Applying geographic profiling used in the field of criminology for predicting the nest locations of bumble bees

Abstract

We tested whether geographic profiling (GP) can predict multiple nest locations of bumble bees. GP was originally developed in the field of criminology for predicting the area where an offender most likely resides on the basis of the actual crime sites and the predefined probability of crime interaction. The predefined probability of crime interaction in the GP model depends on the distance of a site from an offender's residence. We applied GP for predicting nest locations, assuming that foraging and nest sites were the crime sites and the offenders’ residences, respectively. We identified the foraging and nest sites of the invasive species Bombus terrestris in 2004, 2005, and 2006. We fitted GP model coefficients to the field data of the foraging and nest sites, and used GP with the fitting coefficients. GP succeeded in predicting about 10–30% of actual nests. Sensitivity analysis showed that the predictability of the GP model mainly depended on the coefficient value of buffer zone, the distance at the mode of the foraging probability. GP will be able to predict the nest locations of bumble bees in other area by using the fitting coefficient values measured in this study. It will be possible to further improve the predictability of the GP model by considering food site preference and nest density.