ECMWF COPERNICUS REPORT Copernicus Atmosphere Monitoring Service Validation report for the inverted CO 2 fluxes, v18r1 – version 1.0 D1.4.1-2017_v0 (Evaluation and Quality Control document for CO 2 flux inversion 2017 (Period covered 1979 to year 2017) v0 Issued by: CEA / Frédéric Chevallier Date: 23/12/2018 REF.: CAMS73_2018SC1_D73.1.4.1-2017-v0_201812_Validation inverted CO2 fluxes_v1
20
Embed
Validation report for the inverted CO fluxes, v18r1 ...€¦ · Jubany, Antartica, AR (JBN) 1994-2009 WDCGG/ ISAC IAA Jungfraujoch, CH (JFJ) 2004-2018 NOAA/ Univ. Of Bern Jungfraujoch,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ECMWF COPERNICUS REPORT
Copernicus Atmosphere Monitoring Service
Validation report for the inverted CO2 fluxes, v18r1 – version 1.0
D1.4.1-2017_v0 (Evaluation and Quality Control document for CO2 flux inversion 2017 (Period covered 1979 to year 2017) v0
Issued by: CEA / Frédéric Chevallier
Date: 23/12/2018
REF.: CAMS73_2018SC1_D73.1.4.1-2017-v0_201812_Validation inverted CO2
fluxes_v1
Copernicus Atmosphere Monitoring Service
Author 2 of 20
This document has been produced in the context of the Copernicus Atmosphere Monitoring Service (CAMS). The activities leading to these results have been contracted by the European Centre for Medium-Range Weather Forecasts, operator of CAMS on behalf of the European Union (Delegation Agreement signed on 11/11/2014). All information in this document is provided "as is" and no guarantee or warranty is given that the information is fit for any particular purpose. The user thereof uses the information at its sole risk and liability. For the avoidance of all doubts, the European Commission and the European Centre for Medium-Range Weather Forecasts has no liability in respect of this document, which is merely representing the authors view.
Copernicus Atmosphere Monitoring Service
Author 3 of 20
Contributors
CEA Frédéric Chevallier
Copernicus Atmosphere Monitoring Service
Author 4 of 20
Table of Contents
1. Introduction 5
2. Inversion configuration 5
3. Evaluation 11
3.1 Fit to the Globalview+_v4.0 11 3.1.1 Assimilated data 12 3.1.2 Unassimilated data 12
3.2 Fit to TCCON GGG2014 14
3.3 Country and annual scale CO2 budgets 14
Acknowledgements 17
References 17
Copernicus Atmosphere Monitoring Service
Author 5 of 20
1. Introduction The inversion system that generates the CAMS global CO2 atmospheric inversion product is called PyVAR. It has been initiated, developed and maintained at CEA/LSCE within CAMS and its precursor projects GEMS/MACC/MACC-II/MACC-III (Chevallier 2018a, and references therein). Here, we synthesize the evaluation of version 18r1 that was released in November 2018. Version 18r1 covers the years between January 1979 and August 2018. It mainly improves compared to the earlier v17r1 shown in Chevallier (2018b) by the assimilation of more sites and of data for the second and third trimesters of 20181. It is strictly identical to v17r1 for the years before 2000. The presentation of the evaluation procedure is primarily based on the fit of the inversion posterior simulation to whole databases: Obspack Globalview+_v4.0 of Cooperative Global Atmospheric Data Integration Project (2018) and the Total Carbon Column Observing Network (TCCON) GGG2014 archive (Wunch et al. 2011). In addition, time series at national annual scale are shown and briefly discussed. Section 2 describes the PyVAR-CO2 configuration that was used and Section 3 presents the evaluation synthesis.
