Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.



No description

Max R

on 19 October 2016

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of 2015_EGU

T h a n k s
is important for, e.g., carbon cycle scientist, climate modelers, and policy makers,
quantifies the annual net carbon uptake of the European biosphere,
is measured in units of GtC/a (=PgC/a).
The satellite inferred European carbon sink
The European
land sink...
The global land sink
Air sampling networks
have the potential to reduce existing uncertainties,
provide global coverage and
more measurements but with
less precision and accuracy and
variable sampling,
can be used in existing inversion schemes using a
column-average observation operator
Regional satellite inversion
M.Reuter, M.Buchwitz, M.Hilker, J.Heymann, O.Schneising, D.Pillai, H.Bovensmann, J.P.Burrows, H.Bösch, R.Parker, A.Butz, O.Hasekamp, C.W.O'Dell, Y.Yoshida, C.Gerbig, T.Nehrkorn, N.M.Deutscher, T.Warneke, J.Notholt, F.Hase, R.Kivi, R.Sussmann, T.Machida, H.Matsueda, and Y.Sawa
Area: 10¹³m²
1GtC corresponds to
A cube of graphite with 762m edge length
A layer of graphite with 0.04mm thickness (1/3 of a human hair)
The European
From atmospheric concentrations to surface fluxes with global mass conserving inversion models
Global inverse models like CarbonTracker use roughly 100 ground based in situ measurement sites
Inversion procedure
Peylin et al. 2013 (IPCC, AR5)
Summarizes the results of 11 in situ based global inversions
Result for Europe: 0.40+/-0.42GtC/a
Uncertainty still large, e.g., due to sparseness of measurements
The European land sink in the literature (in situ)
Satellite data ...
The European land sink in the literature (satellite)
Satellite data sets
Background model
Retrieval biases
Aggregation error
Integration time
and statistical setup
The background model contributes the a priori fluxes and consistent atmospheric concentrations
It can introduce artificial bias patterns
Replacement of CT2011 with MACC v11.2 results in an RMSD of 0.22GtC/a
The used convection scheme influences the particle transport of STILT
Switching from vertical wind speeds to Convective Available Potential Energy (CAPE) parametrization results in an RMSD of 0.09GtC/a
A posteriori uncertainty
The a posteriori uncertainty accounts for errors defined in the measurement and a priori covariance matrices
This uncertainty estimate is incomplete and other error terms are considered via an ensemble approach
The a posteriori uncertainty of annual averages is typically 0.15GtC/a
The atmospheric transport model used for STILT can influence the results
Using the ERA Interim instead of the NCEP/NCAR reanalysis results in an RMSD of 0.32GtC/a
We derive Europe-wide, monthly flux increments
This can be interpreted as “hard constraint”
Hard constraints can result in spatial or temporal aggregation errors
Splitting the domain in N/S, E/W and in the middle of the months results in an RMSD of 0.08GtC/a on average
Persistent retrieval biases in mean wind direction (W/E) can be misinterpreted as source or sink
Splitting the SCIA data set in two sub sets by wind direction results in flux differences <0.1GtC/a
Variations due to the statistical setup are typically <±0.1GtC/a
Flux results level off well before 480h integration time
The total inversion uncertainty includes:
The ensemble spread of 25 inversion setups
The uncertainty due to the statistical setup and potential retrieval biases
The ensemble median sink in 2010 is 1.0±0.3GtC/a
How can we show that the optimized fluxes are more realistic than the a priori fluxes?
We compute optimized concentrations from the optimized fluxes and compare them with independent measurements
In situ
In situ measurements from NOAA’s cooperative air sampling network are assimilated by CT2011
An improvement cannot be expected
Largest increments at stations with large European surface influence
Slightly degraded seasonal cycle
TCCON column measurements are not assimilated in CT2011
Optimized concentrations are in slightly better agreement with TCCON than CT2011 conentrations
Marginal improvements of station-to-station biases (below TCCON's accuracy)
Slightly improved seasonal cycle with larger amplitude than in CT2011
CONTRAIL aircraft measurements are not assimilated in CT2011
Optimized concentrations are in better agreement with CONTRAIL than CT2011 concentrations
Upper troposphere
Only marginally differences between optimized and CT2011 concentrations because SI is small in high altitudes
Lower troposphere
Improved seasonal cycle with larger amplitude than CT2011
The CO2 absorbed in Europe must come from somewhere
Two potential explanations
Global models like CarbonTracker apply strict mass conservation
I.e., changed fluxes in European have to be compensated elsewhere
Emissions are underestimated
Less is absorbed elsewhere
Fossil fuel emissions are prescribed in global inversion models
Anthropogenic emission data bases show considerable uncertainties (outside Europe and North America)
Emission data bases for China vary by 0.38GtC/a (Guan et al., 2012)
Satellite data suggest a larger increase of emissions in East Asia compared to emission inventories (Schneising et al., 2013; Reuter et al., 2014)

