Predicting range shifts of pikas (Mammalia, Ochotonidae) in China under scenarios incorporating land-use change, climate change, and dispersal limitations

Two of the most important forces affecting biodiversity are land-use change (LUC) and global climate change (GCC). Previous studies have modeled their impacts on species separately and together, but few have done so for multiple species with dispersal limitations incorporated into the models.

We integrate species distribution models plus a dispersal model to predict LUC and GCC impacts on the ranges of five species of pikas in the Qinghai-Tibet Plateau region of China. Pikas are sensitive to land-use and climate change, and have limited dispersal abilities.

The predicted impacts of LUC and GCC on pikas vary between species as well as between LUC and GCC projections. Incorporation of dispersal limitations appreciably restricts the amount of colonized habitat. For all five species, the amount of habitat abandoned or colonized when LUC and GCC are modeled together is less than the sum of LUC and GCC modeled separately. Three of the five species experience a net increase in occupied habitat by 2080 relative to their current ranges under all modeled projections. However, relative to a “Dispersal Only” baseline scenario that assumes no environmental change but continued range expansion into suitable, unoccupied habitat, all five species suffer a net loss of occupied habitat by 2080 under some or all projections.

Predictions of future distributions of species based solely on LUC or GCC, as well as predictions assuming additive impacts, can be misleading. Inclusion of dispersal limitations in models markedly alters predicted future distributions of species. The use of a “Dispersal Only” scenario provides a different and perhaps more accurate way to gauge net impacts to species. Future work should consider incorporating all these parameters to better predict the impacts of LUC and GCC on biodiversity.

Methods
The uploaded occurrence records were collected from online sources (Global Biodiversity Information Facility, iNaturalist, GenBank), museum collections, field records provided by scientists, and the literature. These occurrence records have been through the following quality controls: 1. To avoid potential spatial autocorrelation among occurrence records or uneven sampling, we thinned the records using the package spThin, with a thinning distance of 10-km (Aiello-Lammens et al. 2015); 2. We then removed occurrence records with potentially erroneous coordinates by checking: (a) if the coordinates fell within the species’ IUCN polygons; (b) if the corresponding elevation fell within the elevational range described by the IUCN Red List (after adding a buffer of 100 m); and (c) if the species’ habitat types as described in the IUCN Red List appear in that grid cell according to the LUH2 data (habitat types described in the IUCN Red List and in LUH2 were reclassified as forest vs. non-forest for matching purposes).