Building global range maps for terrestrial insects with species range kernels and modeled aggregations of georeferenced records

Abstract

Range maps for most terrestrial invertebrates generally are not accessible as spatial data. My objective was to build global range maps for 11,844 terrestrial insect species, specifically pollinator species with at least 400 georeferenced records each, by applying species range kernels, which contained about 99% of the records, and polygons of aggregated georeferenced records. Depending on aggregation distance, point aggregation generates a range of polygons, in terms of shape and area. To illustrate, for 225 odonate species that had range maps developed by experts, I selected one of six aggregation distances ranging from 200 km to 5000 km based on resemblance to range maps and fidelity to the georeferenced records. Then to predict aggregation distances for insects without accessible ranges, I modeled variables related to georeferenced records, kernels, and aggregated polygons for six different aggregation distances, based on supervised classification of aggregation distances for 4405 species with ranges. Lastly, I examined the species richness gradients of the species ranges, recognizing incompleteness of sampling. Predicted aggregation distances were predominantly middle distances of 500 km or 1000 km, for 86% of the insect species. Given a six class, complex problem with no true answer of species ranges, model accuracy for aggregation distance was 0.66. Generally, misclassification was only off by one distance, resulting in either better fit in shape or greater connection and coverage. For odonates, the species richness patterns were similar ( r = 0.91) between expert range maps and data range maps, which helped to correct sampling deficiencies in georeferenced records. Range maps for invertebrates, without previously spatially available range maps, expand underlying sampling gradients and offer at least provisional range maps, filling the gap of missing spatial data.