Spatial distribution pattern analysis using variograms over geographic and feature space

Abstract

Variogram analysis is effective in revealing the spatial distribution patterns of geographic variable(s). Yet, existing variograms mostly focus on the impact of spatial correlation on the variable variation and neglect the impact of geographic environment. In this paper, we proposed a variogram analysis method that develops variograms over both geographic and feature space to indicate the variable’s spatial distribution patterns. To achieve this, besides existing variogram over geographic space, an innovative variogram over the feature space is designed to quantify the changing trend of target geographic variable’s variation under different similarities of geographic environment between locations. From the parameters of the variograms over geographic and feature space, a set of indices are designed to quantitatively reflect the impact from spatial correlation and geographic environment on spatial distribution patterns in a comparable manner. Experiments are conducted in four areas representing different spatial patterns with soil property being the target geographic variable. Experimental results show that the effectiveness of the proposed variogram analysis method can differentiate the spatial distribution patterns of the soil property in the experimental areas. The effectiveness of the proposed variogram analysis method is affected by sample representativeness and covariate selection.