Skip to main content

New models for wild ungulates occurrence and hunting yield abundance at European scale

EFSA Journal logo
Wiley Online Library

Meta data

Disclaimer: The present document has been produced and adopted by the bodies identified above as authors. This task has been carried out exclusively by the authors in the context of a contract between the European Food Safety Authority and the authors, awarded following a tender procedure. The present document is published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.


The goal of this report is i) to model the occurrence and hunting yield (HY) density of wild ungulates not only for widely distributed species in Europe, but also for those ones which have a constrained distribution and ii) to compare the output of occurrence with observed HY. Random Forest function was used for modelling occurrence of species. We used occurrence data available from the past 30 years, and HY data (period 2015‐2020) from records collected by ENETWILD. Like previous models based on HY, the response variable was the maximum number of wild ruminants annually hunted in 2015‐2020 hunting seasons divided by the area (km2) of the corresponding administrative unit (HY density). Models based on HY were statistically downscaled to make predictions to 10x10km squares. Occurrence data models indicated a good predictive performance for most species, showing that the model framework proposed have improved results in comparison to previous models. The transferability of models into new regions was limited by the exposure of species to environmental conditions. As for HY models, the calibration plots showed a good and linear predictive performance for widely distributed species, as well as constrained distributed species. Overall, our results were consistent with the expected abundance distribution of widely distributed species. The removal of zeros on the validation datasets affected the calibration plots of all regions, showing a better predictive performance when zeros were removed for widely distribution species, but the opposite was evidenced for species with limited distributions. We conclude that (i) the importance of co‐correlation variables when variable importance is inferenced from random forest model results, (ii) manipulation presence and absence locations could yield further improvement in occurrence model outputs, and (iii) HY model projections displayed good abundance patterns for most of species, showing that the three frameworks proposed were a good approximation for modelling the distribution of wild ungulates HY, although it should be explored how to improve the results when distribution is patchy.