Galactic Dynamics

Abstract. Total bacteria, fungal spore and yeast counts were compared with ultraviolet-light-induced fluorescence (UV-LIF) measurements of ambient aerosol at the summit of the Puy de Dome (PdD) mountain in central France (1465 m a.s.l), which represents a background elevated site. Bacteria, fungal spores and yeast were enumerated by epifluorescence microscopy (EFM) and found to number 2.2 to 23 L −1 and 0.8 to 2 L −1 , respectively. Bacteria counts on two successive nights were an order of magnitude larger than in the intervening day. A wide issue bioaerosol spectrometer, version 3 (WIBS-3) was used to perform UV-LIF measurements on ambient aerosol sized 0.8 to 20 μm. Mean total number concentration was 270 L −1 (σ = 66 L −1 ), found predominantly in a size mode at 2 μm for most of the campaign. Total concentration (fluorescent + non-fluorescent aerosol) peaked at 500 L −1 with a size mode at 1 μm because of a change in air mass origin lasting around 48 h. The WIBS-3 features two excitation and fluorescence detection wavelengths corresponding to different biological molecules, although non-biological interferents also contribute. The mean fluorescent particle concentration after short-wave (280 nm; associated with tryptophan) excitation was 12 L −1 (σ = 6 L −1 ), and did not vary much throughout the campaign. In contrast, the mean concentration of particles fluorescent after long-wave (370 nm; associated with NADH) excitation was 95 L −1 (σ = 25 L −1 ), and a nightly rise and subsequent fall of up to 100 L −1 formed a strong diurnal cycle in the latter. The two fluorescent populations exhibited size modes at 3 μm and 2 to 3 μm, respectively. A hierarchical agglomerative cluster analysis algorithm was applied to the data and used to extract different particle factors. A cluster concentration time series representative of bacteria was identified. This was found to exhibit a diurnal cycle with a maximum peak appearing during the day. Analysis of organic mass spectra recorded using an aerosol mass spectrometer (AMS; Aerodyne Inc.) suggests that aerosol reaching the site at night was more aged than that during the day, indicative of sampling the residual layer at night. Supplementary meteorological data and previous work also show that PdD lies in the residual layer/free troposphere at night, and this is thought to cause the observed diurnal cycles in organic-type and fluorescent aerosol particles. Based on the observed disparity between bacteria and fluorescent particle concentrations, fluorescent non-PBA is likely to be important in the WIBS-3 data and the surprisingly high fluorescent concentration in the residual layer/free troposphere raises questions about a ubiquitous background in continental air during the summer.


Introduction
An improved understanding of the carbon sources and sinks at a regional scale globally is essential to predict the future rate of atmospheric CO 2 increases and to plan an international CO 2 management strategy (Ciais et al., 2010).But these fluxes remain quantitatively uncertain.The full range of results in past studies spans budgets with northern terrestrial uptake of 0.5 to 4 PgC yr −1 , and tropical terrestrial emissions of −1 to 4 PgC yr −1 (Stephens et al., 2007;Peylin et al., 2002;Gurney, 2004).Some studies show increasing sinks in tropical forest plots (Baker et al., 2004).Rising temperatures may already decrease the efficiency of terrestrial carbon uptake in the Northern Hemisphere (Piao et al., 2008), while larger net sinks were found over northern and southern continents than the results of the TransCom-3 study for 1992-1996 (Feng et al., 2011).

