<p>Water-level fluctuations drive plant species composition and abundance in freshwater wetlands. Receding levels stimulate seed-bank germination and increase plant diversity, while increasing levels drown flood-intolerant species, reducing diversity. Earlier short-term studies have quantified aspects of these dynamics but have inadequately incorporated long-lived perennial emergent species with dense rhizome networks. Long-term studies of natural wetlands undergoing multiple water-level cycles are necessary to clarify plant responses. We examine a 50-year vegetation dataset from Cecil Bay on northern Lake Michigan that included the full range of water-level fluctuation observed since 1918, focusing on the emergent marsh zone. These data are well suited to test hypotheses about flood tolerance of coastal wetland species, species response to changing water levels, and long-term wetland stability across water levels. Our analyses reveal that the entire marsh maintained a surprisingly stable breadth throughout the 50 years, with consistent emergent composition defining the outer wetland. Characteristic species defined the emergent communities, while three common emergent plants were found throughout the wetland. Notably, analyses demonstrated that the conventional model of vegetation change with water-level fluctuation underestimated tolerance of rhizomatous emergent plants to long-term flooding, a capacity critical for maintaining coastal wetlands. We present the <i>Persistent Rhizome Recurrent Seed Bank Model</i>, a new conceptual model that incorporates our findings.</p>