{"id":"47ac4477-4c8e-42e2-bcab-566ccce57aeb","arxiv_id":"2505.11512","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"French forests' CO2 sink declined by nearly half from its 2008 peak by 2022, with mortality and harvest increases concentrated in northern regions.","lead":"A new analysis of French forest inventory data shows the national carbon sink peaked in 2008 and had fallen by about half by 2022. It matters because the official sink is used in France's climate neutrality budget, and the decline suggests the land sector will miss its targets.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's own Fig. 1b shows a direct stock-change sink about half the CITEPA/UNFCCC value used for the headline, so the 37.8 MtCO2/yr figure and its mortality/harvest attribution are not yet established.","rationale":"The reader identified the temporal continuity of NFI protocols as the weakest assumption, which the authors themselves flag. My stress-test finds a sharper, more internal problem: the paper's own Fig. 1b contradicts the headline CITEPA series by a factor of about two in recent years, and the authors state that the UNFCCC report likely overestimates the real sink. This is not merely a continuity concern; it is an unresolved quantitative inconsistency inside the evidence base used for the central claim. The protocol-change issue reinforces it, because the decline begins with the new semi-permanent design and the mortality signal appears with the 2015 protocol extension. Both concerns point to the same conclusion: the exact sink value and the attribution of the decline to mortality versus harvest are conditional. The paper is otherwise useful as an outlook and its recommendations on monitoring are reasonable, so rejection is not warranted. The reader's CONDITIONAL verdict already captures this, so no verdict change is needed.","tokens_in":36265,"tokens_out":4384,"duration_ms":49673,"concrete_test":"Compute 5-year block averages of the direct stock-change series in Fig. 1b for 2013-2017 and 2017-2022, using only NFI plots with both visits and identical biomass conversion factors, and compare these to CITEPA's net sink for the same blocks. If the factor-of-two gap survives block averaging, the official sink is biased and the mortality/harvest decomposition must be re-derived from stock differences; if the gap collapses, it was a sampling or timing artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on the CITEPA/UNFCCC flux time series. The paper itself reports that a direct year-on-year stock-change estimate from the same NFI plots gives a sink about two times smaller than the UNFCCC report in recent years, and concludes that 'the current report to UNFCCC very likely overestimates the real carbon sink in the French forests.' If that is true, the headline value of 37.8 MtCO2 yr-1 in 2022 is not the best estimate, and the decomposition of the 2013-2017 decline into 51.5% mortality and 48.5% harvest, computed within the same CITEPA flux framework, is similarly suspect. The authors attribute the gap to latest campaigns not yet included in CITEPA, but this is an assertion, not a reconciliation. Separately, the authors concede that the 2013-2017 decrease coincides with the 2010 semi-permanent plot protocol and the 2015 mortality-tracking extension, and 'it cannot be ruled out that the sink decrease seen in Fig. 1a could be a transient methodological artifact.' Both issues bear directly on the magnitude and mechanism of the claimed decline. The direction of the decline may survive, but the quantitative headline and the attribution do not yet have a secure basis.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript analyzes the carbon balance of Metropolitan French forests from 1990 to 2022 using the official CITEPA/UNFCCC reporting time series, which is fed by National Forest Inventory (NFI) data, with additional external data for harvest before 2010. The authors report that the forest CO2 sink rose from 49.3 MtCO2 yr-1 in 1990 to a peak of 74.1 MtCO2 yr-1 in 2008, then declined rapidly to 37.8 MtCO2 yr-1 in 2022. They decompose the 2013-2017 decline into roughly equal contributions from increased natural mortality (51.5%) and increased harvest (48.5%), and present regional trajectories, including two northern regions becoming net CO2 emitters. A separate analysis of NFI plot-level mortality over 2010-2023 identifies species, regions, and tree-size classes with rising mortality. The paper concludes with a discussion of remote-sensing and data-integration tools to supplement the NFI, and with recommendations for improving future forest carbon accounting. Important caveats are acknowledged throughout, most notably that the sink decrease could be a transient methodological artifact of NFI protocol changes and that a direct stock-change estimate yields a sink about half the reported UNFCCC value.","tokens_in":36517,"tokens_out":4510,"duration_ms":44560,"significance":"If