Use global models to estimate most sensitive elements of terrestrial carbon to climate change for tropics, mid-latitudes, and polar regions
Overall Performance Measures
1st Quarter Metric - COMPLETED
One of the largest potential carbon-climate feedbacks in the terrestrial system is via thawing and decomposition of permafrost, due to the enormous quantity of carbon (C) in permafrost and its temperature dependent stability. However, the nitrogen (N) released by enhanced decomposition may provide a stabilizing feedback as vegetation growth and carbon uptake is stimulated by increased nitrogen availability. Relative importance of permafrost carbon stability and carbon-nitrogen interactions are evaluated.
The feedback between climate warming and the terrestrial carbon cycle was quantified through a series of simulation experiments covering the 450-year time period 1850-2300. Analysis of these experiments focused on the response of high-latitude ecosystems and permafrost distributions to warming as forced by observed climate change drivers for the historical period (1850-2004), and by drivers from the strongly warming Reference Concentration Pathway 8.5 scenario (RCP8.5) for the period 2005-2300. These experiments were based on the Coupled Model Intercomparison Project Phase 5 (CMIP5) protocol, but used the Community Land Model CLM4.5 in an offline mode, forced with climate output from a previous set of CMIP5 simulations performed with the fully-coupled Community Climate System Model version 4 (CCSM4). Carbon cycle-climate feedbacks were quantified by comparing results from three different simulation modes: 1) radiatively-coupled simulations, in which the climate system is responding to increased concentrations of radiatively-active gases through the greenhouse mechanism; 2) biogeochemically-coupled simulations in which the terrestrial ecosystem is responding to increased atmospheric CO2 concentrations; and 3) fully-coupled simulations in which both radiative and biogeochemical coupling mechanisms are active. To quantify the role of the N cycle in climate-carbon feedbacks these simulations were repeated but with N supply controlled to maintain a pre-industrial (1850) level of N limitation on plant growth throughout the simulation period (the C-only case). To determine the sensitivity of model results to the fate of deep soil C, additional experiments were performed in which a parameter controlling turnover of deep carbon pools was varied. Results show 50% loss of permafrost area by year 2100 for the warming scenario RCP8.5, with near-total loss of permafrost area by year 2300. Compensating effects of rising temperature and rising CO2 concentration lead to modest total C soil gains in the permafrost region (~20 Pg C) by year 2150, with modest soil C losses (also ~20 Pg C) by year 2300. In these experiments, C-N interactions limit C accumulation in permafrost regions due to limited N availability, causing a net reduction of ~35 Pg C in soils and vegetation compared to the C-only case. The sensitivity of deep C to warming also has a strong influence on total C loss over the permafrost region by year 2300, with total C losses in year 2300 differing by ~140 Pg C between the most sensitive and least sensitive settings for deep C dynamics tested here.
2nd Quarter Metric - COMPLETED
Comparison of terrestrial ecosystem models with observations, focusing on the interactions between ecosystem behavior and climate conditions over the seasonal cycle at mid-latitudes, suggests that enhanced drought stress is likely to be the main driver of mid-latitude climate–carbon feedbacks. Terrestrial ecosystem–climate feedbacks can be separated into those associated with concentration–carbon feedbacks (i.e., carbon storage increases in response to elevated levels of atmospheric CO2) and climate–carbon feedbacks (i.e., carbon losses or gains driven by climate change)
(Friedlingstein et al. 2006). For mid-latitude forests, extensive past and ongoing work by the Free Air Carbon Dioxide Enrichment community has placed constraints on the magnitude of the former class of feedbacks, while long-term time series from Ameriflux can be used to investigate the latter class of feedbacks, as ecosystems respond to year-to-year variations in temperature and precipitation. In this context, a key driver of the strength of climate–carbon feedbacks at mid-latitudes is the response of ecosystems to drought stress. Increasing drought stress may result from earlier snow melt, increases in spring runoff, greater mid-summer evaporative demand, and longer growing seasons expected in the next several decades as a consequence of climate warming. Here we evaluated the ability of Earth system models to correctly capture the seasonal cycle of atmospheric carbon dioxide (CO2) and gross primary production (GPP) as compared with NOAA and Ameriflux observations. We found that the onset of photosynthesis in spring occurs earlier in CMIP5 models than in the observations. As a consequence, transpiration fluxes (the process whereby plants take up soil water and release it into the atmosphere through their leaves) deplete soil moisture reserves too rapidly in the ESMs, reducing available water in mid-summer. This bias in the representation of photosynthesis may contribute to observed positive summer temperature (and negative summer precipitation) biases observed in the models. Although the plant ecosystem response to water and temperature stress is inferred from model bias (or error), the results suggest that corrected models would have plant growth that enhances mid-latitude drying and warming under warmer climate conditions (i.e. positive feedbacks to drought stress).