2. Inversion configuration The transport model in PyVAR-CO2 is the global general circulation model LMDZ in its version LMDZ5A (Locatelli et al. 2015), that uses the deep convection model of Emanuel (1991). This version has a regular horizontal resolution of 3.75o in longitude and 1.875o in latitude, with 39 hybrid layers in the vertical. The inferred fluxes are estimated in each horizontal grid point of the transport model with a temporal resolution of 8 days, separately for day-time and night-time. The state vector of the inversion system is therefore made of a succession of global maps with 9,200 grid points. Per month it gathers 73,700 variables (four day-time maps and four night-time maps). It also includes a map of the total CO2 columns at the initial time step of the inversion window in order to account for the uncertainty in the initial state of CO2. The prior values of the fluxes combine estimates of (i) gridded annual anthropogenic emissions (EC-JRC/PBL EDGAR version 4.2, CDIAC and GCP), (ii) monthly ocean fluxes (Landschützer et al. 20162), 3-hourly (when available) or monthly biomass burning emissions (GFED 4.1s until 20163 and GFAS afterwards) and climatological 3-hourly biosphere-atmosphere fluxes taken as the 1989-2010 mean
1 Measurements after August 2018 are used to constrain the year 2018 better, but fluxes for those months are not publicly distributed. 2 This database covers the period 1982-2015. We use the monthly values for the years 1982 and 2015 before and after it, respectively. 3 Before 1997, a monthly climatology of this database is used.
Copernicus Atmosphere Monitoring Service
Author 6 of 20
of a simulation of the ORganizing Carbon and Hydrology In Dynamic EcosystEms model (ORCHIDEE, Krinner et al. 2005), version 1.9.5.2. The mass of carbon emitted annually during specific fire events is compensated here by the same annual flux of opposite sign representing the re-growth of burnt vegetation, which is distributed regularly throughout the year. The gridded prior fluxes exhibit 3-hourly variations but their inter-annual variations over land are only caused by anthropogenic emissions. This feature was explicitly demanded by some users who wanted the interannual signals in the inverted natural fluxes to be strictly driven by the atmospheric measurements. Over land, the errors of the prior biosphere-atmosphere fluxes are assumed to dominate the error budget and the covariances are constrained by an analysis of mismatches with in situ flux measurements (Chevallier et al. 2006, 2012): temporal correlations on daily mean Net Carbon Exchange (NEE) errors decay exponentially with a length of one month but night-time errors are assumed to be uncorrelated with daytime errors; spatial correlations decay exponentially with a length of 500 km; standard deviations are set to 0.8 times the climatological daily-varying heterotrophic respiration flux simulated by ORCHIDEE with a ceiling of 4 gC∙m-2 per day. Over a full year, the total 1-sigma uncertainty for the prior land fluxes amounts to about 3.0 GtC∙yr-1. The error statistics for the open ocean correspond to a global air-sea flux uncertainty about 0.5 GtC∙yr-1 and are defined as follows: temporal correlations decay exponentially with a length of one month; unlike land, daytime and night-time flux errors are fully correlated; spatial correlations follow an e-folding length of 1000 km; standard deviations are set to 0.1 gC∙m-2 per day. Land and ocean flux errors are not correlated. Observation uncertainty in the inversion system is dominated by uncertainty in transport modelling and is initially represented from the variance of the high frequency variability of the de-seasonalized and de-trended CO2 time series of the daily-mean measurements at a given location. The values are then adjusted, first by inflating all error variances by the number of measurements at a given location within each calendar day, then by averaging consecutive measurements and defining the resulting error variance as the average of the individual error variances. Version 18r1 analyzed 39.8 years of surface measurements, from January 1979 to October 2018. The assimilated measurements are surface air-sample measurements of the CO2 dry air mole fraction made at 118 sites over the globe. These data are a carefully-selected subset of four large living databases of atmospheric measurements:
• the NOAA Earth System Research Laboratory Observation Package (https://www.esrl.noaa.gov/gmd/ccgg/obspack/, Cooperative Global Atmospheric Data Integration Project, 2018, and NOAA Carbon Cycle Group Obspack Team, 2018),
• the World Data Centre for Greenhouse Gases archive (WDCGG, https://gaw.kishou.go.jp/),
• the Réseau Atmosphérique de Mesure des Composés à Effet de Serre database (RAMCES, http://www.lsce.ipsl.fr/),
• the Integrated Carbon Observation System- Atmospheric Thematic Center (ICOS-ATC, https://icos-atc.lsce.ipsl.fr/).
The detailed list of selected sites is provided in Tables 1 and 2 and their location is displayed per year in Figure 1. The irregular space-time density of the measurements implies a variable constraint on the inversion throughout the 39.8 years, which is documented by the associated Bayesian error statistics.
