Simulation of the terrestrial sink

Stefan Gerber, in collaboration with Lars Hedin, Michael Oppenheimer and Steve Pacala, has completed implementation of nitrogen (N) dynamics in the Princeton-GFDL LM3V land model. The new model resolves processes, dynamics and feedbacks that have not previously been captured and represents a substantial improvement over all earlier and existing dynamic global vegetation models. A critical motivation for the completion of this model was to understand whether nutrients hamper the potential of land systems to sequester anthropogenic carbon.

Gerber and colleagues in the Pacala Group investigated land carbon exchange over the period of global industrialization. Their model results show a net carbon efflux from land sources over a large part of the past 200 years, before the land turned into a sink around 1970 (Figure 22). The net sink is the sum of large losses from land-use activities (200 PgC over 200 years) and a compensatory residual sink (150 PgC). Their model results agree with tracer methods and landuse reconstructions. Model results point to CO2 fertilization as the primary mechanism for the residual sink (130 PgC over 200 years), while nitrogen deposition contributes with about 20 PgC.

On the other hand nitrogen supply has critically inhibited carbon uptake. Model results suggest another 50 PgC could have been sequestered, particularly in nitrogen limited boreal regions, if nitrogen supplies were sufficient.

A critical feedback in the model is the capacity of tropical forest to adapt nitrogen fixation in order to alleviate nitrogen limitation. This observed mechanism, when scaled up across the tropics is the most important mechanism that leads to a strong CO2 fertilization feedback and subsequent carbon sequestration in these forests. Modeled sink magnitudes for natural tropical forests agree with estimates of pantropic carbon sequestration.

Figure 22. Net carbon fluxes with (C‐N) and without (C‐only) N constraints for the past 200 years, compared against the budget derived from oceanic tracer deconvolution, atmospheric 13C/12C deconvolution and modeled oceanic uptake.

Temporal shifts in terrestrial uptake of atmospheric CO2

Last year, the Sarmiento Group detected an abrupt increase in the net land uptake of CO2 after 1988. In this study, the net land flux was estimated as the balance of relatively well-known components of the carbon budget: fossil fuel emissions, the observed growth rate in the atmosphere, and the oceanic uptake from state of the art ocean models. A suite of ocean models was used to represent uncertainties in the temporal variability of oceanic uptake.

Due to the large variability in the net land uptake, it was not possible to determine the exact timing and nature of the increase robustly by visual inspection. Thus, a sophisticated statistical methodology was developed and applied to objectively determine the nature and timing of the shift in the net land uptake. Using this method, the researchers have shown that the abrupt shift in the mean net land uptake occurred between 1986-1993 (with a probability of about 60%), with 1988 being the most likely time for the shift. The analysis further showed that that the magnitude of the shift was approximately 0.8 Pg C/yr and that the risk that this shift was a false detection is very small (approximately 2%). The increase in uptake seems to be linked to the atmospheric growth rate, and ENSO variability and volcanic eruptions were shown to have been insufficient to cause the shift. Additional efforts are necessary to further investigate this step change in terrestrial carbon uptake and its causes. The NASA Carbon Cycle Science Program has awarded a grant to the Sarmiento Group and collaborators at JPL and UCLA for the detection and attribution of rapid large-scale shifts in the terrestrial carbon cycle.


Data assimilation system to constrain terrestrial carbon cycle modeling

A number of terrestrial ecosystem models can reproduce observed broad-scale patterns in plant biomass and net primary production (NPP). However, predictions of future terrestrial carbon stocks and fluxes differ widely among models, and terrestrial NPP is one of the largest sources of uncertainty in predicting future atmospheric CO2, in part due to parameterizations of plant physiology and carbon allocation. Ecosystem models currently require numerous parameters which are typically assigned one of many possible values from the literature. An alternative approach is to fit the models to data using formal, quantitative methods. This method is more objective, but also more computationally challenging.

To enable the fitting of forest data, the Pacala Group has developed an automated optimization system for the NOAA-GFDL land model, LM3V. The optimization system runs LM3V, compares the output to forest inventory or eddy covariance data, calculates new vegetation parameters using the Gauss-Newton algorithm, and then repeats the above steps until the algorithm converges. Predictions from the baseline and optimized models were compared to U.S. Forest Service FIA inventory data in the eastern United States, and forest biomass and productivity were found to be overly sensitive to soil moisture in the baseline model. The data-fitting has revealed a weakness in the model’s representation of homeostatic mechanisms that stabilize the productivity of real forests that is likely is widespread among terrestrial ecosystem models. This model weakness may be addressed by increasing the number of plant functional types and/or allowing for within-functional-type plasticity.


Incorporating fire in carbon-cycle simulations

Fires are a major disturbance in terrestrial ecosystems and a large source of carbon to the atmosphere. While climate is generally considered to be the dominant control on the distribution and timing of global fires, human activities shape global fire distribution directly through practices such as land-clearing, crop and pasture management, and fire suppression, and indirectly through anthropogenically driven climate change.

Steve Pacala and colleagues have developed a new fire model for use in coupled carbon cycle-climate models. A novel feature of the fire model is that the number of fires, burned area, and emissions are determined not only by the climate and by population density, but also by the history of land use changes. Figure 23 shows the burned area (in units of km2/year) from the fire model (black) compared to those of observations (gray shading). With this model, in addition to fully incorporating fire as a process in the carbon cycle, the group aims to better understand the role that humans play in causing and controlling the distribution and timing of fires across the entire planet.

Figure 23. Burned area (in units of km2 /year) from the fire model (black) compared to those of observations (gray shading).