The Sarmiento and Pacala Groups work to continuously improve a “Carbon Observing System” that uses observational data and models to monitor both short and long time scale changes in the land and ocean carbon sinks, and to provide predictions for the future. This year’s research has examined the importance of CO2 fertilization, nitrogen fixation, seasonality, and ocean circulation on CO2 uptake.


Nitrogen limitation of CO2 fertilization

Steve Pacala, Stefan Gerber, Lars Hedin and colleagues have developed the first terrestrial model that is capable of reproducing the magnitude of the change in the land’s carbon inventory since the Industrial Revolution, current biospheric CO2 sequestration, and its broad geographic distribution. The researchers reconcile the terrestrial carbon budget with biospheric nitrogen supply using a state of the art dynamic vegetation model that explicitly take into account the dynamics of secondary vegetation and includes a mechanistic representation of biological nitrogen (N) fixation.

While the biosphere’s carbon budget remained close to neutral since the onset of industrialization, a considerable residual or “missing” terrestrial sink compensates for emissions due to land-use practices. The carbon flux into the residual terrestrial sink ranged between -0.2 and 4.3 PgC yr-1 in the 1980s and 1990s and resulted in a total accumulation of 61 to 141 Pg C between 1800 and 1994 (Sabine,et al., 2004). Sustained terrestrial carbon sequestration in terrestrial ecosystems worldwide, however, might be in conflict with nutrient availability, particularly nitrogen, and terrestrial models suggest wide-spread reduction in terrestrial uptake as CO2 rises due to limited N supply.

Integrated over the historic period, the team found a terrestrial net carbon loss of about 58 PgC in agreement with estimates that stem from ocean, and atmosphere and fossil fuel budgeting. Accounting for emissions from land-use, model results suggest a residual terrestrial sink of ~150 PgC over the same time period, in agreement with budget estimates. Further the simulated distribution of recent carbon sequestration across latitudes can be reconciled with inverse approaches that estimate terrestrial source/sink based on a network of CO2 measurements across the globe.

Their results, however, suggest that N restrictions reduced carbon uptake by 51 PgC between 1800 and 2000 and hampers current sequestrations by 0.9 PgC yr-1. The uptake of anthropogenic carbon is caused by the capacity of tropical trees to alleviate nitrogen limitation through upregulation of biological N fixation, supporting increased productivity from CO2 fertilization. In contrast, productivity in wide areas of temperate and boreal forest exhibits little response to increasing levels of atmospheric CO2 (see Figure 15 on following page).

Figure 15. Modeled carbon uptake into the residual terrestrial carbon sink, integrated between the years 1800 and 2000. Top: Results obtained by taking into account changes in CO2, anthropogenic N deposition and climate change after 1950. Bottom: Reduction in carbon uptake if N deposition rates are kept at pre‐industrial levels. CO2 fertilization in temperate and boreal forest occurs mainly in areas that experience high levels of anthropogenic nitrogen pollution.


CO2 uptake by U.S. forests

An alternative mechanism to CO2 fertilization that may explain some or all of the observed land sink in some regions of the world is forest regrowth. This mechanism is thought to be particularly important in the eastern U.S., where many previously forested areas are regrowing following agricultural abandonment and wide-spread logging during the previous century. Whether the land-sink is due primarily to CO2 fertilization or to forest regrowth has important policy implications. If the land sink is due to CO2 fertilization (as assumed by most global ecosystem models), then the terrestrial biosphere should continue to sequester atmospheric CO2 well into the future, providing a large natural offset to anthropogenic emissions. In contrast, if the land sink is due to forest regrowth, the offset would disappear in the future as the biomass of recovering forests equilibrates.

