Wetlands are crucial ecosystems which provide several functions, beneficial both to human beings and to the environment. Despite such importance, quantitative approaches to many aspects of wetlands are far from being adequate, above all the interaction between rainfall, vegetation, soil moisture and groundwater depth. Starting from a previously developed model for below-ground stochastic water level fluctuations, we extend it to consider the case of waterlogging. The extended model is now suitable for describing the long-term probability distribution of water table depth in temporarily inundated wetland sites, whose hydrologic input is dominated by stochastic rainfall. The extended model performs well when compared to real data collected in the Everglades National Park (Florida, US), confirming its capability to capture the stochastic variability of wetland ecosystems.