Figure 1- Location of the assimilated measurements over the globe for each year in v18r1.
Table 1 - List of the continuous sites used in v18r1 together with the period of coverage (defined as the period between the first sample and the last one), and the data source. Each station is identified by the name of the place, the corresponding country (abbreviated) and the code used in the corresponding database provider. Note that only a subset of the data at each site is selected, based on local time and also excluding outliers.
Locality (indentifier) Period Source
Alert, Nunavut, CA (ALT) 1988-2018 NOAA/ EC
Amsterdam Island, FR (AMS) 1981-2018 ICOS/ LSCE
Argyle, Maine, US (AMT) 2003-2018 NOAA/ ESRL
Baring Head, NZ (BHD) 1979-2017 NOAA/ NIWA
Bratt’s Lake Saskatchewan, CA (BRA) 2009-2018 NOAA/ EC
Barrow, Alaska, US (BRW) 1979-2018 NOAA/ ESRL
Cambridge Bay, Nunavut Territory (CBY) 2012-2018 NOAA/ EC
Candle Lake, Saskatchewan, CA (CDL) 2002-2010 NOAA/ EC
Churchill, CA (CHL) 2011-2016 WDCGG/ EC
CHM
Monte Cimone, IT (CMN) 1996-2018 WDCGG/ IAFMS
Copernicus Atmosphere Monitoring Service
Author 8 of 20
Cape Ochi-ishi, JP (COI) 1995-2002 WDCGG/ NIES
Chapais, Quebec, CA (CPS) 2012-2018 NOAA/EC
Cape Point, SA (CPT) 1993-2017 NOAA/ SAWS
Egbert, Ontario, CA (EGB) 2005-2016 NOAA/ EC
Estevan Point, British Columbia, CA (ESP) 2009-2018 WDCGG/ EC
Esther, Alberta, CA (EST) 2010-2018 NOAA/ EC
East Trout Lake, Saskatchewan, CA (ETL) 2005-2018 NOAA/ EC
Fraserdale, CA (FSD) 1990-2018 NOAA/ EC
Hateruma, JP (HAT) 1993-2002 WDCGG/ NIES
Hidden Peak (Snowbird), Utah, US (HDP) 2006-2015 NOAA/ NCAR
Hohenpeissenberg, DE (HPB) 2015-2018 NOAA/ Scripps
Hegyhatsal tower, 115m level, HU (HUN0115) 1994-2017 NOAA/ HMS
Inuvik,Northwest Territories, CA (INU) 2012-2018 NOAA/ EC
Trinidad Head, California, US (THD) 2002-2017 NOAA/ ESRL Hydrometeorological Observatory of Tiksi, RU
(TIK) 2011-2018 NOAA/ ESRL
Trainou 180m agl, FR (TR3) 2006-2017 LSCE
Tromelin Island, F (TRM) 1998-2007 LSCE
Tierra Del Fuego, Ushuaia, AR (USH) 1994-2018 NOAA/ ESRL
Wendover, Utah, US (UTA) 1993-2018 NOAA/ ESRL
Ulaan Uul, MN (UUM) 1992-2018 NOAA/ ESRL
Sede Boker, Negev Desert, IL (WIS) 1995-2018 NOAA/ ESRL
Mt. Waliguan, CN (WLG) 1990-2018 NOAA/ ESRL Ny-Alesund, Svalbard, Norway and Sweden
(ZEP) 1994-2018 NOAA/ ESRL
3. Evaluation
3.1 Fit to the Globalview+_v4.0 We have run the LMDZ global transport model using the surface fluxes inferred by the inversion as boundary conditions and now compare it with all data in Globalview+_v4.0 for the period 2000-2017 (Globalview+_v4.0 stops in December 2017). A small subset of this data has been assimilated (see Section 2) and the rest is independent from the inversion. We discuss both parts in the following subsections. All model equivalents to individual Globalview+_v4.0 data are publicly available from http://dods.lsce.ipsl.fr/invsat/CAMS/v18r1_GV+4.0.txt or on request to [email protected].
