an evaluation of m5 model tree vs. artificial neural network for estimating mean air temperature as based on land surface temperature data by modis-terra sensor

the use of satellite data in an estimation of air temperature (ta) near the earth’s surface has turned into an effective way for a large area of high spatial and temporal resolution. throughout the present study, artificial neural network (ann) as well as m5 model tree were employed to estimate ta in khuzestan province (south west of iran), using satellite remotely sensed land surface temperature (ts) data acquired through the modis-terra sensor. the input variables for the models consisted of the daytime and nighttime modis ts as well as extraterrestrial solar radiation. a total of 365 images of mod11a1 ts product for the year 2007, covering the area under study were collected from the land processes distributed active archive center (lp daac). the results indicated that coefficient of determination (r2) for both models exceeded 0.96. however, ann model estimations of air temperature were more accurate than rmse with the respective r2 values of 1.7 and 0.97 oc.