Component fluxes of the Carbon balance of Europe and their - - PowerPoint PPT Presentation

Component fluxes of the Carbon balance of Europe

and their uncertainties

Laboratoire des Sciences du Climat et de l’Environnement Gif sur Yvette, France Philippe Ciais

Fossil fuel emissions drive the increase of CO2

Canadell et al. PNAS, press

Land and ocean absorb ≈ 55 % of emissions on average

Land sink is the most uncertain term

Everybody’s tired. Let’s make a deal …

Uncertainties are Large But Don’t be affraid of The biosphere !

Atmospheric and Ecosystem Observations to constraint the C budget of large regions

Atmospheric Concentration Networks Ecosystem Flux Networks

The knowledge challenge

Starting point European ecosystems carbon budget

• 600
• 500
• 400
• 300
• 200
• 100

100 200 300 400 500 600

Forest, woodland Grass- land Crop- land Peat use

Tg C/year

Bottom up Ecosystem estimate Top down Atmospheric estimate

Janssens et al., 2003 Science

Reminder : Emissions = 1200 TgC y-1

• Estimates of NPP, HR for major European ecosystems
• Estimates of NBP (carbon balance)
• Underlying processes
• Trends
• Variability

CO2

Photosynthesis Plant respiration Soil and litter respiration Disturbance (fire, harvest)

Short- term carbon uptake

NPP

Medium- term carbon storage

NEP

Long-term carbon storage

NBP The Net carbon balance (NBP) Is the small difference of two larger gross fluxes : NBP = (NPP) - (Soil and litter respiration + disturbance)

Goals of this study

• Estimates of NPP, HR of major

European ecosystems

• Associated estimates of NBP
• Underlying processes
• Variability
• Trends

Two different planets

• Global Biogeochemical models
• > This presentation

– NOT Simple – INdirect link to data – DONT account WELL for management – HARD to compare with each others – Dynamic evolution of ecosystem flux in response to weather, climate and atmospheric changes – Account for (some) of ecosystem heterogenity – CAN account for ALL ecosystems (if you’re not afraid of simplifications)

• Emission factors
• > This Meeting

– Simple – Directly based on data – Account for management – Can be compared with each

• thers

– ‘static’ Do not account for ecosystem flux variability and trends in a changing world – ‘rigid’ Do not account very well for ecosystem heterogenity – DONT account for ALL ecosystems

Biogeochemical models

This is a forest

Emission factors

Flux = EF * Area This is a forest

Example of interannual variability, the 2003 heatwave impact on Europe‘s carbon balance: A 0.5 PgC loss event

The knowledge challenge

Ciais et al. Nature 2005

Models without data are like fantasy…. … But data without a model often look like chaos

Need to combine both

Forests

Forest

• Data and Method

– National forest inventory diameter increment (sample based inventories) – NPP is modelled using expansion factors and leaf + root turnover rates – NBP is modelled as biomass increment + soil carbon balance

Evolution of stocks vs. NPP since 1950

Compare : Odum Paradigm (one forest)

Litterfall does not tail with NPP for conifers, and moderately increases with NPP for broadleaved Carbon saving sylvivulture

Contribution of climate and CO2

Estimated with biogeochemical models < 50%

CO2 + Climate effects

Forests: conclusions

• NPP = 280 gCm-2 y-1
• NBP = 50 gCm-2 y-1
• NPP possibly underestimated

– Biogeochemical models ≈ 370 - 550 gCm-2 y-1 – Ecological data ≈ 400-600 gCm-2 y-1

• Stocks increased by 75% since 1950
• Carbon stocks have remained

proportional to NPP in nearly all European countries

Factors driving the forest C balance

• Changes in forest area alone do not explain

the stock increase

• Management-related factors

– Juvenile age distribution – Harvesting less than the increment (policy favoring high forest stands) – Better nitrogen recycling in soils

• Biogeochemical factors
• N deposition (Magnani et al. 2007)
• CO2 and climate effects <50%

Croplands

Croplands

• No pan-European soil C inventory
• NPP is inferred from FAO national yield

statistics

• NBP is the modelled soil C balance driven by:

– Recent past changes in farmers practice, – cultivar species, – CO2 and climate – [assumes no C inputs to soils ; 100% oxidation of harvest]

ORCHIDEE-STICS European model of cropland carbon cycling

Leaf Area Index Root density profile Canopy height Nitrogen stress on photosynthesis Date of harvest Amount of irrigation Harvest index Temperature Precipitation SWdown radiation Air humidity Wind LWdown radiation

