Long-Term Assessment of Cattail (Typha spp.) Dynamics and Wetland Restoration in the Everglades Using Object-Based Landsat Time Series and Markov Modeling

Abstract

Assessing long-term vegetation response to restoration requires spatially consistent monitoring approaches that capture both observed change and likely future trajectories. In the Florida Everglades, nutrient enrichment contributed to the expansion of cattail (Typha spp.) within Water Conservation Area 3A (WCA-3A), raising concerns about restoration effectiveness. This study integrates a multi-temporal Landsat time series (2004–2024), object-based image analysis (OBIA), support vector machine (SVM) classification, and Markov transition modeling to quantify and project vegetation change in WCA-3A. Object-based segmentation reduced classification fragmentation in heterogeneous marsh landscapes, while SVM improved separation among spectrally similar vegetation classes. Classification accuracies remained consistently high across sensors (overall accuracy 90–93%; kappa 0.80–0.87). Cattail cover peaked at over 11% in the late 2000s but declined to 1.6% by 2024, coinciding with the expansion of native sawgrass and open marsh communities. Transition probabilities derived from object-level change detection were used to project vegetation patterns to 2034 under continued management conditions, indicating sustained suppression of cattail dominance if current controls persist. By directly linking object-based classification with probabilistic forecasting, this study advances beyond static land-cover mapping and provides a transferable framework for evaluating restoration trajectories in large managed wetlands.