Distribution and factors affecting wild boar (Sus scrofa) damage in a lowland area in northern Italy

Damage to croplands by wild boar is globally taking dramatic proportions and causing considerable economic losses. Conflicts between wild boars and agricultural activities show themselves in different ways: damage, in fact, can be caused by rooting looking for food (invertebrates, bulbs or tubers), by direct feeding on whole plants or vegetative parts, and by trampling [1]. In Italy, the wild boar has recently colonized intensively cultivated lowlands, including western Po Plain. In this work, we first provide a general description of wild boar damage to croplands in “Lomellina”, an intensively cultivated portion of the western Po Plain (province of Pavia, northern Italy). We then used these data to build a distribution model, to identify which factors determine the distribution of damage and to define the areas where the risk of damage is highest. Wild boar damage data were collected over a period of 3 years, from 2013 to 2015. Two different types of data were used: requests of refund arrived at the Wildlife Service of the Province of Pavia, and data obtained through direct meeting with landowners and/or farmers; thus, it was possible to identify which types of crops were sensitive to damage. To evaluate the existence of monthly differences in the number of events, a Chi-square goodness-of-fit test was performed. The areas with the highest presence of damage events were defined with a 50% Kernel Analysis. To build the model, we analyzed the relationship between damage and 11 different predictors: distance from water, distance from woodlands, distance from continuous hedges and rows, distance from discontinuous hedges and rows, distance from urban settlements, human population density, distance from primary roads, distance from secondary roads, distance from railways, area and fractal dimension of damaged fields. For modeling, we randomly selected the same number of pseudo-absences as there were presences. The model was built following a use-versus-availability approach, with a Binary Logistic Regression Analysis [2]. “Damage” was set as the binary dependent variable (damage locations = 1, random locations = 0), and landscape variables previously listed were set as predictors. The response variable was modelled for dependence on predictor variables using the model selection procedure based on the Akaike Information Criterion (AIC) [3]. For all possible models, we calculated AIC values corrected for small sample size (AICc). Models were ranked and scaled by the differences with minimum AICc (ΔAICc) and Akaike weights (ωi) for each i-model [4]. The relative importance of predictor variables (ω) was measured by the sum of Akaike weights of the models in which each variable appeared [4]. The model containing all the variables with a ω value ≥ 0.50 was considered the best one [5]. Damage events almost exclusively involved maize, and, a lesser extent, rice, sorghum, wheat and soya bean. In all years, monthly damage distribution was not random (Chi-square test, P always < 0.05), with a peak in May and a minimum in late autumn and in winter. From the spatial analysis of damage distribution, we observed a different distribution of damage among years; in 2013 damage occurrence was mainly concentrated in the Eastern part of the study area, bordering the “Ticino Valley Regional Park”. In 2014, the distribution of damage was quite uniform in the whole study area, whereas in 2015 the highest proportion of damage was recorded in the Western part of the study area, coinciding with the Special Protection Area “Risaie della Lomellina”. The Binary Logistic Regression model had an excellent predictive power (AUC = 0.96). The risk of damage was highest in fields close to woodlands (which ensure the availability of cover), far from urban settlements and primary roads, and where human population density is low. Furthermore, the model showed a positive effect of the distance from continuous hedges and rows. The geometry of the fields did not have a significant effect on the risk of damage, probably because of the high homogeneity of field structure in the study area. The analysis of boar-induced damage, together with the identification of factors that increase the risk of damage, provides information contributing to the development of an effective plan for managing wild boar populations, which could be important not only to prevent damage to croplands, but also to reduce any adverse effect that boars have on physical and biological components of ecosystems.