Using four decades of individual tree growth records from the U.S. Forest Service’s inventory program, Steve Pacala, Jeremy Lichstein and colleagues have now shown that the dominant trend in the eastern U.S. is growth decline, not growth enhancement. Thus, not only are trees not growing faster in the eastern U.S. now compared to the 1970s and 1980s (as expected from CO2 fertilization), but they are in fact growing more slowly. This implies that the land sink in the eastern U.S. is due entirely to forest regrowth, and is occurring despite the fact that individual trees are growing slower now than in the past. The mechanism causing the growth declines is unclear, but may be related to nitrogen-limitation of tree growth, tropospheric ozone pollution, and/or soil acidification due to nitrogen deposition.


Temporal shifts in terrestrial uptake of atmospheric CO2

The Sarmiento Group has discovered a possible increase in net land uptake of CO2 of 0.9 (0.8 to 1.0) Pg C/yr after 1988/89 based on an analysis of the atmospheric CO2 growth rate. In this approach, the net land flux is 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. Due to substantial uncertainties in the temporal variability of oceanic uptake, a suite of ocean models was used.

The predominant signal in the inferred net land flux is one of very large interannual variability. However, using a low-pass Butterworth filter, it is possible to resolve a clear shift towards greater land uptake for the period after 1988/89 relative to the period before it (Figure 16a). Furthermore, the cumulative land sink clearly shows a deviation occurring after 1988/89 (Figure 16b). The variability may also have been increased after 1988/1989.

Figure 16. The (a) annual and (b) cumulative land carbon sink shown from 1960 onwards. The annual land carbon sink also shows smoothed lines filtered with a 5 year Butterworth filter. The straight lines in (b) area drawn in by hand to provide a guide to the eye showing that the slope increases after 1988/89. A negative flux indicates uptake of CO2 by the terrestrial biosphere.

To identify what causes this shift, it is important to determine the type of shift, when exactly it occurs, and its magnitude with associated uncertainty. Change point methods allow one to detect when the shift occurs and to estimate its magnitude, and also allow determining the nature of the shift (shift in the mean, in the trend or in the variance of a time series). Claudie Beaulieu has developed a general change point method that allows detecting step-like changes in the mean, the variance or in the coefficients of a multiple regression model. This new technique is currently being tested using Monte Carlo simulations to evaluate its power of detection and the risk of false detection one encounters when applying it.

Meanwhile, the group is also analyzing changes in CO2 concentrations, in the atmospheric growth rate of CO2 and in the net land uptake of CO2 at several atmospheric observing stations to verify whether this change is occurring in several regions and if there is a latitudinal gradient in the time the shift occurs. The group is also testing whether the shift can be explained by ENSO mechanisms and/or by the volcanic eruptions. One possible explanation for this increased terrestrial uptake is an observed increase in shortwave radiation accompanied by an intensification of the hydrological cycle in the early 1990’s. The group’s plan to test this hypothesis is described in “Future Plans.”


Impact of interannual variability on terrestrial CO2 uptake

To date, considerations of the response of the terrestrial biosphere to anthropogenic climate change have focused on how changes in the mean physical climate state may impact changes in the global carbon cycle. For example, researchers have used coupled climate models to predict whether changes in precipitation over the Amazon over the next century will result in large losses of CO2 from the terrestrial biosphere to the atmosphere. However, an important modeling result of the IPCC AR4 study is that the changes in the mean state of the physical climate system may be expected to be associated with changes in variance. Thus there is a need for processoriented studies that can complement the fully coupled modeling studies by pointing out the relative importance of changes in the mean and changes in variance for the carbon cycle.

To address this question, Keith Rodgers in the Sarmiento Group (in collaboration with Claudie Beaulieu, David Medvigy and Ni Golaz from the Pacala Group) is conducting a modeling study of the processes regulating the rate of exchange of CO2 between the terrestrial biosphere and the atmosphere using the LM3V dynamical vegetation model (developed jointly by the Pacala Group at Princeton and GFDL). The goal of these simulations is to determine whether terrestrial ecosystems are sufficiently nonlinear that changes in the variance (in the absence of changes in the mean) for the physical forcing can result in a change in the time-mean carbon exchange over large scales. Current work is focusing on the period 1977-1996, when it is known that the phase of the Pacific Decadal Oscillation was relatively stationary, facilitating a focus on interannual variability.