ORCHIDEE

Global Vegetation Model

STICS

Crop Model Wheat / Corn

Soil texture Soil depth Land cover

Carbon fluxes (hrly) GPP, NPP, yield Carbon pools (daily) Water and energy fluxes (hrly) Soil water content (daily)

• Crop phenology
• Root dynamics
• Nitrogen cycle
• Above-ground carbon dynamics
• Land atmosphere Exchange
• Allocation and growth
• Mortality
• Soil organic matter decomposition

Input practice data (annual) Input land data (fixed) Input climate data (hourly) Output fields

Model integration over 1901-2000

1900-1950 Organic fertilisers 2 t ha-1 yr-1 (32 kgN ha-1 yr-1) 1951-2000 Mineral fertilisers from 32 to 150 kgN ha-1 yr-1 1951-2000 Increase of harvest index (linear) from 0.25 to 0.45 1951-2000 Irrigation from 0 to 200 mm yr-1 (maize only) 1900-1980 Short cycle varieties for wheat & maize 1981-2000 Long cycle varieties for wheat & maize 1900-2000 Increase of soil carbon pools' turnover by 10% 1901-2000 from 300 to 370 ppm

S3

Organic fertilisers 2 t ha-1yr-1 1901-2000 from 300 to 370 ppm

S2

Organic fertilisers 2 t ha-1yr-1 1900 from 300 to 370 ppm

S1

Organic fertilisers 2 t ha-1yr-1 1900 until equilibrium 300 ppm

CNT Farming practice

Sowing date Crop Variety (short / long season length) Nitrogen fertilizer amount Fertilizer type (organic / mineral) Irrigation (yes / no)

Climate Atmospheric [CO2]

Simul ation

Model evaluation against historical yield data

Modelled cropland soil C balance

‘Best’ case

If no mineral fertilizers If ploughing X2 If no increase in harvest index

Large sensitivity to, and legacy of past practice

Derivative of C(t) = NBP

CO2 CO2 + Climate CO2 + Climate + Practice

1901 2000 - 1901

Changes in cropland NPP and soil C

Gervois et al. submitted

Cropland NPP, NBP distribution during the 1990s

Croplands: Conclusions

• NPP = 660 gC m-2 y-1
• NBP = 8.4 gC m-2 y-1
• Soil C is a small sink,

– Agrees with crop model analysis Smith et al. 2005 – Disagrees with former CESAR model results – Regional inventories show moderate to large sources

• Uncertainties on cropland NBP due to

– Ploughing impacts on soil C turnover rates – Crop rotation – Historical changes in varieties – Straw and residues fate

Factors driving the croplands C balance

• NBP mostly determined by past practice change
• Rebuilding stocks as a rebound from intensive

ploughing 30 years ago

• Rising CO2 has a <10% effect in increasing

yields, but a 30% effect in increasing in water use efficiency

• Spring dryness trends have negatively impacted

yields in Iberian Peninsula, but this effect is compensated by practice changes

 Area and number of animals

• 20% of Geographic Europe’s area (140 Mha)
• 140 millions of sheep / 140 millions of cows

 Fluxes

• CO2
• annual yield production

~ 1 - 6 t C ha-1 y-1

• animal respiration

~ 1 t C y-1

• N2O
• soil emission

0.1-1 t Ceq ha-1 y-1

• CH4
• animal emission

0.3-1.5 t Ceq y-1

 Very few continental scale estimates and large uncertainties

• carbon balance

→ empirical model

• soil N2O emissions

→ emission factor

• animals CH4 emissions

→ emission factor

Importance of european grasslands

= 101 ± 133 TgC y-1 (Janssens, 2001) = 259 ± 75 Gg N2O y-1 (Freibauer, 2003) = 6.8 Mt CH4 y-1

Grasslands

Grasslands

• No pan-European NPP data (virtually impossible

to measure)

• No soil C inventory to measure NBP
• GPP and NPP modelled at continental scale ;

but estimating NBP was not tackled

• NBP/GPP ratio estimated at 9 eddy-covariance

sites and used in upscaling NBP = (NBPsite/GPPsite)*GPPmodel

PaSim grassland model

photosynthesis substrate allocation lamina stem ear Carbon and Nitrogen BIOMASS mortality Carbon and Nitrogen litter (structural / metabolic) decomposition soil carbon and nitrogen cutting yield grazing animals CH4 autotrophic respiration active slow passive heterotrophic respiration CO2 root N uptake N fertiliser N deposition nitri/ denitrification volatilization leaching N2O NO3- NH3 N2O N2 N fixation urine dung