Z. H. Chen et al.: Improved simulation of regional CO 2 surface concentrations
Where and when atmospheric CO 2 is absorbed by land ecosystems and oceans is a major issue for the global carbon cycle.Optimized estimates of surface sources and sinks have been produced in different ways.One is a top-down way.For example, CO 2 in the atmosphere is affected by surface fluxes.Information about regional carbon sources and sinks can be derived from variations in observed atmospheric CO 2 concentrations via inverse modeling with atmospheric tracer transport models (Gurney et al., 2002).Another is a bottom-up way.For example, the land-atmosphere fluxes can be simulated by different dynamic global vegetation models (DGVMs) (Sitch et al., 2008).Terrestrial carbon cycle model VEgetation-Global-Atmosphere-Soil (VEGAS) is one of the DGVMs that was developed to simulate the net landatmosphere fluxes and has been described by Zeng (2003).The land-atmosphere flux simulated by VEGAS agrees well with the CO 2 growth rate observed at Mauna Loa both in terms of interannual amplitude and phase (Zeng et al., 2005).
The GEOS-Chem atmospheric transport model has been widely used in the assimilation of CO 2 and inverse of CO 2 flux.It has been used to evaluate the influence of reduced carbon emissions on the distribution of atmospheric CO 2 and described in early studies (Suntharalingam, 2004(Suntharalingam, , 2005)).The land-atmosphere fluxes in GEOS-Chem include monthly biomass burning CO 2 emissions, annual inventory of biofuel burning 3-hourly net ecosystem productivity (NEP) for 2000 (Olsen, 2004), and annual climatology based on TransCom CO 2 inversion results in Nassar et al. (2010).The differences between CO 2 model simulation using surface fluxes and observations have been used to improve our understanding of the global surface fluxes.There were different methods to compare CO 2 model results and observations in earlier studies.The mean annual meridianal/longitudinal gradient observation is compared with model values (Bousquet et al., 1999;Kaminski et al., 1998).Latitudinal distribution of the sources and sinks of CO 2 from the concentration gradient has been discussed (Tans et al., 1989(Tans et al., , 1990)).The air samples in flasks were grouped into latitude bands to aid determination of the sources and sinks (Tans et al., 1989).Previous studies have adjusted the CO 2 surface flux via minimizing the distance between the modeled/optimized values and the observational data at each station (Enting, 2002;Peylin et al., 2002;Bousquet, 2000;Baker et al., 2006;Gurney et al., 2002;Rödenbeck et al., 2006).
However, one persistent problem in using modelobservation comparisons for this goal relates to the issue of compatibility.Observations at a single station reflect all underlying processes of all scales.These processes cannot be fully resolved by model simulations at the grid points nearest to the station due to the lack of spatial or temporal resolution or missing processes in the model.In this article we propose a new technique to evaluate the regional surface fluxes by comparing the regional CO 2 concentration from model simulations with observations, rather than the difference at every single observational station.Several stations in one region were grouped according to the regional temporal characteristics of the seasonal cycle, which have been derived from a new atmospheric CO 2 observation data set from GLOBALVIEW-CO 2 2010.The averaged concentration of CO 2 at all stations in one region represents the regional CO 2 concentration in this region.
To validate the usefulness of the new evaluation method about regionally averaged CO 2 concentrations, we compared two simulations using two different emission inventories with observations.One emission inventory is the original surface fluxes in GEOS-Chem, including the NEP from Carnegie-Ames-Stanford Approach (CASA).Another new emission inventory, including the land-atmosphere fluxes from VEGAS, was selected to reproduce CO 2 concentrations in this study.The land-atmosphere fluxes from VEGAS were used in the GEOS-Chem model, replacing all the current inventories except anthropogenic emissions and ocean fluxes.
The outline of this paper is as follows: Sect. 2 introduces the data.Section 3 describes the grouping of observation stations in one region and demonstrates the temporal and spatial variability in CO 2 .Section 4 presents the differences between the modeled regional CO 2 concentrations with fluxes from CASA and the modeled results with fluxes from VE-GAS.We present conclusions in Sect. 5.

GLOBALVIEW CO 2 data
GLOBALVIEW-CO 2 (GLOBALVIEW-CO2, 2010) is a product of the Cooperative Atmospheric Data Integration Project.The project is coordinated and maintained by the Carbon Cycle Greenhouse Gases Group of the National Oceanic and Atmospheric Administration, Earth System Research Laboratory (NOAA ESRL).Gaps in the data are filled by extrapolation from marine boundary layer measurements.Flask samples of whole air enable highly accurate and precise measurements of atmospheric CO 2 concentrations (Conway et al., 1994) This data product includes more than 300 extended records derived from observations made by 22 laboratories from 15 countries in the period 1 January 1979 to 1 January 2010.Data in the files with a sea qualifier that contain a statistical summary of the average seasonal pattern by month were used to analyze the seasonal cycle of the observation stations.Data in the files with an ext qualifier that contain synchronized smoothed values were compared with model results.Where there are several measurements at different altitudes for the same station we only use the lowest in altitude.This gives a total of 108 measurements that were used.