the reported decline in the French forest carbon sink is robust, the result is policy-relevant: it directly challenges the national carbon-neutrality trajectory (SNBC) and illustrates how inventory-based estimates should feed into LULUCF reporting. The paper is valuable as a synthesis of official French data and as a clearly written call for improved monitoring, including deadwood measurement, harvest-time-series reconstruction, and integration of remote sensing. The authors are commendably explicit about the limitations of the underlying data, including non-documented interpolation, biased pre-2010 harvest statistics, and the absence of measured deadwood and soil changes. However, the central quantitative claims—the magnitude of the recent sink, the 2008 peak, and the 51.5%/48.5% attribution—rely on the CITEPA/UNFCCC flux framework that the paper itself calls into question. The significance of the paper therefore depends on the authors resolving these internal inconsistencies in a revision.","major_comments":[{"comment":"The paper presents the UNFCCC/CITEPA time series as the headline result (37.8 MtCO2 yr-1 in 2022), but Fig. 1b shows that a direct year-on-year stock-change estimate from the same NFI plots gives a sink about two times smaller in recent years, and the authors state that 'the current report to UNFCCC very likely overestimates the real carbon sink in the French forests.' This internal inconsistency is load-bearing: if the UNFCCC estimate is an overestimate, then the headline sink value, the regional net-emitter claims, and the 51.5%/48.5% decomposition are all computed on a biased basis. The explanation that 'data from the latest inventory campaigns were not yet included' is plausible but is an assertion, not a reconciliation. The revision should either use the more conservative direct stock-change estimate as the primary result or provide a quantitative reconciliation of the two curves, including confidence limits for the stock-change method.","section":"§1, Fig. 1b"},{"comment":"The authors themselves write that 'it cannot be ruled out that the sink decrease seen in Fig. 1a could be a transient methodological artifact' because 2013 is the median year of the first revisit campaign under the 2010 semi-permanent plot protocol, and they later note that the mortality increase coincides with the 2015 extension to actual mortality tracking. Since the entire 'rapid decline from 2013 to 2017' narrative and its attribution rest on this period, the manuscript needs a sensitivity analysis that isolates the protocol change: for example, restrict the analysis to the post-2010 homogeneous protocol, use only the direct stock-change estimate, or quantify the effect of the harvest-measurement and mortality-tracking changes on the reported fluxes. Without such a test, the claimed 43.05% decrease between 2013 and 2017 remains potentially a methodological artifact, as the text concedes.","section":"§1 and §2, protocol-change caveat"},{"comment":"The long-term (1990-2008) increase and the 2008 peak are based partly on a non-documented CITEPA interpolation for 1990-2005 and on pre-2010 harvest data that the text describes as historically underestimated by about 50% and therefore taken from biased external sources (AGRESTE and energy-wood surveys). These caveats mean the early rise from 49.3 to 74.1 MtCO2 yr-1 is not established with the same data quality as the post-2010 period. The paper should at minimum state the sensitivity of the 2008 peak and the reported linear increase rate (2.45% yr-1) to these assumptions, or present a reconstruction that uses only the internally consistent NFI stock data, as suggested in the authors' own recommendation (i).","section":"§1, pre-2010 data quality"},{"comment":"The attribution of the 2013-2017 sink decline to 51.5% mortality and 48.5% harvest is not robust because the NFI does not distinguish salvage harvesting from planned harvest; trees that die and are removed before the second visit are counted as harvest, not mortality. The text itself notes that 'massive mortality can therefore drive harvest increases without being identified as such.' Additionally, deadwood carbon changes are not measured except after cyclones, and the soil carbon term is modeled with a neutrality assumption for established forests. These omissions directly affect the decomposition that is central to the paper's mechanistic claim. The revision should either present the mortality and harvest contributions as a range under alternative classification assumptions (e.g., moving all post-disturbance salvage into mortality) or clearly label the 51.5%/48.5% split as conditional on the current reporting conventions.","section":"§1 and §3, flux attribution and missing pools"},{"comment":"The mortality rate formula is written as M = [1 - (N_t1/N_t0)^(1/t)] × 100, with N_t1 defined as the number of trees alive at