Determination of the Siberian Carbon Balance by Top-Down and - - PowerPoint PPT Presentation

Determination of the Siberian Carbon Balance by “Top-Down” and “Bottom-Up” Approaches -

• I. The Global Context: The Carbon Cycle in the Climate

System

Martin Heimann Max-Planck-Institute for Biogeochemistry PF 100164, D-07701 Jena, Germany www.bgc-jena.mpg.de/~martin.heimann martin.heimann@bgc-jena.mpg.de

1. Overview of the carbon cycle 2. Fundamental scientific questions 3. Prognostic comprehensive models of the oceanic carbon system and of the terrestrial biosphere 4. Coupled models of the carbon cycle - climate system

History of the Atmospheric CO2 Concentration

The Global Carbon Cycle in the Climate System

Global CO2 Budget [PgC a-1]

1980s 1990s

• 1. IPCC TAR [Prentice et al., 2001]

Atmospheric increase +3.3 ± 0.1 +3.2 ± 0.1 Fossil fuel and cement production +5.4 ± 0.3 +6.3 ± 0.4 Ocean –1.9 ± 0.6 –1.7 ± 0.5 Land (Net) –0.2 ± 0.7 –1.4 ± 0.7 Changes in landuse +1.7 (+0.6 to +2.5)

• Residual (Land)

–1.9 (–3.8 to +0.3)

• 2. Ocean correction [LeQueré et al., 2003]

Ocean –1.8 ± 0.8 –1.9 ± 0.7 Land (Net) –0.3 ± 0.9 –1.2 ± 0.8

• 3. FAO Statistics + Model [Houghton, 2002]

Changes in landuse +2.0 (+0.9 to +2.8) +2.2 (+1.4 to +3.0) Residual (Land)

• 2.3 (–4.0 to –0.3)
• 3.4 (–5.0 to –1.8)
• 4. Remote sensing + Model [De Fries et al., 2002]

Changes in landuse +0.6 (+0.3 to +0.8) +0.9 (+0.5 to +1.4) Residual (Land)

• 0.9 (-3.0 to 0)
• 2.1 (-3.4 to -0.9)

Units: PgC a-1, positive: flux into atmosphere

Basic Questions on Terrestrial and Oceanic Sinks

z Mechanisms - which process is responsible for the CO2 uptake? z Stability - how stable are these sinks? z Permanence - how permanently is the carbon stacked away? z Vulnerability - how vulnerable are the sink processes, e.g. with respect to climate feedbacks or anthropogenic impacts? z Attribution - can we uniquely attribute the sinks to a particular driver? z Spatio-temporal quantification - can we accurately quantify the sinks in time and space?

Tools

z Observation network (eddy covariance flux towers, atmospheric measurements, tracer measurements, inventories, remote sensing, etc.) z Process studies (in situ, manipulative, etc.) z Top-down methods z Bottom-up methods (process models, statistical extrapolations) z Data assimilation z Prognostic coupled carbon cycle - climate system models

Carbon Flows in Terrestrial Biosphere

Modelled Processes in a Terrestrial Dynamical Vegetation Model (DGVM)

Characteristics of Terrestrial Dynamical Vegetation/Carbon Cycle Models

z State variables: Biomass in different pools (above-ground, below-ground, woody, herbaceous, soil pools), leaf area, fractional vegetation cover by different functional vegetation types, nutrients (N, P), temperature and water in different vertical levels in canopy and soil z Plant physiology (e.g. photosynthesis) z Land surface water and energy balance z Biology (e.g. growth, allocation, phenology, senescence, diversity, nutrient interactions): based on empirical relations and optimization principles z Current developments: inclusion of age structure z System of non-linear ordinary first order differential equations z Grid-based (typical 0.5° lat-lon) with no interaction between grid cells

Example: LPJ Model: Sitch et al., GCB, 2003

Coupled Carbon Cycle - Climate Model Simulations

Emissions CO2 Atmosphere Vegetation Soils Ocean Carbon Atmosphere GCM Ocean GCM

Radiative forcing Land surface parameters Climate

Carbon Cycle Model Climate Model

Climate Change Scenarios with Coupled Climate- Carbon Cycle Models

Emissions [PgC a-1] CO2 Concentration Globally averaged near surface temperature Cox et al. 2001, Dufrene et al., 2001