Other regions may have also large uncertainties
However, its unlikely that all unaccounted emissions travel around the world only to be absorbed in Europe
The sparseness of in situ sites could hinder the models to discriminate between the European and Eurasian regions
Basu et al., 2013
The global inversion model used by Basu et al. (2013) finds a large sink in the Eurasian boreal TRANSCOM region and a weaker sink in Europe.
The model shifts the sink towards Europe when assimilating satellite XCO2 measurements
Schneising et al., 2011
This hypothesis is supported by Schneising et al. (2011) suggesting that the Eurasian boreal forests are a weaker sink than expected from CarbonTracker
This study
Inverting only SCIAMACHY measurements in the surrounding of assimilated in situ sites results in a considerably smaller sink (0.58±0.37GtC/a) compared to the full data set (0.90±0.33GtC/a)
IPCC, AR5, 2013
Basu et al. 2013
European sink: 1.0 - 1.3GtC/a
Satellite: GOSAT (RemoTeC)
Discussion: none
Takagi et al. 2014
European sink: >1.0GtC/a
Satellite: GOSAT (NIES, PPDF, ACOS, RemoTeC, UoL-FP)
Discussion: none
Nassar et al. 2011
European sink: 1.3GtC/a
Satellite: TES
Discussion: none
Chevallier et al. 2014
European sink: 1.2 - 1.8GtC/a
Satellite: GOSAT (ACOS, UoL-FP)
Artifacts probably due to
Retrieval biases or
Transport errors
The SI quantifies how much the concentration in air changes when the surface flux changes
STILT is a particle dispersion model designed to calculate SI
We use STILT to calculate the European SI in ppm/(GtC/a)
Air masses have to be followed only a couple of days
This reduces potential issues due to long range transport
No influence from potential large scale biases
Information only from inner-European gradients
Limit satellite data
to Europe
CarbonTracker (CT) is a global inversion model
CT concentrations are consistent with CT fluxes
Plot satellite-CT ΔXCO2 vs. SI
If slope is zero, CT fluxes are correct
The slope equals the flux increment
The offset equals the Bias between Satellite and CT
Satellite-Model vs. SI
Account only for immediate European surface influence
Some math
The satellite - CT difference can be described by the following matrix equation:
Dimension: (n,1)
n ≈ 150000 for 8 years SCIA
Dimension: (n,1)
Dimension: (m,1)
m = 192 (96 monthly fluxes and biases)
A priori state vector
(CT fluxes, no bias)
This is what we want to know!
Dimension: (n,m)
Includes surface influences and
Unity for biases
State vector
Measurement vector
We use the optimal estimation formalism to solve for Δx and get optimized fluxes
This equation can also be used to simulate optimized concentrations
Largest error reduction during growing season because
Many measurements
Largest a prior uncertainty
Consistently larger sink
Larger year-to-year variations may imply larger ecosystem sensitivity which is qualitatively consistent with Schneising et al. 2014
Weakest sink in 2003 (heat-wave / drought, Ciais et al., 2005)
Similar to results of Nassar et al. 2011
Consistently larger sink derived from 5 independent satellite data sets
CT2011: 0.4GtC/a, Satellite: 1.0±0.3GtC/a
The difference is large compared to the modeled sink but similar to the IPCC inter-model spread (Peylin et al. 2013)
Uncertainties have been calculated from the a posteriori errors, ensemble spread and sensitivity studies
Similar seasonal cycle for model and satellite (except for RemoTeC)
Largest sink in June
Strong respiration in dormant season cannot entirely be excluded but this is unlikely because the a priori is relatively well known
Future satellites will provide more data points with better precision and accuracy plus simultaneous chlorophyll fluorescence measurements which will reduce the existing uncertainties
Future inversion methods would profit from an optimization for specific characteristics of satellite data (e.g., by fitting temporal or spatial large scale biases)
Our result suggest that Europe is a considerably larger carbon sink (1.0±0.3GtC/a) than previously estimated (0.4±0.4GC/a)
The uncertainties of the presented results originate in large parts from uncertainties and the sparseness of the satellite data:
Where exactly in Europe is the carbon sink
Uncertainties of 0.3GtC/a are still very large
Full transcript