Session: Using Machine Learning to Quantify and Improve Earth System Predictions
Machine learning to predict peatland greenhouse gas emissions
Wednesday, August 4, 2021
Link To Share This Presentation: https://cdmcd.co/w3QEP5
Yuanyuan Huang, CSIRO Oceans and Atmosphere, Aspendale, Australia, Philippe Ciais, CNRS, France, Yiqi Luo, Department of Biological Sciences, Northern Arizona University, Flagstaff, AZ, Dan Zhu, Beijing University, China, Ying-Ping Wang, CSIRO Marine and Atmospheric Research, Victoria 3195, Australia, Cunjing Qiu, LSCE, Daniel S. Goll and Min Jung Kwon, LSCE, France, Bertrand Guenet, Bioemco, CNRS, Paris, France, David Makowski, INRAE, AgroParisTech, University Paris‐Saclay, UMR MIA 518, 75231 Paris, France, Inge De Graaf, University of Freiburg, Jens Leifeld, Agroscope, Climate and Agriculture Group, Reckenholzstrasse 191, 8046, Zurich, Switzerland., Jing Hu, University of Florida, Laiye QU, State Key Laboratory of Urban and Regional Ecology, Ecology, Research center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing, China
Background/Question/Methods Water table drawdown across peatlands increases carbon dioxide (CO2) and reduces methane (CH4) emissions. The net climatic effect remains unclear. Results/Conclusions Based on observations from 130 sites around the globe, we found a positive (warming) net climate effect of water table drawdown. Using a machine-learning based upscaling approach, we predict that peatland water table drawdown driven by climate drying and human activities will increase CO2 emissions by 1.13 Gt yr-1 and reduce CH4 by 0.27 Gt CO2-eq yr-1, resulting in a net greenhouse gas (GHG) source of 0.86 Gt CO2-eq yr-1 by the end of the 21st century under the RCP8.5 climate scenario. This net source drops to 0.73 Gt CO2-eq yr-1 under RCP2.6. Our results point to an urgent need to preserve pristine and rehabilitate drained peatlands to decelerate the positive (warming) feedback among water table drawdown, GHG emissions and climate change.