Effect of assimilating CO2 observations in the Korean Peninsula on the inverse modeling to estimate surface CO2 flux over Asia

Minkwang Cho; Hyun Mee Kim

doi:10.1371/journal.pone.0263925

Abstract

To investigate the impact of two CO₂ observation datasets obtained from the Korean Peninsula on the surface CO₂ flux estimation over Asia, the two datasets are assimilated into the CarbonTracker (CT) inverse modeling system and the estimated surface CO₂ fluxes are analyzed. Anmyeon-do (AMY) and Gosan (GSN) sites in the Korean Peninsula have observed surface CO₂ mole fraction since the late 1990s. To investigate the effect of assimilating the additional Korean observations on the surface CO₂ flux estimation over Asia, two experiments are conducted. The reference experiment (CNTL) only assimilates observations provided by National Oceanic and Atmospheric Administration (NOAA), while the other experiment (EXP1) assimilates both NOAA observations and two Korean observation datasets. The results are analyzed for 9 years from 2003 to 2011 in Asia region because both AMY and GSN datasets exist almost completely for this period. The annual average of estimated biosphere CO₂ flux of EXP1 shows more flux absorption in summer and less flux emission from fall to spring compared to CNTL, mainly on Eurasia Temperate and Eurasia Boreal regions. When comparing model results to independent CO₂ concentration data from surface stations and aircraft, the root mean square error is smaller for EXP1 than CNTL. The EXP1 yields more reduction on uncertainty of estimated biosphere CO₂ flux over Asia, and the observation impact of AMY, GSN sites on flux estimation is approximately 11%, which is greater than other observation sites around the world. Therefore, the two CO₂ observation sets in the Korean Peninsula are useful in reducing uncertainties for regional as well as global scale CO₂ flux estimation.

Citation: Cho M, Kim HM (2022) Effect of assimilating CO₂ observations in the Korean Peninsula on the inverse modeling to estimate surface CO₂ flux over Asia. PLoS ONE 17(2): e0263925. https://doi.org/10.1371/journal.pone.0263925

Editor: Min Huang, George Mason University, UNITED STATES

Received: August 31, 2021; Accepted: January 30, 2022; Published: February 18, 2022

Copyright: © 2022 Cho, Kim. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data underlying the results presented in the study are available from the cluster of the National Center for Meteorological Supercomputer of the Korea Meteorological Administration (KMA) (http://super.kma.go.kr).

Funding: This study was supported by a National Research Foundation of Korea (NRF) grant funded by the South Korean government (Ministry of Science and ICT) (Grant 2021R1A2C1012572) and the Yonsei Signature Research Cluster Program of 2021 (2021-22-0003).

Competing interests: The authors have declared that no competing interests exist.

1. Introduction

The annual mean surface air temperature of the earth has increased since the second industrial revolution in the late 19^th century. Global mean surface temperature (GMST) of the last decade (2011–2020) was approximately 1.09°C higher than that of the preindustrial period (1850–1900) [1]. The Paris Agreement adopted in 2015 aims at limiting the GMST rise below 2°C, or 1.5°C if possible, compared to the pre-industrial levels [2]. However, Intergovernmental Panel on Climate Change (IPCC) special report has stated that the temperature increase could surpass 1.5°C at the middle of this century without additional emission decrease, even if the current nationally determined contributions (NDCs) are achieved [3]. Therefore, it is important to estimate the precise sources and sinks of greenhouse gases in order to support the emission policies based on the scientific information as well as to manage the risks from the climate change.

CO₂ is the most abundant component among the greenhouse gases in the troposphere. Since [4] attempted to estimate surface CO₂ flux by assimilating observed CO₂ mole fraction data in a model, many researchers have studied surface CO₂ flux optimization using data assimilation (DA) approach. Surface CO₂ flux estimation is an inverse modeling problem, known as an under-deterministic problem, which finds the solutions from relatively small number of observations. Thus, it needs prior surface CO₂ flux information and more importantly observation data accumulated in long time period, so as to estimate the surface CO₂ fluxes more precisely. Observations from the surface (i.e., flask observation and in situ observation) are commonly used in the inverse modeling. The spatial density of the surface CO₂ observations is relatively high in North America and Europe, giving more reliable surface CO₂ flux information of the two continents [5]. On the other hand, Asia, Africa, and South America have relatively low observation densities to well enough constrain the surface CO₂ fluxes. To cover the deficiency of observation density in Asia, inverse modeling researches have been conducted by assimilating tower observations in Siberian region and aircraft observations over the globe into inverse models [6–9]. Satellite column CO₂ (XCO₂) data have also been used to supplement the spatial coverage of surface CO₂ observations. The Greenhouse Gases Observing Satellite (GOSAT; [10, 11]) XCO₂ data have been used in various inverse modeling studies [12–21]. The Orbiting Carbon Observatory-2 satellite (OCO-2; [22, 23]) XCO₂ data also have been used in several studies [19, 15, 24–26]. Despite broad spatial coverage of satellite XCO₂ data, the surface CO₂ flux estimation using satellite XCO₂ data shows large uncertainty depending on observation coverage, number of data, and retrieval algorithm [15, 17, 19]. Those studies have shown some improvements in optimizing surface CO₂ flux in Asia, but more observation data are required to more precisely estimate the surface CO₂ fluxes in terms of their spatial and temporal patterns as well as total annual budgets in different ecoregions over Asia.

Meanwhile, each country involved in the Paris agreement is obliged to set up its own contribution to greenhouse gas emission reduction and report its annual emission inventory to United Nations Framework Convention on Climate Change (UNFCCC) to check how much reduction is achieved. World Meteorological Organization (WMO) has established Integrated Global Greenhouse Gas Information System (IG3IS) platform, which combines traditional inventory reporting and inverse modeling results to support countries to make the inventory report with less uncertainties. [27] applied CO₂ mole fraction data observed from two surface in situ observatories and shipboard across New Zealand to an inverse model, and revealed that the modeled CO₂ flux absorption from indigenous forests in New Zealand is stronger than the absorption calculated from the inventory report. In the UK, CO₂ observation from the tower network were assimilated into an inverse model and the estimated CO₂ flux from the biosphere seemed to be zero balanced, different from the inventory report [28]. The regional scale observation networks have been established in Switzerland and Paris, France, in order to assist regional greenhouse gas emission estimation. As it becomes more important to secure enough observations for surface CO₂ flux optimization, more Asian observations are necessary to be utilized in inverse modeling.

In this study, CO₂ observation data from Anmyeon-do (AMY) and Gosan (GSN), located in the Korean Peninsula, are assimilated into the CarbonTracker (CT), and the effect of the two observation datasets on Asian surface CO₂ flux estimation is investigated. AMY and GSN sites have accumulated observation data over approximately more than 9 years, which could provide flux information over the Korean Peninsula and its surrounding Asian regions. Model and observations used are introduced in section 2. In section 3, the estimated CO₂ fluxes in Asia from the 9-year (2003–2011) assimilation experiments using two additional observation datasets and the effects of assimilating AMY and GSN data into the inverse modeling are discussed. Finally, section 4 presents the summary and conclusions.

2. Methodology

2.1 Inverse modeling system

CT is a global scale inverse model [29] developed to estimate the surface CO₂ fluxes using CO₂ observations as a constraint. To assimilate CO₂ mole fraction observations, CT adopts an ensemble Kalman filter (EnKF) DA method. Because the number of observation sites is too sparse to cover the whole globe, prior flux information needs to be given in advance. First guess of CO₂ flux is presented as a combination of four different prior flux information as below: (1) where F_bio, F_ocn, F_fire and F_ff represent CO₂ flux from terrestrial ecosystem activity (CASA GFED v3.1: [30, 31]), atmosphere-ocean CO₂ exchange [32], biomass burning from the forest fire [30, 31], and fossil fuel combustion (Carbon Dioxide Information and Analysis Center [CDIAC] database: [33]), respectively. Since CO₂ emissions from fossil fuel use and forest fire are prescribed, it is important to find precise biosphere flux and ocean flux estimates to optimize the total CO₂ fluxes. In CT, scaling factor λ_r is updated through DA and used for optimizing the two CO₂ flux components (i.e., biosphere and ocean fluxes). Each scaling factor corresponds to the specific ecosystem called ecoregion. The 126 ecoregions in the land among 209 ecoregions that combines Transcom regions (i.e., 11 regions) and ecosystem types (i.e., 19 types), and the 30 ecoregions in the ocean [34] are paired with scaling factors and those scaling factors update the flux values. These total 156 ecoregions are shown in CarbonTracker website (https://gml.noaa.gov/ccgg/carbontracker/CT2013B_doc.pdf). Among 209 ecoregions mentioned above, the unrealistic combinations of Transcom regions and ecosystem types (e.g., mangrove in the Eurasia Boreal) are not included in the 126 ecoregions in the land.

Transport Model 5 (TM5: [35]), an offline atmospheric chemical transport model, converts surface CO₂ flux from Eq (1) into model CO₂ concentration (i.e., mole fraction). TM5 then calculates advection, convection, and vertical diffusion of CO₂ using ERA-Interim reanalysis data from the European Centre for Medium-Range Weather Forecasts (ECMWF). TM5 works as an observation operator calculating model CO₂ concentration corresponding to the observed CO₂ concentration at the time and space of which observation occurs.

