Application of land surface temperature inversion based on emissivity mixture analysis at sub-pixel scale in geothermal exploration

Abstract

The detection of thermal anomaly at land surface is crucial in the exploration of geothermal resources. Since field observations are costly and only limited to certain area, the requirement of land surface temperature (LST) inversion estimating accurately in geothermal exploration makes the remote sensing of LST a prevailing topic. The information about the spatial distribution of land surface emissivity (LSE) is a vital land surface temperature inversion parameter. LSE being applied to convert heat energy into radiant energy plays a key role in the measure of land surface inherent efficiency. This study proposes a new method for improving the estimation of LSE. Linear spectral mixture analysis (LSMA) model is applied to thermal infrared band data from Landsat-7 Enhanced Thematic Mapper plus (ETM+) remote sensing sensor. The effective emissivity of one single pixel is regarded as the combination of sub-pixel level emissivity and corresponding area proportion. In this paper, the first step is to conduct georeferencing and atmospheric correction on the imaging data of Nanjing in order to establish a geographic coordinate system and remove the effect of water vapor in the atmosphere. Due to the complexity of the land surface type and its physical conditions, surface pixel is assumed to be divided into three fundamental land cover components after MNF (minimum noise fraction) rotation and PPI (purity pixel indices) calculation, namely vegetation,rock and soil.This kind of dividing method is a variant of V-I-S model proposed by Ridd. The validation of LSMA showes that almost root-mean-square error (RMSE) value is less than 0.02 which can get a precise result. Representative emissivity values of three fundamental land cover are assigned to each component which is derived from ASTER Spectrum Laboratory and MODIS UCSB Emissivity Library. Following pixel effective emissivity calculation, the mono-window algorithm is used to extract LST map. Analyses and evaluations are conducted on thermal state of Tangshan area with thermal anomalies being represented by thermal field variance (TFV) index in LST map. Two possible geothermal fields are found to be located under the zone of Tangshan. Some investigations in field survey reveal a close correlation between the geothermal anomalies and the distribution of hotsprings. The feature of geothermal anomalies also indicates that the geothermal anomaly areas in Tangshan are subjected to the development of fault structure. It can be well explained that the fault structure serve as channels for thermal transmission with which the subterranean heat water can be efficiently transferred to land surface and detected as LST anomalies in thermal infrared images. This study shows that the estimation of LSE is effective and TIR remote sensing can be used in geothermal exploration accurately.