Modeling the land carbon fluxes
The net ecosystem exchange (NEE) is simulated by DGVMs and equals the heterotrophic respiration (RH) subtracted from the net primary productivity (NPP).Simulated landatmosphere fluxes are between −1.52 PgC yr −1 (Lund-Postdam-Jena (LPJ) model) and −2.75 PgC yr −1 (Sheffield-DGVM (SHE) model) for the 1990s.The DGVMs simulate a greater land carbon uptake, which is in agreement with IPCC estimates (Sitch et al., 2008).The land fluxes are defined as the sum of photosynthesis, ecosystem respiration and biomass burning.The terrestrial carbon model VEGAS is described in Zeng (2003).It was run at 2.5 • × 2.5 • resolution and forced by precipitation and temperature, the seasonal climatologies of radiation, humidity, and wind speed.The driving data of precipitation for VEGAS come from a combination of the Climate Research Unit (CRU; New et al., 1999;Mitchell and Jones, 2005) data set for the period of 1901-1979and the Xie and Arkin (1996) Hansen et al. (1999), adjusted by CRU climatology of .A fire module includes the effects of moisture availability, fuel loading, and plant functional type dependent resistance.Unique features of VEGAS include a vegetation height dependent maximum canopy, which introduces a decadal timescale that can be important for feedback into climate variability and a decreasing temperature dependence of respiration from fast to slow soil pools.Specially, two lower soil pools have weaker temperature dependence of decomposition due to physical protection underground in VEGAS (Q10 value of 2.2 for the fast pool, 1.35 for the intermediate pool, and 1.1 for the slow pool.The monthly land-atmosphere fluxes simulated by VEGAS are regridded offline to the GEOS grids (2 • × 2.5 • ) in this study, which is equal to the magnitude of NEE.The seasonal cycle of landatmosphere fluxes from VEGAS is shown in Fig. 1.A positive flux indicates a flux of CO 2 from the land to atmosphere and negative is uptake by the land.
Monthly mean NEP fluxes for 2000 from CASA are constructed from gross primary production (GPP) and ecosystem respiration (Re) (Olsen, 2004).Inputs to CASA included a 1990 monthly normalized difference vegetation index (NDVI) product derived from the NOAA/NASA Pathfinder data set, surface solar insolation (Bishop and Rossow, 1991), mean temperature and precipitation from the period 1950 to 1980 (Shea, 1986), soil texture (Zobler, 1986), and a land cover classification based on NDVI (DeFries and Townshend, 1994).The response of heterotrophic respiration to surface air temperature is described by using a Q10 function of 1.5 (Raich and Potter, 1995).The net global contribution from CASA is set to 0 PgC yr −1 in order to represent terrestrial fluxes with no anthropogenic interference.The seasonal cycle of NEP from CASA is shown in Fig. 1.
Anthropogenic interferences such as biomass burning were specified as 2.96 PgC yr −1 in GEOS-Chem.To account for the total annual sum of biospheric uptake and emission of CO 2 , the residual annual terrestrial exchange of inverse results from TransCom, a global total of −5.29 PgC yr −1 , was included in the land-atmosphere fluxes (Nassar et al., 2010).The seasonal cycle of total land-atmosphere fluxes used in GEOS-Chem is shown in   Feng et al. (2011) and Nassar et al. (2011).
The land-atmosphere flux from VEGAS in January is 270 TgC less than that from CASA.These differences are distributed over tropical land regions as shown in Fig. 2. The fluxes from VEGAS are smaller than the original landatmosphere flux in GEOS-Chem, especially from June to August (about 460 TgC, 770 TgC, and 180 TgC, respectively).The differences between the flux from VEGAS and that from CASA in July are distributed over the regions of Asia, temperate North America, and tropical South America (Fig. 3), which reaches about 500 TgC in total.

