Modelling uncertainty and spatial dependence: Stochastic imaging

Abstract

The most vibrant area of research in geostatistics is stochastic imaging, that is, the modelling of spatial uncertainty through alternative, equiprobable, numerical representations (maps) of spatially distributed phenomena. These stochastic images are conditioned to a variety of data accounting for their specific measurement scale and reliability.

Any geostatistical prediction is built on a prior model of spatial correlation that ties data to unsampled values and, equally importantly, unsampled values at different locations together. Since a major goal in the exercise of mapping is to display organization in space, spatial correlation is a necessity. As for uncertainty it is so pervasive that it is imperative to account for it.