A Markov chain Monte Carlo-based particle filter approach for spatiotemporal modelling of an environmental process

Abstract

The knowledge of the global solar surface irradiance (SSI) incident on the earth's surface and its spatiotemporal distribution is important to numerous solar-based applications. However, spatiotemporal modelling of SSI (a non-linear process) from Earth observation data is unfortunately not straightforward. From a signal processing perspective, it is a non-stationary, non-linear/non-Gaussian dynamical inverse problem. In this paper, we propose an MCMC particle filter approach that combines satellite images and in situ data for space-time-referenced SSI modelling. We propose original observation and transition functions taking advantage of the characteristics of the involved types of data. A simulation study of solar irradiance is conducted in parallel with this method and a map of SSI potential in French Guiana for the year 2012 is provided.