Complex patterns of plant flammability traits across a soil moisture gradient

Justin P. Wright, Biology, Duke University, Durham, NC, Anita Simha, Biology Department, Duke University, Durham, NC, Rachel M. Mitchell, School of Earth and Sustainability, Northern Arizona University, Flagstaff, AZ and Steven Anderson, Duke University, Durham, NC

Background/Question/Methods
In fire-adapted systems, the distribution of fire-related plant traits can both respond to and affect fire intensity. Here we examine whether patterns in the distribution of fire traits are predictable at the level of traits, species, or communities. Specifically we ask three questions: 1) Are key flammability traits correlated with each other? 2) Are species flamability traits predicted by soil moisture, woodiness, or species responses to fire? and 3) Are the community weighted means (CWM) of fire traits related to soil moisture? We sampled vegetation along 32 moisture gradients in the sandhills of NC, collecting over 2000 samples across 93 species across 3 years. We then conducted single-species litter burn-trials measuring 7 traits associated with flammability (leaf dry matter content, time to ignition, maximum flame height, maximum flame temperature, burn time, % consumption, and smolder time), and used Principal Components Analysis to examine patterns of correlations between traits. At the species-level we modeled fire traits as a function of lambda (change in cover from years before and after a fire), soil moisture, and growth form (woody vs. herbaceous). At the community level, we modeled the community weighted mean fire traits as a function of soil moisture.
Results/Conclusions
We found that there are two independent axes of variation in these fire traits. On one axis, maximum flame height was negatively related to time to ignition and smolder time. On an orthogonal axis, maximum temperature, leaf dry matter content, burn time, and % consumption were all positively correlated with each other. At the species level, there was a significant three way interaction between the mean soil moisture at which a species was found, its responses to previous fire and its woodiness. Together, the three drivers predicted species’ positions along the first axis of the PCA of fire traits, while for the second axis there was a marginally significant relationship with woodiness. At the community level, a different pattern emerges. Community weighted means of traits were not significantly related to soil moisture. The contrasting patterns of species level and community level trait distributions along the fire intensity gradient suggest the existence of filters on the local abundance of species in this pyrogenic landscape.