Air temperature estimation based on environmental parameters using remote sensing data

This study is aimed at estimating monthly mean air temperature (Ta) using the MODIS Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), latitude, altitude, slope gradient and land use data during 2001-2015. The results showed that despite some spatial similarities between annual spatial patterns of Ta and LST, their variations are significantly different, so that the Ta variation coefficient is four times the one of the LST. Our analysis indicated that while in winter latitude is the key factor in explaining the distribution of the differences LST-Ta, in other seasons the role of slope and vegetation become more prominent. After obtaining the spatial patterns of LST and Ta, we estimated Ta using regression models in spatial resolution of 0.125˚. The lowest estimation error was found in the months of November and December with a high explanatory coefficient (R2) of 70% and a standard error of 1 ° C. On the other hand, the maximum error was obtained from May to August with R2 between 59 to 63% and a standard error of 1.6 ° C which is significant at the 0.05 level. In addition, result of evaluation of individual months showed that estimation of Ta is more accurate at the cold months of the year (November, December, January, February, and March). With considering different land uses, the highest R2 was related to waters and urban areas (96 to 99%) in warm months, and the lowest R2 was for mixed forest and grassland (between 15 and 36%) in cold months.