Pith. sign in

REVIEW 5 major objections 5 minor 3 references

An outlook on the Rapid Decline of Carbon Sequestration in French Forests and associated reporting needs

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read 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.

desk verdict 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. read the letter →

arxiv 2505.11512 v1 pith:7QBS2O2U submitted 2025-05-03 physics.ao-ph physics.data-an

P . Ciais , C. Zhou , P . Schneider , M. Schwartz , N. Besic , C. Vega , J.-D. Bontemps This is my paper · ORCID
classification physics.ao-phphysics.data-an
keywords FrenchforestscarbonsinkforestmortalitynationalinventoryaccountingclimatechangeharvestLULUCF
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

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.

What carries the argument

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.

What would settle it

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.

Watch

Extended reading notes

Core claim

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.

Load-bearing premise

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.

Editorial extensions

If this is right

  • 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.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • 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.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

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.

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 (5)
  1. [§1, Fig. 1b] 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.
  2. [§1 and §2, protocol-change caveat] 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.
  3. [§1, pre-2010 data quality] 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).
  4. [§1 and §3, flux attribution and missing pools] 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.
  5. [§3, mortality rate formula] 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.
minor comments (5)
  1. [§1, text around Fig. 1a] 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.
  2. [§2, Grand Est mortality] The text contains an unresolved placeholder: 'mortality increased by *% per year after 2015.' Provide the actual percentage or remove the quantitative claim.
  3. [§1, Fig. 1c vs concluding remarks] 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.
  4. [§1, NFI weights] 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.
  5. [Various] 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'.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the paper interprets external NFI/CITEPA inventory data and does not fit parameters or rename outputs as predictions; self-citations are complementary and non-load-bearing.

full rationale

The paper's central claims—the decline of the French forest CO2 sink from 74.1 MtCO2 yr-1 in 2008 to 37.8 MtCO2 yr-1 in 2022 and the roughly equal attribution of the 2013-2017 decline to mortality (51.5%) and harvest (48.5%)—are presented as analyses of CITEPA and National Forest Inventory data, not as outputs of a model fitted to those data. No equation in the paper defines a predicted quantity in terms of the target quantity, and no fitted parameter is renamed as a prediction. The comparison in Fig. 1b between the UNFCCC-reported sink and a direct year-on-year stock-change estimate is an internal consistency check, not a circular construction; if anything, it challenges the headline estimate rather than entailing it. The paper's substantial citations of the authors' own prior remote-sensing work (e.g., Schwartz et al. 2023) support the complementary monitoring discussion and the height/biomass maps, but the sink estimates do not depend on those maps. The acknowledged limitations—protocol changes around 2010 and 2015, non-documented CITEPA interpolation for 1990-2005, biased pre-2010 harvest data, and the possibility that the sink decrease is a transient methodological artifact—are data-quality and interpretation risks, not circular reasoning. The paper is self-contained as an interpretation of externally sourced inventory statistics, so no circularity step can be exhibited. The honest finding is no significant circularity; the score reflects only the presence of many self-citations, which are not load-bearing for the core claim.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new theory, fitted parameters, or entities. It interprets official NFI/CITEPA statistics, so the central claim inherits all assumptions behind that accounting chain: sampling representativeness, cross-protocol comparability, the undocumented CITEPA interpolation, the neutrality of soil carbon, model-based wood products, and external harvest statistics before 2010. The two-fold discrepancy between the official net sink and the direct stock-change method (Fig. 1b) is an unresolved load-bearing uncertainty that affects all national conclusions.