The feedback between the terrestrial carbon and hydrological cycles at mid-latitudes was evaluated using a collection of historical simulation experiments performed for the Coupled Model Intercomparison Project Phase 5 (CMIP5). The timing of spring photosynthetic onset in CMIP5 models was compared with atmospheric CO2 mole fraction observations from NOAA Global Monitoring Division (GMD) and eddy covariance observations from Fluxnet. Diagnostics of the phase of the annual cycle of atmospheric CO2 from the Community Earth System Model (CESM) indicated the model exhibited early drawdown of CO2, suggesting the onset of the growing season occurred too early in the simulation (Keppel-Aleks et al. 2013). This is likely a result of photosynthesis submodels lacking a representation of cold hardening and frost recovery processes in leaves. To explore if other models showed similar behavior, the terrestrial net ecosystem exchange (NEE) carbon fluxes from a collection of CMIP5 models were individually transported through the Goddard Earth Observing System Chemical Transport Model (GEOS-Chem) model using prescribed contemporary ocean and fossil fuel fluxes (Bey et al. 2001). The resulting atmospheric CO2 seasonal cycles from these simulations were sampled at NOAA GMD flask monitoring station locations and compared with observations for the time period 1995–2005. Results indicated that atmospheric CO2 also is drawn down too early in springtime in most CMIP5 Earth system models and that the strength of this early season uptake bias varies as a function of latitude in the Northern Hemisphere.
To investigate the causes and consequences of these timing biases, model carbon and energy fluxes were extracted for gridcells corresponding to Ameriflux and Fluxnet sites in North America and Eurasia, between 35°N and 45°N, where the timing biases were strongest, during the times for which observations were available. Comparisons with Fluxnet sites confirmed that early onset of photosynthesis and accompanying increases in GPP are the primary cause for early NEE uptake and atmospheric CO2 drawdown. Comparisons of model energy fluxes with Fluxnet indicated that most models exhibited an early peak in latent heat flux (i.e., water losses to the atmosphere) and significant overestimate of sensible heat flux (i.e., rates of atmospheric heating) in mid- to late-summer. This pattern is consistent with plants using their water supplies too rapidly in the models, and, as a consequence, heating up and having a reduced efficiency of growth. Comparisons of gridded model precipitation with Global Precipitation Climatology Project version 2 (GPCP2) observations and surface air temperature with Climate Research Unit (CRU) observations for the month of August indicated the models had low bias in precipitation and a significantly high bias in surface temperature in the north-central regions of North American and Eurasia, suggesting the biases in photosynthesis may feedback to influence regional precipitation and surface air temperatures in the models.
3rd Quarter Metric - COMPLETED
The availability of forest nutrients, such as nitrogen and phosphorus to support both new plant growth and the activity of soil microorganisms, is a primary factor regulating feedbacks between the terrestrial carbon cycle and climate change for tropical forest ecosystems. Previous research on the interactions between carbon and nitrogen cycles highlighted the importance of that interaction in temperate and high-latitude regions and suggested that phosphorus cycling mechanisms would impose even more important regulating influence on the terrestrial carbon cycle response to climate in tropical regions, especially in tropical forests growing on deeply weathered soils. The influences of nitrogen and phosphorus availability on tropical carbon cycle and carbon-climate feedbacks are evaluated and quantified. The availability of nutrients limits plant growth, and therefore, reduces carbon dioxide uptake that would result from carbon dioxide increase; however, warming enhances the availability of nutrients through accelerated decomposition of soil organic matter. The first effect dominates, so that nutrient restrictions overall reduce the carbon uptake of the system under climate change.
