Space-time geostatistics

Abstract

Conventional geostatistics deals with variables that vary exclusively in space. However, many branches within the earth and environmental sciences, including soil science, frequently have to deal with variables that vary not only in space, but also in time. Recently, additional effort has been made to develop spatiotem poral statistical models and to apply spatiotemporal kriging. This chapter reviews the main approaches to extending conventional geostatistical methods to the space-time domain. Whenever possible, one should try to explain part of the temporal variation by including drift functions that represent dynamic process knowledge. The residual may then be modeled as a realization of a stationary random function, which will usually have geometric as well as zonal anisotropies. Space-time interpolation is performed using standard kriging algorithms. The theory is illustrated with an example of space-time kriging of soil water content.