SmarteR Approach for the Mapping of Invasive Plant Species

Abstract

Invasive species distribution has caught global attention due to its detrimental effects on biodiversity. Species that have become invasive often come from climatic conditions similar to those found in other parts of the world. Species distribution modelling (SDM), also referred to as habitat suitability mapping, ecological niche modelling, or climate envelope modelling can be used to map potential sites of invasion probability. Some of the major SDM techniques that fall under this category are the Bioclim model, Domain model, and Maxent model. A more sophisticated set of R tools has made it simpler to estimate a species’ habitat range based on its geolocation. For SDMs to be successful, it must be possible to accurately describe the distribution of the species in question. Based on climate data and other environmental criteria such as elevation, surface water, soil moisture, and land use as well as other human-induced variables, SDMs attempt to map anticipated species distributions or habitat suitability. We describe here two major approaches of classical SDM and machine learning techniques for mapping invasion. Machine learning comprises algorithms that do not rely on rule-based programming; instead, it learns from the data. It is increasingly being used to build predictive habitat suitability maps using binary response data and environmental predictors. Various models that utilize machine learning include Generalized linear model (GLM), Support vector machines (SVM), Gradient boosting machine (GBM), k-nearest neighbor (kNN), and Random Forest (RF). This chapter will discuss various modelling techniques with practical examples along with model evaluation using R language and platform.