The feedbacks between climate, carbon cycle, and nutrient availability in the tropics were quantified through a pair of global simulation experiments representing a pre-industrial steady-state (circa 1850 AD) for the terrestrial biosphere with respect to decadal-scale carbon fluxes. Steady-state with respect to carbon flux means that the global simulation was carried out until the net flux of carbon between land and atmosphere (the net ecosystem carbon exchange, or NEE) was approximately zero as both a global integral and for individual model grid cells, when averaged over a 10-year period. These simulations were performed on a 0.5° x 0.5° grid, using surface weather fields (temperature, precipitation, shortwave and longwave incident radiation, humidity, and wind speeds) from reanalysis products for the period 1901-1920. Reanalysis products are based on historical observed atmospheric and surface conditions, processed by a numerical weather prediction code to generate gridded surfaces with global coverage.
Atmospheric CO2 concentration, rates of mineral nitrogen deposition, and land cover states for these steady-state simulations are based on observed or inferred conditions circa 1850 AD. These simulations used the Community Land Model (CLM4) in offline mode (driven by reanalysis weather fields, as opposed to coupled, with an atmospheric model) and included a new representation of the global-scale phosphorus cycle, based on recently published work. To quantify the sensitivity of the terrestrial carbon cycle and carbon-climate feedbacks to multiple nutrient limitations, we performed one steady-state simulation using the standard CLM4 representation of coupled carbon and nitrogen cycles (CLM4-CN), as implemented for the recent Coupled Model Intercomparison Project Phase 5 (CMIP5), and a second, independent steady-state simulation using a modified version of CLM that includes coupled carbon, nitrogen, and phosphorus cycles (CLM-CNP). Photosynthetic carbon uptake by terrestrial ecosystems is reduced by 22% globally due to the influence of limited phosphorus availability, with disproportionately large reductions for core tropical forest regions (~30% reductions over the Amazon region, ~40% reductions over Central Africa).
Global-scale evaluation of the relative limitations of the carbon cycle due to nitrogen and phosphorus availability confirm that phosphorus limitation dominates over most of the tropics and sub-tropics, with nitrogen limitation dominating over most of the temperate and high-latitude regions. Distinct spatial and temporal patterns of nitrogen and phosphorus limitation are evident in the South American, African, and Asian tropics. Phosphorus limitation reduces the sensitivity of the global net carbon flux to variation in temperature and precipitation, with the most significant influence over tropical forest regions with the strongest degree of predicted phosphorus limitation (e.g., Central Africa).
4th Quarter Metric - COMPLETED
Feedbacks between climate and the carbon cycle are important determinants of future climate system state, but are complex and driven by multiple interacting factors. Carefully designed simulation experiments allow a factor-by-factor dissection of the feedback components, including a quantification of interaction terms. A series of land model simulations over the historical period (1850-2009) are evaluated and feedback components are quantified at both global and regional scales. The simulations include prognostic carbon and nitrogen cycles, and consider rising atmospheric CO2 concentration, rising anthropogenic mineral nitrogen deposition, changes in land use and land cover, and varying climate, taken as single and combined forcings. Regional analyses of the climate driving factors show that precipitation anomalies are important drivers of land carbon stocks for the historical period. Results suggest that land has switched from a net carbon source to a net carbon sink in recent decades.
The contributions of multiple forcing factors to changes in terrestrial carbon stocks over the historical period 1850-2009 were quantified through a series of global-scale simulations, performed in an offline mode where the land model is connected to a gridded source of historical surface weather data as opposed to an active coupling with an atmospheric simulation model. Simulations were performed on a 0.5° x 0.5° grid, using surface weather fields (temperature, precipitation, shortwave and longwave incident radiation, humidity, wind speeds) from reanalysis products for the period 1901-2009. A repeating cycle of reanalysis forcing from the period 1901-1920 was used to drive simulations from 1850-1900. Reanalysis products are based on historical observed atmospheric and surface conditions, processed by a numerical weather prediction code to generate gridded surfaces with global coverage. Atmospheric CO2 concentration, rates of mineral nitrogen deposition, and land cover states were either held fixed based on observed or inferred conditions circa 1850, or were prescribed to follow their observed or inferred historical trajectory from 1850-2009. These simulations used the Community Land Model (CLM4) with coupled and fully prognostic carbon and nitrogen cycles.