Fifty years of community change in a Chihuahuan Desert ecosystem: using the U.S. National Vegetation Classification to link the past to the future

Background/Question/Methods

Uncertainty in plant community dynamics models can arise not only because of insufficient information on ecosystem behavior, but also because the entities themselves—the plant communities—may be ill defined.  This is particularly acute in semi-arid ecosystems where we have less history of vegetation classification.  For example, arid-land state-transition models depict changes of community types over time (changes of state) as driven by specific processes, but how those communities are defined is often ad-hoc and relative to a particular locale or project goal.  As a result, they may potentially be idiosyncratic and not generalizable over a broader area.  We suggest that if datasets are linked to well-defined vegetation units such as those recognized in the U.S. National Vegetation Classification (USNVC) then modeling outcomes about the future will be more robust and more consistently communicated.  To explore this, we used a high-resolution decadal dataset from Big Bend National Park in west Texas (the Ecological Survey of Big Bend Area) and standard multivariate techniques to characterize vegetation change over a fifty-year period in the context of the USNVC hierarchy from the plant association on up.  We then compared the resulting statistical models of plant associations and their dynamics to existing regional state-transition/successional models for agreement on community definitions and dynamics at different USNVC hierarchy levels. 

Results/Conclusions

We identified 10 plant associations through time that could be attributed to four different Groups in the USNVC hierarchy.  We also detected different speeds and trajectories of vegetation recovery by plant association out of the 1950’s drought that were correlated to landform/soil type and initial conditions.  Given this high diversity of composition and responses within a relatively small dataset, fitting our outcomes into the existing dynamics models was problematic. Some models were generalized at higher than even the Group level of USNVC or were not cross-walkable, making comparisons highly generalized with broad sideboards on outcomes.  Others that provided specific information on species composition and soil types were more amendable to comparison to the degree that we could relate those entities to known quantities in the classification.  We recommend that given the diversity of community-level responses that can occur at a site, specifying plant association, groups, and macro-groups from the USNVC hierarchy may significantly aid consistent interpretation of models and confidence in their predictions for the future.