Attributing Changes in Gross Primary Productivity from 1901 to 2010

Wednesday, 17 December 2014: 3:28 PM
Christopher R Schwalm1, Deborah N Huntzinger1, Anna M Michalak2, Robert B Cook3, Bassil ElMasri4, Daniel J Hayes3, Maoyi Huang5, Andrew R Jacobson6, Atul K Jain4, Huimin Lei7, Chaoqun Lu8, Hanqin Tian8, Kevin M Schaefer9 and Yaxing Wei3, (1)Northern Arizona University, Flagstaff, AZ, United States, (2)Carnegie Institution for Science, Washington, DC, United States, (3)Oak Ridge National Laboratory, Oak Ridge, TN, United States, (4)University of Illinois at Urbana, Urbana, IL, United States, (5)Pacific NW Nat'l Lab-Atmos Sci, Richland, WA, United States, (6)NOAA Boulder, Boulder, CO, United States, (7)Tsinghua University, Beijing, China, (8)Auburn University at Montgomery, Auburn, AL, United States, (9)University of Colorado, National Snow and Ice Data Center, Boulder, CO, United States
Model-based studies are foundational to perform diagnosis (has there been a change?) and attribution (what caused this change?) in the context of global environmental change. Here we employ a dual method approach using machine learning and simulation differencing across an ensemble of terrestrial biosphere models (TBM) to attribute changes in gross primary productivity (GPP) from 1901 to 2010. The simulations are taken from the Multi-scale Synthesis and Terrestrial Model Intercomparison Project (MsTMIP). For each TBM MsTMIP prescribes a semi-factorial set of five runs (globally at 0.5º spatial resolution) where time-varying controls on carbon metabolism are sequentially enabled. MsTMIP has a constrained simulation protocol –driving data, vegetation cover, boundary conditions, and steady-state spin up protocol are all standardized– such that only model structure varies and ensemble spread addresses process uncertainty. Applying this dual method to MsTMIP simulation output we attribute changes in GPP to changes in climate, land cover/land use change, atmospheric CO2, nitrogen deposition, near-surface air temperature, precipitation, and downwelling shortwave radiation as well as climate sigma (irreducible climate noise) and nonlinearity (interactions). Globally, the key factor associated with the Anthropocene, namely the sustained increase in atmospheric CO2, dominates changes in GPP across the full time period. Climate factors are of secondary importance and, along with land cover/land use change, may act to decrease GPP depending on decade and reference period. Despite differences in model structure attribution results across the full ensemble are generally consistent. Spatial morphologies, replicating the same dual approach by grid cell, exhibit high variability but with an atmospheric CO2 fertilization effect dominating the tropical zone. Our results suggest that the modern era of global warming, when viewed through the prism of GPP attribution, reaches back at least until the early 1900s.