Hadley IPSL

+5°C +3°C 1000ppm 750ppm

Simulated CO2 - Uptake by Land and Ocean

Cox et al. 2001, Dufresne et al., 2001 Friedlingstein et al., in press, 2003

Simulated Changes in Terrestrial Carbon Pools 1860-2100 in Hadley Model

Cox et al. 2001

Feedback Analysis

Hadley Model

Quantitative Feedback Analysis

Friedlingstein et al., 2003

Conclusions I

z Global carbon cycle behaves up to now almost like a linear system in response to the anthropogenic perturbation z First coupled climate model simulations demonstrate drastically different, but potentially significant positive feedbacks between carbon cycle and physical climate system z Non-linear effects become apparent during early 21st century - reduce CO2 sinks on land and ocean z Need for the development of comprehensive monitoring strategy for: y Early detection significant carbon cycle modifications (e.g. permafrost decline) y Model evaluation y Political motivation: corroboration of Kyoto target compliance

• 1. Methodologies
• 2. Top-down
• 3. Bottom-up
• 4. Outlook

Determination of the Siberian Carbon Balance by “Top-Down” and “Bottom-Up” Approaches -

• II. Estimating Regional Carbon Balances

Terrestrial Carbon Observing System - Siberia (TCOS-Siberia, EVK2-CT 2001-00131) 2002-2004

Principal Investigators: z MPI BGC Jena, Germany (Heimann, coordination) z LSCE, Saclay, France (Ciais) z IUP, University of Heidelberg, Germany (Levin) z RUG, Groningen, Netherlands (Meijer) z UNITUS, Viterbo, Italy (Valentini) z University of Amsterdam, The Netherlands (Dolman) z IPEE-RAS, Moscow, Russia (Varlagin) z IFOR-RAS, Krasnojarsk, Russia (Shibistova) z IBPC-RAS, Yakutsk, Russia (Maximov) z PIG-RAS, Cherskii, Russia (Zimov) z UNI.BIAL, Bialystok, Poland (Chilmonczyk) z UNI.FB.FBS, Ceske Budejovice, Czech Republic (Santruckova)

Methodology of CarboEurope Project (IP FP6)

Top-Down Approach: Inverse Modeling of Atmospheric CO2

Ingredients: z Observations at atmospheric station network + observation errors z 3d atmospheric transport model z A priori source/sink distribution + error covariances z Additional constraints z Inversion method

Global Atmospheric Background Station Network (2001)

Globalview (95) AEROCARB (20) TCOS-Siberia (8)

Near Surface Annual Mean CO2 Concentration Induced by Fossil Fuel CO2 Emissions (5.4 PgC a-1) [GFDL Model Simulation]

90 180 270 360 60 30

• 30
• 60

lon lat 2 4 6 8 10 12 ppmv

Continental Monitoring Locations of TCOS-Siberia

AEROCARB EUROSIBERIAN CARBONFLUX TCOS Siberia NOAA-CMDL (flasks) NOAA, MISU (continuous)

Aircraft profiles Surface sampling Tall Towers

Zotino/Bor (60N, 90E)

Continental Measurements: Terrestrial Carbon Observing System - Siberia

Profiling and sampling of lower troposphere (up to 3000m) by light aircraft Direct CO2 flux measurements in key ecosystems by means of eddy covariance technique

Typical Daytime Summer Profiles Zotino, 60°N, 90°E

Observed Seasonal Cycle of CO2 in Lower Troposphere above Zotino, 60°N, 90°E

Year ppm

Seasonal Cycle of CO2 at Zotino (60°N,90°E)

Zotino, PBL Zotino, free Troposphere (>~3km) North Atlantic Surface, 60°N (ICE, STM, NOAA-CMDL) ppm

Atmospheric Model Nesting

Meteorology CO2 Transport model CO2 Source Global Mesoscale 1/2° Mesoscale 1/6° local ECMWF TM3 TURC Canopy-CBL Model REMO 0.5° REMO 1/6° TURC

B,I B,I

Concentration of CO2 in PBL (~300m) REMO 0.5° Regional Model Simulation

Main monitoring sites

Footprint Analysis TCOS - Siberia Sites ( dC/dQ, relative units)

Z=500m Z=2500m

Monthly Mean W-E Composite Concentration Gradient at 60°N, July 1998

REMO (0.5°x0.5°x19L), TM3 (4°x5°x19L)

• 20

20 40 60 80 100 120 Longitude

• 4
• 2

2 4 m p p Composite Gradients at 60N, TM3, REMO, July

3000m 250m

Top-Down Approach: Inverse Modeling of Atmospheric CO2

Ingredients: z Observations at atmospheric station network + observation errors z 3d atmospheric transport model z A priori source/sink distribution + error covariances z Additional constraints z Inversion method

Inverse Modeling Strategy - Bayesian Approach

Transport Model T Space of Concentrations C(x,t) Space of Sources Q(x,t) Null Space A Priori Source A Priori Source Uncertainty “Optimal” Solution Concentration Uncertainty Solutions Compatible with Observations+Uncertainty Observations

