Modelling biodiversity in the Grand River Watershed using a species distribution prediction approach
This research examined and evaluated the use of GARP (Genetic Algorithm for Rule-Set Prediction) in combination with GIS (Geographic Information Systems), to effectively model biodiversity on the sub-watershed scale in Southern Ontario. To accomplish this, 11 environmental layers (based on topography, land cover, and climate) and georeferenced point data on 29 nationally ranked species at risk (plants, birds, reptiles, and amphibians) were obtained. Results indicated that five layers, (elevation, average summer precipitation, average annual temperature, average January temperature, and average July temperature) produced the most accurate results predicting species distributions (93.4%, 88.2%, 91.6%, 87.8%, and 92% accuracy, respectively). Using the spatial statistic measure AUC (Area Under Operating Characteristic Curve), an accuracy of 0.92 (max 1.0) using the best five layers combined in one model was reached compared to 0.91 using all 11 layers. Within the study area, a hotspot of biodiversity was projected stretching from Cambridge to approximately 15 km south of Brantford. These results aid in understanding the critical layers in prediction modelling, saving time and resources for future researchers in this area. Identifying the most biodiverse regions in the study area provides valuable information for land managers, as these areas often receive the highest priority for conservation. Lastly, this modelling approach can be applied to determine how species distributions will shift in response to climate change.