|Abstract or Summary
- Question: Most results of restoration efforts are species-specific and/or site-specific and
therefore are not general enough to be easily applied to other species and other sites. Our
research addresses the issue of species-specific results by investigating the feasibility of
using plant traits instead of taxonomic species to characterize species responses to restoration
treatments. Specifically, we test the explanatory power of plant traits, one necessary prerequisite
for the development of predictive and general plant trait models.
Location: Ten remnant bunchgrass prairie sites in the Pacific Northwest of North America
(Oregon and Washington, USA; British Columbia, Canada).
Methods: We developed two types of quantitative models for each of 10 prairie restoration
sites: 1) plant trait models, which related plant traits to species field responses following
restoration management treatments, and 2) species identity models, which related species
taxonomic identity to species field responses following restoration management treatments.
Species identity models determined the maximum amount of variability of field responses
that can be explained by differences in individual species’ responses to management
treatments. Plant trait models determined what proportion of this explanatory power can be
attributed to plant traits. This approach contrasts with approaches often used in other plant
trait studies that describe how traits vary with environmental conditions.
Specifically, we used these two models to address the following questions: 1) How much
of the variability in field responses of plants to restoration management treatments is
explained by plant traits? 2) How well do plant traits explain the variability of field
responses following restoration management treatments compared to models relating field responses to species identity? Our approach was to measure two aspects of explanatory
power: R2 (variability explained) and AIC (a measure of model fit that accounts for
parsimony, i.e., how well a model fits the data with relatively few explanatory variables).
Results: 1) The plant trait models (relating plant traits with plant field responses) explained
much of the variability within each of the ten restoration sites, with R2 values ranging
between 31% and 69%. 2) The species identity models (relating species taxonomic identity
with plant field responses) explained between 47% and 74% of variability in field
performance (R2). Thus, the plant trait models explained nearly as much variability as the
species identity models.
In seven out of nine sites, the plant trait models were superior to the species identity
models as measured by AIC; that is, the trait models did well at explaining variability with
less model complexity (i.e., fewer explanatory variables).
Conclusion: Development of general and predictive plant trait models is a multi-step
process. Strong explanatory power by plant trait models, both on an absolute scale and as
compared to species identity models, supports the feasibility of using plant traits instead of
species taxonomic identity as a common language to characterize plant field responses to
restoration treatments. Such high explanatory power is one necessary pre-requisite for the
development of predictive and general plant trait models.
Our results also indicate that the plant trait models are site-specific even though all sites
were upland bunchgrass remnant prairies. We discuss the next steps in the development of
more general and predictive models: incorporating environmental factors into the plant trait
models to address the issue of site-specificity and testing the power of these models to
predict vegetation responses.
- Clark, D. L., Wilson, M., Roberts, R., Dunwiddie, P. W., Stanley, A. and Kaye, T. N. (2012), Plant traits – a tool for restoration?. Appl Veg Sci, 15: 449–458. doi:10.1111/j.1654-109X.2012.01198.x