Model-R: A Framework for Scalable and Reproducible Ecological Niche Modeling

Spatial analysis tools and synthesis of results are key to identifying the best solutions in biodiversity conservation. The importance of process automation is associated with increased efficiency and performance both in the data pre-processing phase and in the post-analysis of the results generated by the packages and modeling programs. The Model-R framework was developed with the main objective of unifying pre-existing ecological niche modeling tools into a common framework and building a web interface that automates steps of the modeling process and occurrence data retrieval. The web interface includes RJabot, a functionality that allows for searching and retrieving occurrence data from Jabot, the main reference on botanical collections management system in Brazil. It returns data in a suitable format to be consumed by other components of the framework. Currently, the tools are multi-projection, they can thus be applied to different sets of temporal and spatial data. ModelR is also multi-algorithm, with seven algorithms available for modeling: BIOCLIM, Mahalanobis distance, Maxent, GLM, RandomForest, SVM, and DOMAIN. The algorithms as well as the entire modeling process may be parametrized using command-line tools or through the web interface. We hope that the use of this application, not only by modeling specialists but also as a tool for policy makers, will be a significant contribution to the continuous development of biodiversity conservation analysis. The Model-R web interface can be installed locally or on a server. A software container is provided to automate the installation.