B32A-07
Process-based upscaling of surface-atmosphere exchange
Wednesday, 16 December 2015: 11:50
2006 (Moscone West)
Trevor F Keenan1, Iain Colin Prentice2, Josep Canadell3, Christopher A Williams4, Han Wang2, Michael R Raupach5, George James Collatz6, Tyler Davis7, Benjamin Stocker8 and Bradley John Evans1,9, (1)Macquarie University, Sydney, Australia, (2)Northwest A&F University, Yangling, China, (3)CSIRO Ocean and Atmosphere Flagship Canberra, Yarralumla, Australia, (4)Clark University, Worcester, MA, United States, (5)Australian National University, Canberra, Australia, (6)NASA Goddard Space Flight Center, Greenbelt, MD, United States, (7)USDA-ARS, Robert W. Holley Center for Agriculture and Health, Ithaca, NY, United States, (8)University of Bern, Bern, Switzerland, (9)Terrestrial Ecosystem Research Network Ecosystem Modelling and Scaling Infrastructure, Macquarie University 2109 and University of Sydney 2006, NSW, Australia, Sydney, Australia
Abstract:
Empirical upscaling techniques such as machine learning and data-mining have proven invaluable tools for the global scaling of disparate observations of surface-atmosphere exchange, but are not based on a theoretical understanding of the key processes involved. This makes spatial and temporal extrapolation outside of the training domain difficult at best. There is therefore a clear need for the incorporation of knowledge of ecosystem function, in combination with the strength of data mining. Here, we present such an approach. We describe a novel diagnostic process-based model of global photosynthesis and ecosystem respiration, which is directly informed by a variety of global datasets relevant to ecosystem state and function. We use the model framework to estimate global carbon cycling both spatially and temporally, with a specific focus on the mechanisms responsible for long-term change. Our results show the importance of incorporating process knowledge into upscaling approaches, and highlight the effect of key processes on the terrestrial carbon cycle.