An Improvement on Land Surface Temperature Determination by Producing Surface Emissivity Maps

Emissivity mapping of the Earth’s surface is the prerequisite to thermal remote sensing. A precise determinationof a surface's temperature is dependent upon the availability of precise emissivity data for that surface. The presentstudy area is a part of sugarcane plantation fields in the west part of Khuzestan province. In this work, TemperatureEmissivity Separation algorithm (TES) was applied to five different ASTER L1B images. It was found out that TESmethod overestimates temperature in all the five thermal bands, and underestimates the emissivities as compared tothe laboratory values. The differences in the emissivity values (as compared to laboratory values) varied from 10% inband 10 to 3% in band 14. The main reasons for these discrepancies were a lack of proper calibration of the thermalbands, the possible presence of radiometric noises in the calculation of the emissivity Maximum MinimumDifferences (MMD) as well as mixed pixels. To overcome these uncertainties in the TES algorithm, an ImprovedTES method (ITES) was introduced. In the ITES method, the surface exiting thermal fluxes were simulated. Theemissivities of four different reference surfaces, along with air temperature measured at nearby weather stations(believed to represent LST of full vegetated pixels) and the band 14 temperature, were employed as inputs. Theresults show noticeable improvements in the predicted emissivity to around 1% for band 10 and less than 1% forbands 13 and 14 as compared to the corresponding laboratory values. The root mean square error (RMSE) ofemissivities for full vegetation cover was less than 0.015 and less than 0.01 for partial vegetated cover, bare soil, andsea water surface. Finally, emissivity maps for one sample image, employing the five thermal bands, were produced.It is believed that these maps can be used in other satellite images as layers of emissivity values for the purpose of aproper estimation of surface temperatures.