Comparison of Land Surface Temperature Retrieval using Remote Sensing Imagery: Classification-Based versus Normalized Difference Vegetation Index-Based Emissivity Methods
摘要
Land surface temperature (LST) is a key parameter in identifying drought conditions, and is widely used in drought-related formulas. In addition, the type of LST measurement used, including from both single-channel and multi-channel methods (SCM and MCM, respectively), affects the LST result. The land surface emissivity (LSE) contributes to the determination of the LST, relying on several different methods, such as classification- and Normalized Difference Vegetation Index (NDVI)-based emissivity methods (CBEM and NBEM, respectively). In this study, we examined differences in the LST values obtained using SCM and MCM by applying emissivity values from CBEM and NBEM to MODIS data. Measurement of the LST was performed using the Artis and Carnahan SCM (SCMAC) and six MCMs, including Sobrino et al. (MCMSob), Qin et al. (MCMQin), Mao et al. (MCMMao), Becker and Li (MCMBL), Wan and Doizer (MCMWD), and Coll et al. (MCMColl). We found that the CBEM and NBEM emissivity did not affect the SCMAC because bands 31 and 32 did not produce significant differences in the LST values. The average error value (K) for the SCMAC exceeded 7 K, but validation indicated that the SCMAC provided adequate LST values above 310 K. Differing from the other models, the MCMQin produced underestimated LST values, with an error of 14 K. By contrast, the MCMWD aligned closely with MOD11A2, with an error of 2 K, making it the preferred option for measuring LST using MODIS imagery. Therefore, the MCM is the most suitable method for LST retrieval from MODIS imagery.