Achieving downscaling of Meteosat thermal infrared imagery using artificial neural networks

Abstract

This study presents the successful application of artificial neural networks (ANNs) for downscaling Meteosat Second Generation thermal infrared satellite imagery. The scope is to examine, propose, and develop an integrated methodology to improve the spatial resolution of Meteosat satellite images. The proposed approach may contribute to the development of a general methodology for monitoring and downscaling Earth’s surface characteristics and cloud systems, where there is a clear need for contiguous, accurate, and high-spatial resolution data sets (e.g. improvement of climate model input data sets, early warning systems about extreme weather phenomena, monitoring of parameters such as solar radiation fluxes, land-surface temperature, etc.). Moderate Resolution Imaging Spectroradiometer (MODIS) images are used to validate the downscaled Meteosat images. In terms of the ANNs, a multilayer perceptron (MLP) is used and the results are shown to compare favourably against a linear regression approach.

Acknowledgements

The authors would like to express special thanks to the anonymous reviewers for their constructive comments that helped to improve the completeness and clarity of the article.