3.1.1 Assimilated data Figure 2 shows the posterior root mean square (RMS) and bias of the model-minus-measurement difference as a function of the corresponding error statistics that we have assigned at each assimilated data from Globalview+_v4.0. Measurement error is negligible here and the assigned error statistics refer to transport model errors and to representation errors (see Section 2). As expected, the inversion fits the assimilated data within the standard deviation of the assigned observation uncertainty. Biases are usually less than 1 ppm in absolute value. Figure 2 - Statistics of the differences between the posterior inversion simulation and individual assimilated surface measurements from Globalview+_v4.0 as a function of the assigned observation error standard deviation for each measurement site. The statistics cover the period 2000–2017.
3.1.2 Unassimilated data Among the unassimilated Globalview+_v4.0 dry air mole fraction measurements, we focus on two subgroups. The first one contains all continuous or flask aircraft measurements in the free troposphere. The free troposphere is simply defined here as the atmospheric layer between 2 and 7 km above sea level (asl). The second subgroup contains all unassimilated flask measurements made from ships or ground-based stations. It is important to note that the unassimilated flask measurements made from ships or ground-based stations correspond either (i) to data that have been rejected by the inversion quality-control, or (i) to data collected at sites that have relatively short time series (we impose a minimum of five-year worth of measurements), or (i) to sites for which the transport model is found less reliable, for instance because of the proximity to local sources, because of complex orography or because of coastal wind circulation patterns. We use these leftover data in this section, but still exclude all data at five sites (BKT, BSC, CRI, LJO, OTA) that show many misfit values larger than 10 ppm. We focus on the period 2000-2016, but statistics including data from earlier decades are marginally different.
Copernicus Atmosphere Monitoring Service
Author 13 of 20
Figure 3 - Model-minus-aircraft biases in 15o latitude bins and in three longitude bands (Western hemisphere, central hemisphere and Eastern hemisphere). The top figure presents the aircraft statistics, while the bottom one shows the statistics for the unassimilated surface flask data. The statistics cover the period 2000–2016.
The biases (Figure 3) are within 1.6 ppm at the surface and within 0.5 ppm in the free troposphere, even when splitting the globe into three longitude bands. The simulation for the Eastern hemisphere appears to be less accurate than for the two other hemispheres which can be explained by the lesser density of the assimilated data there (Figure 1). In all three hemispheres and both at the surface and the free troposphere, there is no obvious latitudinal trend (except a small trend for aircraft that decreases with latitude, but the amplitude of the trend remains within 0.3 ppm), and therefore no obvious flaw of the model vertical mixing (Stephens et al., 2007). Standard deviations vary with the fraction of land masses in a given latitude, as expected. They reach up to 2 ppm in the free troposphere and 6 ppm at the surface. The value at the surface is expectedly larger than the largest one found for the assimilated data (Figure 2). When taking all free tropospheric aircraft data together, the posterior simulation deviates from the measurements by 0.0±1.3 ppm (bias ± standard deviation), which is within the specification (key performance indicator) of the CAMS CO2 inversion. Figure 4 – Same as Figure 3, but for the standard deviation.
Copernicus Atmosphere Monitoring Service
Author 14 of 20
3.2 Fit to TCCON GGG2014 Figure 5 shows the misfit statistics for the column retrievals at each TCCON station. For the comparison, the model has been convolved with the retrieval averaging kernels. All available TCCON station records are shown for the sake of completeness, but sites Pasadena, JPL and Paris are located in urban areas that are not well represented at the horizontal resolution of the transport model (3.75o in longitude and 1.875o in latitude): in this case the statistics logically show negative model biases larger than -1 ppm. Apart from these urban stations, absolute biases are less than 1 ppm at all sites. The standard deviation is usually about 1 ppm, but it reaches 2 ppm at the Zugspitze mountain site. We note that the model usually fits TCCON retrievals better than the satellite retrievals presented by Wunch et al. (2017).
3.3 Country and annual scale CO2 budgets The aggregation of the inversion results at country scale is based on the country mask from http://themasites.pbl.nl/tridion/en/themasites/hyde/. The resulting annual CO2 budgets for v18r1 (Figure 6, Figure 7, Figure 8, and Figure 9) are rather similar to those presented for v17r1 (Chevallier, 2018b). We still note an increased uptake in 2011 for Australia, that is consistent, although smaller in amplitude, with other studies using different types of measurements (satellite XCO2 retrievals, satellite observations of vegetation activity, …) that reported an anomalous uptake in Australia during this particular La Niña episode (Poulter et al. 2013, Detmers et al. 2015, Ma et al. 2016).
