Titolo della tesi: Optimizing modelling tools to anticipate biological invasions – on and beyond Species Distribution Models
Invasive Alien Species (IAS) are non-native taxa that cause ecological and/or economic damage. They are one of the primary causes of loss of biodiversity and ecosystem services, though they can be managed. However, conservation efforts are often short on manpower and resources. It is therefore crucial to computationally model the potential distributions of IAS. This thesis focuses on optimizing existing modelling techniques for this purpose. One such technique is known as Species Distribution Modelling (SDM), which works by relating the environmental conditions where the species occurs to those where it does not, or, more commonly, to those at randomly sampled locations, which are known as background points (BP). This allows SDMs to simulate the realized niche of and habitat suitability for a species.
Because non-native species can become invasive without our knowledge, the first chapter of this thesis compares which environmental variables most strongly predict the distributions of non-native species and their native congenerics. Subsequently, as SDMs quantify an environmental niche, we designed novel ways of environmentally sampling these variables using BP and compared them to the typical geographic way in terms of SDM performance. The clues gathered in the previous two chapters were then applied in the third to a well-known IAS, Heracleum mantegazzianum Sommier & Levier. Finally, the probability of invasion success in a new area was studied using the Niche Margin Index (NMI), which has been shown in previous studies to be positively correlated with establishment probability and measures the (dis)similarity between the native climatic niche and the local environmental conditions. We researched if different ways of quantifying the native niche affected the NMI values.
It was found that non-native species are more strongly associated with anthropogenic environmental variables and less pris ne land cover types than their native counterparts, though climate was the strongest predictor in both species groups. Sampling BP stratified randomly and fully randomly in environmental space yielded more stable and accurate SDMs than the classic method. The use of environmentally stratified BP may facilitate SDM over ng, however. Finally, the highest NMI values were obtained when the native niche was quantified using whole maps of the native range, followed by those obtained when a thresholded SDM of the native range was used and finally by those obtained using only native occurrences.
These findings provide novel guidelines and considerations for using SDMs and analyzing the probability of IAS establishment in order to manage biological invasions.