Nitrogen outputs Nitrogen Inputs Carbon Inputs CO2 emissions CH4 emissions

yield decomposition

GPP Soil HR Export

NPP NBP

CH4 N2O Import Run at equilibrium, thus HR equalling NPP Tested against 9 eddy coveriance site level GPP and NEE Management (grazing / cut) intensity distribution is calculated High N inputs / zero N inputs end-members

From site to Europe

PaSim

Hourly values of

• irradiance
• temperature
• pressure
• humidity.
• wind speed

Climatology from ECMWF Grassland fractional coverage Combined CORINE. PELCOM

• Soil texture (Zoebler)
• Water content parameters

(FAO) dates of harvest. animal stocking rate and grazing periods. dates of application and amount of N- fertilizers

Land cover map Soil data Management drivers Equilibrium run at a spatial resolution of 1 degree

• GPP
• Respirations
• N2O emissions
• CH4 emissions

Model outputs

Automatic optimal management module

Cut meadows Pastures

Modelled NPP and N2O emissions

Vuichard et

• al. GBC

2007ab

Grasslands: Conclusions

• NBP = 104 gC m-2 y-1
• NPP = 714 gC m-2 y-1

– Pastures = 1220 – Meadows = 520

• NPP is high-biased in the PaSim model

(absence of extensive management)

• Large uncertainties in site NBP upscaling

Soussana et al. 2007 Vuichard et al. 2007

• Fire emissions

– Regrowth after fire is already counted in inventories

• Fluxes involving lateral transport

– VOC emissions and deposition – Geological carbon cycle and river transport – Trade of food and wood products

Other component fluxes

Fire emissions

• Van der Werf et al. 2006
• CASA carbon stocks *

MODIS burned areas

• Average over 1997-2005
• Cropland 3.3 gC m-2 y-1
• Forest 2.3 gC m-2 y-1
• Grassland ≈ 0
• Total emission = 11.4 TgC y-1

• Plant BCOV emission model and anthropogenic emissions inventories
• Oxidation in PBL, transport and deposition calculated using 3D

Chemistry Transport model LMDZ-INCA

• Net carbon source = RCC escape - deposition = 76 TgC y-1

Lateral flux : Reduced Carbon Compounds budget

• Estimated from FAO agricultural statistics
• Net carbon source of 24 TgC y-1

Lateral : Trade of wood and food

River transport to estuaries mouth

• Includes DIC, DOC and POC
• Freshwater and estuaries CO2 outgassing
• Carbon Burial in sediments
• Rock weathering
• A net atmospheric carbon sink of 113 TgC y-1
• Fluxes estimated from river data synthesis of Abril et al.

and Meybeck et al.

Net Biome Productivity (summary)

• Grasslands is a sink 2 times larger than

forests (per unit area)

• Croplands are a smaller C source than

formerly estimated

• Both results cannot be verified by data

NBP = Forest + Crop + Grass - fires = 334 + 11 + 48 - 11 = 382 TgC y-1

Net Carbon Balance

NCB = NBP + COVs COVs + Trade + Rivers = 382 - 76 - 24 + 113 = 395 TgC y-1

New estimate of the European C budget

• 600
• 500
• 400
• 300
• 200
• 100

100 200 300 400 500 600

Forest, woodland Grass- land Crop- land Peat use

Tg C/year

Bottom up Ecosystem estimate Top down Atmospheric estimate

Lateral effects

sink source

Reminder : Emissions = 1200 TgC y-1

395 TgC y-1

Uncertainties

• Not assessed rigorously

– Comparison with available data – Model parameter sensitivity tests – Spread of independent studies

• Forest NBP -> soils -> 30%
• Cropland NBP -> 50%
• Grassland NBP -> 50%

New Net Carbon Balance estimate

Comparison with inversions

Careful experimentalist Dary Modeler

Unknown Truth

Thanks to …

• S. Piao, G. van der Werf, S. Luyssaert, S.

Zaehle, G.J. Nabuurs, J. Grace, P. Smith, J.F. Soussana, I. Janssen, I. Levin, M. Jung, M. Reichstein, E.D. Schulze, N.Vuichard, M. Vetter, G. Churkina, C. Beer, A.D. Friend, A. Hastings, U. Karstens, D. Papale, M.-J. Schelhaas, N. Viovy, M. Wattenbach, M. Heimann, R. Valentini and CARBOEUROPE members

Recent trends in the European carbon balance

Atmospheric CO2 trends over Europe

A decrease in northern hemisphere terrestrial sink ?

13C latitudinal gradients

inverted in double deconvolution (Miller et al.

• pers. com.)