Figure 5 - Statistics of the difference between the posterior model and individual TCCON measurements, ordered by increasing latitude indices in the LMDZ model. A site may appear several times if several instruments have been used over time there. The statistics cover the period 2004-2018.
Figure 6 - National-annual-scale time series of the total natural flux in v18r1 (green) with its 1-σ uncertainty (yellow), of the LULUCF emission reported to UNFCCC (blue) and of the energy sector emission reported to UNFCCC in 2018 (grey). Positive values denote sources to the atmosphere (emissions), while negative values denote storage in soils and vegetation (sink). The model grid points associated to each country appear in red on the global maps.
Copernicus Atmosphere Monitoring Service
Author 16 of 20
Figure 7 – Continued.
Figure 8 – Continued.
Copernicus Atmosphere Monitoring Service
Author 17 of 20
Figure 9 – Continued.
Acknowledgements The author is very grateful to the many people involved in the surface and aircraft CO2 measurements and in the archiving of these data that were kindly made available to him by various means. TCCON data were obtained from the TCCON Data Archive, operated by the California Institute of Technology from the website at http://tccon.ornl.gov/. Obspack data were obtained from https://www.esrl.noaa.gov/gmd/ccgg/obspack/. Mass fluxes for the LMDZ transport model have been provided by Y. Yin, Y. Wang, R. Locatelli and P. Bousquet. Some of this work was performed using HPC resources of DSM-CCRT and of CCRT under allocation A0030102201 made by GENCI (Grand Équipement National de Calcul Intensif).
References Carbontracker Team (2018), Compilation of near real time atmospheric carbon dioxide data; obspack_co2_1_NRT_v4.2_2018-04-06; NOAA Earth System Research Laboratory, Global Monitoring Division. http://doi.org/10.15138/G3RP8K Chevallier, F., N. Viovy, M. Reichstein, and P. Ciais: On the assignment of prior errors in Bayesian inversions of CO2 surface fluxes. Geophys. Res. Lett., 33, L13802, doi:10.1029/2006GL026496, 2006.
Chevallier, F., T. Wang, P. Ciais, F. Maignan, M. Bocquet, A. Arain, A. Cescatti, J.-Q. Chen, H. Dolman, B. E. Law, H. A. Margolis, L. Montagni, and E. J. Moors: What eddy-covariance flux measurements tell us about prior errors in CO2-flux inversion schemes. Global Biogeochem. Cy., 26, GB1021, doi:10.1029/2010GB003974, 2012. Chevallier, F., Description of the CO2 inversion production chain. CAMS deliverable CAMS73_2015SC3_D73.1.5.6_201803_CO2 inversion production chain_v1. http://atmosphere.copernicus.eu/, 2018a. Chevallier, F., Validation report for the inverted CO2 fluxes, v17r1. CAMS deliverable CAMS73_2015SC3_ D73.1.4.2-1979-2017-v1_201806. http://atmosphere.copernicus.eu/, 2018b. Cooperative Global Atmospheric Data Integration Project. (2018). Multi-laboratory compilation of atmospheric carbon dioxide data for the period 1957-2017; obspack_co2_1_GLOBALVIEWplus_v4.0_2018-08-02 [Data set]. NOAA Earth System Research Laboratory, Global Monitoring Division. https://doi.org/10.25925/20180802 Detmers, R. G., O. Hasekamp, I. Aben, S. Houweling, T. T. van Leeuwen, A. Butz, J. Landgraf, P. Köhler, L. Guanter, and B. Poulter, Anomalous carbon uptake in Australia as seen by GOSAT, Geophys. Res. Lett., 42, 8177–8184, doi:10.1002/2015GL065161, 