The EnKF scheme used in CT is the ensemble square root Kalman filter (EnSRF) adopted from [36]. EnSRF separately updates ensemble mean and ensemble perturbation, as follows: (2) (3) where is the ensemble mean of the state vector x, which is scaling factor being updated in the DA system in this study. Subscripts a and b are analysis and background, respectively, and y^O is the observation vector. H is a linear observation operator projecting the model state vector onto the observation space, and the TM5 works as H in CT. K and are Kalman gain and reduced Kalman gain, respectively. (4) (5) where , P^b is the model’s background error covariance, and R is the observation error covariance. P^bH^T and HP^bH^T is calculated using the equations below: (6) (7) where N is number of ensemble members.

It is necessary to prevent the sampling error amplification by limited ensemble members. Covariance localization technique [37] is conducted in CT to exclude the effect of the remote observations to the surface CO₂ flux estimation, as the remote observations from the flux location are barely correlated with the flux concerned. Since no physical relationship exists between scaling factors, the correlations are calculated between scaling factor deviations of ensemble members and corresponding modeled flux deviations. If the correlation values fail the significance test, the scaling factor is not updated. Marine Boundary Layer (MBL) observations are exempt for the localization as MBL sites captures the flux signal from a distance [34].

On each week of the simulation period, the EnKF assimilates observations from the most recent week to update the scaling factors of the five weeks including the past four weeks as well as the most recent week. As [38] denoted, the time lag scheme helps to consider that the observations can contain the signal of sources or sinks away from the observing sites. Previous studies using a five-week time lag [7, 29, 39–41] showed that the five-week time lag is appropriate to optimize the surface CO₂ flux in North America, Europe, and Asia.

In CT, the scaling factor for the upcoming analysis week is predicted by a simple model as follows: (8) where is a prior scaling factor for the upcoming analysis week t; and are posterior scaling factors of week t-2 and t-1, respectively. λ^p is a fixed value 1, so that the scaling factor returns to 1 when there are no assimilated observations.

2.2 CO₂ observation

In this study, two surface CO₂ mole fraction observation datasets from AMY (https://gaw.kishou.go.jp/search/file/0039-2014-1001-01-01-9999) and GSN (https://gaw.kishou.go.jp/search/file/0052-2025-1001-01-01-9999) sites in the Korean Peninsula are additional observations that are assimilated in CT. AMY station has been operated by Korea Meteorological Administration (KMA) since 1998 and a year later has been designated as a regional global atmospheric watch (GAW) station. GSN station has started observation in 2002 by National Institute of Environmental Research (NIER), and KMA has taken over the operation since 2012 with new acronym JGS (Jeju Gosan). AMY and GSN data are obtained relatively remote from anthropogenic sources such as factories or residential area, which are appropriate to represent the CO₂ concentration in Northeast Asia region. The method used for observing the two datasets is non-dispersive infrared analyzer (NDIR), which is able to log quasi-continuous CO₂ concentration.

In this study, AMY and GSN data are assimilated together with other observation datasets from Observation Package (ObsPack) product in CT. The ObsPack product, provided by National Oceanic and Atmospheric Administration (NOAA) Earth System Research Laboratory (ESRL) [42], is a collection of CO₂ observations around the world. Diverse research institutes including NOAA, the Commonwealth Scientific and Industrial Research Organization (CSIRO), the National Center for Atmospheric Research (NCAR), and Environment and Climate Change Canada (ECCC) have provided observed data for ObsPack production. Most ObsPack data are obtained by averaging observed values between 12–16 local standard time (LST) since the TM5 model shows good performance in simulating well-mixed atmospheric layer of daytime. For observation sites located at the mountaintops, observations between 00–04 LST are averaged because there is less chance of local biogenic or anthropogenic CO₂ inflow from the downslope during the nighttime [34]. Daily mean AMY and GSN data are also obtained by averaging 12–16 LST data, following the ObsPack data.

Model-data mismatch (MDM) (i.e., observation error) for ObsPack is prescribed based on observation type and geographic characteristics in CT. When assimilated, the observation error for two datasets needs to be prescribed. MDM of both AMY and GSN is set to 3 ppm based on several verifications conducted in [43]. Beside AMY and GSN, Tae-ahn Peninsula (TAP) data are already included in the ObsPack. Note that TAP’s MDM is 5 ppm, following [41].

For verification of the results, independent CO₂ observations in Asia that are not assimilated in CT are used. Those independent observations are aircraft observation data from the National Institute for Environmental Studies (NIES) Japan, called The Comprehensive Observation Network for Trace gases by Airliners (CONTRAIL) (http://doi.org/10.17595/20180208.001; [44, 45]), and surface observations from World Data Centre for Greenhouse Gases (WDCGG, https://gaw.kishou.go.jp/). Tables 1 and 2 and Fig 1 present information of the observations in Asia, used for assimilation and verification in this study. Note that only Asian ObsPack stations are depicted here. The CONTRAIL data used for verification shown in Fig 1 is from Nov. 2005 to Dec. 2011. As Fig 1, all figures including map image in this study were produced using NCL [46].

Download:

Fig 1. Observation sites in Asia, used either in data assimilation (○: NOAA ObsPack sites in Asia; ●: AMY and GSN) or in verification (◇: Observation sites of JMA, NIES, and ESRL).

CONTRAIL observations are marked as squares (thick line: locations of ascending/descending mode; thin line: level mode). The dashed box represents the nested domain in TM5 transport model.

https://doi.org/10.1371/journal.pone.0263925.g001

Download:

Table 1. Information of observation sites in Asia used for assimilation in this study.

https://doi.org/10.1371/journal.pone.0263925.t001

Download:

Table 2. Information of observation sites in Asia used for independent verification in this study.

https://doi.org/10.1371/journal.pone.0263925.t002

2.3 Experimental framework

CT2013B version is used in this study, which is able to simulate the surface CO₂ flux from 2000 to 2012. Two experiments are conducted to investigate the impact of AMY and GSN observations on surface CO₂ flux estimation. EXP1 experiment assimilates all available observations (AMY, GSN observations, and ObsPack datasets), while CNTL assimilates only ObsPack data. The TM5 model runs on a two-way nested grid with a 3°⨯2° outer domain on the globe and a 1°⨯1° nesting domain centered on Asia (Fig 1). The experimental period is from 2002 to 2011 because both AMY and GSN datasets exist almost completely for this period. The experimental results are analyzed for 9 years from 2003 to 2011 except for the first year (i.e., 2002) as a spin-up. More details about the experimental settings are summarized in Table 3.

Download:

Table 3. Experimental framework for estimating surface CO₂ flux in Asia.

https://doi.org/10.1371/journal.pone.0263925.t003

3. Results

3.1. Characteristics of surface CO₂ flux in Asia

3.1.1 9-year average surface CO₂ flux distribution.

Fig 2 shows the 9-year average surface CO₂ flux distribution within the nested domain shown in Fig 1. Fig 2A shows the surface CO₂ flux calculated by prior fluxes (i.e., with no DA), Fig 2B and 2C are the CNTL and EXP1 results, Fig 2D is the average flux difference between EXP1 and CNTL, and Fig 2E shows the spatial distribution of the “Mixed forest” ecoregion in Asia where the AMY site is located. DA enhances the absorption and emission of surface CO₂ fluxes compared with prior fluxes (Fig 2A–2C). In particular, there are strong CO₂ flux absorptions in the Indochina Peninsula, inland China, the Korean Peninsula, Japan, and southern Siberia, and strong CO₂ flux emissions in inland India and a small part of southern China (Fig 2B and 2C). The average flux distributions of CNTL and EXP1 show generally similar patterns (Fig 2B and 2C), but EXP1 shows more absorption than CNTL does in inland southern China, the Korean Peninsula, and Japan (Fig 2D). Locations such as the border region between northern Thailand and China represent greater emissions in EXP1 than in CNTL (Fig 2D). Mixed forest areas coincide well with the areas where the DA effects are obvious (Fig 2B, 2C and 2E), which is due to background error covariance in the EnKF in CT. As each scaling factor is assigned to respective ecoregions, the background error covariance matrix shows correlations between ecoregions in different Transcom regions. Since the dynamical model in Eq (8) does not include an error term, the background error covariance is set to a prior covariance structure and not predicted with the dynamical model [29]. According to [34], the same ecoregions among five different Transcom regions (North American Boreal, North American Temperate, Eurasia Boreal (EB), Eurasia Temperate (ET), and Europe) have correlations although those correlations become small for distant ecoregions. These correlations allow observations in a certain ecoregion to update scaling factors connected to the same ecoregion concerned, through the DA. It explains how AMY data affects more on the specific areas.

Download:

Fig 2.

