Inundation-driven community transitions in saltmarsh environments: Insights from multi-scale classification models

This novel study presents vegetation community maps for saltmarshes along the Petites-Bergeronnes River, Québec (Canada), derived from supervised classification models of varying spatial and temporal resolutions. Random forest classification was performed using medium-resolution (3 m) PlanetScope satellite imagery, incorporating 12 images from the snow-free season (May-October 2023). High-resolution (0.2 m) multispectral imagery from the Government of Québec, Canada, supported a single-date (August 24, 2023) classification for the same region. This marks the first temporal feature approach to saltmarsh classification in the St. Lawrence Estuary, and the first direct comparison of community-level saltmarsh classification contrasting high spatial- versus high temporal-resolution. For the first time, tidal inundation dynamics were explicitly incorporated during classification, using inundation time as a predictor variable. This improved overall classification accuracy to 83% for the temporal imagery model and 96% for the high-resolution model. Moving beyond regional mapping, community extents were compared with tidal inundation data to identify thresholds for eco-geomorphic transitions and develop stress functions for saltmarsh evolution modeling. A region-independent threshold of ∼11% inundation time marked the majority transition from pioneer to high marsh vegetation. Mudflat-to-pioneer transitions were region-dependent (16-29% inundation time), indicating that the lower elevation limit of establishment is not solely dictated by inundation stress. As the first inundation thresholds identified independently of tidal datums for saltmarshes dominated by S. alterniflorus, S. pumilus, and P. maritima, these findings represent a significant advancement for saltmarsh eco-geomorphic studies across diverse geographies.