Fundamental influence of carbon-nitrogen cycle coupling on climate-carbon cycle feedbacks Peter Thornton NCAR, CGD/TSS Collaborators: Keith Lindsay, Scott Doney, Keith Moore, Natalie Mahowald, Mariana Vertenstein, Forrest Hoffman, Jean-Francois Lamarque, Nan Rosenbloom, Beth Holland, Johannes Feddema

Overview Review carbon-nitrogen coupling hypotheses Summarize results from offline simulations New results from fully-coupled simulations Review disturbance history hypotheses Preliminary results from fully-coupled simulations with landcover change

Recent findings from field experiments Norby et al., Finzi et al., Hungate et al., Gill et al., 2006 (Ecology) CO2 fertilization shifts N from soil organic matter to vegetation and litter Some sites show progressive N limitation Reich et al., 2006 (Nature) Low availability of N progressively suppresses the positive response of plant biomass to elevated CO2 Perennial grassland: CO2 x N x (sp. diversity) van Groenigen et al., 2006 (PNAS) Soil C accumulation under elevated CO2 depends on increased N inputs (fixation or deposition).

Brief history of coupled C-N modeling Pre-1983: Decomposition models Parton et al., 1983, 1987, 1988, 1989, etc. CENTURY: soil organic matter and vegetation model, coupled C, N, P, and S dynamics McGuire et al., 1992 TEM: C, N interactions control net primary production (N.Am.) Rastetter et al., 1992, 1997 MEL: Effects of CO2, N deposition, and warming on carbon uptake Schimel et al., 2000 VEMAP: Terrestrial biogeochemistry model intercomparison Thornton et al., 2002 Biome-BGC: C, N, and disturbance history control net ecosystem carbon exchange (evaluation against eddy flux obs.) Liu et al., 2005 IBIS: NPP estimates improved when N dynamics added to a C-only model (evaluation against remote sensing NPP est.)

Brief history of coupled climate – carbon cycle modeling Cox et al., 2000 HADCM3 / MOSES / TRIFFID: Strong positive carbon-climate feedback. Joos et al., 2001 BERN-CC: Dufresne et al., 2002 LMD5 / SLAVE: Moderate positive carbon-climate feedback Thompson et al., 2004 CCM3 / IBIS: Strong negative CO2 feedback Fung et al., Doney et al., 2005 CCM3 / LSM / CASA’: moderate CO2 feedback, weak climate feedback Friedlingstein et al., 2006 C4MIP Phase 2: Eleven models, wide range of CO2 and climate (temperature) sensitivities.

Convergence toward a fully coupled carbon-nitrogen-climate model CENTURY MEL TEM VEMAP Biome-BGC IBIS+N Convergence toward a fully coupled carbon-nitrogen-climate model Carbon-Climate Hadley IPSL C4MIP Phase 2 Carbon-Nitrogen-Climate CCSM/CLM-CN ‘86 ‘90 ‘88 ‘94 ‘92 ‘96 ‘00 ‘98 ‘04 ‘02 ‘06

litterfall & mortality Carbon cycle Nitrogen cycle Atm CO2 Internal (fast) External (slow) photosynthesis denitrification Plant N deposition assimilation respiration litterfall & mortality Litter / CWD Soil Mineral N Mention the hypothesis of demand-based competition between plants and heterotrophs Mention plant stoichiometric constraints Mention that warming increases PSNS, respiration, and N mineralization – contrast tropics and boreal zone N fixation decomposition Soil Organic Matter mineralization N leaching

Schematic of terrestrial nitrogen cycle Image source: US EPA

C-N coupling: hypotheses Smaller CO2 fertilization effect, compared to C-only model Reduced ecosystem base state Stoichiometric constraints on new growth Progressive N-limitation: carbon accumulation in litter and soil leads to increased nitrogen immobilization Reduced carbon cycle sensitivity to temperature and precipitation variability Internal N cycling links plant and microbial communities

C-N coupling: hypotheses (cont.) Climate x CO2 response Expect changes over time in land carbon cycle sensitivity to variability in temperature and precipitation, forced by land carbon cycle response to increasing CO2. Different fertilization responses for increased CO2 and increased mineral N CO2: increased storage in vegetation carbon N: increased storage in vegetation and soil carbon

Offline simulation protocol Drive CLM-CN with 25 years of hourly surface weather from coupled CAM / CLM-CN. Spinup at pre-industrial CO2, N deposition, landcover Transient experiments (1850-2100) Increasing CO2 Increasing N deposition Increasing CO2 and N deposition Repeat experiments in C-only mode Supplemental N addition eliminates N-limitation Test dependence on base state

Evaluation of fluxes and states Thornton and Zimmermann, J. Clim. (in press)

Land biosphere sensitivity to increasing atmospheric CO2 (L) Veg C (Pg C) GPP (PgC/y) L (2100) CLM-C 1014 177 1.44 CLM-C2 771 146 1.25 CLM-CN 653 102 0.35 % change Veg C GPP L CLM-C2 -24% -18% -13% CLM-CN -35% -42% -76% CN : C2 1.5 2.3 5.8 Difference in beta at 2100 is not primarily due to altered base state CLM-C CLM-CN (CO2)  C4MIP models  C4MIP mean CLM-C2 Thornton et al., GBC (in review)

Spatial distribution of L 2000 2100 C-N C-N Biggest differences in tropics and above 40N. Relationship to base state. Negative beta due to increased fire under higher CO2 C-only C-only Thornton et al., GBC (in review)

NEE sensitivity to Tair and Prcp (at steady-state) Tair response not primarily due to altered base state, but prcp response is consistent with altered base state % change VegC GPP Tair Prcp CLM-C2 -24% -18% -17% -32% CLM-CN -35% -42% -77% -58% CN : C2 1.5 2.3 4.5 1.8 Thornton et al., GBC (in review)

