Feasibility literature review: do global crop-yield projections omit a soil-degradation feedback?

18 October 2026

Working question: Do current global crop-yield projections omit a soil-degradation feedback, and how large is it?

0. Executive summary

Verdict: feasible, but the headline hypothesis is narrower and riskier than the framing suggests.

1. The core idea is not new in kind, only in scale and method. Four strands have already estimated an "underestimation term":

   • Site scale: Basso et al. 2018, 8 sites, multi-model. Adding SOC/N feedbacks deepens losses at +3 °C by about 4.5 percentage points. This is the only figure IPCC AR6 WGII Ch. 5 cites.
   • National scale: Qiao et al. 2022, Nature Climate Change. Soil degradation worsens the projected RCP8.5 cereal loss in China by 14%. The full text shows this uses gradient-boosted trees on a soil-quality index with CMIP5 climate. The "degradation" case is a stylised scenario in which all farmland is set to low quality, not a modelled trajectory.
   • Regional multi-model: Couëdel et al. 2024 and 2026, 15–16 soil–crop models in sub-Saharan Africa. Soil-fertility decline (about −24%) outweighs single climate factors.
   • Global statistical, improvement direction: Deng et al. 2023 (STOTEN) and He et al. 2025 (Eur. J. Agron.). SOC is the dominant soil control on how sensitive yields are to warming.

   No study found in this review does the global, process-based, degradation-direction version tied to texture- or mineralogy-normalised thresholds and benchmarked against the GGCMI phase 3 / CMIP6 projections. That is the defensible gap.

   The gap is now verified from the GGCMI phase 3 protocol. In Jägermeyr et al. 2021 (full text, Methods):

   • Management is held at 2015 levels after 2015.
   • All models are spun up to soil and carbon-pool equilibrium.
   • EPIC-IIASA re-initialises all soil attributes, including SOM fractions, every year.
   • Erosion is excluded, and nutrients other than N are non-limiting.
   • Soils come from a single static, cropland-weighted dominant HWSD topsoil per 0.5° cell (GGCMI phase 3 soil-input document).

   The flagship projections therefore start from equilibrium soils and carry no management-driven degradation trajectory.

2. Evidence against a large term is substantial and must be designed in, not argued away.

   • The SOC×climate interaction was small in the strongest multi-model test (Couëdel et al. 2026). Most of the feedback there is an additive fertility-decline effect.
   • SOC's marginal yield effect is small where nitrogen is non-limiting. Ma et al. 2023 found it one-fifth as effective as N. Hijbeek et al. 2016 (Plant Soil) found +1.4% (n.s.) across 20 European long-term trials, and Xu et al. 2025 attribute about 12% of global yield variance to soil health against about 70% to management.
   • Trial-derived SOC benefits did not appear on real farms (Zhao et al. 2026, Nature Food, causal inference).
   • SOC's drought-buffering benefit is largest under moderate water supply, not extreme drought (King et al. 2026, GCB). This contradicts Kane et al. 2021 at the extremes.
   • Even the most favourable global crop-model estimate puts SOC buffering at only 0.5–1.1 %/°C for maize, against 6.6–7.5 %/°C projected losses, and the effect weakens under SSP5-8.5 (He et al. 2025).

3. Threshold concept is the weakest link.

   • The SOC:clay 1/13 boundary comes from soil-structure quality in 161 Swiss soils (Johannes et al. 2017). It was never shown to mark a yield breakpoint.
   • It has been heavily criticised at national and continental scale:
     • Poeplau & Don 2023: 94% of German Chernozems classed degraded.
     • Rabot et al. 2024 (France): the ratio is a poor predictor of bulk density and aggregate stability, and would class 63% of cropland as unhealthy.
     • Mäkipää et al. 2024 (EU-wide LUCAS): a single 1:13 threshold is inadequate outside its derivation conditions.
     • Feeney et al. 2024.
   • Farmer-perceived SOM thresholds (Hijbeek et al. 2017, 1,452 farmers) are highest for sandy soils (sand 1.2–4.7%, loam 0.6–2.6%, clay 1.0–2.4% SOM). This is the opposite of what a clay-normalised ratio implies.
   • It is physically questionable in sesquioxide-rich and allophanic tropical soils. Mineral capacity per kg silt+clay is about 46 g C for 1:1 clays vs about 82 for 2:1 clays (Six et al. 2024).
   • No study located shows non-linear "collapse" of yield at a texture-normalised SOC threshold. The study that defined the thresholds found the SOC–structure relationship to be linear (Johannes et al. 2017). The only critical-level review found the quantitative evidence "slight" (Loveland & Webb 2003).

