Estimation of Solar Radiation Using Land Surface Temperature MODIS Sensor Data and Neural Network Model

Estimation the amount of radiation reaching the Earth's surface (Rs) is an important factor in the energy balance models simulation of plant growth and evapotranspiration estimation. Most Estimation models to radiation reaching the Earth's surface use satellite data and they are based on land surface temperatures. In this study, the Accuracy of solar radiation estimation is investigated Using four different models of neural networks (with the names of ANN1,ANN2, ANN3, ANN4) with the inputs Including products land surface temperature MODIS sensor (models 1 and 2, and models 3 and 4 are based on MOD11A1 MYD11A1 products, respectively), extraterrestrial radiation (Ra) and relative sunshine (n / N). The results show that four neural network models are able to estimate the amount of radiation reaching the Earth's surface with good correlation (R2>. 85). However, models based on MOD11A1 products have a higher accuracy than models based on MYD11A1 products. Neural network model of ANN1 (based on MOD11A1 products, relative sunshine and extraterrestrial radiation (Ra)) with the coefficient of determination (R2) equal to. 9332 and the root mean square error (RMSE) equal to 1.4448 MJ per square meter per day is more accurate on the estimation of solar radiation than other models. The results also showed that the Neural network model ANN2, comparing with Hargreaves and Samani models based on air temperature and extraterrestrial radiation, is More accurate in estimating of solar radiation.