Using a multi-season SAR-optical approach, we mapped historical changes in extents of invasive Phragmites australis and Typha spp. in wetland ecosystems of Lake Huron from circa 1998, 2010, and 2018. Field data on dominant vegetation cover, water depth, and presence of invasive plants from >500 locations were used to train supervised machine learning algorithm Random Forest. We applied hybrid change detection between the map years using radiometric and categorical data. Integrating radiometric change magnitude information, derived from a change vector analysis into traditional categorical change can improve accuracy by reducing commission error. Final hybrid-change layers were used to assess wetland change trends from 1998-2010 and 2010-2018 for the overlapping mapped extents and for coastal wetlands. Change results found from 1998 to 2010 there were decreases in most wetland types (> 8,000 ha) and high conversation to Phragmites (5000 ha increase). From 2010 to 2018, as water levels increased, the trend flipped with wetland areas increasing. Major trends included wetlands shifting type across hydrology zones (shrub to emergent, emergent to floating), wetland conversion to Phragmites, and Phragmites flooding out to open water. Linking our change analysis to the shifting zone of coastal influence allows us to more accurately quantify the gain, loss, and change dynamics in wetland ecotypes.