Average biosphere fluxes (g C m^-2 yr ^-1) for 2003–2011 period over the nested Asian domain: (a) prior, (b) CNTL, (c) EXP1, and (d) difference between EXP1 and CNTL. (e) Spatial distribution of the “Mixed forest” ecoregion in Asia, where AMY site is located.

https://doi.org/10.1371/journal.pone.0263925.g002

3.1.2 Annual and average surface CO₂ flux.

Fig 3 shows annual and average surface CO₂ fluxes on the globe, the land, and the ocean. Compared to prior flux, global CO₂ flux uptake by land vegetation and ocean in CNTL and EXP1 is approximately 2 Pg C yr^-1 greater (Fig 3A). Most of this CO₂ uptake difference between the experiments and prior flux comes from the CO₂ absorption by the terrestrial vegetation (Fig 3B), while CO₂ flux absorptions from the ocean in the CNTL and EXP1 are only slightly different from the prior flux (Fig 3C). When comparing CNTL and EXP1 results, EXP1 shows slightly more (less) biogenic (oceanic) CO₂ flux absorption than CNTL does. The interannual variation of CNTL and EXP1 during the analysis period is very similar, indicating that the two Korean observation datasets assimilated in CT did not interrupt the consistency in the global surface CO₂ flux variability. The prior fluxes show the greatest uncertainties on the globe, the land, and the ocean, followed by CNTL and EXP1 (Fig 3A–3C). The average uncertainty of the prior flux for 9 years on the globe decreases by 22.5% in CNTL and 24.2% in EXP1 (Fig 3A). The decreases of uncertainties in CNTL and EXP1 compared to the prior flux uncertainty are greater on the land than the ocean (Fig 3B and 3C).

Download:

Fig 3.

Annual and average biosphere and ocean CO₂ fluxes (Pg C yr^-1) from the prior (black), CNTL (gray), and EXP1 (white) with their uncertainties aggregated over the (a) whole globe, (b) land, and (c) ocean.

https://doi.org/10.1371/journal.pone.0263925.g003

Fig 4 is the same as Fig 3 but for the three Transcom regions in Asia. In the EB and ET regions, the surface CO₂ fluxes from the prior flux are very small, while CNTL and EXP1 result in absorbing large amounts of surface CO₂ fluxes. Compared to CNTL, EXP1 generally estimates more biogenic CO₂ flux absorption for every Transcom regions in Asia. CNTL and EXP1 estimate -1.08 Pg C yr^-1 and -1.27 Pg C yr^-1 of CO₂ flux for EB (Fig 4A), while -0.43 Pg C yr^-1 and -0.61 Pg C yr^-1 of CO₂ flux for ET (Fig 4B). In particular, compared to CNTL, EXP1 shows larger biogenic surface CO₂ flux absorption in ET since the added two Korean CO₂ observation sites are located in ET. The biogenic CO₂ flux absorption in EB is also affected by the two observation datasets in ET owing to the background error covariance structure described in section 3.1.1. Assimilation of AMY and GSN datasets results in negative CO₂ fluxes greater in Asia, which implies the possibility of the enhanced CO₂ absorption as well as the weakened CO₂ respiration. The prior fluxes show the greatest uncertainties in EB, ET, and Tropical Asia (TA), followed by CNTL and EXP1 (Fig 4A–4C). The average uncertainty of the prior flux for 9 years in EB decreases by 19.8% in CNTL and 21.8% in EXP1 (Fig 4A), and that in ET decreases by 18.7% in CNTL and 23.9% in EXP1 (Fig 4B). In TA, CNTL and EXP1 show very similar uncertainties (Fig 4C).

Download:

Fig 4.

Annual and average biosphere CO₂ fluxes (Pg C yr^-1) from the prior (black), CNTL (gray), and EXP1 (white) with their uncertainties aggregated over the (a) Eurasia Boreal, (b) Eurasia Temperate, and (c) Tropical Asia.

https://doi.org/10.1371/journal.pone.0263925.g004

In EXP1, the EB region has the lowest flux absorption in 2005 and 2006 and the greatest flux absorption in 2008 and 2011 (Fig 4A). The ET has the lowest CO₂ uptake in 2003 and 2005 and the greatest in 2011 (Fig 4B). This interannual variability of biogenic CO₂ flux seems to be affected by the climate events. It is known that El Nino enhances the CO₂ sources while La Nina intensifies the CO₂ sinks [47, 48]. Based on the ENSO ONI index from NCEP (https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php), the strong La Nina events occurred during 2007–2008 and 2010–2011, and the biogenic CO₂ flux absorption estimated from the CT increased at the same period. Meanwhile, weak El Nino events occurred during 2004–2005 and late 2009, and the biogenic CO₂ flux absorption from CT weakened during that period. Additionally, extreme drought conditions occurred in 2003 in all of the northern midlatitudes [49] result in reduced uptake of CO₂ [6]. Therefore, assimilating CO₂ observation datasets in Korea reflects the climate effect on the surface CO₂ exchange in EB and ET (Fig 4A and 4B). TA has very small CO₂ uptake and emission of less than 0.3 Pg C yr^-1, irrespective of the experiments (Fig 4C). Surface CO₂ flux estimates over the TA region have been known to have high uncertainty because there are little observations for the inverse modeling to represent the signal of source and sink [39, 50, 51]. [52, 53] showed that the near-neutral CO₂ flux in tropical region is due to the balance between the CO₂ release from deforestation and the CO₂ uptake by the intact tropical forests. [48] showed that the carbon budget in South Asia and Southeast Asia is close to neutral, with weak signs of carbon sink.

3.1.3 Monthly and weekly aggregated surface CO₂ flux.

Fig 5 shows the time series of monthly surface CO₂ fluxes averaged in the analysis period (i.e., 2003–2011) for the individual Transcom regions in Asia. In both CNTL and EXP1, a distinct seasonal variation pattern is found in the EB and ET regions (Fig 5A and 5B), in which flux absorption occurs in summer and flux emission occurs from autumn to spring.

Download:

Fig 5.

Monthly biosphere CO₂ fluxes (Pg C yr^-1) of the prior (black), CNTL (blue), and EXP1 (red), averaged for 2003–2011 period, with their uncertainties over the (a) Eurasia Boreal, (b) Eurasia Temperate, and (c) Tropical Asia.

https://doi.org/10.1371/journal.pone.0263925.g005

Over the EB region, where vegetation activity is very active, approximately -12 Pg C yr^-1 is estimated to be absorbed to the surface every summer with a large surface CO₂ flux uncertainty (Fig 5A). Compared to CNTL, CO₂ flux emission in spring in EXP1 decreases in the EB region. The ET region shows large difference between CNTL and EXP1 (Fig 5B). Compared to CNTL, EXP1 shows stronger flux absorption in the summer and weakened flux emission in winter and spring. In particular, CO₂ flux emission in spring in EXP1 is reduced to less than half of CNTL. The uncertainties of surface CO₂ flux estimation in CNTL and EXP1 are the greatest in summer. In the TA region, there is little difference between CNTL and EXP1 and there is no distinct seasonal variation although there are CO₂ flux absorption in spring and fall and release in summer (Fig 5C).

Therefore, the assimilation of AMY and GSN observations in CT enhances the monthly surface CO₂ flux absorption in summer in Asia region (especially EB and ET), and decreases the CO₂ emission in spring and winter season. The seasonal surface CO₂ fluxes would be changed if the seasonal and diurnal variations of MDM are considered. The effect of MDM variations in estimating the surface CO₂ fluxes over Asia would be a future study.

Figs 6 and 7 show the weekly cumulative fluxes of each year and their differences from 9-year average values calculated for EB and ET, respectively. The EB region shows seasonal variation every year, absorbing CO₂ strongly in summer and emitting CO₂ in spring and winter (Fig 6A and 6C). CNTL shows a decrease in summer CO₂ flux absorption in 2003, 2004, 2006, and 2007, whereas shows strong above average CO₂ flux absorption in summer of 2008, 2009, and 2011 (Fig 6B). In EXP1, the flux absorption is much more active during the whole analysis period since averaged weekly cumulative flux is shifted to the negative direction (Fig 6C). From 2007 to 2011, the flux absorption in EXP1 is similar to or greater than the 9-year averaged flux (Fig 6D).

Download:

Fig 6.

Weekly cumulative biosphere CO₂ fluxes (Pg C) of each year over Eurasia Boreal region: (a) CNTL and (c) EXP1. Weekly cumulative biosphere CO₂ flux differences (Pg C) in each year from the 9-year average for 2003–2011 period over Eurasia Boreal region: (b) CNTL and (d) EXP1.

https://doi.org/10.1371/journal.pone.0263925.g006

Download:

Fig 7. Same with Fig 6 except for Eurasia Temperate (ET) region.

https://doi.org/10.1371/journal.pone.0263925.g007

The ET region shows seasonal variation similar to the EB region, but its magnitude decreases by half (Fig 7A and 7C). In CNTL, the flux absorption is weaker than the 9-year averaged flux absorption in 2003, 2005, and 2010, whereas the flux absorption is stronger and emission is weaker in 2006, 2007, and 2009 (Fig 7B). The average cumulative CO₂ flux for EXP1 is lower than the CNTL, indicating EXP1 uptakes more CO₂ flux than CNTL does (Fig 7D). In EXP1, strong spring CO₂ uptake occurred in 2007 and from 2009 to 2011, which made the average cumulative flux in spring period close to zero. In 2010, CO₂ uptake in EXP1 decreased from summer to fall, which is similar to the CNTL result, but EXP1 showed more CO₂ flux absorption in spring and early winter.

