Predicting the risk of wild boar damage to rice paddies using presence-only data in Chiba Prefecture, Japan

Abstract

Population sizes of wild boar (Sus scrofa) and the damage they cause to crops have been increasing in Japan. Reliable techniques are needed to estimate the potential for damage at the landscape scale. Here, we predict the risk of damage to rice (Oryza sativa) paddies by wild boar in Chiba Prefecture, Japan, by means of three different modelling methods – Maxent, generalised linear model (GLM) and generalised additive model (GAM) – using a combination of presence-only damage data obtained from a local agency and environmental information derived from publicly available databases. We used damage locations in 2007 and 2008. To validate the models, we calculated the area under the curve (AUC) and the correct classification rate (CCR) using independent data obtained in field surveys. The three methods gave similar results, indicating that we could construct a predictive model with high accuracy from presence-only data. Among these three, Maxent showed the closest fit, with 0.78 AUC and 72.6% CCR values. Because it is important to estimate the risk of damage to reduce future damage and costs, these damage prediction methods using presence-only data, which administrative agencies can obtain with no cost, this could assist local governments in formulating damage control plans.