the first visit and N_t0 as the number that survived between visits. With these definitions, N_t1 > N_t0, so the ratio exceeds unity and the formula yields a negative mortality rate. The subscripts appear to be reversed; the intended Kohyama et al. (2018) formulation uses (survivors / initial) raised to the power 1/t. This is not a mere notational slip because Section 3's mortality estimates, including the reported fold-changes for Jura, Vosges, and ash, depend on the correct calculation. Please correct the formula and confirm that the numerical results were computed with the standard expression.","section":"§3, mortality rate formula"}],"minor_comments":[{"comment":"There is an internal numerical inconsistency: the main text says the sink 'reached a value of 38.8 MtCO2 yr-1 in 2022' in one place, then states 'the sink kept stable to a low value of 37.8 MtCO2 yr-1' in another; the abstract uses 37.8. Please harmonize the reported 2022 value and verify the source table.","section":"§1, text around Fig. 1a"},{"comment":"The text contains an unresolved placeholder: 'mortality increased by *% per year after 2015.' Provide the actual percentage or remove the quantitative claim.","section":"§2, Grand Est mortality"},{"comment":"The concluding remarks refer to 'Fig. 1b' when discussing the SNBC-2/SNBC-3 target revision, but the SNBC trajectories are shown in Fig. 1c. Please correct the cross-reference.","section":"§1, Fig. 1c vs concluding remarks"},{"comment":"The statement that the NFI weights are 'not publicly available to reproduce the calculation' is an important transparency limitation; consider adding this to the data-availability statement and discussing the implications for reproducibility of the reported confidence intervals.","section":"§1, NFI weights"},{"comment":"Placeholder references '(REF)' appear in the text for NFI accuracy and for the Sentinel-2 bark-beetle detection citation; these must be completed before submission. Minor language issues include 'decreased quickly' twice in one sentence, 'offsetted', and inconsistent usage of 'Mton CO2' versus 'MtCO2'.","section":"Various"}],"recommendation":"major_revision","confidential_remarks":"This manuscript addresses a timely and policy-relevant topic, and the authors have been unusually candid about the data limitations. However, the central quantitative claims are internally contradicted by the paper's own direct stock-change comparison and by its explicit methodological-artifact caveat. The revision needs to either re-center the paper on the more defensible direct stock-change estimate or provide a rigorous reconciliation, plus a sensitivity analysis around the 2010/2015 protocol changes. I would not reject the paper outright, because the qualitative decline in the sink and the regional divergence may well survive such an analysis, and the recommendations are valuable. Please also ensure the placeholder values and inconsistent numbers are fixed, as they currently appear in the public version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nThe short version: this is an important data synthesis about the French forest carbon sink, and the broad narrative — sink peaked around 2008, then declined sharply, with mortality playing a big role — is plausible and worth attention. But the specific headline numbers (74.1 to 37.8 MtCO2/yr, mortality 51.5% vs harvest 48.5%) are not on solid ground, and the paper itself says so in places. The reader's conditional verdict is fair; I'd actually be a bit more blunt.\n\nWhat's new: the integrated 1990–2022 trajectory with regional and species-level mortality decomposition is not in the earlier papers. The authors pull together CITEPA/UNFCCC reporting with NFI tree-level data, and the mortality analysis (Section 3) is a useful contribution. They also give a clear reading of the SNBC targets against the sink decline, which is policy-relevant. The paper is honest about the data's limits — they flag the protocol change risk, the pre-2010 harvest bias, the modeled soil term, the deadwood gap. That candor is a real strength.\n\nWhere it wobbles: the stress-test concern is not pedantry. The paper's own Fig 1b shows a direct stock-change estimate from the same NFI plots giving a sink about half the CITEPA/UNFCCC number in recent years. The authors hand-wave that as \"latest campaigns not yet included\" but that's not a reconciliation. If the direct stock-change method is even close to right, the headline 37.8 is too high, and the mortality/harvest decomposition computed within the CITEPA framework inherits the problem. They also concede, in the text, that the 2013–2017 drop coincides with the 2010 protocol change and the 2015 mortality extension, and \"cannot be ruled out\" as a transient artifact. That is a big concession to leave embedded rather than to resolve. So the direction of the decline might survive, but the magnitude and the attribution to mortality vs harvest are not yet established.