Overall, more weekly cumulative CO₂ absorption is simulated for the terrestrial biosphere in Asia, and the flux differences are more diverse when assimilating the two Korean observation datasets in CT.

3.2 Verification with independent observations

3.2.1 Verification with surface observations.

Table 4 summarizes bias and RMSE of model CO₂ concentrations with respect to observed CO₂ concentrations, and correlation coefficient between model CO₂ concentrations and observed CO₂ concentrations, for seven independent surface CO₂ observation sites in Asia. CNTL shows positive bias for every site except RYO, implying that the CNTL generally overestimates observed CO₂ concentration at evaluation sites. EXP1 also shows positive bias for all sites, but their absolute values are smaller than those of CNTL except RYO, which indicates that EXP1 estimates more accurate CO₂ concentrations than CNTL.

Download:

Table 4. Bias (ppm), RMSE (ppm), and correlation coefficient of model CO₂ concentrations of CNTL and EXP1 with respect to observed CO₂ concentrations at seven independent surface observation sites in Asia.

https://doi.org/10.1371/journal.pone.0263925.t004

The monthly model CO₂ concentrations in CNTL and EXP1 are mostly overestimated compared to the monthly observed CO₂ concentrations and the biases are relatively smaller in winter than in summer, indicating better performance of CT in winter (not shown). EXP1 shows smaller biases than CNTL during November to April period except January.

The RMSE of CNTL is smaller than that of EXP1 for COI, HAT, RYO, MNM, and YON sites, which are located on islands or seaside. In contrast, the RMSE of EXP1 is smaller than that of CNTL for LLN and SDZ sites which are located inland. Averaging over all sites, the RMSE of EXP1 (3.279 ppm) is slightly less than that of CNTL (3.304 ppm). Pearson correlation coefficient shows similar results to the RMSE results. SDZ and LLN have higher correlation values in EXP1 than CNTL, while the other sites have higher correlation coefficient values in CNTL than EXP1.

The evaluation with the independent surface CO₂ observations indicates that, by assimilating AMY and GSN site observations into CT, the bias of model CO₂ concentration could be reduced and the model CO₂ concentration with DA could be more accurate than that without DA, particularly over inland vegetation region. In terms of monthly verification, the bias of monthly model CO₂ concentration could be reduced for the winter to early spring seasons, by assimilating AMY and GSN site observations into CT.

3.2.2 Verification with CONTRAIL aircraft observations.

In this section, the model CO₂ concentrations for each experiment are verified with respect to the independent CONTRAIL observations, which are not assimilated in any of the experiments. CONTRAIL observations are categorized with two types of observation mode: ascending/descending mode and level mode (Fig 8). The vertical observation (ascending/descending mode) is conducted while ascending from/descending to the airport. Bin numbers 3 to 8 represent the ascending/descending mode where the observations are conducted at four different levels (level 1: 625–575 hPa, level 2: 525–475 hPa, level 3: 425–375 hPa, and level 4: 275–225 hPa). The 6 bins drawn with thick boundaries represent ascending/descending mode, and their locations are New Delhi (bin 3), Bangkok (bin 4), Singapore (bin 5), Jakarta (bin 6), Incheon (bin 7), and Tokyo (bin 8). The other bins (bin numbers 17 to 35) contain the observations from the upper atmospheric layer of 275–225 hPa (level mode). The observations in the stratosphere are excluded from the analysis. The method of separating CONTRAIL data into several bins follows [7].

Download:

Fig 8. CONTRAIL bins filled with dark gray represent that RMSEs of CNTL are smaller than those of EXP1, while bins filled with dots represent that RMSEs of EXP1 are smaller than those of CNTL.

The numbers are assigned on bins to identify the observational regions of CONTRAIL data.

https://doi.org/10.1371/journal.pone.0263925.g008

Table 5 shows RMSE of model CO₂ concentrations for each experiment with respect to the CONTRAIL observations. The average RMSEs for total CONTRAIL data are smaller in EXP1 (1.506 ppm) than in CNTL (1.545 ppm), and those for the ascending/descending mode are also smaller in EXP1 (1.581 ppm) than in CNTL (1.643 ppm). At bins 3, 4, 7 and 8, EXP1 shows smaller RMSEs than CNTL does at each level. At bin 5, EXP1 shows smaller RMSE than CNTL does only at level 4. At bin 6, CNTL shows smaller RMSEs than EXP1 does at level 1, 2 and 4. In case of the level mode, the RMSE of EXP1 (1.414 ppm) is also smaller than that of CNTL (1.424 ppm).

Download:

Table 5. RMSE (ppm) of model CO₂ concentrations of CNTL and EXP1 with respect to observed CO₂ concentrations at the CONTRAIL bins in the nested domain.

https://doi.org/10.1371/journal.pone.0263925.t005

When considering the whole levels for the ascending/descending mode observation, the bins 3, 4, 5, 7 and 8 have the smaller RMSEs in EXP1, while the bin 6 has the smaller RMSE in CNTL (Fig 8). For the level mode observation, bins 24–27 and 29–35 have the smaller RMSEs in EXP1, and the other level mode bins (i.e., 17–23 and 28) have the smaller RMSEs in CNTL. The bins showing smaller RMSEs in EXP1 are located over ET regions including Korea and Japan, TA regions, and Northwest Pacific Ocean. The bins showing smaller RMSEs in CNTL are located over the Siberian regions (EB) and near the equator.

Overall, the evaluation with the independent CONTRAIL observations shows that assimilating additional Korean observations into the CT led to more accurate surface CO₂ flux estimations over Asia.

3.3 Uncertainty reduction

Fig 9 shows the uncertainty reduction rate of each experiment for the estimated posterior surface CO₂ flux compared to the prior surface CO₂ flux, averaged over the analysis period. The area of maximum uncertainty reduction in CNTL appears on parts of inner China, Mongolia, and central Asia regions, showing approximately 40% of reduction. It is followed by Siberian region near 60°N latitude, showing approximately 28% of uncertainty reduction after the optimization (Fig 9A). For EXP1, the uncertainty reduction is approximately 59% on Eastern China, Korea, and Japan, where the location coincides with the mixed forest ecoregion of ET (Fig 9B). Uncertainty reduction rate around 60°N Siberian region is still relatively high in EXP1 compared to CNTL, nearly 35% of uncertainty decreases. Some parts of India show 27% uncertainty reduction in EXP1.

Download:

Fig 9.

Average uncertainty reduction (%) for 2003–2011 period: (a) CNTL, (b) EXP1, and (c) the difference between EXP1 and CNTL.

https://doi.org/10.1371/journal.pone.0263925.g009

The difference of the uncertainty reduction between EXP1 and CNTL is shown in Fig 9C. Compared to CNTL, EXP1 shows more uncertainty reduction over Eastern China, Korea, Japan, and India, located in ET region. In Siberia, there is no distinct difference in uncertainty reduction between CNTL and EXP1. EB and TA regions also show little differences between EXP1 and CNTL. Therefore, the uncertainties in the estimated surface CO₂ flux over ET are reduced by adding the two Korean observation datasets in CT.

3.4 Influence matrix

An influence matrix is used to measure the influence of assimilated observation data on the model result. [6, 41, 54] calculated the influence of surface CO₂ observation data assimilated on the estimated surface CO₂ flux in CT. The diagonal component of the influence matrix, known as self-sensitivity or observation impact, represents the impact of the observation on model value on each observation site.

Fig 10 shows the self-sensitivities of the surface CO₂ observations assimilated in EXP1, averaged over the analysis period 2003–2011. AMY, GSN, and some overlapped sites are marked as circles with different size and color. The spatial density of observation sites is high in the North America and the Europe. The number of observation sites is relatively smaller in Asia, Australia, other continents, and oceans. Large self-sensitivities are found around observation-sparse regions. The self-sensitivities are fairly evenly distributed where observation sites are dense, although some sites in Alaska, Western America, and Northern Europe show large self-sensitivities. The global average self-sensitivity in EXP1 is 6.14%. The self-sensitivities of AMY and GSN are 11.71% and 11.38%, respectively, which are larger than the global average. Given that the self-sensitivity of TAP is 2.7%, the relatively large self-sensitivities in AMY and GSN imply that AMY and GSN observations play a more important role in producing the optimized surface CO₂ flux. AMY and GSN observations also help estimate surface CO₂ flux in Asia with low observation density.

Download:

Fig 10. Average self-sensitivities at surface observation sites in the ObsPack and added sites in Korean Peninsula (AMY, GSN) for 2003–2011 period.

The black box with dotted line is the nesting domain. The overlapping observation sites located at the same place or placed closely are distinguished by different sizes of circles.

https://doi.org/10.1371/journal.pone.0263925.g010

4. Summary and conclusion

In this study, two CO₂ observation datasets from AMY and GSN sites in the Korean peninsula are introduced in CT DA system and the effect of the observations on surface CO₂ flux estimation in Asia is investigated for the 9-year period from 2003 to 2011. The annual average surface CO₂ flux uptake on the East Asia is enhanced in EXP1 experiment in which AMY and GSN observations are assimilated. By assimilating observations from the AMY and GSN, ET regions including the Korean peninsula, Japan, and inland China show stronger CO₂ absorption in summer, while weakened CO₂ emission in spring and autumn. EB regions also show the similar pattern.

