A two-step species distribution modeling to disentangle the effect of habitat and bioclimatic covariates on Psacothea hilaris, a potentially invasive species

Abstract

Integrating host plants in distribution modeling of phytophagous species and disentangling the effect of habitat and bioclimatic variables are key aspects to produce reliable predictions when the aim is to identify suitable areas outside species’ native range. To this aim, we implemented a framework of Species Distribution Model aimed at predicting potential suitable areas of establishment for the beetle Psacothea hilaris across the world. Since habitat (including host plants) and bioclimatic variables affect species distribution according to processes acting at different geographical scales, we modeled these variables separately. For the species native range, we fitted a habitat (HSM) and a bioclimatic (BSM) suitability model calibrated on a local and a large scale, respectively; the overall suitability map was obtained as the spatial product of HSM and BSM projection maps. ROC, TSS and Cohen’s Kappa obtained in validation confirmed a good predictive performance of modeling framework. Within HSM, host plants played a substantial effect on species presence probability, while among bioclimatic variables, precipitation of the warmer quarter and isothermality were the most important. Native HSM and BSM models were used to realize an overall suitability map at world scale. At global scale, many areas resulted suitable for habitat, some for bioclimate, and few for both conditions; indeed, if the species would not be able to modify its bioclimatic niche, it might not be considered a major invasive species. However, the high rate of range expansion documented for P. hilaris in Northern Italy, a poorly suitable bioclimatic area, suggests a plasticity of the species that requires increasing the level of attention to its invasive potential.