Nature Conservation Research: Заповедная наука (Nov 2019)
Brown bear (Ursus arctos) habitat suitability and distribution modelling in the southern taiga subzone using the method of maximum entropy
Abstract
The article presents results of the brown bear habitat suitability and distribution modelling, conducted in the Central Forest Nature Reserve and its buffer zone (West-European Russia, Tver region) using the MaxEnt approach. The basic rules for performing such a study, approaches and modelling techniques were reviewed briefly. Vegetation indices, morphometric characteristics of the relief, proximity rasters and land cover types were used as predictors. The occurrence points of the studied species were recorded on permanent routes using a GPS navigator during 2008–2018. Eight models with different combinations of input data (occurrence points and environment parametres) were chosen as final set. We used two main modelling approaches. In the first approach, for modelling we used only occurrence points reflecting the relationships of the species with the habitat (feeding places mainly). On the basis of them, habitat suitability models were built. In the second approach, we used all occurrence points of the studied species. On the basis of them, distribution models were built. The set of brown bear occurrence points registered by forest rangers and researchers was used as an independent test data set. The scenarios of the influence of anthropogenic food sources (abandoned apple orchards and oat fields) were modelled separately. The obtained AUCtest values ranged from 0.61 to 0.73. The maximum TSStest was 0.50. The continuous Boyce index ranged from 0.63 to 0.99. The models correctly recognised from 68% to 82% of independent points. The predictor of anthropogenic food sources largely contributed to all models, where it was presented, and highly distorted the overall picture of suitability and distribution. In other cases, grasslands, NDVI, and young deciduous forests had the highest contribution. In the study area, brown bears preferred grasslands, concentrated on moraine-kama ridges, which provided them with food throughout the wakeful period, as well as forest glades, scarce forests, young deciduous and mixed forests with dense undergrowth and nemoral spruce forests. The study area of the partially disturbed buffer zone was more suitable for brown bears than the intact areas of the Protected Area (66–67% and 51% of suitable habitats, respectively). We identified main omissions in the applied method of occurrence point's registration. This could lead to incorrect estimates of the contribution of some predictors (underestimations of boreal spruce forests, raised bogs and floodplain meadows).
Keywords