2015. Emanuel, K.: A Scheme for Representing Cumulus Convection in Large-Scale Models, J. Atmos. Sci., 48, 2313–2329, doi:10.1175/1520-0469(1991)048<2313:ASFRCC2.0.CO;2, 1991. Krinner, G., Viovy, N., de Noblet-Ducoudré, N., Ogée, J., Polcher, J., Friedlingstein, P., Ciais, P., Sitch, S., and Prentice, I. C.: A dynamic global vegetation model for studies of the coupled atmosphere-biosphere system, Global Biogeochem. Cy., 19, GB1015, doi:10.1029/2003GB002199, 2005. Landschützer, P., N. Gruber and D.C.E. Bakker. 2015. A 30 years observation-based global monthly gridded sea surface pCO2 product from 1982 through 2011 (NCEI Accession 0160558). Version 2.2. NOAA National Centers for Environmental Information. Dataset. doi:10.3334/cdiac/otg.spco2_1982_2011_eth_somffn Locatelli, R., Bousquet, P., Hourdin, F., Saunois, M., Cozic, A., Couvreux, F., Grandpeix, J.-Y., Lefebvre, M.-P., Rio, C., Bergamaschi, P., Chambers, S. D., Karstens, U., Kazan, V., van der Laan, S., Meijer, H. A. J., Moncrieff, J., Ramonet, M., Scheeren, H. A., Schlosser, C., Schmidt, M., Vermeulen, A., and Williams, A. G.: Atmospheric transport and chemistry of trace gases in LMDz5B: evaluation and implications for inverse modelling, Geosci. Model Dev., 8, 129-150, doi:10.5194/gmd-8-129-2015, 2015. Ma, X., A. Huete, J. Cleverly, D. Eamus, F. Chevallier, J. Joiner, B. Poulter, Y. Zhang, L. Guanter, W. Meyer, Z. Xie, G. Ponce-Campos, 2016: Drought rapidly disseminates the 2011 large CO2 uptake in semi-arid Australia. Scientific Reports, 6. doi: 10.1038/srep37747.
NOAA Carbon Cycle Group ObsPack Team. (2018). Multi-laboratory compilation of atmospheric carbon dioxide data for the year 2018; obspack_co2_1_NRT_v4.3_2018-10-17 [Data set]. NOAA Earth System Research Laboratory, Global Monitoring Division. https://doi.org/10.25925/20181017 Poulter, B., D. Frank, P. Ciais, R. B. Myneni, N. Andela, J. Bi, G. Broquet, J. G. Canadell, F. Chevallier, Y. Y. Liu, S. W. Running, S. Sitch and G. R. van der Werf: Contribution of semi-arid ecosystems to interannual variability of the global carbon cycle. Nature, doi:10.1038/nature13376, 2014 Stephens, B. B., et al.: Weak northern and strong tropical land carbon uptake from vertical profiles of atmospheric CO2, Science, 316, 1732–1735, 2007. Wunch, D., Toon, G. C., Blavier, J.-F. L., Washenfelder, R. A., Notholt, J., Connor, B. J., Griffith, D. W. T., Sherlock, V., and Wennberg, P. O., The Total Carbon Column Observing Network, Phil. Trans. R. Soc. A:2011369 2087-2112, doi10.1098/rsta.2010.0240, 2011. Wunch, D., Wennberg, P. O., Osterman, G., Fisher, B., Naylor, B., Roehl, C. M., O'Dell, C., Mandrake, L., Viatte, C., Kiel, M., Griffith, D. W. T., Deutscher, N. M., Velazco, V. A., Notholt, J., Warneke, T., Petri, C., De Maziere, M., Sha, M. K., Sussmann, R., Rettinger, M., Pollard, D., Robinson, J., Morino, I., Uchino, O., Hase, F., Blumenstock, T., Feist, D. G., Arnold, S. G., Strong, K., Mendonca, J., Kivi, R., Heikkinen, P., Iraci, L., Podolske, J., Hillyard, P. W., Kawakami, S., Dubey, M. K., Parker, H. A., Sepulveda, E., García, O. E., Te, Y., Jeseck, P., Gunson, M. R., Crisp, D., and Eldering, A.: Comparisons of the Orbiting Carbon Observatory-2 (OCO-2) XCO2 measurements with TCCON, Atmos. Meas. Tech., 10, 2209-2238, https://doi.org/10.5194/amt-10-2209-2017, 2017.