Independent surface and aircraft CO₂ observations are used for the verification of the experimental results. Assimilating two additional observation datasets into CT reduced the root mean square error of modeled CO₂ concentration with respect to independent CO₂ observation concentration, and enhanced uncertainty reduction when optimizing surface CO₂ flux in Asia region. The regions with small RMSEs are consistent with the regions with significant uncertainty reduction, which include the Korean Peninsula, southern inland China, eastern China, and Japan.

Self-sensitivities at AMY and GSN are relatively high, which indicates that the two observation sites in Korea (AMY, GSN) are considerably important in estimating surface CO₂ flux in Asia. The use of CO₂ observations in the Korean Peninsula is expected to greatly contribute not only to the estimation of surface CO₂ flux in Asia at various scales, but also to the elaboration of the national emission inventory.

Acknowledgments

The authors appreciate the reviewers’ valuable comments. The authors are grateful to National Institute of Meteorological Sciences at KMA for CO₂ mole fraction data at AMY and GSN sites, NOAA ESRL for providing observations and resources to run CarbonTracker, and NIES Japan for CONTRAIL data. The authors also acknowledge atmospheric CO₂ measurement data providers and cooperating agencies at China Meteorological Administration, Japan Meteorological Administration, Commonwealth Scientific and Industrial Research Organization, Environment and Climate Change Canada, Instituto de Pesquisas Energeticas e Nucleares, Lawrence Berkeley National Laboratory, National Institute of Environmental Research, and National Center for Atmospheric Research. The simulations were primarily conducted using the cluster of the National Center for Meteorological Supercomputer of the Korea Meteorological Administration (KMA).