assumptions (6)
  • domain assumption NFI design-based estimates are unbiased at national and regional scales despite non-public plot weights.
    Section 1 states plot weights 'are not publicly available to reproduce the calculation', so the unbiasedness claim cannot be independently verified; the paper reports ~4% national accuracy from a reference.
  • domain assumption Flux estimates across the 1990-2022 period are comparable despite major protocol changes in 2005, 2010, and 2015.
    The authors explicitly write that 'it cannot be ruled out that the sink decrease seen in Fig. 1a could be a transient methodological artifact' and that pre-2010 harvest from external data is biased.
  • domain assumption CITEPA's non-documented interpolation of NFI data for 1990-2005 is accurate enough for trend analysis.
    Section 1: 'A non-documented interpolation method was used by CITEPA to elaborate a regional and nationwide overview of forest fluxes.'
  • domain assumption Soil and litter carbon in established forests does not change, and deadwood carbon changes are negligible.
    Section 1: 'the inventory makes a neutrality assumption that there is no storage in soil carbon'; deadwood is not measured except after the two cyclones.
  • domain assumption Growth estimated from 5-year tree cores and recruitment approximates actual growth.
    Introduction: growth is not derived from circumference remeasurement but from coring at first visit, plus recruitment at the revisit.
  • domain assumption The 2.5-year assumed growth period for trees that are cut or dead at revisit is reasonable.
    Introduction: 'with the assumption that they were still growing during 2.5 years before being cut'.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An outlook on the Rapid Decline of Carbon Sequestration in French Forests and associated reporting needs." pith.science (2026). https://pith.science/paper/7QBS2O2U

@misc{pith2026250511512,
  author       = {Pith},
  title        = {Pith review of: An outlook on the Rapid Decline of Carbon Sequestration in French Forests and associated reporting needs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7QBS2O2U}},
  note         = {Machine review of arXiv:2505.11512}
}
read the original abstract

In this study, we present and discuss changes in carbon storage in the French forests from 1990 to 2022, derived from CITEPA statistics on forest carbon accounting, fed by National Forest Inventory (NFI) data collected through an extensive network of measurement sites across Metropolitan France, and other data sources as regards forest removals. The NFI is designed to provide statistical estimations of growing stock, gains and losses at the national or subnational levels but is unable, in its classical form, to provide detailed spatial outlook, such as on abrupt losses during fires, droughts and insect attacks. A continuing removal of CO2 from the atmosphere by the French forests occurred from 1990 to 2022, because harvest and mortality CO2 losses remained smaller than CO2 removals by forest growth and the increase in forest area (70,000 ha per year but insignificant in terms of increased carbon stocks at present). The CO2 removal by forests was 49.3 MtCO2 yr-1 in 1990, increased to reach a peak of 74.1 MtCO2 yr-1 in 2008 and then quickly decreased down to 37.8 Mton CO2 yr-1 in 2022. After 2017, the sink remained low and mortality rates stayed larger than during any of the previous years. This recent period is marked by climate shocks such as summer droughts and heatwaves in 2015, 2018, 2022, 2023. The full impacts of the droughts in 2022 and 2023 are not yet covered with full precision, as some of the sites measured by the national inventory before those droughts are still pending a second visit. The different regions of France show contrasted trajectories. Southern Mediterranean regions where forests have a low harvest rate have experienced a lower increase in mortality and a sustained CO2 uptake.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

  1. [2]

    Changes in carbon stocks at regional scale 9 Fig. 2. Same as Fig. 1a but for 13 administrative regions of France indicated in the map. In this section, to gain insight on the contribution of each region to the nation-wide reduction of the carbon sink and increase of carbon losses, we analyze regional trends. Fig 2 shows carbon gains and losses of the 13 d...

    work page 2008

  2. [3]

    Grande Région ECOlogique

    Changes in tree mortality across regions, species, and height classes More insight into tree mortality across different regions, species, and tree size classes was gained by analyzing the NFI data collected during first visits between 2010 and 2018, with re-visits occurring five years later from 2015 to 2023. Following a revision of the sampling protocol ...

    work page 2010

  3. [4]

    Early detection of bark beetle infestation in Norway spruce forests of Central Europe using Sentinel-2,

    New tools to monitor forests using satellite and ground observations Ground-based observations, such as the French National Forest Inventory surveys, provide high-quality data for monitoring forest dynamics at the national or regional scale (Fig 1, 2). However, these data have several limitations. First, the temporal resolution for forest flux measurement...

    work page arXiv 2023

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.