Determining groups of observational stations
We grouped several observation stations in one region based on the seasonal cycle at each station in our study.The stations in one region were grouped based on the amplitude and phase of the seasonal cycles at each station.The average of CO 2 at all stations in one region represents the regional CO 2 concentrations.The amplitude and phase of the seasonal cycle at each station in one group are similar, while the average amplitude and phase of the seasonal cycle for each group are different from that of other groups.There are 36 stations on the land and 72 stations on the ocean (see Table A1).These stations were classified into 26 groups.A map of all grouped stations is shown in Fig. 4.  All stations on the land show similar seasonal cycles.The concentration of CO 2 decreases during summer and autumn and increases during spring and winter.The difference between minimum and maximum values is greater than 6 ppm for most stations on the land.We cannot split the land based on the seasonal cycles at stations on the land because the phase of seasonal cycles at all stations on the land is similar; for example, CO 2 at all stations on the land decreases in autumn and increases in spring.The land was divided into 11 regions based on the TransCom land regions (shown in Fig. 4).The land region north of 40 • N in North America is called boreal North America (L1), and the region south of 40 • N in North America is called temperate North America (L2).The region north of 5 • S in South America is called South America tropical (L3), and the region south of 5 • S in South America is called South America temperate (L4).The other seven land regions are northern Africa (L5), southern Africa (L6), Eurasian boreal (L7), Eurasian temperate (L8), tropical Asia (L9), Australia (L10), and Europe (L11).The stations in each land region were grouped.The magnitude of the amplitude of the seasonal cycles at different stations in one land region may be different.To represent the regional CO 2 concentration for the land regions, the averages of seasonal cycles at more than two stations with similar amplitudes were required in one land region.There are more than 2 stations with similar amplitudes of seasonal cycles in only 5 land regions (North America, temperate North America, Eurasian boreal, Eurasian temperate, and Europe).Therefore, the regional CO 2 concentrations of these 5 land regions were used to evaluate the observation-model differences of CO 2 .
The amplitude and phase of seasonal cycles at stations on the ocean are different.For example, CO 2 decreases in April for one region while in August for another region.The stations on the ocean were grouped based on the amplitude and phase of seasonal cycle.The stations on the ocean were grouped into 15 groups, and the ocean was divided into 15 regions in this study.The 11 ocean basis regions were chosen to approximate circulation features such as gyres and upwelling regions in the TransCom study (Gurney et al., 2002).

Seasonal cycles of stations on the land
The seasonal cycles at all stations in 5 land regions are shown in Fig. 5.The annual mean has been removed.The average minimal value for each region is smaller than −7 ppm (−11.5 ppm for North American boreal (L1), −7.1 ppm for North American temperate (L2), −10 ppm for Eurasian boreal (L7), −8.7 ppm for Eurasian temperate (L8), −8.1 ppm for Europe (L11)).Seasonal cycles of atmospheric CO 2 are caused primarily by the terrestrial biosphere moving from being a net source of carbon to the atmosphere (mainly in winter) to becoming a net sink (mainly in summer), where net carbon uptake or release is determined by the balance between photosynthesis and respiration, which vary in response to temperature and precipitation anomalies.Studies have shown the seasonal cycle of atmospheric CO 2 in the Northern Hemisphere (NH) is in phase with the ecosystems (e.g., Randerson et al., 1997).The geographic domain from which surface fluxes influence the measured seasonal variation in gas concentration can be assessed through a footprint analysis (Randerson et al., 1997).The fluxes in this domain could be adjusted according to the differences between the modeled regional CO 2 concentrations and observations.The difference in seasonal amplitude of all groups in the NH can be an important constraint for further improving our understanding of the surface fluxes in the NH.

