Evaluating the correlation between ground information and satellite spectral data by geostatistical tools

Satellite information opened new scenarios for planet surface mineral exploration. Hyperspectral information brought by sensors on board potentially help identifying and measuring concentrations of an element if an accurate calibration is done, based on available ground sampling. Most popular uses of satellite images refer to 2D problems and most calibrations refer to the spectral properties of the objects to be discovered and characterized (Follador M., 2005). Before calibration, images are affected by standard preprocessing, for instance for filtering unwanted effects, and for enhancing the information considered useful. The general problem for mineral exploration and reserves characterization is the spatial distribution of the target variable, with limited and sparse in situ information. Satellite images provide auxiliary information, which can be used, when correlation is found with the target variable. In this case the expected result should improve the estimation or the representation of spatial distribution of the sampled variable. Independently of the variable at hand (grades, discovery probability, …) and of the spatial distribution model (estimation, simulation), Geostatistics allows to tackle the central problem: finding meaningful correlations and modelling the unknown surface distribution of the interest variable by including satellite data as auxiliary information (Chiles et al., 2012; van der Meer, 1994). Three issues need attention when considering satellite images: a) the different support of direct and auxiliary information, being pixel data refereed to a surface, in contrast with the punctual ground data b) the need of 3D modelling c) the space-time nature of the satellite information.