Review of the manuscript: "Investigating the differences in calculating global mean surface CO2 abundance: the impact of analysis methodologies and site selection" by Zhendong Wu et al. (Atmospheric Chemistry and Physics, https://doi.org/10.5194/egusphere-2023-1173)

The paper proposes an analysis of the average concentration of CO2, and its growth rate, by comparing several observation networks, and time series filtering. A comparison is also proposed with CO2 concentrations simulated by an atmospheric transport model, after a phase of assimilation of surface observations. A comparison is also made between the average concentration obtained from the surface observation network and the total amount of CO2 in the atmosphere, as simulated by the model after assimilation. Estimating these values is of course important for monitoring atmospheric radiative forcing, both for scientists and policy makers. The study therefore deserves to be published after adding few points for discussion, as detailed below. Overall, the paper is clearly written, but sometimes lacks precision and quantitative values. I think that certain recommendations, such as not extrapolating measurement series, deserve to be more clearly stipulated.

I am concerned by the fact that, every year, slightly different global CO2 estimates emerge from several networks, as detailed in this study. Even if these values differ only slightly (as the results of this study show), I think it's still a not great to multiply these slightly different estimates. Wouldn't it be possible to make a recommendation to set up a global reference network to calculate unanimously accepted values? By the way, in addition to estimates based on surface networks, every year we now also see values from networks measuring total columns measured from the ground (TCCON) or from space. This aspect is not discussed at all, but it is conceivable that these measurements could provide a more relevant assessment of the atmospheric global average. Could the model not be used to estimate this contribution from the total column measurements?

In many cases, reference is made to the CO2 concentration at Mauna Loa as a proxy for global CO2. The advantage of relying on 1 or 2 stations (MLO and SPO, for example) is that it avoids the problems of changing the configuration of the global network, and enables a fast calculation. The disadvantage is that you are stuck if the reference station fails. Having said that, I would have been interested to see a comparison of average concentrations and growth rates considering only these 2 stations.

Figure 4b shows a maximum divergence of methods over the last few months of 2020, which is confirmed by Table 1, where the U(Gatm) uncertainty in 2020 is about 3 times greater than in previous years. This bias is important in view of the high demand for these estimates in near-real time. You mention the problem in the discussion as a result of a side-effect of the filtering procedures. Could you propose alternative to reduce this side effect ?

Data filtering: In addition to station selection, it was not clear for me if you apply a filter on the day/night periods. Mountain stations like Manua Loa are traditionally selected only during the night, while continental stations on the plains are generally selected during the day to increase the representativeness of the time series. In the discussion you mention the higher concentrations when adding continental sites, but clearly the offset will be quite strongly different if you include or not the night time CO2 accumulation in continental surface stations. Could you elaborate on this aspect ?

Few more specific comments:

Title ‘Global mean surface CO2’ : For temperature measurements the elevation is normalized (e.g 10 m above ground level). This is not the cas for CO2, for which we rather avoid measuring the concentration close to the surface. Consequently It would be more accurate to refer to the ‘global marine boundary layer CO2’

Lines 47-48: please provide a reference for the conversion GtC yr-1 to ppm.yr- 1

Line 75 : “ … hundreds of stations stations coordinated by WMO GAW: really ?”

Line 84: “…i.e. the full troposphere (up to ~8-15 km altitude) and the stratosphere or the regions of the world with substantial observational gaps”

Line 125 : “CTE compare well…”: could you more precise ?

Figure 4 & Line 249: Same trend with and without continental sites.  Figure 4A shows that CTEobs-NOAA differences change quite markedly before and after 2000, which is not the case for GAW-NOAA scenarios. In particular, strong winter differences emerge with the development of continental stations. it's a little disturbing that this fairly clear shift between the two networks isn't reflected in long-term trends. I imagine that the difference would be seen on the trend of annual concentrations, but not when looking at annual growth rates, as this is a transient change. Could you discuss this issue ?

Line 253: “red and blue lines”: actually, it is red and green on Figure 5

Line 256: earlier detection of Gatm change: can you quantify how much earlier ?

Line 306-307: ‘The NOAA network tracks atmospheric CO2 change better’: I would rather say that the result based on the NOAA netwotk comes closer to the CTE estimate.

Lines 405-407: The conclusion need to be rephrased for clarity.

Wu et al. (2023) provide a valuable analysis focusing on the impact of continental site inclusion when calculating global CO2 growth rates.  Employing the CTE model, the authors conduct synthetic tests to ascertain the accuracy of various growth rate estimate methods. The study is a valuable contribution to our understanding of the sampling error in the growth rate of atmospheric CO2 and is generally well-conceived. Nonetheless, the paper would benefit from clarifications and adjustments to enhance its readability and coherence.

Main Comments:

Presentation quality: The primary analysis of the paper focuses on the impact of including the continental sites for calculating the global CO2 growth rate. The study compares growth rate estimates from three sets of observations using, in essence, the NOAA's growth rate method:

1. NOAA: MBL sites only
2. WDCGG: MBL and some continental sites
3. CTE: MBL and a more extensive inclusion of continental sites

Given the many tests conducted and the slight variations between them, I recommend presenting this information in a table. Please specify in the table what is being compared to what is in each test to enhance the clarity of the methodology.

"The semi-NOAA method": The authors introduce a method called "semi-NOAA," adding unnecessary complexity to the presentation. The approach is not new, mainly the NOAA approach on an observation set including continental sites. Referring to all the filtering and fitting procedures as components of the original NOAA method would be more effective. Subsequently, the authors could delineate any variations they are implementing compared to the standard NOAA and WDCGG methods.

Minor Comments:

1. I suggest modifying the abstract to clearly state the study's purpose: to evaluate the impact of using continental sites in CO2 growth rate calculations. It drifts off by introducing the "semi-NOAA" method, which I do not think is the main point of this work.
2. It needs to be clarified how CTE is precisely used. CTE is sometimes a network, a growth rate, and a transport/inversion model run. Please use more clear terminology to differentiate. State this information in a table.
3. The study mainly addresses monthly and multi-decadal scales. I suggest adding an analysis on annual growth rates, which have been the scales that NOAA and WDCGG report the growth rates.
4. Throughout the manuscript, excessive use of parentheses interrupts the reading flow. Consider using tables to present some of the information the reader can refer to easily.
5. Many sentences are unnecessarily long and could be divided into shorter, more readable sentences.

Technical Corrections:

1. Line 82: The term "biased" seems unfair when referring to NOAA's estimate.
2. Lines 184-189: These lines could be made clearer to understand.
3. Line 328: Explain the acronym IVA.