Seasonal cycles of stations on the ocean
The ocean was divided into 15 regions based on the seasonal cycles of CO 2 , including Pacific Ocean region (O1-O7), Atlantic regions (O8-O11), Indian regions (O12-O13), northern ocean (O14), and Southern Ocean (O15) in this study.
The stations within the Pacific Ocean north of 5 • S were classified into 5 different groups (O1, O2, O3, O6 and O7), and the stations within the Atlantic Ocean were classified into 2 groups (O8 and O9).Though the seasonal cycles of the ocean regions north of 5 • S were similar to that of the land groups in Northern Hemisphere, there are different where there are more than 2 stations (L1: boreal North America, L2: temperate North America, L7: Eurasian boreal, L8: Eurasian temperate, L11: Europe.5 regions are shown in Fig. 4. Broken line denotes the seasonal values for all stations in one region; solid line denotes the grouped average value of each region).The stations within the region North American boreal (L1) are labeled with "L1" in the "group" columns of Table A1.The way to find the stations in other regions is similar.amplitudes (as shown in Fig. 6).The amplitudes of groups O1 and O6 are larger than 10 ppm, and the amplitudes of O2 and O7 are much less than that of other northern regions, while the amplitude of group O3 is much less than 6 ppm.The amplitude of group O9 is less than that of group O8.The amplitude is less in the Southern Hemisphere (SH), since the Southern Hemisphere has less mid-latitude vegetation to absorb and release CO 2 seasonally (Randerson et al., 1997).
The South Pacific region between 5 • S and 35 • S was divided into two subregions (O4 and O5) according to the different seasonal cycles of CO 2 measured at stations in these regions.Though the amplitude is smaller than 1.4 ppm, the CO 2 seasonal cycle of the groups is clear in these regions.Generally there is an increase period and a decrease period for one seasonal cycle.While CO 2 increases from April to June and from October to December for the South Pacific tropics (O4), CO 2 decreases from January to April and from August to October for the South Pacific temperate (O5).
www.atmos-chem-phys.net/13/7607/2013/Atmos.Chem.Phys., 13, 7607-7618, 2013 Fig. 6.Seasonal cycles of observational stations of 15 ocean groups in ocean regions (15 regions are shown in Fig. 4. Broken line denotes the seasonal values for all stations in one region; solid line denotes the regional averaged value of each region).The stations within the region northeast Pacific (O1) are labeled with "O1" in the "group" columns of Table A1.The way to find the stations in other regions is similar.
The seasonal cycles are more complicated in the Indian Ocean north of 35 • S. They were classified into two groups with different seasonal cycles (O12 and O13).The average seasonal cycles of these two regions are different from other ocean regions.The North Indian Ocean (O12) shows a consistent decrease from February to November.The concentrations of stations within the South Indian Ocean (O13) range from −1.5 ppm to 1.5 ppm during the first half year and show an increase (about 1 ppm) in the second half year.The South Atlantic was divided into 2 regions (O10 and O11) with different amplitudes.The minimum and maximum values are −0.9 ppm and 0.7 ppm for the Atlantic tropics (O10), while they are −0.3 ppm and 0.3 ppm for the South Atlantic temperate (O11).
The concentrations of CO 2 at stations within the ocean south of 5 • S are mainly influenced by the oceanic sources and sinks, and the amplitudes of seasonal cycles are not more than 2 ppm (O4, O5, O10, O11 and O15), which is much smaller than that of the NH.It is evident that the seasonal anomalies of CO 2 are positive in NH winter (January) and negative in NH summer (August).Inversely, the seasonal variations are positive in the southern hemispheric winter (August) and negative in the southern hemispheric summer (January) south of 35 • S (Fig. 6).Seasonal signals observed in all subtropical regions of the NH and SH show that the CO 2 concentration decreases southward in summer and vice versa in winter (Metzl et al., 2006).An increase of the seasonal cycle for Southern Ocean occurs in September, while the seasonal anomalies of CO 2 in the Northern Hemisphere are negative at the same time.The two seasonal cycles of the Southern Ocean (O15) and the Northern Hemisphere are out of phase.Northern Hemisphere terrestrial ecosystems contribute substantially to the seasonal cycle at many stations in the Southern Hemisphere.Because of lags in transport and nonoverlapping growing seasons, some components from the northern and southern hemispheres are out of phase with one another.Thus, an increase in seasonal cycle of NEP fluxes from terrestrial uptake in the Northern Hemisphere could drive a decrease in the amplitude of the seasonal cycle of atmospheric CO 2 at stations in the Southern Hemisphere (Randerson et al., 1997).

Simulation results and comparison with observations
We use the GEOS-Chem model (Suntharalingam, 2004(Suntharalingam, , 2005) ) to describe the relationship between 3-D atmospheric CO 2 concentrations and surface CO 2 fluxes.A detailed description of the original emission inventory is given in Nassar et al. (2010).Our model simulation was initialized with a uniform global distribution of 375 ppm on 1 January 2004 and integrated forward to 1 January 2006 using the original emission inventory.The modeled CO 2 distribution on 1 January 2006 was the initial concentration for two simulations with the original emission inventory (ori) and the new emission inventory (new) from 1 January 2006 to 1 January 2007.Both model simulations were run at a horizontal resolution of 2 • latitude × 2.5 • longitude.Figures 7 and 9 show differences between the model results with the original inventory and the results with the new inventory during 2006.

