Surface temperature estimated with Landsat 8 images and geostatistical tools in the northwestern Sao Paulo state

Abstract

In the northwestern side of Sao Paulo state, irrigated crops are replacing natural vegetation, bringing importance for the development and applications of tools to quantify the energy and water balances. Remote sensing together with geostatistical tools are suitable for these tasks, being the surface temperature (T-0) one of the radiation balance modelling input parameters. However, due to the importance of high both spatial and temporal resolutions to capture the dynamics of water and vegetation conditions, when the thermal bands are absent in several high-resolution satellites, applications on water resources studies are limited. This paper aimed to test the Moving Average (MA) and the Nearest Point (NP) geostatistical interpolation methods for estimate T-0 with and without the Landsat 8 (L8) thermal bands by using a net of agrometeorological stations. In the case of using the L8 satellite thermal radiances, the Plank. s low was applied to its bands 10 and 11. Without these bands, T-0 was retrieved as residue in the radiation balance. Up scaling the satellite overpass T-0 to daily scale resulted in a root mean square error (RMSE) of only 1.72 and 1.74 K when compared with values resulted from the MA and NP applications with the residual method, respectively. However, the MA method seemed to be more suitable than the NP one, being concluded that the coupled use of high spatial resolution images without a thermal band and interpolated weather data throughout the MA method is suitable for large-scale energy and water balance studies.