Exploring alternatives
Trimming of ERMs by landcover and elevation is regularly promoted as a means to trim ERMs, but it is unknown if simple elevation and landcover trimming correct biases effectively. We tested diversity patterns generated via original ERMs versus those from analysis of bat point data with and without trimming and with published models (Hughes 2017). Point data were clipped for Eurasia and minimum convex polygons (MCPs) created in ArcMap for species with at least five points. Filters were created for each species based on elevation and landcover, both using IUCN assessment data exclusively and that based on extracting environmental data from points, and these were then paired with associated environmental data to clip species range on a per species basis.
We used point data to extract elevation from 1km-resolution dem, with min, max, mean and standard deviation per species from summary statistics. Species exclusively <1000m=lowland, 1000-2000m=mid, >2000m=high, between these ranges ranked accordingly: lowland, low-mid, low-high, mid, high. DEMs were reclassed to corresponding elevation bands. IUCN assessment listings of elevational preference were recorded. A “integrated” status was determined based on comparing the point-based with IUCN-based assessments (when species were assessed in IUCN and had sufficient point data): where only one assessment was given it was retained, where the two agreed it was retained, and where they differed we used the point-based data given higher precision and transparency.
For habitat intactness, we collated IUCN assessments and data extracted from point data. For IUCN assessments we used keywords to assay disturbance tolerance. Habitat listings which referenced roosting in buildings, houses, tunnels were assigned as generalists. Species listed in cultivated areas, paddies, plantations, agriculture were assigned as semi-intact and those listing forest and no other “disturbed” habitats assigned as intact. For point data we classified population layers to under 50 people per kilometer as intact, 51-100 as semi-intact and over 100 per km as generalist. From point data species with over 50% of localities in the generalist category were listed as generalists, and species with at least 75% of records in the under 50 people were classed as intact. The IUCN and point generated categories were then compared, where the two categories differed we selected the “final” classification based on further searches or actual experience with the species listed.
For richness mapping, we joined the elevation field based on species names, split into five elevation categories, each of which was then clipped by a polygon layer of the appropriate elevation bands and merged. This was repeated for the MCP layer and ERM layers. The ERM layer was run twice, for the “integrated” assessment data using the “integrated” category, and once for IUCN elevation assessments. These were then merged to form three species elevation trimmed species collations (one MCP, two ERM). Layers were then joined to intactness categories, and split into three categories prior to trimming with the appropriate intactness filter (intact, semi-intact, generalist). These were then merged before using the count overlap toolbox to count the number of species overlapping in any given area. This enabled comparison of trimmed and untrimmed layers to a previously published Maxent layer (Hughes 2017) to assess how useful these alternate approaches are.