Mapping land surface temperature distribution over Jiashan County based on multi-source & multi-resolution remote sensing and meteorological data records

In the present work, we investigate the relationship between the ground surface temperature and some selected factors that influence its spatial distribution over the land surface. These factors include the Land use land cover (LULC) types, surface air temperature, wind velocity, and relative humidity. A quantitative analysis taking into account the respective individual influencing weights of each of these factors can help in achieving better accuracy in mapping ground surface temperature distribution. We herein processed a LANDSAT ETM+ image of Spring Season (M 11th, 2009), a QUICKBIRD image acquired on June 30th 2010, and meteorological data collected from 8 meteorological stations distributed over Jiashan County, covering 507km2. Each of the variables including ground surface temperature, air surface temperature, wind velocity, relative humidity, NDVI, NDWI, NDBI was used in Bayesian networks analysis in order to quantify its individual importance in controlling ground surface temperature distribution. Based on structure learning, parameter learning and inference of conditional probability tables for each parent nodes, it was found that the spatial distribution of ground surface temperature is strongly influenced by the types of LULC. We then used the Junction tree inference engine to extract the marginal probabilities of each LULC type in terms of its contributing weight on the spatial distribution of ground surface temperature. A temperature index was then derived which allows a high accuracy of ground surface temperature mapping by taking into account the influence of LULC types.