Document Type


Journal/Book Title/Conference

Journal of Hydrology




Elsevier B.V.

Publication Date


First Page


Last Page



The main objective of this study was to investigate whether dynamically downscaled high resolution (4-km) climate data from the Weather Research and Forecasting (WRF) model provide physically meaningful additional information for reference evapotranspiration (E) calculation compared to the recently published GridET framework that uses interpolation from coarser-scale simulations run at 32-km resolution. The analysis focuses on complex terrain of Utah in the western United States for years 1985–2010, and comparisons were made statewide with supplemental analyses specifically for regions with irrigated agriculture. E was calculated using the standardized equation and procedures proposed by the American Society of Civil Engineers from hourly data, and climate inputs from WRF and GridET were debiased relative to the same set of observations. For annual mean values, E from WRF (EW) and E from GridET (EG) both agreed well with E derived from observations (r2 = 0.95, bias < 2 mm). Domain-wide, EW and EG were well correlated spatially (r2 = 0.89), however local differences ΔE=EW-EG were as large as +439 mm year−1 (+26%) in some locations, and ΔE averaged +36 mm year−1. After linearly removing the effects of contrasts in solar radiation and wind speed, which are characteristically less reliable under downscaling in complex terrain, approximately half the residual variance was accounted for by contrasts in temperature and humidity between GridET and WRF. These contrasts stemmed from GridET interpolating using an assumed lapse rate of Γ = 6.5K km−1, whereas WRF produced a thermodynamically-driven lapse rate closer to 5K km−1 as observed in mountainous terrain. The primary conclusions are that observed lapse rates in complex terrain differ markedly from the commonly assumed Γ = 6.5K km−1, these lapse rates can be realistically resolved via dynamical downscaling, and use of constant Γ produces differences in E of order as large as 102 mm year−1.