it has been found that the total volume of water simulated using downscaled climate projections for the entire historical baseline period for all six watersheds is 4% less than the observed amount representing a very good estimation in terms of percent error volume (PEV). However, in the case of individual watersheds, analysis of total annual water volumes showed that estimated total annual water volumes were higher than the observed for Big Cottonwood, City Creek, Millcreek and lower than the observed total annual volume of water for Little Cottonwood, Red Butte Creek, and Parleys Littledell, demonstrating similar characteristics obtained from the calibration results. Seasonal analysis showed that the models can capture the flow volume observed for Big Cottonwood, City Creek and Red Butte Creek during the peak season, and the models can capture the flow volume observed for all the watershed satisfactorily except Big Cottonwood during the dry season. Study results indicated that the dynamically downscaled climate projections used in this study performed satisfactorily in terms of stream runoff, total flow volume, and seasonal flow analyses based on different statistical tests, and can satisfactorily capture flow patterns and flow volume for most of the watersheds considering the uncertainties associated with the study.

Keywords

• water supply
• Jordan River basin
• streamflow prediction
• climate projections
• spatial and temporal variations