4. Raw CMIP6 cSoil is unusable for croplands. This confirms the planned design choice.

   • There is no cropland management in ESMs.
   • Inputs are NPP-dominated, and "false priming" artificially narrows the spread (Varney et al. 2022, 2023).
   • Data-constrained turnover flips the sign of 21st-century soil C change from +19–30 PgC to −19 to −45 PgC (Ren et al. 2024).
   • Offline models also disagree on sign depending on the carbon-input assumption: RothC global gains (Gottschalk et al. 2012) vs LPJmL cropland losses in all scenarios (Herzfeld et al. 2021).

5. Likely result: a globally production-weighted underestimation term that is small (low single-digit % of projected loss) in high-input breadbaskets and material in low-input tropical and dryland systems. That is a credible and publishable finding, but it fits Nature Food or Nature Communications as a spatially resolved "where it matters" result rather than a Science-level "projections are wrong" claim.

6. The transformer baseline needs a head-to-head test. Under spatially blocked validation, gradient-boosted trees beat transformers on cropland SOC (R² ≈ 0.66 vs ≈ 0.21; Fourel et al. 2026). The baseline's value must come from calibrated uncertainty and threshold-exceedance probabilities, and it should be benchmarked against tree ensembles and SoilGrids on cropland-only spatial folds.

7. Recommended re-framing: test "Does omitting soil-state dynamics bias projected yield impacts, where, and through which pathway (additive fertility loss vs altered climate sensitivity)?" Use Couëdel's reset-vs-continuous decomposition at global scale. Treat SOC:clay as one of several normalised indices and report a null result as a bound.

1. Evidence base and limits

The literature search covered OpenAlex, Crossref, Semantic Scholar, publisher pages, and references associated with key papers. Nature- and Science-family journals were searched separately. Bibliographic details and DOIs were checked against Crossref. The final annotated bibliography contains 183 entries.

The review distinguishes full-text findings from abstract, secondary-source, and metadata-only evidence. Full-text readings focused on relevant methods, results, limitations, quantities, and qualifications; they do not imply that every page was read. The bibliography preserves the evidence basis for each annotation.

At the time of this assessment, nine papers had not been examined in full. These included Zhao 2026, Bradford 2019, Wuepper 2019, Hijbeek 2026, and Hijbeek 2017. Claims attributed to them in this assessment retain their stated abstract-level limits. The later historical review incorporates additional full texts. This is a substantial core collection, not an exhaustive census of the literature.

Several bibliographic distinctions matter: Prout's soil-carbon-to-clay paper appeared online in 2020 and in a 2021 issue; Deng's global crop-yield study is from 2023; GSoilCPM is Zhang et al. 2025; and Feeney's contextual benchmarks are from 2024. Hijbeek's farmer-perceived thresholds come from the 2017 Ecological Indicators paper, which is distinct from the organic-input meta-analysis published online in 2016 in Plant and Soil.

2. Thematic synthesis

A. SOC–crop yield relationships

• Shape: positive but saturating, with optima that disagree by about 4×.
  • Oldfield et al. 2019 (global meta-analysis, maize and wheat): yields level off near 2% SOC (about 20 g/kg). About two-thirds of maize and wheat land sits below that.
  • Ma et al. 2023 (Nat. Geosci., 13,662 controlled trials): optima of 43–44 g/kg (maize), 13 g/kg (wheat), 31–32 g/kg (rice).
  • Vendig et al. 2023 (Nat. Sustain., cover-crop meta-analysis): direct SOC yield benefits only below 9.2 g/kg.
  • Zhao et al. 2026 (Nat. Food): trials show gains to 30–40 g/kg, but farms show at most a 5% gain up to 10 g/kg, after which yields fall back.
  • Contradiction: the same quantity ("optimum SOC") ranges from about 9 to about 44 g/kg depending on crop, data source and method. Any adopted yield-sensitivity function inherits this uncertainty and should be propagated, not fixed.
• Magnitude relative to nitrogen.
  • Ma et al. 2023: SOC sequestration is about one-fifth as effective as N fertilisation. With current practices, global production would rise only 0.7% once other constraints are met.
  • Hijbeek et al. 2017 (Plant Soil; 20 European long-term trials): organic inputs add +1.4 ± 1.6% (n.s.) to yield when NPK is sufficient.
  • Vonk et al. 2020 (1,264 European farms): no consistent SOM–yield correlation (positive for wheat in the Continental climate, negative in the Atlantic climate).
  • Xu et al. 2025 (Nat. Food): soil health explains about 12% of global production variation vs about 70% for management.
  • Implication: in fertiliser-intensive systems, a falling SOC level can largely be compensated by N. Any "omitted term" will be damped there.
• Causality.
  • Most evidence is observational and cross-sectional (Oldfield, Ma, Deng 2023, Kane 2021, Qiao 2022).
  • Reverse causation is plausible: higher yields produce more residue inputs and so higher SOC (Janzen 2006; Moinet et al. 2023).
  • The only explicitly causal-inference study (Zhao et al. 2026) finds the trial relationship does not transfer to farms, because excess N, farmer knowledge and infrastructure override it.
  • Long-term experiments give within-site causal contrasts: Kätterer et al. 2024 (Ultuna, +16% maize per % SOC where not N- or pH-limited); Laub et al. 2025 (four Kenyan trials, where yield and SOC trends are tightly coupled at adequate-rainfall sites but not at dry sites); Poulton et al. 2018 (Rothamsted).
• Moinet et al. 2023 (GCB) reviewed 21 meta-analyses and found yield effects of increasing SOC "inconsistent", ranging from negative to positive. This is the strongest general counter-argument to "SOC loss → yield loss".

