A comparison between GARP model and SVM regression to predict invasive species potential distribution: the case of Miconia calvescens on Moorea, French Polynesia

* Corresponding author. ** This work was partially supported by the Research Department of the Government of French Polynesia. Abstract Biological invasion is one of the main drivers for global biodiversity loss. In the islands of French Polynesia, the worst invasive alien species is arguably the small tree Miconia calvescens. The knowledge of miconia distribution and dynamics is critical for monitoring and control efforts. However, classical direct remote sensing methods are vain because miconia invades rainforest understory. In this study, we introduce the possibility to use a Support Vector Machine (SVM) regression for ecological niche modelling and we compare it with the commonly used Genetic Algorithm for Rule-set Production (GARP). Both models integrate several environmental layers extracted from a Digital Elevation Model (DEM) such as elevation, slope, aspect, windwardness and wetness. SVM regression achieves an overall accuracy of 92.5% and outperforms significantly the GARP model. These results may be explained by the ability of SVM regression to integrate heterogeneous data and to be trained on small sets of pixels. We suggest that SVM regression can be used as an effective tool to predict invasive (but also native) species potential distribution.