GEOGRAPHICALLY WEIGHTED REGRESSION TO ANALYZE THE LAND SURFACE TEMPERATURE OVER MEDELLIN, COLOMBIA

Abstract

Rural areas have different thermal properties than urban materials such as concrete and asphalt, which show higher evidence of the urban heat island (UHI) effect. The intensity of an UHI is related to the land cover configuration and physical factors affecting the urban terrain. In this research a multiple regression analysis was carried out to examine the spatial patterns of the UHI phenomena in Medellin, Colombia. Landsat 7 ETM+ from 2010 and WorldView-2 from 2011 and 2012 satellite images were used. The regression analysis was based on the Ordinary Least Squares and the Geographically Weighted methods using average information from the 344 of the 537 barrios situated in the Aburra valley, the place where Medellin is located. Results showed that the valley’s rugged relief influences UHI spatial patterns and that this phenomena depends not only on the presence of hard surfaces or vegetation, but also on physical factors such as solar exposure and average height of the terrain.

Key Words: geographically weighted regression, satellite imagery, urban heat island.