B. SOC as a buffer against climate stress

• Supportive:
  • Qiao et al. 2022 (NCC, 12,115 site-years, China): high-quality soils cut yield variability by 15.6% and improve the climate-change outcome by 1.7%. Degradation exacerbates the RCP8.5 loss by 14%.
  • Kane et al. 2021 (ERL, 12,376 US county-years): under severe drought, +1% SOM ≈ +2.2 Mg/ha and −36% insurance payouts. The effect weakens in clay soils.
  • Williams et al. 2016: soil water-holding capacity reduces yield volatility.
  • Iizumi & Wagai 2019: small SOC increases raise drought tolerance on 70% of harvested area.
  • Deng et al. 2020 and Feng et al. 2022 (China): soil properties explain about 72% of maize temperature sensitivity, with SOC the largest share.
  • Deng et al. 2023 (global random forest): SOC is the dominant explainer of how yield responses to warming vary in space.
  • He et al. 2025 (9 GGCMs × 37 CMIP6 GCMs, 3 SSPs; full text): soil properties explain 51–59% of the spatial variation in temperature sensitivity. But the buffering magnitude is modest: SOC improvement cuts warming losses by only 0.5–1.1 %/°C (maize) and 1.3–2.5 %/°C (soybean), against projected losses of 6.6–7.5 %/°C (maize). The benefit shrinks under SSP5-8.5.
  • Teng et al. 2024 and Nouri et al. 2021: conservation agriculture improves stability under warming and extremes.
• Contrary or qualifying:
  • King et al. 2026 (GCB, 67,000 US county-years, 1981–2020): SOC's yield benefit is largest under moderate water supply, not drought. SOC cannot compensate for extreme water scarcity. This contradicts Kane et al. 2021 at the drought extreme. It directly challenges the "SOC × drought breakpoint" premise in the proposed analysis.
  • Minasny & McBratney 2018: +1% OC raises available water capacity by only about 1.2 mm per 100 mm of soil, a weak water mechanism.
  • Hendrawan et al. 2023 (global, 0.5°): precipitation, water-holding capacity, acidity, PET and fertiliser explain crop drought sensitivity. SOC is not among the leading drivers.
  • Rigden et al. 2020 (Nat. Food): soil moisture × VPD drives damage, so water supply (not storage capacity) is often limiting.
• Interaction with N and irrigation.
  • Buffering effects are weaker where fertiliser is intensive (Xu 2025; Zhao 2026) and are largely irrelevant under full irrigation.
  • Folberth et al. 2016 show edaphic effects matter less "when water and nutrients are supplied exogenously".
  • The analysis must stratify by N-input and irrigation (MIRCA, Adalibieke et al. 2023 N dataset).
• Methodological caution. Several "SOC lowers climate sensitivity" results (He 2025; part of Deng 2023) are derived from crop-model or machine-learning partial dependence, not observed interactions. Using them to estimate an omission in crop models risks circularity.