References

1. IPCC. Summary for Policymakers. In: Masson-Delmotte V, Zhai P, Pirani A, Connors SL, Péan C, Berger S, et al., editors. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Forthcoming 2021.
- View Article
- Google Scholar
2. UNFCCC, Conference of the Parties (COP). Adoption of the Paris Agreement. Proposal by the President. Paris Climate Change Conference–November 2015. COP 21. 2015. 32 p. FCCC/CP/2015/L.9/Rev.1. Available from: http://unfccc.int/resource/docs/2015/cop21/eng/l09r01.pdf.
- View Article
- Google Scholar
3. IPCC. Summary for Policymakers. In: Masson-Delmotte V, Zhai P, Pörtner HO, Roberts D, Skea J, Shukla PR, et al., editors. Global warming of 1.5°C. An IPCC Special Report on the impacts of global warming of 1.5°C above pre-industrial levels and related global greenhouse gas emission pathways, in the context of strengthening the global response to the threat of climate change, sustainable development, and efforts to eradicate poverty. Geneva, Switzerland: World Meteorological Organization; 2018. 32 p.
4. Enting IG, Trudinger CM, Francey RJ. A synthesis inversion of the concentration and δ13C of atmospheric CO₂. Tellus B Chem Phys Meteorol. 1995;47:35–52.
- View Article
- Google Scholar
5. van der Laan-Luijkx IT, van der Velde IR, van der Veen E, Tsuruta A, Stanislawska K, Babenhauserheide A, et al. The CarbonTracker Data Assimilation Shell (CTDAS) v1.0: implementation and global carbon balance 2001–2015. Geosci Model Dev. 2017;10(7):2785–800.
- View Article
- Google Scholar
6. Kim J, Kim HM, Cho CH, Boo KO, Jacobson AR, Sasakawa M, et al. Impact of Siberian observations on the optimization of surface CO₂ flux. Atmos Chem Phys. 2017;17(4):2881–99.
- View Article
- Google Scholar
7. Kim H, Kim HM, Cho M, Park J, Kim D. Development of the Aircraft CO2 Measurement Data Assimilation System to Improve the Estimation of Surface CO2 Fluxes Using an Inverse Modeling System. Atmosphere. 2018b;28(2):113–21. (in Korean with English abstract).
- View Article
- Google Scholar
8. Zhang HF, Chen BZ, van der Laan-Luijkx IT, Machida T, Matsueda H, Sawa Y, et al. Estimating Asian terrestrial carbon fluxes from CONTRAIL aircraft and surface CO₂ observations for the period 2006–2010. Atmos Chem Phys. 2014;14(11), 5807–5824.
- View Article
- Google Scholar
9. Maksyutov S, Machida T, Mukai H, Patra PK, Nakazawa T, Inoue G, et al. Effect of recent observations on Asian CO₂ flux estimates by transport model inversions. Tellus B Chem Phys Meteorol. 2003;55(2):522–9.
- View Article
- Google Scholar
10. Kuze A, Suto H, Nakajima M, Hamazaki T. Thermal and near infrared sensor for carbon observation Fourier-transform spectrometer on the Greenhouse Gases Observing Satellite for greenhouse gases monitoring. Appl Opt. 2009;48(35):6716–33. pmid:20011012
- View Article
- PubMed/NCBI
- Google Scholar
11. Yokota T, Yoshida Y, Eguchi N, Ota Y, Tanaka T, Watanabe H, et al. Global Concentrations of CO2 and CH4 Retrieved from GOSAT: First Preliminary Results. SOLA. 2009;5:160–3.
- View Article
- Google Scholar
12. Takagi H, Saeki T, Oda T, Saito M, Valsala V, Belikov D, et al. On the Benefit of GOSAT Observations to the Estimation of Regional CO2 Fluxes. SOLA. 2011;7(0):161–4.
- View Article
- Google Scholar
13. Saeki T, Maksyutov S, Saito M, Valsala V, Oda T, Andres RJ, et al. Inverse Modeling of CO2 Fluxes Using GOSAT Data and Multi-Year Ground-Based Observations. SOLA. 2013;9(0):45–50.
- View Article
- Google Scholar
14. Basu S, Krol M, Butz A, Clerbaux C, Sawa Y, Machida T, et al. The seasonal variation of the CO₂ flux over Tropical Asia estimated from GOSAT, CONTRAIL, and IASI. Geophys Res Lett. 2014;41(5):1809–15.
- View Article
- Google Scholar
15. Chevallier F, Remaud M, O’Dell CW, Baker D, Peylin P, Cozic A. Objective evaluation of surface- and satellite-driven carbon dioxide atmospheric inversions. Atmos Chem Phys. 2019;19(22):14233–51.
- View Article
- Google Scholar
16. Deng F, Jones DBA, O’Dell CW, Nassar R, Parazoo NC. Combining GOSAT XCO₂ observations over land and ocean to improve regional CO₂ flux estimates. J Geophys Res Atmos. 2016;121(4):1896–913.
- View Article
- Google Scholar
17. Houweling S, Baker D, Basu S, Boesch H, Butz A, Chevallier F, et al. An intercomparison of inverse models for estimating sources and sinks of CO₂ using GOSAT measurements. J Geophys Res Atmos. 2015;120(10):5253–66.
- View Article
- Google Scholar
18. Feng L, Palmer PI, Parker RJ, Deutscher NM, Feist DG, Kivi R, et al. Estimates of European uptake of CO₂ inferred from GOSAT XCO₂ retrievals: sensitivity to measurement bias inside and outside Europe. Atmos Chem Phys. 2016;16(3):1289–302.
- View Article
- Google Scholar
19. Wang H, Jiang F, Wang J, Ju W, Chen JM. Terrestrial ecosystem carbon flux estimated using GOSAT and OCO-2 XCO₂ retrievals. Atmos Chem Phys. 2019;19(18):12067–82.
- View Article
- Google Scholar
20. Byrne B, Liu J, Lee M, Baker I, Bowman KW, Deutscher NM, et al. Improved Constraints on Northern Extratropical CO₂ Fluxes Obtained by Combining Surface‐Based and Space‐Based Atmospheric CO₂ Measurements. J Geophys Res Atmos. 2020;125(15):e2019JD032029.
- View Article
- Google Scholar
21. Jiang F, Wang H, Chen JM, Ju W, Tian X, Feng S, et al. Regional CO₂ fluxes from 2010 to 2015 inferred from GOSAT XCO₂ retrievals using a new version of the Global Carbon Assimilation System. Atmos Chem Phys. 2021;21(3):1963–85.
- View Article
- Google Scholar
22. Crisp D, Atlas RM, Breon FM, Brown LR, Burrows JP, Ciais P, et al. The Orbiting Carbon Observatory (OCO) mission. Adv Space Res. 2004;34(4):700–9.
- View Article
- Google Scholar
23. Eldering A, O’Dell CW, Wennberg PO, Crisp D, Gunson MR, Viatte C, et al. The Orbiting Carbon Observatory-2: first 18 months of science data products. Atmos Meas Tech. 2017;10(2):549–63.
- View Article
- Google Scholar
24. Liu J, Bowman KW, Schimel DS, Parazoo NC, Jiang Z, Lee M, et al. Contrasting carbon cycle responses of the tropical continents to the 2015–2016 El Nino. Science. 2017;358(6360): eaam5690. pmid:29026011
- View Article
- PubMed/NCBI
- Google Scholar
25. Miller SM, Michalak AM, Yadav V, Tadić JM. Characterizing biospheric carbon balance using CO₂ observations from the OCO-2 satellite. Atmos Chem Phys. 2018;18(9):6785–99.
- View Article
- Google Scholar
26. Crowell S, Baker D, Schuh A, Basu S, Jacobson AR, Chevallier F, et al. The 2015–2016 carbon cycle as seen from OCO-2 and the global in situ network. Atmos Chem Phys. 2019;19(15):9797–831.
- View Article
- Google Scholar
27. Steinkamp K, Mikaloff Fletcher SE, Brailsford G, Smale D, Moore S, Keller ED, et al. Atmospheric CO₂ observations and models suggest strong carbon uptake by forests in New Zealand, Atmos Chem Phys. 2017;17(1):47–76.
- View Article
- Google Scholar
28. White ED, Rigby M, Lunt MF, Smallman TL, Comyn-Platt E, Manning AJ, et al. Quantifying the UK’s carbon dioxide flux: an atmospheric inverse modelling approach using a regional measurement network. Atmos Chem Phys. 2019;19(7):4345–65.
- View Article
- Google Scholar
29. Peters W, Jacobson AR, Sweeney C, Andrews AE, Conway TJ, Masarie KA, et al. An atmospheric perspective on North American carbon dioxide exchange: CarbonTracker. Proc Natl Acad Sci U S A. 2007;104(48):18925–30. pmid:18045791
- View Article
- PubMed/NCBI
- Google Scholar
30. van der Werf GR, Randerson JT, Giglio L, Collatz GJ, Kasibhatla PS, Arellano AF Jr. Interannual variability in global biomass burning emissions from 1997 to 2004. Atmos Chem Phys. 2006;6(11):3423–41.
- View Article
- Google Scholar
31. van der Werf GR, Randerson JT, Giglio L, Collatz GJ, Mu M, Kasibhatla PS, et al. Global fire emissions and the contribution of deforestation, savanna, forest, agricultural, and peat fires (1997–2009). Atmos Chem Phys. 2010;10:11707–35.
- View Article
- Google Scholar
32. Jacobson AR, Fletcher SEM, Gruber N, Sarmiento JL, Gloor M. A joint atmosphere-ocean inversion for surface fluxes of carbon dioxide: 2. Regional results. Global Biogeochem Cycles. 2007;21(1):GB1020.
- View Article
- Google Scholar
33. Boden TA, Marland G, Andres RJ. Global, Regional, and National Fossil-Fuel CO2 Emissions, Carbon Dioxide Information Analysis Center (CDIAC). Oak Ridge National Laboratory, US Department of Energy, Oak Ridge. US Department of Energy Office of Scientific and Technical Information. 2010. https://doi.org/10.3334/CDIAC/00001_V2010
34. CarbonTracker. CarbonTracker Documentation CT2013B release. 2015. Online at: https://gml.noaa.gov/ccgg/carbontracker/CT2013B/CT2013B_doc.pdf.
- View Article
- Google Scholar
35. Krol M, Houweling S, Bregman B, van den Broek M, Segers A, van Velthoven P, et al. The two-way nested global chemistry-transport zoom model TM5: algorithm and applications. Atmos Chem Phys. 2005;5:417–32.
- View Article
- Google Scholar
36. Whitaker JS, Hamill TM. Ensemble Data Assimilation without Perturbed Observations. Mon Weather Rev. 2002;130(7):1913–24.
- View Article
- Google Scholar
37. Houtekamer PL, Mitchell HL. A Sequential Ensemble Kalman Filter for Atmospheric Data Assimilation. Mon Weather Rev. 2001;129:123–37.
- View Article
- Google Scholar
38. Bruhwiler LMP, Michalak AM, Peters W, Baker DF, Tans PP. An improved Kalman smoother for atmospheric inversions. Atmos Chem Phys. 2005;5(10):2691–702.
- View Article
- Google Scholar
39. Kim J, Kim HM, Cho CH. Application of Carbon Tracking System based on Ensemble Kalman Filter on the Diagnosis of Carbon Cycle in Asia. Atmosphere. 2012;22(4):415–27. (in Korean with English abstract)
- View Article
- Google Scholar
40. Kim J, Kim HM, Cho CH. The effect of optimization and the nesting domain on carbon flux analyses in Asia using a carbon tracking system based on the ensemble Kalman filter. Asia Pac J Atmos Sci. 2014a;50(3):327–44.
- View Article
- Google Scholar
41. Kim J, Kim HM, Cho CH. Influence of CO₂ observations on the optimized CO₂ flux in an ensemble Kalman filter. Atmos Chem Phys. 2014b;14(24):13515–30.
- View Article
- Google Scholar
42. Masarie KA, Peters W, Jacobson AR, Tans PP. ObsPack: A framework for the preparation, delivery, and attribution of atmospheric greenhouse gas measurements. Earth Syst Sci Data. 2014;6(2):375–84.
- View Article
- Google Scholar
43. Cho M, Kim HM. Impact of assimilating CO₂ observations in the Korean Peninsula on surface carbon flux estimation in East Asia and comparison of the estimated carbon flux with emission inventory data. Poster session presented at: European Geosciences Union General Assembly; 2019 Apr 7–12; Vienna, Austria.
- View Article
- Google Scholar
44. Machida T, Matsueda H, Sawa Y, Nakagawa Y, Hirotani K, Kondo K, et al. Worldwide measurements of atmospheric CO₂ and other trace gas species using commercial airlines. J Atmos Ocean Technol. 2008;25 (10):1744–54.
- View Article
- Google Scholar
45. Machida T, Ishijima K, Niwa Y, Tsuboi K, Sawa Y, Matsueda H, et al. Atmospheric CO2 mole fraction data of CONTRAIL-CME. Version 2017.1.0 [dataset on the Internet]. Tsukuba-city (Japan): Center for Global Environmental Research, NIES, 2018 Feb 8 [cited 2021 Dec 12]. https://doi.org/10.17595/20180208.001
46. The NCAR Command Language. Version 6.6.2 [software]. Boulder (CO): UCAR/NCAR/CISL/TDD. 2019 [cited 2021 Dec 12]. Available from: http://dx.doi.org/10.5065/D6WD3XH5.
47. Gérard JC, Nemry B, Francois LM, Warnant P. The interannual change of atmospheric CO₂: Contribution of subtropical ecosystems? Geophys Res Lett.1999;26(2):243–6.
- View Article
- Google Scholar
48. Cervarich M, Shu S, Jain AK, Arneth A, Canadell J, Friedlingstein P, et al. The terrestrial carbon budget of South and Southeast Asia. Environ Res Lett. 2016;11(10):105006.
- View Article
- Google Scholar
49. Knorr W, Gobron N, Scholze M, Kaminski T, Schnur R, Pinty B. Impact of terrestrial biosphere carbon exchanges on the anomalous CO₂ increase in 2002–2003. Geophys Res Lett. 2007;34(9):L09703.
- View Article
- Google Scholar
50. Gurney KR, Law RM, Denning AS, Rayner PJ, Baker D, Bousquet P, et al. Towards robust regional estimates of CO₂ sources and sinks using atmospheric transport models. Nature. 2002;415:626–30. pmid:11832942
- View Article
- PubMed/NCBI
- Google Scholar
51. Kim H, Kim HM, Kim J, Cho CH. Effect of data assimilation parameters on the optimized surface CO2 flux in Asia. Asia-Pacific Journal of Atmospheric Sciences. 2018a;54:1–17.
- View Article
- Google Scholar
52. Peylin P, Law RM, Gurney KR, Chevallier F, Jacobson AR, Maki T, et al. Global atmospheric carbon budget: results from an ensemble of atmospheric CO₂ inversions. Biogeosciences. 2013;10(10):6699–720.
- View Article
- Google Scholar
53. Gaubert B, Stephens BB, Basu S, Chevallier F, Deng F, Kort EA, et al. Global atmospheric CO₂ inverse models converging on neutral tropical land exchange, but disagreeing on fossil fuel and atmospheric growth rate. Biogeosciences. 2019;16(1):117–34. pmid:31708981
- View Article
- PubMed/NCBI
- Google Scholar
54. Park J, Kim HM. Design and evaluation of CO2 observation network to optimize surface CO2 fluxes in Asia using observation system simulation experiments. Atmos Chem Phys. 2020;20(8):5175–95.
- View Article
- Google Scholar

[ref1] 1. IPCC. Summary for Policymakers. In: Masson-Delmotte V, Zhai P, Pirani A, Connors SL, Péan C, Berger S, et al., editors. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Forthcoming 2021.
View Article
Google Scholar

[2] View Article

[3] Google Scholar

[ref2] 2. UNFCCC, Conference of the Parties (COP). Adoption of the Paris Agreement. Proposal by the President. Paris Climate Change Conference–November 2015. COP 21. 2015. 32 p. FCCC/CP/2015/L.9/Rev.1. Available from: http://unfccc.int/resource/docs/2015/cop21/eng/l09r01.pdf.
View Article
Google Scholar

[5] View Article

[6] Google Scholar

[ref3] 3. IPCC. Summary for Policymakers. In: Masson-Delmotte V, Zhai P, Pörtner HO, Roberts D, Skea J, Shukla PR, et al., editors. Global warming of 1.5°C. An IPCC Special Report on the impacts of global warming of 1.5°C above pre-industrial levels and related global greenhouse gas emission pathways, in the context of strengthening the global response to the threat of climate change, sustainable development, and efforts to eradicate poverty. Geneva, Switzerland: World Meteorological Organization; 2018. 32 p.