\n\nThere are also small signs of a draft: \"REF\" placeholders, a missing number in \"increased by *% per year\", and some garbled sentences near the end. Those are fixable but they add to the sense of haste.\n\nWho is this for: people working on European forest carbon accounting, LULUCF policy, and anyone citing the French sink in assessments. That's a wide audience, and the paper deserves a serious referee. I would not desk-reject it, but I'd push for a major revision that reconciles the two sink estimates and either confirms or brackets the 2013–2017 decline as protocol-driven.\n\nRecommendation: engage with it, but treat the headline as an upper bound until the stock-change discrepancy is resolved.","headline":"A useful, unusually candid data synthesis, but the headline sink decline and its mortality/harvest split rest on an unreconciled discrepancy with the direct stock-change estimate.","tokens_in":37062,"tokens_out":2021,"would_cite":false,"duration_ms":20894,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"France's forests removed 37.8 MtCO2 in 2022, down from a 74.1 MtCO2 peak in 2008, and the paper attributes most of the loss to rising mortality and harvests.","keywords":["French forests","carbon sink","forest mortality","national forest inventory","carbon accounting","climate change","harvest","LULUCF"],"falsifier":"Recompute the national forest carbon budget using only plots measured under the consistent post-2010 semi-permanent protocol, with harvests retrospectively reconstituted from stock differences as the paper recommends; if the abrupt 2013-2017 sink drop disappears under that homogeneous protocol, the decline is an artifact rather than a real biospheric response.","tokens_in":36078,"feed_emoji":"🌲","tokens_out":7783,"duration_ms":72314,"temperature":0.7,"pith_summary":"The paper argues that French forests, a long-standing land carbon sink, removed 74.1 MtCO2 per year from the atmosphere in 2008 but only 37.8 MtCO2 per year in 2022, a decline concentrated in 2013-2017. It attributes the drop almost equally to rising natural tree mortality (51.5% of the decrease) and to increased harvest (48.5%), and shows that two northern regions now emit more carbon than they absorb. The authors present this decline as the best-supported reading of official forest carbon accounting while cautioning that a change in the national forest inventory's sampling protocol in 2010-2015 could in principle produce part of the apparent drop as a methodological artifact.","feed_headline":"France's forest CO2 sink has halved since 2008","feed_subtitle":"Rising tree mortality and harvests drove the drop; two northern regions are now net emitters.","key_machinery":"The load-bearing object is the national forest inventory's semi-permanent plot network: since 2010 roughly 7,000 new plots and 7,000 five-year revisits are measured each year, with individual trees tracked for survival since 2015. The carbon accounting breaks the net sink into gains from growth of established forests, new forest creation, modeled soil and litter change, and wood products, minus losses from harvest, natural mortality, and storm damage. Tree mortality is computed with the standard rate formula $M = \\left[1 - \\left(N_{t1}/N_{t0}\\right)^{1/t}\\right] \\times 100$, where $N_{t0}$ and $N_{t1}$ are survivor counts at two visits $t$ years apart. The entire argument depends on treating these inventory-derived fluxes as a continuous 1990-2022 time series despite the protocol changes.","core_discovery":"The central discovery is a halving of France's forest CO2 sink over 14 years. Using national forest inventory measurements aggregated into official carbon reports, the authors find that annual CO2 removal by forests rose from 49.3 MtCO2 in 1990 to a peak of 74.1 MtCO2 in 2008, then fell to 37.8 MtCO2 in 2022. They decompose the decline into three phases: a slow decrease from 2008 to 2013, a rapid 43% drop from 2013 to 2017 driven about equally by increased natural mortality and increased harvest, and a stable but historically low sink afterward. Regionally, Hauts-de-France and Grand Est have become net CO2 emitters, while lightly managed southern regions and the Landes plantations retain a sink. The authors also show that post-2017 mortality is concentrated in drought-sensitive species such as Norway spruce, silver fir, and ash, and in large trees, and they stress that the full effects of the 2022-2023 droughts are not yet visible in the inventory's five-year revisits. They explicitly flag that the abrupt decline coincides with the first full revisit cycle of the new semi-permanent plot protocol and with a change in mortality tracking in 2015, so a transient methodological artifact cannot be ruled out.","pith_inferences":["If the protocol-artifact caution is set aside, the same inventory-design problem affects any country whose national