A new forest model

The Pacala group has developed a new forest model (the perfect plasticity approximation, or PPA) that scales from individual tree growth and mortality rates to community-level and ecosystemlevel dynamics. A critical aspect of the model is its representation of height-structured competition for light, a mechanism that is absent from most global vegetation models. From an evolutionary perspective, this mechanism explains the existence of wood, which is the most abundant source of carbon in the terrestrial biosphere. Previous work has demonstrated that the PPA model accurately predicts the dynamics of forest communities in the northern midwest of the U.S.

More recently, the researchers have made significant progress on two fronts towards implementing a global version of the PPA model that can be coupled to global ecosystem models: (1) They have developed physiological and allocational submodels that quantify growth and mortality rates of individual trees as a function of plant traits (wood density; allocation to roots, leaves, and reproduction) and environmental conditions (disturbance rate, soil nitrogen and water availability). These analyses will soon lead to a new biome model that predicts the characteristics of the optimal plant as a function of climate and soil characteristics. (2) They are developing an optimization system that will allow them to fit parameters in global ecosystem models using formal, quantitative methods. This technology will allow us to fine tune the parameters in a biodiversity rich global vegetation model that incorporates the PPA forest model.


Temporal changes in air-sea CO2 fluxes

An important goal of global carbon cycle research is to identify how ocean circulation changes over recent decades associated with anthropogenic climate change may be impacting the exchange of CO2 between the atmospheric and ocean carbon reservoirs. To this end, Joseph Majkut and Jorge Sarmiento have been evaluating how air-sea CO2 fluxes have changed in an ocean circulation model (GFDL’s MOM4) coupled to an ocean biogeochemistry model forced with NCEP-1 reanalysis fluxes. This work has been included in a recent study that integrated physical and economic models as well as observational data to estimate the anthropogenic carbon budget through 2008. A number of alternatives to the NCEP-1 data product now exist and differ in their representation of the structures and amplitudes of decadal trends in the surface winds and temperatures, and the group is currently investigating the implications of these differences for the ocean carbon cycle.

Keith Rodgers is conducting a separate study to understand how changes in the climate system’s seasonal cycle could impact air-sea CO2 fluxes and global climate. This work focuses on the latest generation of models that have been developed in preparation for the next IPCC study (IPCC AR5).

The main first result is that the anthropogenic transient in the carbon cycle does in fact reveal substantial changes in seasonality for the models that have been evaluated thus far. A comparison of the integrated fluxes for the GFDL and IPSL models is shown in Figure 17. For the global fluxes, both models exhibit an increase in uptake over the period 1861-2099. However, when the hemispheres are considered separately the amplification of the seasonality becomes clearer. For the extratropical fluxes, increased/decreased positive uptake is associated with winter/summer conditions. For both the North and the South, there is a tendency for increased uptake in winter over the course of the models runs. For summer, the trend is weaker, and can even be in the opposite sense of the winter trend. Importantly, both models exhibit large increases in seasonality, although there are important differences between the models. This work highlights for the first time that the full seasonal cycle must be taken into account to understand the changes in CO2 uptake.

Figure 17. A comparison of air‐sea fluxes over 1861‐2099 for concentration scenarios (SRES_A2) with two IPCC‐class Earth System Models. Shown are the global mean flux (top two panels), the Northern Hemisphere flux in the extratropics over 15°N‐65°N (middle panels), and the flux for the Southern Hemisphere extratropics over 80°S‐ 15°S (bottom panels). For both models the monthly mean fluxes are shown in black, with positive fluxes indicating a net uptake of CO2 by the oceans, and a 12‐month running mean is shown in red.