Comparisons of regionally averaged CO 2 concentrations for land regions
The CO 2 seasonal cycles were simulated by the model with original and new emission inventories.The largest difference between the model results and observations for runs with the original emission inventory is 17.5 ppm, about 4.5 % of observation values.The difference for the simulation with the new emission inventory is below 8.4 ppm, about 2.2 %.The largest differences for both simulations appear in region L11, which indicates there may be large uncertainties for these two CO 2 surface fluxes in Europe.
The difference of the regional CO 2 concentration between model results with the new emission inventory and observations is less than 2 ppm for North American boreal (L1) from January to June, which is smaller than 1 % of observations.The difference is about 2 ppm during July, August, and December, which suggests that there are large uncertainties in North America for the new inventories during these periods.The difference between simulations with the original emission inventory and observations reaches 6 ppm from April to May.
The largest difference between the simulation with fluxes from VEGAS and observations is 2.8 ppm for the North American boreal (L1), 2.9 ppm for North American temperate (L2), 3.1 ppm for Eurasian boreal (L7), 3.5 ppm for Eurasian temperate (L8), and 4.3 ppm for Europe (L11), which is smaller than that of CASA (5.8 ppm,6.3 ppm,14.5 ppm,10.9 ppm,13.1 ppm,respectively).The spread of the regional CO 2 of observations for each region is shown in Fig. 7, which is determined by the concentrations of stations in the region.
The root-mean-square difference (RMSD) of regionally averaged value between model results with fluxes from VE-GAS and observation is reduced by 0.24-0.63ppm for 5 land regions.The RMSD between two simulations and observations for each station ranges from 0 to 2 ppm.As shown in Fig. 8, the largest RMSE between the simulations with fluxes from VEGAS and observations for regional CO 2 concentrations is 0.2 ppm, which is smaller than the value with fluxes from CASA (0.4 ppm).
The new emission inventory can be used as good prior fluxes in the forward model and be adjusted in future inverse models from the above comparisons of 5 land regions.

Comparisons of regionally averaged CO 2 concentrations for ocean regions
The seasonal cycles of CO 2 concentration at stations on the ocean are also influenced by the change of emission inventories on land.The difference between the simulations with the new inventories and observations ranges from 0.02 ppm to 2 ppm (0.7 ppm to 4 ppm for the old inventories) for the South Pacific temperate (O5) during 2006.It can be deduced that the regional CO 2 concentration of the ocean regions could be improved through the improvement of the land fluxes.The largest difference (about 8 ppm) for runs with new inventories appears in April 2006 for the Indian tropical region (O12).It is a high value (about 387 ppm) for observations in April 2006, while the simulated result with the new emission inventory is 379 ppm (Fig. 9).Fluxes that contribute to the concentration of this region should be improved for this new emission inventory.
There are differences between both model results and observations from January to April for the South Indian temperate (O13).The peak-to-trough amplitude of the regional CO 2 concentration for this region is no more than 2 ppm in 2006, while the spread of the observed concentrations in this  4).The error bar represents the spread of the observations.region is larger than 2 ppm for all months in 2006.Unfortunately, there are too few observations in the adjacent land regions.Some more observations are very necessary for these regions in the future.
There is still a large positive bias (about 5 ppm) for North Atlantic temperate (O8, O9) from July to September.It is necessary to improve the fluxes in this region or the surrounded land regions.For the South Pacific tropics (O4) and South Pacific temperate (O5), it is difficult to simulate the two increase phases and two decrease phases in the seasonal cycle of observations (Fig. 6).It could be effective for improving the fluxes in the ocean regions to match observations because the seasonal cycle simulated by the land fluxes is characterized by one increase and one decrease phase.
The concentrations of CO 2 at stations on the ocean are influenced by the change of emission inventories on land.As shown in Fig. 10, the RMSD of regionally averaged value between model results using fluxes from VEGAS and observation is less than the results using fluxes from CASA by 0.15-0.53ppm for northeast Pacific, South Pacific and Southern Ocean (O1, O4, O5, O12, O13 and O15).There is phases in the seasonal cycle of observations (Fig. 6).It could be effective for improving the fluxes in the ocean regions to match observations because the seasonal cycle simulated by the land fluxes are characterized by one increase and one decrease phase.little improvement for North Pacific and northern ocean (O2, O6, and O14).It is convenient to evaluate the regional model results according to the comparisons of regionally averaged values.

Conclusions
We grouped several observation stations in one region according to the phase and amplitude of seasonal cycles of measured CO 2 .The regionally averaged values contain less small-scale "noise" that models often cannot resolve and are used to evaluate the regional model results.The differences of regionally averaged values between observations and model results reflect the uncertainties of the flux in the region where the grouped stations are.
We compared regionally averaged values between model results with two land-atmosphere flux from CASA (VE-GAS) and observations.Results show that the differences between the modeled regionally averaged values of CO 2 concentrations with fluxes from VEGAS and observations have improvements in most regions.There is still large uncertainty in the Atlantic and North Atlantic, Indian Ocean, and South Pacific tropics.
The regional CO 2 surface fluxes can be estimated by different methods.It is very useful for evaluating regional surface fluxes by comparing the CO 2 regionally averaged values from modeled results with observations.The differences of regionally averaged values between observations and model results can be used to estimate the uncertainty of regional fluxes and to optimize the regional fluxes with inverse methods in future work.

