Habitat suitability model for european wild rabbit (Oryctolagus cuniculus) with implications for restocking

Abstract

The European wild rabbit, Oryctolagus cuniculus, is a key species in the Iberian ecosystems. However, its populations are seriously depleted. Therefore, the development of habitat suitability models could be an important step towards the establishment of management and recovery plans. An habitat suitability model was developed for the rabbit in the 4,606 ha of the Peneda-Gerês National Park (northern Portugal). The framework was based on 3 steps: (1) construction of an habitat suitability model using the logistic regression of presence/absence of rabbits on habitat variables, (2) location of suitable habitat on a map, with the help of a Geographic Information System (GIS), by applying the above-obtained regression coefficients to the digitalized data of a land-cover map, and (3) delineation of the best potential sites for restocking. In June-July 2000, presence/absence of rabbits was determined by the presence/absence of pellets, warrens and individuals on 86 100-m radius random circular plots, i.e. 3.14 ha each representing the area occupied by a family group in summer. The initial dataset was divided into 56 sampling plots to build the model and 30 sampling plots (15 with rabbits and 15 without rabbits) for cross-validation. Measured habitat variables were % cover of tall scrubland, % cover of rocks, and interspersion (an index measuring the amount of intermixture of forage and shelter patches). These variables were calculated for all pixels of a digital land-cover map, using a moving 100-m radius window approach. Plots on which wild rabbits were present had a mean 10+3.4%(SE) percentage of rocks, 16.3+3.8% percentage of tall scrubland and an interspersion index of 0.9+0.09 (vs 1.3+0.7, 2.7+1.5 and 0.4+0.05 for “absence” plots, P < 0.05). The logistic model revealed an overall prediction success of 86.7%. The percentage of true correctly-predicted positives was 93.3% and the percentage of true correctly-predicted negatives was 80.0% (cut-off level at p = 50%). Model results showed that 34.5% of the area provided suitable rabbit habitat (P > 50%), distributed over 106 patches. Three clusters of suitable patches were delineated, assuming a critical threshold distance of 200 m between neighbour patches (distance equal to the mean dispersal ability of rabbits) for their inclusion into the same group. In that way, an area of 508.7 ha (11% of the study area) including the best potential site for restocking was delineated. Results suggested that management efforts should be aimed at promoting the connection between the three habitat-patch clusters. The framework presented provides a method for accurate and rapid assessment of habitats suitable for wild rabbit and could assist managers in identifying the best potential restocking sites.