The Relative Importance of Spatial and Temporal Autocorrelation

Abstract

Spatial scientists wish to understand when space matters in studies dealing with phenomena evolving through time. Predictive models often exploit temporal and spatial autocorrelation latent in space–time datasets to make forecasts for these landscapes. In many cases, temporal autocorrelation offers a superior description of dynamics in a geographic landscape. However, especially after some perturbation event, the memory captured by temporal autocorrelation may become seriously corrupted, with one consequence being that spatial autocorrelation furnishes a superior description of, and forecasts for, a given phenomenon. This chapter discusses the relative importance of these two types of autocorrelation, providing examples of when each type dominates in a geographic landscape.