Relative role of parameter vs. climate uncertainty for predictions of future Southeastern U.S. pine carbon cycling

Tuesday, 15 December 2015: 09:30
2004 (Moscone West)
Annika Jersild1, R. Quinn Thomas1, Evan Brooks1, Robert O Teskey2, Randolph Hamilton Wynne3, David Arthur4, Carlos Gonzalez5, Valerie A Thomas1, Thomas D. Fox1 and Luke Smallman6, (1)Virginia Polytechnic Institute and State University, Blacksburg, VA, United States, (2)University of Georgia, Athens, GA, United States, (3)Virginia Tech, Blacksburg, VA, United States, (4)Arizona State University, Tempe, AZ, United States, (5)University of Florida, Ft Walton Beach, FL, United States, (6)University of Edinburgh, Edinburgh, United Kingdom
Predictions of the how forest productivity and carbon sequestration will respond to climate change are essential for assisting land managers in adapting to future climate. However, current predictions can include considerable uncertainty that is often not well quantified. To address the need for better quantification of uncertainty, we calculated and compared parameter and climate prediction uncertainty for predictions of Southeastern U.S. pine forest productivity. We used a Metropolis-Hastings Markov Chain Monte Carlo-based data assimilation technique to fuse regionally widespread and diverse datasets with the Physiological Principles Predicting Growth model (3PG) model. The datasets incorporated include biomass observations from forest research plots that are part of the Pine Integrated Network: Education, Mitigation, and Adaptation project (PINEMAP) project, photosynthesis and evaporation observations from loblolly pine Ameriflux sites, and productivity responses to elevated CO2 from the Duke Free Air C site. These spatially and temporally diverse data sets give our unique analysis a more accurately measured uncertainty by constraining complimentary components of the model. In our analysis, parameter uncertainty was quantified using simulations that integrate across the posterior parameter distributions, while climate model uncertainty was quantified using downscaled RCP 8.5 simulations from twenty different CMIP5 climate models. Overall, we found that the uncertainty in future productivity of Southeastern U.S. managed pine forests that was associated with parameterization is comparable to the uncertainty associated with climate simulations. Our results indicate that reducing parameterization in ecosystem model development can improve future predictions of forest productivity and carbon sequestration, but uncertainties in future climate predictions also need to be properly quantified and communicated to forest owners and managers.