NEE sensitivity to Tair and Prcp (at steady-state) C-N C-N Strong and potentially cancelling spatial patterns of uptake/release under changes in tair and prcp. Shifting dominance to N.Hem response in both T and P C-only C-only Thornton et al., GBC (in review)

NEE sensitivity to Tair and Prcp: effects of rising CO2 and anthropogenic N deposition Carbon-only model has increased sensitivity to Tair and Prcp under rising CO2. CLM-CN has decreased sensitivity to both Tair and Prcp, due to increasing N-limitation. Thornton et al., GBC (in review)

Fertilization responses to CO2 and mineral N deposition (period 2000 – 2100) Tot C (PgC) % Veg C % Lit C % SOM C C-N: CO2 204 79 13 8 C-N: Ndep 50 56 31 C-only: CO2 843 66 14 20 C-only(2): CO2 740 21 C-N gives qualitative match to observations from field experiments Changing the base state has negligible effect on partitioning of fertilization response Important because of order-of-magnitude differences in turnover times for vegetation and SOM pools, regionalization of sink permanence. Thornton et al., GBC (in review)

Summary of offline results C-N coupling results in large decrease in CO2 fertilization of land ecosystems. Sensitivities of net carbon flux to temperature and precipitation variation are damped by C-N coupling. C-N coupling reverses sign of trend in carbon-climate sensitivity under increasing CO2. Model agrees with experimental studies on contrasting effects of CO2 and N fertilization

Fully-coupled climate-carbon-nitrogen simulations CCSM3 framework: CAM + CICE + POP Doney/Moore/Lindsay ocean ecosystem + CLM-CN 1000-year stable pre-industrial control Historical + A2 (1890-2100), forced with fossil fuel emissions and nitrogen deposition. Coupled vs. fixed radiative effects of atmospheric CO2.

Land biosphere sensitivity to increasing atmospheric CO2 (L) Offline results Results from fully-coupled runs: Land sensitivity to increasing CO2 is not significantly different between offline and fully-coupled simulations Negative climate feedback of land carbon cycle

Changes in land carbon stocks: 1870-2100 Pool Change PgC % of total Plant 300 89% CWD 20 6% Litter 1 <1% Soil 17 5% Total 338 (25%)

NEE sensitivity to Tair and Prcp (at steady-state) CCSM3 Both C-only and C-N have the same sign for global mean response to tair and prcp, but smaller magnitude (buffering) Land carbon cycle sensitivity to variation in temperature and precipitation is not significantly different between offline and fully-coupled simulations

Climate-carbon cycle feedbacks CO2-induced climate change (warmer and wetter) leads to increased land carbon storage

Climate-carbon cycle feedbacks Vegetation C CWD C Climate change feeds back to carbon cycle in part through change in nitrogen availability Litter C Soil C

Climate-carbon cycle feedbacks Climate change impact on land carbon uptake closely related to changes in precipitation

Climate-carbon cycle feedbacks Fractions of 1870-2100 anthropogenic emissions in land, ocean, and atmosphere pools CCSM3.1-BGC C4MIP mean uncoupled coupled Land-borne fraction 0.15 0.18 0.29 0.21 Ocean-borne fraction 0.22 0.25 Airborne fraction 0.63 0.61 0.46 0.54

What about disturbance history and landcover change?

Results: NEE response to disturbance, changing [CO2]atm, and Ndep shows strong interaction effects

Fully-coupled simulations with prescribed landcover changes (prognostic fluxes) Using subset of data from Johannes Feddema Annual time slices from 1870-2100, originally on half-degree grid. Historical data from Ramankutty and Foley (1999), Goldewijk (2001), integration with present-day MODIS landcover. Landcover change for 2000-2100 based on SRES A2 scenario. New prognostic component in CLM-CN to handle carbon and nitrogen fluxes and mass balance associated with landcover change. Conversions to/from all plant functional types. Tracking two wood product pools in each land gridcell (10-yr and 100-yr turnover times). Fossil-fuel emissions, N deposition, radiatively coupled

Influence of landcover change on land carbon flux and global pools Grass Tree +200 ppmv net (by 2100) Shrub Agric +250 ppmv from land -50 ppmv from ocean

GPP falls and then recovers during landcover transition Plant respiration increases, due to replacement of woody with herbaceous biomass NPP falls and remains lower (GPP is offset by plant respiration) Soil respiration rises in response to conversion fluxes, then falls in response to reduced NPP

Carbon cycle impacts of landcover change Transient fluctuations in GPP as new ecosystem is established Reduced NPP due to increased plant respiration Depends on tree-to-ag vs. grass-to-ag transition and regional climate (biggest effect for tropical forest conversion) Increased soil respiration due to more labile litter inputs. Reduced potential for CO2 fertilization of land ecosystems, due to loss of area of woody vegetation.

Changes in vegetation carbon pools due to landcover change

Comparison to previous results (preliminary) IPCC TAR includes estimates of fluxes due to landcover change, for the A2 scenario, which are more than an order of magnitude smaller than those predicted here for the 21st century. Appear to ignore the mechanism of reduced sink potential. Gitz and Ciais (2003): EMIC. Landcover flux much larger than IPCC, smaller than CCSM3-BGC.

Conclusions Carbon-nitrogen cycle coupling has a fundamental influence on climate-carbon cycle dynamics Mechanistic treatment of landcover change suggests this signal will be a major factor influencing future atmospheric CO2 concentrations, and regional patterns of climate change. Next generation models should explicitly include the factors driving fossil fuel emissions and landcover change.