Assessing bias in evapotranspiration and soil moisture estimates due to the use of modeled solar radiation and dew point temperature data

Abstract

Solar radiation and dew point temperature are important input variables for many crop growth ecological, hydrological, and meteorological models. It is also well known that solar radiation and dew point temperature (and relative humidity) data are not readily available for most locations over the globe. To use the above models, estimation of these input variables is often required. However, there is a nearly complete absence of scientific literature quantifying the impacts of these estimated input variables on final outputs. To address this void in the literature, the present study investigates the impacts of modeled solar radiation and dew point temperature (for relative humidity estimation) data on estimated evapotranspiration (ET) and soil moisture (SM). Two reliable solar radiation and dew point temperature models are used to simulate both of these input variables. Subsequently, measured and estimated solar radiation and dew point temperature are used as inputs to a soil water-energy balance model to estimate ET and SM. This model is applied to three contrasting hydroclimatic locations which are representative of east to west declining precipitation gradient of the North American Great Plains. In addition, at each of these locations the model was applied for three land uses, including, irrigated corn, rainfed corn, and grass. The simulations were conducted for the period 1990–2001 at a daily time resolution. A detailed quantitative assessment of impacts of modeled input data is completed by comparing reference (simulations conducted using measured input data) with estimated (simulations with modeled input data) ET and SM. The study finds that for the complete time series of ET, d- index, coefficient of efficiency ( E), RMSE, MAE, and r 2 range from 0.93 to 0.96, 0.72 and 0.82, 0.92 to 1.31, 0.41 to 0.64, and 0.77 to 0.89 mm, respectively. Likewise, for soil moisture, d- index, E, RMSE, MAE, and r 2 range from 0.64 to 0.95, 0.82 to 1.11, 2.53 to 11.95, 1.62 to 11.24, and 0.33 to 0.88 cm, respectively. Impacts of modeled input data varied from one land use to the other and also from one location to another. It is found that estimated solar radiation tends to be higher and relative humidity tends to be lower compared to observed data. These biases were also a source of disagreement between reference and estimated ET and SM. In summary, this study provides quantitative estimates of potential ranges of errors due to the use of estimated input data in a soil water-energy balance model. It is suggested that this assessment will provide a benchmark to the scientific community during any modeling exercise that uses estimated solar radiation and dew point temperature (and relative humidity).