By Markov simulation modelling we analyzed an almost continuous monitoring series over 27 years of the abundance of the highly endangered amphibious plant species Myosotis rehsteineri at the shore of Lake Constance (Germany). The habitat is characterized by yearly floodings of variable magnitude that produce a zonation of species with different flood tolerance. A transect was established in 1983, perpendicular to the shoreline, and divided into segments (2 x 2m) of different flood duration. Population sizes, expressed as the number of inflorescences, were determined per segment and year. Additionally, we had detailed data on flooding and knowledge on the dynamics of competing species. A Markov model fitted by nonlinear least squares against the time series of spatially structured abundance data proved to be a useful tool to evaluate population dynamics and to assess the effects of varying environmental conditions that are not accessible to experimental control. Water-level fluctuations and competition by Agrostis stolonifera were both found to affect the population dynamics of M. rehsteineri, but without causing a long-term change in population size under present conditions. The modelling enabled us to evaluate different flooding scenarios and thus providing support for future conservation measures, and we found that extinction risk will increase under changing flooding regimes. Our data and the results of the Markov model simulations showed that under strong abundance fluctuations long-term monitoring is indispensable to explore population dynamics of rare and 38 endangered species in their natural context.