|This manuscript is a comprehensive and important update to the annual carbon budget, a dataset which has become one of the most important and widely used outputs in the field of climate science. The authors are to be congratulated on this sustained effort over time.|
Given the breadth of the paper, I have restricted my comments to those areas where I have some reasonable level of expertise. My comments are fairly minor, but could improve the clarity of the manuscript in some places.
1) As a note on the use of 1870 as a reference year, the new IPCC 1.5°C Special Report seems to be defining “pre-industrial temperature” as the 1850-1900 average (rather than 1860-1880 as some have previously done) — this suggests that there may be utility in updating the reference year at some point to 1875 to represent the mid-point of the pre-industrial temperature range.
2) The difference between E_LU for DGVMs vs Book-keeping models in the recent decade is obviously striking (and well discussed), though I didn’t come away with any explanation for the mechanism behind this difference. Is there some change in LU patterns (used to drive DGVMs) that explains the increase from the previous decades? Or is this a reflection of some response to climate change? Some speculation could help here (even if that is all it is at this point).
3) Related to this:
- On page 18 (lines 19-20), note that DGVMs and BK models agree prior to the recent decade (but not for the most recent year / decade)
- On page 18, lines 11-12: Is this loss of additional sink capacity the same as the finding that some have shown that simulated LU emissions in models tend to increase as a function of increasing CO2, since CO2 fertilization acts equally in “agricultural” areas, and maintaining constant cropland/pasture areas in a model results in higher emissions when CO2 is higher compared to when it is lower. I think it is kind of the same process at work but I am not completely sure…
4) Section 3.2: It is worth clarifying here that the values chosen to reflect E_LU and S_LAND in the budgets for the recent decade and for 2017 reflect different (and not necessarily consistent) methods. i.e. E_LU is from Book-keeping models (which do not include lost sink capacity), whereas S_LAND is from DGVM (which does include lost since capacity). It would be worth justifying this choice (notably the choice of using the BK estimate of emissions rather than DGVMs) and noting the inconsistency more explicitly to avoid confusion comparing numbers between Tables 5 and 6.
5) In the 2018 emissions projections for China, US, EU and India, why not add red dots to Figure 5 (similar to the global value)? Also, it is striking to me that US emissions are anticipated to increase by 2.2%, which is a very large departure from the recent decreasing trend. It would be worth noting this on page 39 (lines 3-4).
6) Another potentially missing process (that is not mentioned here at all I don’t think) is the effect of terrestrial weathering. Of course this is a very small carbon flux, but some recent model studies suggest the possibility of small increases over the historical period (due to warming, increased mid/high latitude runoff, vegetation expansion …), which might account for some portion of the cumulative carbon imbalance shown here.