forest inventory changed from periodic to semi-permanent plots, so the French case is a general warning about interpreting recent forest sink declines.","Because deadwood is not measured in the inventory, the current net sink may be temporarily inflated: trees that died recently are counted as mortality loss but their carbon is not yet tracked as deadwood, and its future decay will add a delayed CO2 emission.","The paper's proposal to reconstitute past harvests from stock differences could be tested against independent satellite-based clearcut maps, providing a direct arbitration of whether the 2013-2017 harvest increase was real or an artifact of external statistics.","If mortality stays at post-2017 levels, the French forest sink is roughly half its 2008 value, meaning national climate plans that assume a growing forest sink will need to find offsetting reductions elsewhere."],"forward_implications":["If the decline is real, France's land-use carbon budget targets are out of reach: the earlier target of about -39 MtCO2e per year was already missed, and the revised target of -9 MtCO2e per year for 2024-2028 would require the forest sink to stop falling.","Because natural mortality has tripled and remains elevated, the sink will stay low even if harvest returns to previous levels.","The two northern regions that became net emitters show that national accounting can hide subnational carbon losses.","Since the inventory reports five-year running averages, the carbon losses from the 2022 and 2023 droughts and fires will appear only in later editions, so current official estimates understate recent disturbance impacts.","A direct year-on-year stock-change calculation from annual plots gives a sink about half the size of the official report, implying the official numbers likely overestimate the forest sink."],"supporting_citations":[{"why":"Documents the 2010 shift to semi-permanent five-year revisit plots, the protocol change that complicates long-term flux continuity.","marker":"Bontemps and Bouriaud, 2024"},{"why":"Shows that five-year moving-average inventory estimates lag disturbance events, motivating the paper's caution about the 2013-2017 decline.","marker":"Van Deusen, 2011"},{"why":"Demonstrates that pre-2010 NFI harvest measurements were biased and underestimated by about 50%, forcing reliance on external harvest data.","marker":"Denardou-Tisserand, 2019"},{"why":"Supplies the mortality rate formula used to turn repeated tree censuses into annual mortality percentages.","marker":"Kohyama et al., 2018"},{"why":"Supports attribution of the post-2018 mortality surge in northeastern France to the extreme central European summer drought.","marker":"Schuldt et al., 2020"},{"why":"Establishes the bark-beetle-drought link behind spruce mortality and the conversion of spruce forests from sinks to sources.","marker":"Hlásny et al., 2021"},{"why":"Provides independent inventory scrutiny of growth stagnation, used in the paper to separate growth effects from mortality effects.","marker":"Hertzog et al., 2025"},{"why":"Supplies an independent satellite-based estimate of the 2022 wildfire biomass loss in Nouvelle-Aquitaine that the NFI has not yet captured.","marker":"Vallet et al., 2023"}],"fun_headline_variants":["France's forest CO2 sink halved since 2008 peak","Tree mortality and harvests halve France's forest carbon sink","French forests absorb half as much CO2 as in 2008","Forest deaths and logging cause 49% drop in France's CO2 sink"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes the post-2010 semi-permanent plot fluxes are continuous with pre-2010 inventory campaigns and external harvest statistics; the paper itself states that it cannot be ruled out that the sink decrease is a transient methodological artifact.","fun_headline_variants_meta":{"raw":{"variants":["France's forest CO2 sink halved since 2008 peak","Tree mortality and harvests halve France's forest carbon sink","French forests absorb half as much CO2 as in 2008","Forest deaths and logging cause 49% drop in France's CO2 sink"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000242,"raw_usage":{"total_tokens":1636,"prompt_tokens":1165,"completion_tokens":471,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":781,"completion_tokens_details":{"reasoning_tokens":395}},"tokens_in":781,"tokens_out":471,"duration_ms":4542,"temperature":1.0,"reasoning_tokens":395,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:10:48.664109+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the national forest carbon budget using only plots measured under the consistent post-2010 semi-permanent protocol, with harvests retrospectively reconstituted from stock differences as the paper recommends; if the abrupt 2013-2017 sink drop disappears under that homogeneous protocol, the decline is an artifact rather than a real biospheric response.","supporting_citations":[],"review_version":1}