Ability to predict daily solar radiation values from interpolated climate records for use in crop simulation models

Abstract

Daily solar radiation is a common input in crop growth and development models yet given the sparse coverage it is likely not measured in the area where the growth model is applied. Consequently, solar radiation needs to be estimated from readily available commonly measured meteorological variables. Daily shortwave solar radiation accumulations were estimated using three existing models over the period 1990 through 2000 for 10 locations across the North Central Region of the USA where hourly automated weather stations provided observations of solar radiation. The climate represented by these stations had annual precipitation ranging from 320 mm to 1063 mm and annual mean daily temperatures from 7.5 °C to 14.2 °C. To match the scale of USDA crop yield data, the daily precipitation, maximum temperature, and minimum temperature was estimated for the centroid of the counties, in which the automated weather stations were located using Thiessen polygon area-weighted interpolated daily cooperative weather station network (COOP). Solar radiation at each county and location was estimated using three models. The Mahmood–Hubbard (M–H) model was better than the Bristow–Campbell (B–C), and Weather Generator for Solar Radiation (WGENR) models at estimating daily solar radiation at county centroids across the Midwest. The error associated with the interpolated estimates was at least 1 MJ m −2 worse than the error when no interpolation was necessary (estimates at the automated station sites). The magnitude of this interpolation error is comparable to the differences in daily solar radiation measurements due to station separation.