Anomalies for the period June 1998 to May 2002 relative to climatology (Zheng et al. 2004)

Marine & Continental sites Marine & Continental sites over

• ver Europe

Europe

AZR, IZO, ICE, STM, BAL, BSC

ESRL/NOAA (flasks)

MHD, PUY

LSCE (In-situ)

ORL

LSCE (airborne flasks)

SCH

IUP-UBA (In-situ)

HH1

HMS (In-situ)

PAL

FMI (In-Situ)

Spectacular increase in the annual mean CO2 difference to Mace-Head

Is this signal caused by fluxes or transport ?

+ 15 % y-1 suggests increasing emissions or decreasing sink over Europe at a rate of ≈ 0.3 GtC y-1 (BIG !)

Trend of NEP from ecosystem models 1998-2005

• M. Jung pers.

Communication (Carboeurope model results)

KgC m-2 y-2

More gains More losses

Losses < 0.12 PgC y-1, compensated by gains further East

Mean PBL (day) trends (m/year) from ECMWF operational analyses

Summer JJA Winter DJF

Summer (MJJA) Year

Mean PBL trends (m/year), from in-situ radiosoundings

T 2m Short-Wave Rad Sensible HF Soil Moisture

Trends in surface energy budget (MM5 regional model simulations)

Trends in sensible Heat Flux (% year-1) at eddy-covariance flux towers

+2 to +4 % y-1 in the last four years

• M. Reichstein
• D. Papale pers. Com.

Conclusion on trends

• Fast and large increase in surface CO2

accumulation over Europe in the 1990s

• Linear trend, which excludes the singular

effect of extreme year 2003

• Trend in increasing PBL height can explain

30% of the signal (corroborated by sensible HF increase)

• Rest could be due to decrease in NBP or to

regional re-distribution of emissions

Spatial and temporal effects of climate on the European carbon balance

Spatial distribution of ecosystem fluxes controlled by climate

Lluyssaert et al., from biometric site studies

Reichstein et al., from flux tower networks

 GPP and TER show parallel control by climate,  Temperature in the North, water in the South  Climate controls partly cancel each

• ther in NEP

 NEP controlled by GPP and water  All for middle-aged stands

Models capability to reproduce spatial flux distribution controlled by climate

Boreal forest sites

Models capability to reproduce spatial flux distribution controlled by climate

Southern forest sites

Temporal variations in ecosystem fluxes

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 GPP Reco NEP stdev [kg m-2 yr-1] fed Spatial- Continental Interannual

• S. Piao, Factorial simulations using the ORCHIDEE model

(gc m

• 2

yr

• 1

)

• 150
• 100
• 50

50 100 150

• 150
• 100
• 50

50 100 150

(gc m

• 2

yr

• 1

)

• 150
• 100
• 50

50 100 150

• 150
• 100
• 50

50 100 150

Year

1980 1984 1988 1992 1996 2000

(gc m

• 2

yr

• 1

)

• 120
• 80
• 40

40 80

• 120
• 80
• 40

40 80

CO2 only Temperature only Precipitation only All factors

Slope: 3.406 gc m-2 yr-2 Slope: 0.857 gc m-2 yr-2 Slope: 0.691 gc m-2 yr-2 Slope: 4.413 gc m-2 yr-2 Slope: 2.508 gc m-2 yr-2 Slope: 1.385 gc m-2 yr-2 Slope: 0.807 gc m-2 yr-2 Slope: 4.473 gc m-2 yr-2 Slope: -0.898 gc m-2 yr-2 Slope: 0.529 gc m-2 yr-2 Slope: 0.116 gc m-2 yr-2 Slope: 0.059 gc m-2 yr-2

R2(T) - R2(P) = % of variance explained by temperature (>0) or by rainfall (<0)

• Temperature

controls flux variability in Northern Europe

• Precipitation

controls variability in Southern Europe (< ≈ 53°N)

• NEP sensitivity to

cliate variability is less sensitive to each factor than gross fluxes

Conclusions on variability

• Average GPP and TER spatial distribution :
• Both fluxes are temperature driven in the North, and water driven

in the South [from eddy data]

• Good agreement between site-level data, and models for GPP and

TER spatial distribution

• NEP gradients not clearly related to climate (GPP and TER

compensations)

• GPP and TER temporal interannual variations :
• Smaller, but similar regional response to rainfall and temperature

than mean gross fluxes [from ORCHIDEE model]

• NEE variability appears GPP driven in the South and TER driven in

the North

• More complex influence of climate variable on NEE interannual

variability (GPP, TER) than the widespread observed one in summer 2003

Future work

• Better separation of interannual and

spatial variability from eddy data

• Deconvolution of age, management and

climate effects on gross fluxes

• Improve consistency of estimates
• Comparison with regional inversions

Coupling with uncertainty analysis Extend to regional (subcontinental) analysis

Thank you for your attention