Correlation Between Retrieved Land Surface Temperature from Landsat and in-Situ Measurements for Geothermal Mapping in Taiwan

Abstract:

Land surface temperature (LST) is the key parameter in a wide variety of the studies of evapotranspiration, soil moisture, surface energy balance, and urban heat islands. LST with relatively high accuracy is needed in the assessment of potential area for geothermal resource. Besides, the spatial and temporal patterns of land surface temperature are essential to monitor and define geothermal activity of a specific geographic region. In Taiwan, the geothermal energy shows opportunities both for the renewable energy and economic growth ever since the government’s master project of energy in 2007. Moreover, land surface temperature derived from satellite imagery may supplement or substitute for near-surface air temperature while lacking air temperature data or poor quality data from weather stations (measured at 1.5 to 2.0 m above ground). This study uses 2001 daytime winter data to study the relationship between land surface temperature (LST) and land surface air temperature (LSAT) in central and northern region of Taiwan, with the aims of comprehending the correlation and serving as the validation of satellite retrieved LST. Two data sources are to be used: 1-hour averaged ground measurements from the permanent weather station network of the CWB (Central Weather Bureau of Taiwan); Land Surface Temperature (LST) retrieved from the Landsat-7 ETM+ Satellite Imagery of USGS (U.S. Geological Survey). Differences between air and land surface temperatures are statistically compared. To verify the suitability of LST derived from LANDSAT Imagery, the relationship between LST and air temperature at weather stations was investigated using a difference analysis by computing average relative difference (ARD) and root mean-squared difference (RMSD). A linear regression model was also applied to correct LST for the difference between LST and measured air temperatures. The results suggest that LST is inappropriate to be used as a direct substitute for air temperature at weather stations due to the differences in their absolute values. However, the regression model can lower the difference between LST and estimated air temperatures to a satisfying level.

Keywords: Land Surface Temperature (LST), Land Surface Air Temperature (LSAT), Thermal Remote Sensing, Landsat-7 ETM+, Correlation Analysis, Difference Analysis