Assessment of solar radiation models and temporal averaging schemes in predicting radiation and cotton production in the southern United States

Abstract

Crop models require daily weather input data for solar radiation (I rad), minimum and maximum air temperatures, precipitation, and windspeed; but measured I rad may not be available at some locations, necessitating I rad estimates. A total of 28 scenarios (7 solar radiation models (SRMs) × 4 temporal-averaging schemes (TASs)) were examined to estimate I rad and cotton (Gossypium hirsu- tum L.) yield at 10 U.S. locations. The SRMs showed positive correlations of I rad with daylength and temperature range (Tmax - Tmin), and were relatively accurate in predicting I rad and yield. The I rad esti- mation accuracy depended on SRM, TAS, and location. Temporal averaging smoothed out short-term fluctuations, resulting in decreased temporal scatter in the weather parameters. The combination of Tmin, Tmax, precipitation and wind (TmRnWn) model performed best, and I rad estimation accuracy was highest in Shafter, California, and Maricopa, Arizona. Highest I rad estimation accuracy was obtained with the TmRnWn model, using a double TAS, in Maricopa (r 2 = 0.99). Geographical variability in I rad was observed, showing effects of regional climate on measured I rad and on I rad estimation accuracy. Yield estimation accuracy depended on I rad estimation accuracy and yield response to I rad changes, and depended more strongly on location and management practice (rainfed (RF) versus irrigated (IRR)) than on SRM and TAS. All 7 SRMs performed comparably well in predicting RF and IRR yields. Estimation accuracies for I rad and RF + IRR cotton yields among the 28 scenarios were highest for Shafter and Maricopa (e.g. r 2 > 0.99 for yield). Coupled with crop simulation models, SRMs are use- ful for predicting I rad and crop yields, particularly in regions with unavailable measured I rad data.