There are no established methods for comparing model performance across different geographical regions, species or data sets.
Species distribution modeling (SDM) is a big field. Yet, there are no established methods for model identification and robust model assessment. It is in fact hard to find guidance about best practices in model development and model testing for SDM. We are in search for colleagues who are willing to pick-up the callenge to prepare a good data set and run a method comparison project for SDM. We already have some protocols prepared. If you are interested - send us an email: mecoco.project@gmail.com