COS 120-6
Soil carbon stocks in subtropical pastures are strongly coupled to vegetative productivity over short timescales

Thursday, August 13, 2015: 3:20 PM
321, Baltimore Convention Center
Chris Wilson, Agronomy, University of Florida, Gainesville, FL
T. Trevor Caughlin, School of Forest Resources & Conservation, University of Florida, Gainesville, FL
Sami Walid Rifai, School of Forest Resources and Conservation, University of Florida, Gainesville, FL
Elizabeth Hermanson Boughton, Archbold Biological Station, Venus, FL
Michelle C. Mack, Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ
Luke Flory, Agronomy Department, University of Florida, Gainesville, FL
Background/Question/Methods

Grassland ecosystems occupy at least a fifth of earth’s land area and account for a large proportion of the global soil carbon stock. Increasing the size of soil carbon stocks in managed grasslands could potentially contribute to greenhouse gas mitigation. However, there is considerable uncertainty about the factors that regulate storage and turnover of carbon in these ecosystems. Working in grazed subtropical pastures in Florida, USA we coupled remote sensing datasets with intensive soil carbon sampling across a hydrological gradient in order to study the factors that best predict soil carbon stocks. Specifically, we compared the utility of elevation/hydrology, soil series, soil fertility, and an integrative index of vegetation productivity (enhanced vegetation index, EVI) from a 30-yr Landsat satellite dataset to predict variations in measured soil carbon. Then, we used a Bayesian framework to integrate prior experimental data on bulk soil carbon turnover with our time-series of EVI in order to mechanistically link vegetative productivity to observed soil carbon stocks.

Results/Conclusions

We found that the 30-yr dataset on EVI substantially enhanced predictions of soil carbon stocks (p < 0.0001). A model based on mean EVI alone explained a significant fraction of variance in the measured carbon stocks (r2 = 0.325), while adding all other predictors increased explained variance (adjusted r2 = 0.4682), but examination of regression coefficients suggests that mean EVI is the strongest single predictor. Our Bayesian analysis supports that a substantial fraction of soil carbon turns over rapidly in this system, particularly in the upper soil layers. Plant productivity strongly predicts soil carbon stocks, but only after an initial lag period of several years. We show how this temporal behavior is explained in a model where plant carbon cohorts undergo a two-stage decomposition process, and we illustrate the implications of this model for integrating remote sensing data into predictions of soil carbon. Overall, our findings suggest that total soil carbon stocks in this ecosystem are strongly coupled to vegetative productivity, and thus will be highly sensitive to any deleterious impacts of global change or poor management practices.