CR 29:233-243 (2005)  -  doi:10.3354/cr029233

ABSTRACT: Lack of site-specific global solar radiation (GSR) data is a significant impediment for most crop model applications. First, several empirical methods for estimating GSR from observed maximum and minimum temperature were evaluated using data from urban (National Climatic Data Center [NCDC] data set) and rural (Florida Automated Weather Network [FAWN] data set) sites in Florida. Next, the spatial structure of empirical model parameters was investigated and the possibility of employing spatially interpolated coefficients developed at urban or rural sites to predict solar radiation at rural locations, where most crops are grown and often crop modeling is required to address various issues, was assessed. Finally, the effects of observed and estimated GSR on simulated potential yield and evapotranspiration with CROPGRO-Peanut and CERES-Maize models at 6 crop–location combinations were evaluated. The model developed by Donatelli and Bellocchi (DB) achieved the most accurate estimations of GSR across Florida. The quality of those predictions varied with environment setting (urban-rural) and latitude. RMSE (root mean square error) between observed and fitted daily GSR were in the range of 3.1–4.1 and 3.2–4.9 MJ m^–2 d^–1 for FAWN and NCDC evaluation sub-sets, respectively. Radiation for FAWN sites was efficiently estimated by the coefficients interpolated from rural sites but was substantially overestimated by 17.6%, on monthly basis, when urban sites supplied the model empirical parameters. NCDC sites were site specific and suitable sources of solar radiation at urban environments only. Input of estimated with the DB model GSR into a crop model would be a realistic option at sites where only temperature and precipitation are available.

KEY WORDS: Solar radiation estimation · Air temperature · RadEst · Crop simulation models · Urban–rural stations · Florida