Segmentation of global climate dataset into contiguous spatial units having quantitatively homogeneous climates

Abstract

Discretization of global climate simplifies the spatial variability of climates. The Köppen–Geiger classification (KGC), and more recently, clustering of global climate grids, discretize the landmass into several zones corresponding to qualitatively distinct climate types. However, quantitatively, in terms of local climatograms, such zones have significant climatic variability. In this paper, we propose a segmentation of the global climatic grid as a means to delineate contiguous segments—spatial units characterized by quantitatively homogeneous climates. The level of segments' climatic homogeneity is controlled by a single parameter δ whose value determines a maximum allowable level of segments' inhomogeneity. Segmentation is implemented using a graph-based segmentation algorithm customized to the climatic data. Application of segmentation with three different values of δ is demonstrated on the WorldClim 1.4 2.5 arc minutes resolution data. Major uses for climate segmentation are: (a) precise delineation of highly homogeneous, local climatic units, (b) checking to what degree delineation of vegetation zones can be based on climate alone, and (c) data compression of large climatic grids. The first use is demonstrated using the island of Great Britain as an example. The second use is demonstrated by the juxtaposition of KGC zones with segments. We discuss the third use in the context of using segmentation as a pre-processing to climate clustering.