Common Land Model (CLM) and Land Surface Process (LSP) model simulations are compared to measured values for a 13-day dry-down period with a rapidly decreasing near-surface water table for a marsh wetland community in Florida. LSP was able to provide reasonable estimates without any modifications to the model physics. To obtain reasonable simulations using CLM, the baseline TOPMODEL baseflow generation and the bottom drainage mechanisms were not employed and the lower layers were allowed to remain saturated. In addition, several of CLM’s default wetland vegetation parameters were replaced with grassland parameters. Even after these modifications, CLM underestimated soil water storage. However, both model-simulated soil temperatures showed very good agreement as compared to measured temperatures, capturing both the soil warming during the study period and the diurnal fluctuations. Modeled surface energy fluxes also agreed well with measured values. LSP’s inability to consistently capture latent heat fluxes appears to be linked to its canopy resistance scaling functions. Other minor issues were that CLM’s rooting depth greatly exceeded observed depths and that CLM did not move water in the vadose zone from lower to upper layers during the nighttime as observed in the measurements. Overall, these results suggest that LSP can be applied to characterize a marsh dry down, but that minor modifications could greatly improve results. CLM demonstrated considerable potential, but requires some changes to model physics and default parameters prior to application to wetlands at a subgrid scale.