Fig. 1 .
The original CO 2 fluxes used in this study include 7.8 PgC (anthropogenic emissions), −1.4 PgC (net oceanatmosphere fluxes), and −2.3 PgC (net land-atmosphere fluxes) for 2006.The original global annual net CO 2 flux for 2006 is 4.1 PgC.The new CO 2 fluxes used in this study include 7.8 PgC (anthropogenic emissions), −1.4 PgC (net ocean-atmosphere fluxes), and −1.9 PgC (net land-atmosphere fluxes) for 2006.The new global annual net CO 2 flux for 2006 is 4.5 PgC.There are also little differences between the total fluxes from other inversion results.JENA S99V3.

Fig. 1 .Fig. 1 .
Fig.1.Monthly terrestrial fluxes from dynamic global vegetation models (CASA and VEGAS) and original land-atmosphere fluxes (ORI, including fluxes from CASA, biofuel burning, biomass burning and residual annual biospheric flux ) in GEOS-Chem in 2006.

Fig. 2 .Fig. 2 .
Fig.2.Spatial distribution of difference between terrestrial exchange from CASA and fluxes from VEGAS in January 2006 (positive values denote the fluxes from VEGAS are greater than the fluxes from CASA)

Fig. 2 .Fig. 3 .
Fig.2.Spatial distribution of difference between terrestrial exchange from CASA and fluxes from VEGAS in January 2006 (positive values denote the fluxes from VEGAS are greater than the fluxes from CASA)

Fig. 5 Fig. 5 .
Fig.5 Seasonal cycles of observational stations in 5 land regions where there are more than 2 stations (L1: Boreal North America, L2: Temperate North America, L7: Eurasian boreal, L8: Eurasian Temperate, L11: Europe.5 regions are shown in Fig.4, broken line denotes the seasonal values for all stations in one region, solid line 16 Fig. 5. Seasonal cycles of observational stations in 5 land regionswhere there are more than 2 stations (L1: boreal North America, L2: temperate North America, L7: Eurasian boreal, L8: Eurasian temperate, L11: Europe.5 regions are shown in Fig.4.Broken line denotes the seasonal values for all stations in one region; solid line denotes the grouped average value of each region).The stations within the region North American boreal (L1) are labeled with "L1" in the "group" columns of TableA1.The way to find the stations in other regions is similar.

Fig. 7 .
Fig. 7. Comparisons of regionally averaged values of CO 2 between model results from GEOS-Chem with the original emission inventory (dotted line) and the new emission inventory (dashed line) and GLOBALVIEW-CO 2 (solid line) for 5 land regions (L1: boreal North America, L2: temperate North America, L7: Eurasian boreal, L8: Eurasian temperate, L11: Europe) in 2006 (5 regions are shown in Fig. 4).The error bar represents the spread of the observations.

Fig. 8 .
Fig. 8 Comparisons of RMSD at each station and regionally averaged values between model results and observations for 5 Land regions (L1: Boreal North America, L2: Temperate North America, L7: Eurasian boreal, L8: Eurasian Temperate, L11: Europe) in 2006.Each region is shown in Fig. 4. Triangle (Asterisk) denotes RMSD of regionally averaged values

Fig. 9 .
Fig. 9. Comparisons of regionally averaged values of CO 2 between model results from GEOS-Chem with the original emission inventory (dotted line) and the new emission inventory (dashed line) and GLOBALVIEW-CO 2 (solid line) for 15 ocean regions in 2006 (15 ocean regions are shown in Fig. 4).The error bar represents the spread of the observations.

Fig. 10 Fig. 10 .
Fig. 10 Comparisons of RMSD at each station and regionally averaged values between model results and observations for the ocean regions in 2006.Each region is shown in Fig. 4. Triangle (Asterisk) denotes RMSD of regionally averaged values between model results using fluxes from CASA (VEGAS) and observations, cross (diamond) denotes RMSD of each station between model results using fluxes from CASA (VEGAS) and observations(See the legend in Fig.8).