[ref4] 4. Enting IG, Trudinger CM, Francey RJ. A synthesis inversion of the concentration and δ13C of atmospheric CO₂. Tellus B Chem Phys Meteorol. 1995;47:35–52.
View Article
Google Scholar

[9] View Article

[10] Google Scholar

[ref5] 5. van der Laan-Luijkx IT, van der Velde IR, van der Veen E, Tsuruta A, Stanislawska K, Babenhauserheide A, et al. The CarbonTracker Data Assimilation Shell (CTDAS) v1.0: implementation and global carbon balance 2001–2015. Geosci Model Dev. 2017;10(7):2785–800.
View Article
Google Scholar

[12] View Article

[13] Google Scholar

[ref6] 6. Kim J, Kim HM, Cho CH, Boo KO, Jacobson AR, Sasakawa M, et al. Impact of Siberian observations on the optimization of surface CO₂ flux. Atmos Chem Phys. 2017;17(4):2881–99.
View Article
Google Scholar

[15] View Article

[16] Google Scholar

[ref7] 7. Kim H, Kim HM, Cho M, Park J, Kim D. Development of the Aircraft CO2 Measurement Data Assimilation System to Improve the Estimation of Surface CO2 Fluxes Using an Inverse Modeling System. Atmosphere. 2018b;28(2):113–21. (in Korean with English abstract).
View Article
Google Scholar

[18] View Article

[19] Google Scholar

[ref8] 8. Zhang HF, Chen BZ, van der Laan-Luijkx IT, Machida T, Matsueda H, Sawa Y, et al. Estimating Asian terrestrial carbon fluxes from CONTRAIL aircraft and surface CO₂ observations for the period 2006–2010. Atmos Chem Phys. 2014;14(11), 5807–5824.
View Article
Google Scholar

[21] View Article

[22] Google Scholar

[ref9] 9. Maksyutov S, Machida T, Mukai H, Patra PK, Nakazawa T, Inoue G, et al. Effect of recent observations on Asian CO₂ flux estimates by transport model inversions. Tellus B Chem Phys Meteorol. 2003;55(2):522–9.
View Article
Google Scholar

[24] View Article

[25] Google Scholar

[ref10] 10. Kuze A, Suto H, Nakajima M, Hamazaki T. Thermal and near infrared sensor for carbon observation Fourier-transform spectrometer on the Greenhouse Gases Observing Satellite for greenhouse gases monitoring. Appl Opt. 2009;48(35):6716–33. pmid:20011012
View Article
PubMed/NCBI
Google Scholar

[27] View Article

[28] PubMed/NCBI

[29] Google Scholar

[ref11] 11. Yokota T, Yoshida Y, Eguchi N, Ota Y, Tanaka T, Watanabe H, et al. Global Concentrations of CO2 and CH4 Retrieved from GOSAT: First Preliminary Results. SOLA. 2009;5:160–3.
View Article
Google Scholar

[31] View Article

[32] Google Scholar

[ref12] 12. Takagi H, Saeki T, Oda T, Saito M, Valsala V, Belikov D, et al. On the Benefit of GOSAT Observations to the Estimation of Regional CO2 Fluxes. SOLA. 2011;7(0):161–4.
View Article
Google Scholar

[34] View Article

[35] Google Scholar

[ref13] 13. Saeki T, Maksyutov S, Saito M, Valsala V, Oda T, Andres RJ, et al. Inverse Modeling of CO2 Fluxes Using GOSAT Data and Multi-Year Ground-Based Observations. SOLA. 2013;9(0):45–50.
View Article
Google Scholar

[37] View Article

[38] Google Scholar

[ref14] 14. Basu S, Krol M, Butz A, Clerbaux C, Sawa Y, Machida T, et al. The seasonal variation of the CO₂ flux over Tropical Asia estimated from GOSAT, CONTRAIL, and IASI. Geophys Res Lett. 2014;41(5):1809–15.
View Article
Google Scholar

[40] View Article

[41] Google Scholar

[ref15] 15. Chevallier F, Remaud M, O’Dell CW, Baker D, Peylin P, Cozic A. Objective evaluation of surface- and satellite-driven carbon dioxide atmospheric inversions. Atmos Chem Phys. 2019;19(22):14233–51.
View Article
Google Scholar

[43] View Article

[44] Google Scholar

[ref16] 16. Deng F, Jones DBA, O’Dell CW, Nassar R, Parazoo NC. Combining GOSAT XCO₂ observations over land and ocean to improve regional CO₂ flux estimates. J Geophys Res Atmos. 2016;121(4):1896–913.
View Article
Google Scholar

[46] View Article

[47] Google Scholar

[ref17] 17. Houweling S, Baker D, Basu S, Boesch H, Butz A, Chevallier F, et al. An intercomparison of inverse models for estimating sources and sinks of CO₂ using GOSAT measurements. J Geophys Res Atmos. 2015;120(10):5253–66.
View Article
Google Scholar

[49] View Article

[50] Google Scholar

[ref18] 18. Feng L, Palmer PI, Parker RJ, Deutscher NM, Feist DG, Kivi R, et al. Estimates of European uptake of CO₂ inferred from GOSAT XCO₂ retrievals: sensitivity to measurement bias inside and outside Europe. Atmos Chem Phys. 2016;16(3):1289–302.
View Article
Google Scholar

[52] View Article

[53] Google Scholar

[ref19] 19. Wang H, Jiang F, Wang J, Ju W, Chen JM. Terrestrial ecosystem carbon flux estimated using GOSAT and OCO-2 XCO₂ retrievals. Atmos Chem Phys. 2019;19(18):12067–82.
View Article
Google Scholar

[55] View Article

[56] Google Scholar

[ref20] 20. Byrne B, Liu J, Lee M, Baker I, Bowman KW, Deutscher NM, et al. Improved Constraints on Northern Extratropical CO₂ Fluxes Obtained by Combining Surface‐Based and Space‐Based Atmospheric CO₂ Measurements. J Geophys Res Atmos. 2020;125(15):e2019JD032029.
View Article
Google Scholar

[58] View Article

[59] Google Scholar

[ref21] 21. Jiang F, Wang H, Chen JM, Ju W, Tian X, Feng S, et al. Regional CO₂ fluxes from 2010 to 2015 inferred from GOSAT XCO₂ retrievals using a new version of the Global Carbon Assimilation System. Atmos Chem Phys. 2021;21(3):1963–85.
View Article
Google Scholar

[61] View Article

[62] Google Scholar

[ref22] 22. Crisp D, Atlas RM, Breon FM, Brown LR, Burrows JP, Ciais P, et al. The Orbiting Carbon Observatory (OCO) mission. Adv Space Res. 2004;34(4):700–9.
View Article
Google Scholar

[64] View Article

[65] Google Scholar

[ref23] 23. Eldering A, O’Dell CW, Wennberg PO, Crisp D, Gunson MR, Viatte C, et al. The Orbiting Carbon Observatory-2: first 18 months of science data products. Atmos Meas Tech. 2017;10(2):549–63.
View Article
Google Scholar

[67] View Article

[68] Google Scholar

[ref24] 24. Liu J, Bowman KW, Schimel DS, Parazoo NC, Jiang Z, Lee M, et al. Contrasting carbon cycle responses of the tropical continents to the 2015–2016 El Nino. Science. 2017;358(6360): eaam5690. pmid:29026011
View Article
PubMed/NCBI
Google Scholar

[70] View Article

[71] PubMed/NCBI

[72] Google Scholar

[ref25] 25. Miller SM, Michalak AM, Yadav V, Tadić JM. Characterizing biospheric carbon balance using CO₂ observations from the OCO-2 satellite. Atmos Chem Phys. 2018;18(9):6785–99.
View Article
Google Scholar

[74] View Article

[75] Google Scholar

[ref26] 26. Crowell S, Baker D, Schuh A, Basu S, Jacobson AR, Chevallier F, et al. The 2015–2016 carbon cycle as seen from OCO-2 and the global in situ network. Atmos Chem Phys. 2019;19(15):9797–831.
View Article
Google Scholar

[77] View Article

[78] Google Scholar

[ref27] 27. Steinkamp K, Mikaloff Fletcher SE, Brailsford G, Smale D, Moore S, Keller ED, et al. Atmospheric CO₂ observations and models suggest strong carbon uptake by forests in New Zealand, Atmos Chem Phys. 2017;17(1):47–76.
View Article
Google Scholar

[80] View Article

[81] Google Scholar

[ref28] 28. White ED, Rigby M, Lunt MF, Smallman TL, Comyn-Platt E, Manning AJ, et al. Quantifying the UK’s carbon dioxide flux: an atmospheric inverse modelling approach using a regional measurement network. Atmos Chem Phys. 2019;19(7):4345–65.
View Article
Google Scholar

[83] View Article

[84] Google Scholar

[ref29] 29. Peters W, Jacobson AR, Sweeney C, Andrews AE, Conway TJ, Masarie KA, et al. An atmospheric perspective on North American carbon dioxide exchange: CarbonTracker. Proc Natl Acad Sci U S A. 2007;104(48):18925–30. pmid:18045791
View Article
PubMed/NCBI
Google Scholar

