The carbon cycle response of the Amazon forest during the 2010 drought in dynamic global vegetation models

Monday, 15 December 2014
Anna B Harper1,2, Pierre Friedlingstein3, Stephen Sitch3, Peter Michael Cox3, Chris Jones2, Almut Arneth4, Athanasios Arvanitis4, Philippe Ciais5, Christian Frankenberg6, Atul K Jain7, Etsushi Kato8, Sam Levis9, Nicholas Parazoo6, Benjamin Poulter10, Benjamin David Stocker11, Andy Wiltshire2 and Soenke Zaehle12, (1)University of Exeter, Exeter, EX4, United Kingdom, (2)Met Office Hadley Centre, Exeter, United Kingdom, (3)University of Exeter, Exeter, United Kingdom, (4)Karlsruhe Institute of Technology, Karlsruhe, Germany, (5)CEA Saclay DSM / LSCE, Gif sur Yvette, France, (6)NASA Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, United States, (7)University of Illinois at Urbana, Urbana, IL, United States, (8)NIES National Institute of Environmental Studies, Ibaraki, Japan, (9)National Center for Atmospheric Research, Boulder, CO, United States, (10)Montana State University, Bozeman, MT, United States, (11)Imperial College London, London, United Kingdom, (12)Max Planck Institute for Biogeochemistry, Jena, Germany
Understanding the mechanisms behind the carbon cycle response to extreme events is essential for predicting future behaviour of tropical forests. In this study, we focus on the modelled response during the 2010 drought in the Amazon forest using a suite of dynamic global vegetation models. The potential impacts of drought on the carbon cycle include increased mortality, reduced productivity, and changes to heterotrophic and autotrophic respiration rates. In addition, the 2010 drought was accompanied by substantial biomass burning which also emitted CO2 to the atmosphere. Many of these fires were from human ignitions related to deforestation, which can be particularly difficult for models to represent.

During the 2000’s, the models all showed that the forests of the Amazon basin were a net sink of CO2, with an average net biome productivity, or NBP, of 0.20±0.19 PgC/yr. This is within the range of bottom-up and top-down estimates of NBP in undisturbed forests (0.39+/-0.26 PgC/yr and 0.25+/-0.14 PgC, respectively) [Phillips et al. 2009, Gloor et al. 2012, Gatti et al. 2014]. This sink was reduced during the 2010 drought but the models tend to underestimate the severity of the reduction. The model average NBP during 2010 was -0.12 PgC, compared to -0.48±0.18 PgC from an observational study [Gatti et al. 2014]. In particular, the models simulated positive NBP during the dry season of 2010, indicating a carbon sink even during the peak drought months. This result is in direct contrast to recent estimates of seasonal terrestrial fluxes during 2010 based on atmospheric measurements [Gatti et al. 2014]. We investigate the roles of fire CO2 emissions, GPP sensitivity to drought, and uncertainty in driving data in the modelled responses, utilizing the GOSAT retrievals of chlorophyll fluorescence and the Global Fire Emissions Dataset. Based on this analysis we will identify priorities for continued model development of DGVMs in tropical forests.