Environmental variables have become increasing available with the advent of GIS. These are mostly continuous in space and time. Collecting denser environmental data in discrete space and time domains are rather cost effective and time consuming. However, when the data at each spatial or time index are considered as outcomes of a random variable, stochastic processes become enviable useful to build models and predict the outcomes at locations where data were never collected. The meaningful assumptions include stationarity of the mean and the covariance to ascertain an expression for spatial dependency/autocorrelation. With a stationary process (i.e. constant mean), simple and ordinary kriging is used. Other variants like kriging with external drift, universal kriging and regression kriging also alleviate the challenge of non-stationary mean. These methods are also applicable when temporal indexes rather than spatial indexes are of interest.
Understand the underlying assumptions for spatial stochastics process
Explain the need for the stationarity assumption
Explain the different forms of kriging
Completed (GI-N2K)