The recovery of the otter in the Cevennes (France) : a GIS-based model

The Eurasian otter (Lutra lutra) is an elusive semi-aquatic mammal. Its absence is difficult to prove and censuses can be biased by false-negative observations. A model is proposed to correct this bias by predicting occurrence probabilities where observations are negative. In case of otter signs observation in one place, the probability to find the species in another place is assumed to decrease if the distance between both places increases. This approach is tested on a 15-years otter survey dataset (4,592 data) from the National Cevennes Park (3,200 km2, France). Four methods are compared, using Euclidian distances, decisional distances, cost-weighted distances and basin analyses. The validation is performed on 61 stretches that were prospected each year in the Cevennes. For these stretches, if observations were negative during a year A but positive during the previous and the following year, observations were assumed to be false-negative during this year A and compared to predictions given each modelling method. The modelling proves the interest of describing the landscape and defining spatially weighted distances from places where otter is present to predict the species occurrence where its search is unsuccessful. The results highlight that, despite the high mobility of the otter, its distribution is limited by some watershed limits. Hydrographical basins represent a relevant spatial unit to predict the otter occurrence and to analyse its recovery.