Evaluation of Disaggregation Methods for Downscaling MODIS Land Surface Temperature to Landsat Spatial Resolution in Barrax Test Site

Abstract

Thermal infrared (TIR) data are usually acquired at a coarser spatial resolution (CR) than visible and near infrared (VNIR). Several disaggregation methods have been recently developed to enhance the TIR spatial resolution using VNIR data. These approaches are based on the retrieval of a relation between TIR and VNIR data at CR, or training of a neural network, to be applied at the fine resolution afterward. In this work, different disaggregation methods are applied to the combination of two different sensors in the experimental test site of Barrax, Spain. The main objective is to test the feasibility of these techniques when applied to satellites provided with no TIR bands. Landsat and moderate imaging spectroradiometer (MODIS) images were used for this work. Land surface temperature (LST) from MODIS images was disaggregated to the Landsat spatial resolution using Landsat VNIR data. Landsat LST was used for the validation and comparison of the different techniques. Best results were obtained by the method based on a linear regression between normalized difference vegetation index (NDVI) and LST. An average RMSE = ±1.9 K was observed between disaggregated and Landsat LST from four different dates in a study area of 120 km <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> .