|This paper is clearly intended to challenge the conclusions of the regional study performed by Reuter et al. (2014). They do so by arguing that the large European uptake is primarily due to measurements outside the European domain, which couldn't be assessed properly by a regional modelling framework (as that of Reuter et al.). They then "explain" the rest of the discrepancy by deriving simple bias scenarios that bring the fluxes from the GOSAT inversions into agreement with those from the surface-based inversions. They back away from suggesting that these biases are actually there, however, and only suggest that there are insufficient measurements to prove or disprove any of these results. Which tells us very little indeed. |
From a logical point of view I found the two arguments to be rather at odds with one another: if so much of the excessive uptake is due to measurements outside of Europe, wouldn't one suspect a bias in the measurements outside of Europe, rather than in the region with the densest coverage for TCCON measurements? The HIPPO analysis in Figure 2 shows generally good agreement with all simulations around the European latitudes, but large discrepancies in the tropics. A more interesting and evidence-backed question might be how performing a bias correction in the tropics (to match HIPPO, or based on the newer TCCON sites) might impact the European fluxes.
It is not news that small systematic errors in total column measurements can lead to large biases in retrieved fluxes. A number of studies have already tested this, including Houweling et al. (2010), which was notably missing as a reference. However unless there is some physical reason to suspect that the bias might actually have the structure in time and space that the on-line bias correction produces, it seems more like an assessment of residuals in atmospheric mixing ratios between the two runs rather than solving for a physically meaningful bias. I was also curious as to how realistic the pattern of bias corrections looked: is it very noisy, or rather continuous? Is it correlated with anything that one would expect to cause a retrieval error? Figure 5 shows that there are significantly larger bias corrections during the growing seasons over Eastern Europe than Western Europe, although the former is certainly less well-constrained by measurements. Without this sort of analysis of how plausible such an error is, it seems more like a numerical fact rather than a hypothesis: suppose z is a function of x and y; if we add so much to x and take away so much from y, we can make z approximately equal to w. This doesn't, however, tell us anything about the underlying processes. One could just as easily define the error in PBL height that is required to bring the surface-based fluxes into agreement with GOSAT-based fluxes: it doesn't mean that those errors are real.
Given the transport errors that clearly trouble all inversion systems (see the poor agreement of all inversions in Figure 2 with HIPPO over high austral latitudes as one example), on-line bias corrections seem questionable at best, especially when they are not constrained by any physical drivers. In contrast to this, the retrieval teams have put much time and effort into deriving bias corrections based on independent (TCCON) measurements. Unless you can demonstrate that the on-line bias corrections produces better agreement with independent measurements, it seems better left alone. I agree with other reviewers that it would have been better to leave TCCON out of the reference inversion, and use it for validation instead (although it is, strictly speaking, not entirely independent from the bias corrections of the retrieval teams). Ignoring this suggestion was a mistake, in my opinion.
The manuscript was improved from the first round of reviews, especially in terms of highlighting the fact that there are not sufficient measurements to prove or disprove either theory. Despite this, the authors have chosen to ignore the recommendations to change the title. Given the evidence presented in the current manuscript, publishing it under this title is misleading. Perhaps a more realistic: "Elevated uptake of CO2 over Europe inferred from GOSAT XCO2 retrievals: the jury's still out". Given the lack of convincing evidence to support the conjecture contained in this study, I am very hesitant to recommend this study for publication in ACP.
After the first round of review some of the conclusions were (rightly) scaled back, and some more nuance was added to the discussion. In the response to reviewers the authors suggest that they wanted to use this lack of evidence to prove or disprove either hypothesis as an impetus to improve our validation capacity. If the study were recast in this light, perhaps with suggestions about where these validation measurements could be implemented for maximum impact, it would be a more constructive contribution to the ongoing debate.
Other problems I came across while reading the manuscript:
L25: observation without s
L77: remove first comma, add "the" before global
L84 and 89: model -> modelling
Map of sub-regions would be nice to see, especially with bias pattern...
L167: should be Figure 3?
Why compare CONTRAIL under 3 km and HIPPO under 5 km? How sensitive is the analysis to these cut-offs?
Figure 3: I really only see convincing evidence for the better agreement of the in-situ/TCCON inversion for February for Amsterdam, and maybe the autumn for Moscow. It's hard to say how significant this is, as none of the measurements are presented with error bars, or even standard deviations. This should be fixed. The broken magenta line should be removed: mixing the fluxes from two inversions leads to a meaningless result. Mixing concentration values from different simulations in an inversion (as is done later) makes more sense, as at least the final result is consistent with the input information. This, on the other hand, is like changing the boundary conditions *after* performing a regional inversion, and then discussing the results. It also adds little to the discussion, in my opinion.
L205: Should this be Figure 3?
L228-229: Sentence should be rewritten.
L234-235: mover "errors" before the parentheses.
L243: a few -> are few
L244: have -> has
L245: This sentence is redundant, isn't it? Given L228-229?
L247: are -> is
L264 & L265: "the" before "JPL ACOS team", "University of Leicester"