Comparison of approaches for modelling submerged aquatic vegetation in the Toronto and Region Area of Concern

Abstract

In freshwater aquatic ecosystems, submerged aquatic vegetation (SAV) is critical habitat for may fish species and provides a variety of ecosystem services including nutrient filtration and substrate stabilization. Characterizing habitats and assessing their suitability for fish and other aquatic and terrestrial organisms is an important component of delisting efforts in the Toronto and Region Area of Concern (AOC). The primary objective of this study was to develop a spatial model for SAV within the AOC. A variety of modelling options were explored with a two stage random forest model identified as the most accurate approach; a two stage boosted regression tree model yielded comparable accuracy but was more complicated and processing intensive to implement. The final models for presence (modelled first) and SAV percent cover (applied only where the presence model predicted SAV to occur) incorporated directionally weighted wind fetch, water depth, and clarity (Secchi depth) with relatively high predictive accuracy (87.1% for presence). When applied across the AOC, SAV was primarily found to occur within the Central Waterfront, particularly adjacent to and among the Toronto Islands. Outside of this area, SAV was generally sparse and confined to areas that were protected from wind and wave action from Lake Ontario. Future habitat creation and remediation efforts should therefore focus on creating habitat conducive to SAV establishment.