Species distribution modelling as a conservation tool: the case study of Vipera aspis in a highly exploited region

Vipera aspis is quite common in hilly and mountainous area of Northern Italy but is becoming rare in the Po plain with few exceptions along main rivers. The reasons for such a distribution seem related to the progressive reduction of suitable habitat due to human landscape modifications, rather than to specific ecological needs. In this scenario, the knowledge of the exact distribution of this species and the identification of the factors which still make some planitial areas suitable would be very useful for conservation strategies. So the aims of this work are: i) modelling a distribution map of the potential suitable areas to focus surveys efforts and spatial relations with protected areas; ii) identifying the most important factors which affect suitability. The study area was the plain area of Lombardy that covered about 14100 km2 and was defined by 300 m a.s.l. contour line. The 45 occurrence data came from various research projects from 2000 to 2009. Fourteen non-correlated potential predictors were included in the model: six bioclimatic variables (http://www.worldclim.org/); five environmental (land use; density of cultivated fields); three anthropic (fresh water level of fungicide and organochlorine, roads density). All the variables were resampled to 100 m spatial resolution. Potential distribution was modelled using the maximum entropy method, and performed by the software MAXENT 3.3.1 (http://www.cs.princeton.edu/~schapire/maxent/). We used the bootstrap procedure to generate ten different models each with 25% of localities as test sample. We then averaged single model results. The models fit well (mean training AUC: 0.981 ± 0.008; mean testing AUC: 0.942 ± 0.032). The suitable area at the 10th percentile threshold is 3.4% of the study area and is reduced to its North-Western part, with few and isolated spots elsewhere. Fifty-six percent of suitable cells falls within protected areas, and most of the remaining part is nearby. The most informative variables are land use and roads density, which together explain 56.1% of observed variance. Suitability is positively affected by woods and meadows, and negatively by anthropic structures and increasing roads density. Our results demonstrate that asp viper is a species at risk in the Po plain, due to habitat fragmentation and modifications caused by anthropic activities. The north-western part of the study area plays a key role in its conservation thanks to more natural conditions and to a higher protection level. It is fundamental to preserve this territory and to increase the connectivity among existing natural parks