Validation of Area of Habitat Maps for Terrestrial Mammals

Fellow Prabhat Raj Dahal gave a talk entitled Validation of Area of Habitat Maps for Terrestrial Mammals on 17 June 2021 during the symposium III of the online 100th Annual Meeting of the American Society of Mammalogists.

You can find the programme here.

[Oral presentation] Prabhat R. Dahal, Maria Lumbierres, Stuart H.M. Butchart, Paul F. Donald, Carlo Rondinini (2021) Validation of Area of Habitat Maps for Terrestrial Mammals, Symposium III Global Trends in Mammals, 100th Annual Meeting of the American Society of Mammalogists, online.

Abstract: Area of Habitat (AOH) maps show the distribution of suitable habitat/s of a species inside the geographical and elevational limits of the species. We validated a new set of AOH maps for terrestrial mammals which was produced by using a novel habitat – landcover model. Initially we identified errors in the AOH maps using a logistic model which identified poorly performing models as outliers. The AOH maps were then validated at species level using available point locations of mammal species downloaded from GBIF (Global Biodiversity Information Facility). We selected species with at least 10 points falling inside the geographical range of the species and having coordinate uncertainty of less than 300 m dated between 2019 – 2020. We had point data for 447 mammal species for validation. These points were buffered by 300 m and overlaid on top of the AOH maps to calculate ‘point prevalence’, defined as the proportion of points or their buffers falling in pixels of suitable habitat out of all the point localities of the species in question. We also evaluated ‘model prevalence’ which is the probability of a randomly positioned point falling in the suitable habitat pixel of this AOH. If point prevalence exceeds model prevalence, we assume that the AOH is a better than random reflection of the species’ distribution within its range. We compared the model and point prevalence of 447 mammal species out of which 424 were better than random.