Atmospheric Inversion Problem is Ill-Defined

Computational Aspects: Efficient Computation of Sensitivities dCi/dQj

• Use of Adjoint Atmospheric Transport Model

Example: global, 10-year inversion of surface sources: z Number of monthly observations: ~ 10yr x 12 months x 100 stations =~ 12’000 data points z Number of unknown monthly source values (4°x5° global grid): ~ 72 x 48 x 12 months x 10 yr =~ 400’000 source values z Dimension of transport matrix: 12’000 x 400’000

Net CO2 Sources (Average 1995-2000)

Long-term observation stations NOAA/CMDL gC m-2 a-1

Fossil+Ocean+Land Ocean+Land

Rödenbeck et al., ACP, 2003

Net CO2 Sources (Average 1995- 2000)

Long-term observation stations NOAA/CMDL gC m-2 a-1

Fossil+Ocean+Land Ocean+Land

Rödenbeck et al., ACP, 2003

Temporal Variability of Natural CO2 Sources Estimated from Atmospheric Measurements by Inverse Modeling

Results of different set-ups (No. of observation stations, a priori sources, etc.)

Rödenbeck et al., ACPD, 2003

Regional Estimates Less Certain Due to Poor Density of Observation Network

Rödenbeck et al., ACPD

PgC a-1 PgC a-1

Anomalous Non-Fossil Fuel CO2 Sources 1995-2000

gC m-2 a-1

Drivers of Interannual Variability: Precipitation? El Niño 1997/8

Rödenbeck et al., ACPD, 2003

CO2 Sources and Biomass Burning

CO2 Source Flux (2 setups, Units: PgC a-1) Firecounts (arbitrary scale)

Rödenbeck et al., ACPD, 2003

Preliminary Inversion for Siberia Including First Measurements from TCOS-Siberia

B F/T S Z T/H Y C K

1-spost/sprior

Tools for the Bottom-Up Approach

z Process-based ecosystem models (DGVMs) z Process-based ocean carbon cycle models (OCCMs) z Extrapolation techniques from point measurements (i.e. artificial neural networks) z GIS information (i.e. anthropogenic emissions, land use, vegetation type) z Remote sensing of surface properties from space (i.e. vegetation index, above ground biomass)

PAR, T2m, FCO2 Zotino (90°E, 60°N)

[Pinus Silvestris, age: 200yr, 7-day running means]

10 20 30 40 50 60

• 40
• 30
• 20
• 10

10 20 30

• 0.4
• 0.3
• 0.2
• 0.1

0.1 0.2

1998 1999 2000

PAR T2m FCO2

Wm-2 °C mol m-2 d-1

Evaluation of LPJ-DGVM with Eddy Covariance CO2 Flux Observations (Process- Based Model)

Sitch et al. 2003

Artificial Neural Network for Gapfilling and Upscaling - Validation (“Data Driven Model”)

Drivers: NDVI Temperature Landcover Fuzzy variables (seasons)

Papale and Valentini. 2003

Weekly NEE Diurnal Cycles Bordeaux

Estimated Terrestrial Carbon Flux, July

NEE-ANN NEP-LPJ NET- Inversion

gC m-2 month-1 gC m-2 month-1 gC m-2 day-1

Consistency? Top-Down and Bottom-up Estimates of CO2 Fluxes

gC m-2 a-1

Mean annual total flux 1995-2000 based on inversion of NOAA-CMDL data [Rödenbeck et al., 2003] Ocean flux [Takahashi, 1998] + Fossil fuel emissions 1990 [EDGAR] + LPJ Model 1980-89 [McGuire et al., 2001]

gC m-2 a-1

Top-down and Bottom-up Estimates of Total CO2 Flux (Fossil + Land + Ocean) over Northern Eurasia

Carbon Balance of Siberia in the Global Context

Ciais et al.. 2003 in press

Conclusions II

z Top-down approach: y Provides large-scale constraints y Ill-posed mathematical problem y Atmospheric model error difficult to quantify y Requires high-quality multi-year concentration measurements z Bottom-up approach: y Direct flux observations representative only for a very small region y Errors of up-scaling techniques (GIS, remote sensing data, process models, neural networks) difficult to quantify

Conclusions III - the Way Forward

z “Data assimilation”: consistent merging of top-down and bottom-up method y Combined surface-atmosphere (regional-) model with included ecosystem process model y Optimization of model parameters y Problems: x Nonlinearity of surface ecosystem processes x Possible unknown biological processes, drivers x Appropriate time-space scale representation of observations and model output z Remote sensing of CO2 concentration from space using passive (e.g. near-infrared) or active sensors (lidar techniques) y Advantage: high temporal and spatial coverage of entire globe y Problems x Accuracy (<<1%) x Key terrestrial concentration signals in PBL

Elements of a Global Observing System

High-precision

• bservations of

atmospheric biogeochemical trace species (concentration and fluxes)