A Naive Bayes classifier for modeling distributions of the common reed in Southern Finland

Abstract

The field of species distribution and habitat suitability modeling has witnessed significant advancements in a number of aspects. One area that received much attention is the statistical underlying models in these studies. As data becoming bulky and prediction is often the goal of modeling, Data Mining and Machine Learning methods are becoming favorable in providing the underlying probability models for species distribution studies. Machine Learning encompasses a wide range of classification techniques, among others, with various capabilities. Although a number of techniques were presented and applied in species distribution modeling, many remain still untested. We here examine the potential of the Naive Bayes classification method, a widely and successfully applied technique in a number of fields, for modeling the common reed Phragmites australis distributions. We developed a Naive Bayes classifier to predict occurrences of Phragmites australis in a site on the Southern Finnish coast. We also tested the potential of the classifier to provide input to a cellular automaton for modeling the spread of Phragmites australis. The results suggests that the Naive Bayes classifier has significant potential in predicting species occurrences and providing transition rules for the dynamic modeling of species distributions.