Autocorrelation-Based Regioclassification - a self-calibrating classification approach for choropleth maps explicitly considering spatial autocorrelation

摘要

The definition of classes for choropleth maps is commonly based on nonspatial attribute values, ignoring the spatially autocorrelated nature of almost all geographical data. This 'blindness' toward spatial configuration during the classification process leads to relatively complex and fragmented spatial patterns that confuse visual perception and impair the subsequent cognitive processes involved in map interpretation. This article presents a new approach to cartographic classification of univariate, quantitative polygonal data. The proposed method adapts to the degree of spatial autocorrelation in data by utilizing the Moran's I scatter plot in combination with the Fisher-Jenks algorithm. When data are spatially autocorrelated, the resulting maps are visually less complex than those derived using equivalent nonspatial classification approaches. However, the resulting classes might overlap in the value domain. The cartographic concept that we present therefore combines the advantages of traditional classification with those of our proposed method: it allows the visual assignment of individual polygons to mutually exclusive value ranges, while still preserving visual clarity of patterns.