A comparison of Maxlike and Maxent for modelling species distributions

摘要: Summary Understanding species spatial occurrence patterns and their environmental dependence is one of the fundamental goals in ecology evolution. Often, models are built with presence-only data because absence unavailable. We compare strengths limitations recently developed modelling method, Maxlike, more widely used Maxent. In spite disparities highlighted by developers Maxlike Maxent, we show approximate formal relationships between parameters Maxent for two scenarios to illustrate similarity. Using case studies based on real simulated data, how these similarities manifest practice. We find than differences including coefficient values, predicted distributions, similarity presence–absence models, predictive performance ranking suitability cells. reliably absolute probabilities very large sets landscapes where probability approximately spanned [0,1]. For smaller sets, uncertainty was large, due inherent data. In contrast, constrained predicting relative or rates unless it provided additional information from Both can predict probability. The choice which model use depends partly sampling assumptions, discuss detail. Due ecologists should typically focus interpretations relying rates. remedy a number concerns about avoid some potential pitfalls – particularly related high variance predictions. conclude that both methods similarly valuable understanding species’ distributions terms when specified carefully.