Habitat and spatial thinning improve the Maxent models performed with incomplete data. Issue 6 (8th June 2017)

• Record Type: Journal Article
• Authors: Kiedrzyński, Marcin; Zielińska, Katarzyna M.; Rewicz, Agnieszka; Kiedrzyńska, Edyta
• Is Part Of: Journal of geophysical research. Volume 122:Issue 6(2017)
• Page Start: 1359
• Page End: 1370
• Publication Date: 2017-06-08
• Subjects: species distribution models -- Maxent -- sampling bias -- habitat predictors -- rare species -- Festuca amethystina; Geobiology -- Periodicals; Biogeochemistry -- Periodicals; Biotic communities -- Periodicals; Geophysics -- Periodicals; 577.14
• DOI: 10.1002/2016JG003629
• Languages: English
• ISSNs: 2169-8953

Abstract

Abstract: Species distribution models need adequate sets of data, particularly in the case of range‐restricted species. The problem faced in the modeling of rare species is twofold: a small sample size and the occurrence of sampling biases. The present analysis combines spatial‐ and habitat‐thinning approaches to improve maximum entropy models based on geographically incomplete data of relict and subendemic Festuca amethystina L. grass on Polish territory. The results show that models based on strongly incomplete historic data did not predict the occurrence of all important areas where the species was found in the following decades. However, the introduction of species‐specific thinning allows for more precise prediction of the species range, i.e., the detection of suitable areas on a more local scale. The introduction of habitat thinning caused the diversity of important predictors in model to increase, but spatial thinning decreased the number of significant predictors and made interpretation easier. Additionally, a combination of thinning techniques allowed significant improvements to be made to the model predictions after the experimental addition of a lower number of localities to regions which had previously been poorly recognized. It can be concluded that in the case of incomplete data, the above corrections allow the true range of the species to be predicted after the discovery of a lower number and relatively dispersed new localities. Key Points: Habitat and spatial thinning were applied to the modeling of the environmental niche of the species and the distribution of suitable areas from incomplete data More precise prediction has been achieved; however, models based on very incomplete data did not predict the occurrence of all important areas of the species Corrections allow to more realistic predictions after the addition of a lower number and relatively dispersed new localities