Useful, but can be improved: Methodological Advances for Niche Modeling and Species Distribution Modeling
ecological signal, collection bias, model evaluation, variable importance.
Ecological Niche Modeling and Species Distribution Modeling (ENM and SDM) are widely used in various areas of knowledge that aim to understand niche and species distribution features. The methodological and conceptual development to model the niche and species distribution occurred quickly throughout their brief history, which began in the 1990s, aiming to incorporate more ecological significance and precision to the prediction maps of the models. Therefore, searching for more accurate algorithms became inevitable, as well as for more accurate evaluations. However, the modelers are not unanimous in recommending the methodological and theoretical framework, and this generates inaccuracies in the construction and interpretation of models. In this sense, we still need to move forward conceptually and methodologically, and we propose to achieve this goal of making the ENM and SDM more reliable. For this we seek to answer the following question: Can we build and interpret more robust ENM and SDM, with valid ecological interpretations in their multiple applications? We believe that we indeed can, and we propose new methodologies to: 1) Identify processes, with ecological significance, to better understand the species attributes; 2) Evaluate and propose alternatives to provide a most accurate model evaluation; 3) Identify the effect of the collection bias on the predictive performance of the models and propose solutions; and 4) Propose a protocol, probabilistically based, to identify the most important variables for plant modeling, providing useful ecological interpretations.