A simplified modeling framework to define vegetation functional community distributions based on two limiting factors

Bianca Charbonneau, Oak Ridge Institute of Science and Education, Oak Ridge, TN, Candice Piercy, Environmental Lab, U.S. Army Engineer Research and Development Center, USACE, Vicksburg, MS and Todd M. Swannack, Environmental Laboratory, U.S. Army Engineer Research and Development Center, USACE, Vicksburg, MS

Background/Question/Methods:
Forecasting local plant community distributions is a topic of increasing concern globally surrounding climate change. Plant distributions can affect many ecosystem processes and habitat stability and resiliency in ecogeomorphic habitats, like coastal dunes. Modeling efforts surrounding habitat evolution and change are predicated upon prior knowledge of plant community distributions at simulation onsets. However, these data are often unavailable and limit model usability, making accurately predicting initial plant distributions necessary to establish accurate baseline simulation conditions. Towards this goal, a framework for accurately populating a grid landscape was developed drawing upon the theories of limiting factors, habitat suitability, environmental niche, and universal adaptive strategy. The model populates a simulation landscape of differing plant functional communities based on limiting factors across communities. Input data required is limited surrounding threshold values for limiting factors, which are used deterministically to define suitable habitat space for monospecific or heterospecific community assemblages. The model was tested by predicting coastal dune community distributions across seven US National Seashore landscapes and comparing these placement results to known vegetation distributions. The seashores span the US Eastern Seaboard and Gulf Coasts and thereby contain different species latitudinally that represent different plant functional types, with their distributions predicted to be controlled spatially by cell height and distance to the ocean.
Results/Conclusions:
The case studies revealed that using just two limiting factors had surprisingly high accuracy for predicting plant community distributions in a model domain, and the framework was generalizable across species representing different functional communities. This this was true for dune-builders regardless of the species identity reflecting the community type, Ammophila breviligulata, found in the Mid-Atlantic sites, or Uniola paniculate found along the Gulf coast and southern US eastern seaboard. Inaccuracies in placement primarily stemmed from errors of exclusion and inclusion both of which were affected by the presence and absence of topographic features and their extremes across the landscape. For example, indicating if a swale or certain functional community was missing from the landscape improved placement accuracy. However, model placement of communities naturally missing from a landscape may be indicative of their potential range should they become established. In general, errors of exclusion resulted in higher accuracy than errors or inclusion, with overall accuracy comparable to the accuracy of the Vegetation Mapping Program surveys themselves. This approach is simple, requiring limited data to establish plant community distributions and can be applied across habitats for integration into varied modeling frameworks.