[86] View Article

[87] PubMed/NCBI

[88] Google Scholar

[ref30] 30. van der Werf GR, Randerson JT, Giglio L, Collatz GJ, Kasibhatla PS, Arellano AF Jr. Interannual variability in global biomass burning emissions from 1997 to 2004. Atmos Chem Phys. 2006;6(11):3423–41.
View Article
Google Scholar

[90] View Article

[91] Google Scholar

[ref31] 31. van der Werf GR, Randerson JT, Giglio L, Collatz GJ, Mu M, Kasibhatla PS, et al. Global fire emissions and the contribution of deforestation, savanna, forest, agricultural, and peat fires (1997–2009). Atmos Chem Phys. 2010;10:11707–35.
View Article
Google Scholar

[93] View Article

[94] Google Scholar

[ref32] 32. Jacobson AR, Fletcher SEM, Gruber N, Sarmiento JL, Gloor M. A joint atmosphere-ocean inversion for surface fluxes of carbon dioxide: 2. Regional results. Global Biogeochem Cycles. 2007;21(1):GB1020.
View Article
Google Scholar

[96] View Article

[97] Google Scholar

[ref33] 33. Boden TA, Marland G, Andres RJ. Global, Regional, and National Fossil-Fuel CO2 Emissions, Carbon Dioxide Information Analysis Center (CDIAC). Oak Ridge National Laboratory, US Department of Energy, Oak Ridge. US Department of Energy Office of Scientific and Technical Information. 2010. https://doi.org/10.3334/CDIAC/00001_V2010

[ref34] 34. CarbonTracker. CarbonTracker Documentation CT2013B release. 2015. Online at: https://gml.noaa.gov/ccgg/carbontracker/CT2013B/CT2013B_doc.pdf.
View Article
Google Scholar

[100] View Article

[101] Google Scholar

[ref35] 35. Krol M, Houweling S, Bregman B, van den Broek M, Segers A, van Velthoven P, et al. The two-way nested global chemistry-transport zoom model TM5: algorithm and applications. Atmos Chem Phys. 2005;5:417–32.
View Article
Google Scholar

[103] View Article

[104] Google Scholar

[ref36] 36. Whitaker JS, Hamill TM. Ensemble Data Assimilation without Perturbed Observations. Mon Weather Rev. 2002;130(7):1913–24.
View Article
Google Scholar

[106] View Article

[107] Google Scholar

[ref37] 37. Houtekamer PL, Mitchell HL. A Sequential Ensemble Kalman Filter for Atmospheric Data Assimilation. Mon Weather Rev. 2001;129:123–37.
View Article
Google Scholar

[109] View Article

[110] Google Scholar

[ref38] 38. Bruhwiler LMP, Michalak AM, Peters W, Baker DF, Tans PP. An improved Kalman smoother for atmospheric inversions. Atmos Chem Phys. 2005;5(10):2691–702.
View Article
Google Scholar

[112] View Article

[113] Google Scholar

[ref39] 39. Kim J, Kim HM, Cho CH. Application of Carbon Tracking System based on Ensemble Kalman Filter on the Diagnosis of Carbon Cycle in Asia. Atmosphere. 2012;22(4):415–27. (in Korean with English abstract)
View Article
Google Scholar

[115] View Article

[116] Google Scholar

[ref40] 40. Kim J, Kim HM, Cho CH. The effect of optimization and the nesting domain on carbon flux analyses in Asia using a carbon tracking system based on the ensemble Kalman filter. Asia Pac J Atmos Sci. 2014a;50(3):327–44.
View Article
Google Scholar

[118] View Article

[119] Google Scholar

[ref41] 41. Kim J, Kim HM, Cho CH. Influence of CO₂ observations on the optimized CO₂ flux in an ensemble Kalman filter. Atmos Chem Phys. 2014b;14(24):13515–30.
View Article
Google Scholar

[121] View Article

[122] Google Scholar

[ref42] 42. Masarie KA, Peters W, Jacobson AR, Tans PP. ObsPack: A framework for the preparation, delivery, and attribution of atmospheric greenhouse gas measurements. Earth Syst Sci Data. 2014;6(2):375–84.
View Article
Google Scholar

[124] View Article

[125] Google Scholar

[ref43] 43. Cho M, Kim HM. Impact of assimilating CO₂ observations in the Korean Peninsula on surface carbon flux estimation in East Asia and comparison of the estimated carbon flux with emission inventory data. Poster session presented at: European Geosciences Union General Assembly; 2019 Apr 7–12; Vienna, Austria.
View Article
Google Scholar

[127] View Article

[128] Google Scholar

[ref44] 44. Machida T, Matsueda H, Sawa Y, Nakagawa Y, Hirotani K, Kondo K, et al. Worldwide measurements of atmospheric CO₂ and other trace gas species using commercial airlines. J Atmos Ocean Technol. 2008;25 (10):1744–54.
View Article
Google Scholar

[130] View Article

[131] Google Scholar

[ref45] 45. Machida T, Ishijima K, Niwa Y, Tsuboi K, Sawa Y, Matsueda H, et al. Atmospheric CO2 mole fraction data of CONTRAIL-CME. Version 2017.1.0 [dataset on the Internet]. Tsukuba-city (Japan): Center for Global Environmental Research, NIES, 2018 Feb 8 [cited 2021 Dec 12]. https://doi.org/10.17595/20180208.001

[ref46] 46. The NCAR Command Language. Version 6.6.2 [software]. Boulder (CO): UCAR/NCAR/CISL/TDD. 2019 [cited 2021 Dec 12]. Available from: http://dx.doi.org/10.5065/D6WD3XH5.

[ref47] 47. Gérard JC, Nemry B, Francois LM, Warnant P. The interannual change of atmospheric CO₂: Contribution of subtropical ecosystems? Geophys Res Lett.1999;26(2):243–6.
View Article
Google Scholar

[135] View Article

[136] Google Scholar

[ref48] 48. Cervarich M, Shu S, Jain AK, Arneth A, Canadell J, Friedlingstein P, et al. The terrestrial carbon budget of South and Southeast Asia. Environ Res Lett. 2016;11(10):105006.
View Article
Google Scholar

[138] View Article

[139] Google Scholar

[ref49] 49. Knorr W, Gobron N, Scholze M, Kaminski T, Schnur R, Pinty B. Impact of terrestrial biosphere carbon exchanges on the anomalous CO₂ increase in 2002–2003. Geophys Res Lett. 2007;34(9):L09703.
View Article
Google Scholar

[141] View Article

[142] Google Scholar

[ref50] 50. Gurney KR, Law RM, Denning AS, Rayner PJ, Baker D, Bousquet P, et al. Towards robust regional estimates of CO₂ sources and sinks using atmospheric transport models. Nature. 2002;415:626–30. pmid:11832942
View Article
PubMed/NCBI
Google Scholar

[144] View Article

[145] PubMed/NCBI

[146] Google Scholar

[ref51] 51. Kim H, Kim HM, Kim J, Cho CH. Effect of data assimilation parameters on the optimized surface CO2 flux in Asia. Asia-Pacific Journal of Atmospheric Sciences. 2018a;54:1–17.
View Article
Google Scholar

[148] View Article

[149] Google Scholar

[ref52] 52. Peylin P, Law RM, Gurney KR, Chevallier F, Jacobson AR, Maki T, et al. Global atmospheric carbon budget: results from an ensemble of atmospheric CO₂ inversions. Biogeosciences. 2013;10(10):6699–720.
View Article
Google Scholar

[151] View Article

[152] Google Scholar

[ref53] 53. Gaubert B, Stephens BB, Basu S, Chevallier F, Deng F, Kort EA, et al. Global atmospheric CO₂ inverse models converging on neutral tropical land exchange, but disagreeing on fossil fuel and atmospheric growth rate. Biogeosciences. 2019;16(1):117–34. pmid:31708981
View Article
PubMed/NCBI
Google Scholar

[154] View Article

[155] PubMed/NCBI

[156] Google Scholar

[ref54] 54. Park J, Kim HM. Design and evaluation of CO2 observation network to optimize surface CO2 fluxes in Asia using observation system simulation experiments. Atmos Chem Phys. 2020;20(8):5175–95.
View Article
Google Scholar

[158] View Article

[159] Google Scholar

Figures

Abstract

1. Introduction

2. Methodology

2.1 Inverse modeling system

2.2 CO2 observation

2.3 Experimental framework

3. Results

3.1. Characteristics of surface CO2 flux in Asia

3.1.1 9-year average surface CO2 flux distribution.

3.1.2 Annual and average surface CO2 flux.

3.1.3 Monthly and weekly aggregated surface CO2 flux.

3.2 Verification with independent observations

3.2.1 Verification with surface observations.

3.2.2 Verification with CONTRAIL aircraft observations.

3.3 Uncertainty reduction

3.4 Influence matrix

4. Summary and conclusion

Acknowledgments

References

2.2 CO₂ observation

3.1. Characteristics of surface CO₂ flux in Asia

3.1.1 9-year average surface CO₂ flux distribution.

3.1.2 Annual and average surface CO₂ flux.

3.1.3 Monthly and weekly aggregated surface CO₂ flux.