Hespdiv: an R package for spatially constrained, hierarchical and contiguous regionalization in palaeobiogeography

Abstract

AbstractThe objective determination of boundaries of bioregions is a nontrivial problem. Here we present a new method family, HespDiv, and an algorithm for its implementation in a new R package: hespdiv. The hespdiv algorithm performs iterative hierarchical nonlinear spatial subdivisions of taxic data into topologically contiguous geographic regions. Bioregions obtained in this way can be viewed as realistic and causally important entities belonging to eco‐genealogical Bretskyan hierarchy. The possibilities of the hespdiv algorithm are successfully demonstrated on the Miocene mammal data from the contiguous United States, where a hierarchical set of bioregions was determined. The algorithm works in a decision‐tree‐like manner by generating multiple split‐lines, that each subdivide a study area and data into two parts per iteration per region. The performance of each split‐line is measured using a combination of data generalization and comparison functions that can be custom made or selected from pre‐set subdivision methods. In each iteration, the best split‐line is used to produce a subdivision, until no more adequate split‐lines are found. The process results in a hierarchy of subdivisions of distribution of taxonomic occurrences in space. Benefits of hespdiv include the delineation of spatially contiguous clusters, flexibility, and unique representations of spatial hierarchy tree. The package is shown to be effective in determining significant contiguous spatial structures of biota, so called geobiomes, in sufficiently sampled time intervals and regions. The flexibility and ease to use of the approach allows its application to the whole range of palaeo‐variables and environmental proxies.