Distribution of air temperature around Guyuan Prefecture, south part of Ningxia in China, which has only a few meteorological observation sites, was estimated in 2km×2km mesh size by using the following methods; 1) regression estimation by altitude of each mesh, 2) regression estimation by surface temperature estimated from GMS infrared data, 3) multiple regression estimation by topographical factors of each mesh, which were calculated using altitude mesh data around the study area, 4) multiple regression estimation by NDVI and altitude of each mesh.
Hourly air temperature was estimated by methods 1) and 2). Correlation coefficients by method 1) were low in winter morning, but were sufficiently high in other time. During the daytime, estimation by method 2) was accurate enough. But in most cases, accuracy of surface temperature method was lower than the method by using altitude, which was caused by low resolution of GMS data.
Monthly mean temperature and monthly mean maximum and minimum temperatures were estimated by methods 1), 3) and 4). Correlation coefficients by method 1) were low in winter, especially in monthly mean minimum temperature. For method 3), stepwise regression analysis was used to decide the most effective topographical factors for the monthly mean minimum temperature. As a result, some topographical factors except altitudes were selected in winter and estimation using selected topographical factors was more accurate than using altitude only. This result indicates that the effect of topographical factors for air temperature was larger than that of altitude in winter morning. For method 4), stepwise regression analysis was carried out and NDVI was selected for several months. Estimation using NDVI and altitude was more accurate than using only altitude. These results indicate that the condition of ground vegetation affect the air temperature distribution in summer.