Variation in the accuracy of thermal remote sensing

Thermal infrared (TIR) remote-sensing techniques have been used to estimate land surface temperatures (LSTs) and to study the relationship between land uses and LSTs. Remotely sensed thermal data provide a time-synchronized dense grid of temperature data, and there has been a growing interest on LSTs in various fields, such as urban climatology and global environmental change. It is also important for urban planning and management practices to maintain thermally efficient urban structures. This article focuses on differences between the estimated LSTs and the measured temperatures. Three different methods, the Black Body method, the normalized difference vegetation index (NDVI) Threshold method, and the US Geological Survey (USGS) method, are used to retrieve LSTs; how various land uses affect the accuracy of LSTs is also discussed. Landsat-5 data sets covering the Columbus Metropolitan Area (CMA) are used to obtain LSTs for six different dates between January 2005 and February 2006. The land-use map is derived from Landsat-5 data, dated 1 August 2005. Measured temperatures derived from the five measuring stations, located in or near the CMA, are used to compare with LSTs. The USGS method has the estimated temperatures that are close to the measured temperatures for most cases, and the LSTs in homogeneous areas, including green and agricultural areas, have temperatures closer to the measured temperatures. Statistical results of LSTs in six land uses for seven different dates are also displayed.