Normalizing Satellite Images-Derived Land Surface Temperature Relative to Environmental Parameters Based on the Soil and Vegetation Energy Balance Equations

Land surface temperature plays an important role in the physics of surface atmosphere interactions. It is at the same time a driver and a signature of the energy and mass exchanges over land. Land surface temperature is highly variable in both space and time mainly as a result of the heterogeneity of the meteorological forcing, land cover, soil water availability, surface radiative properties and topography. Therefore, satellite-derived land surface temperature is widely used in a variety of applications including evapotranspiration monitoring, climate change studies, soil moisture estimation, vegetation monitoring, urban climate studies and forest fire detection. The normalization of the surface temperature relative to environmental parameters is essential in scientific studies and management decisions of urban and non-urban areas. For the first time, a normalization method for topography-induced variations of instantaneous solar radiation and air temperature has been applied to satellite land surface temperature. While land surface temperature data are widely used over relatively flat areas, this new approach offers the opportunity for new applications over mountainous areas. As a significant perspective, such a normalization method could potentially be used in conjunction with land surface temperature-based evapotranspiration methods over agricultural and complex terrain, soil moisture disaggregation methods and forest fire prediction models, among others. In practice, when applyingthe normalized land surface temperature as into to energy balance models, the energy balance would be driven by the mean (instead of the spatially-variable) downward radiation within the study area as it is commonly done over flat areas. The aim of the current study is to utilize the physical model based on the soil and vegetation energy balance equations for normalizing the land surface temperature relative to environmental parameters. For this purpose, Landsat 7 satellite bands, AST08 Surface kinetic temperature, MODIS water vapor product, ASTER digital elevation model and meteorological and climatic data sets were used. In the current work, topographic factors, the radiation which reached the surface, albedo, environmental lapse rate and vegetation, were considered as environmental parameters. For calculating the surface temperature, the single channel algorithm was used. Moreover, for calculating the downward radiation to the surface, the albedo of the surface, lapse rate and vegetation; respectively, direct and diffuss radiation of the solar and neighboring surfaces, a combination of Landsat 8 reflective bands, the digital elevation model, and NDVI index, were used. Finally, by creating the energy balance equations for dry bare soil cover, wet bare soil, fully stressed vegetation and unstressed vegetation, the temperature of various coverages was extracted by exploiting Newton's method and by optimizing parameters of the model in both global and local optimizations and combining resultant temperatures; modeled and normalized surface temperature was obtained. The environmental parameters normalization model is calibrated in two main steps using Landsat land surface temperature observations. The first step minimizes the mean difference between observed and modeled land surface temperature. The second step adjusts environmental lapse rate, surface soil dryness index and vegetation water stress index by minimizing the RMSE between Landsat land surface temperature and model-derived land surface temperature. For evaluating the accuracy of the proposed model results, coefficient correlation indexes and RMSE were used between the modeled and observed surface temperature values as well as the variance of normalized surface temperature values. The results of this study indicate that in global optimization, the values ​​of the correlation coefficient, RMSE and variance for AST08 data were 0.89, 2.6 and 6.44, respectively for Landsat 7 data 0.93, 2.08 and 1.1 and in local optimization mode, the values ​​of these criteria for AST08 data were 0.962, 1.61 and 0.71 respectively and for the data of Landsat 7, 0.977, 1.2 and 0.13. The results of the study showed that, in both global and local optimization methods, the performance of Landsat 7 for normalizing the land surface temperature is higher than ASTER. Also, the use of local optimization method for global optimization to estimate the optimal values ​​of the missing parameters increased the accuracy of normalization results. The investigation of results of the relation between the surface temperature and considered environmental parameters in this study before and after the normalization indicate that the effect of environmental parameters on the surface temperature noticeably reduced after the normalization. Results of the current study imply the high efficiency of the proposed model for normalizing the surface temperature relative to environmental parameters. Generally, the results of the research were an indicator of the efficiency of the proposed model for normalizing the surface temperature relative to environmental parameters.