C. SOC thresholds and critical levels

• Absolute thresholds.
  • Loveland & Webb 2003 (full text): the widely held 2% SOC (≈3.4% SOM) critical threshold has only "slight" quantitative support. There may be a desirable range, and there is little quantitative evidence that SOC decline below it reduces yields.
  • Musinguzi et al. 2013: about 2% is widely cited for tropical soils, but N-response thresholds range from 0.5% to 2%, and no universal value exists.
• Texture-normalised (SOC:clay).
  • Dexter et al. 2008 proposed about 1/10 as the clay-complexation limit (1 g OC complexing 10 g clay).
  • Johannes et al. 2017 (full text) defined 1/8, 1/10 and 1/13 structure classes from visual structure scores of 161 Swiss Cambi-Luvisol samples. They found the SOC–pore-volume relation linear at all clay contents, with no optimum, which contradicts Dexter's complexed-carbon theory and argues against threshold collapse. Their abstract misprints the poor-structure boundary as "1:8 or smaller"; the body uses 1:13, and a corrigendum exists (10.1016/j.geoderma.2017.05.009). Prout et al. 2021 and 2022 applied them in England and Wales: about half of arable soils fell below 1/13, and soils ≥1/8 lost the most carbon.
  • Critiques:
    • Poeplau & Don 2023 (2,958 German topsoils): the ratio is "strongly biased"; 94% of Chernozems are classed as degraded.
    • Feeney et al. 2024 (LUCAS, n = 19,855): an observed-to-typical SOC index by pedo-climatic zone outperforms SOC:clay.
    • Rabot et al. 2024 (French RMQS; full text): SOC/clay is a poor predictor of bulk density and aggregate stability, and raw SOC is better. Acidic soils are systematically rated healthy and alkaline soils unhealthy. 1/13 would class 63% of cropland as unhealthy.
    • Mäkipää et al. 2024 (EU-wide LUCAS; full text): one 1:13 threshold is inadequate outside the conditions it was derived from, and the ratio is inconsistent with national GHG-inventory stock changes.
    • Sauzet et al. 2024 (commentary; full text) defend the ratio as a structural-vulnerability index and argue the benchmark alternative can rate depleted soils "good". The disagreement is unresolved.
  • Prout et al. 2022 themselves note that sesquioxide-dominated tropical soils may behave differently.
• Capacity and saturation.
  • Georgiou et al. 2022 (Nat. Commun.): agricultural soils are the most undersaturated relative to mineral capacity.
  • Begill et al. 2023 find no upper MAOC limit (189 German soils). Cotrufo et al. 2023 dispute this. Six et al. 2024 give rules and mineralogy-specific maxima (about 82 vs 46 g C per kg silt+clay). King et al. 2025 (Sci. Adv.) find new MAOC formation promoted by both existing MAOC and saturation deficit. Heinemann et al. 2024 find no saturation under extreme manure.
  • Contradiction flagged: saturation vs no saturation. This matters because capacity-normalised indices assume a ceiling.
  • Ling et al. 2025 (Nat. Commun.): 15 g C/kg switches how N fertiliser partitions new carbon (POC vs MAOC). This is a threshold in carbon dynamics, not in yield.
• Non-linear structural collapse. No study found demonstrates a yield or function "collapse" at a texture-normalised SOC threshold.
  • The closest is Tittonell & Giller 2013: degraded African soils become fertiliser-non-responsive. Vendig et al. 2023 also show benefits only at low SOC.
  • Both support diminishing returns / low-end sensitivity, not a sharp breakpoint at 1/13.
  • Meng et al. 2026 (Nat. Commun.) report a "threshold" in SOC change vs warming rate in Mollisol croplands, but it is a space-for-time statistical feature.

D. Earth system model soil carbon

• Evaluation.
  • Todd-Brown et al. 2013 (CMIP5): about 5-fold spread in global SOC. A simple NPP + temperature model explains 62–93% of ESM spatial SOC but only about 10% of observed SOC.
  • Varney et al. 2022 (CMIP6): still inconsistent. NPP is improved, but turnover time is not, and the NPP–SOC spatial correlation is unrealistically high.
  • Hartley et al. 2021: the observed temperature–SOC decline is about 3× stronger in coarse soils, a pattern an ESM could not reproduce.
  • Koven et al. 2017: temperature-sensitivity patterns are missed.
  • Georgiou et al. 2024: the "protected" pool fraction ranges 16–85% across models.
• Projections.
  • CMIP5: −72 to +253 PgC by 2100 (Todd-Brown 2014).
  • CMIP6: spread less than 50% of CMIP5, but partly an artefact of false priming (Varney 2023). CO2 (βs) dominates up to 2×CO2, and the climate term (γs) becomes increasingly uncertain at high CO2 (Varney 2024).
  • Data-constrained turnover reverses the sign to a loss of 19–45 PgC by 2100 (Ren et al. 2024).
  • Shi et al. 2018 and Wieder et al. 2013 show structural choice can flip the sign of the feedback.
• Cropland management. ESMs generally have no tillage, residue or fertiliser-driven SOC dynamics. Exceptions in DGVMs are LPJ-GUESS (Olin et al. 2015) and LPJmL5-tillage (Herzfeld et al. 2021). Cropland is lumped with urban land in the CMIP5 biome analysis (Todd-Brown 2013).
• Depth and units. cSoil is a whole-column stock (kg C m⁻²; depth varies by model, often about 1 m or full column). SOC:clay needs topsoil concentration (g/kg), which requires bulk density and depth harmonisation. This is a non-trivial conversion and a source of bias.
• Conclusion: the literature strongly supports using raw cSoil only as a sensitivity check.

E. Global gridded crop models (AgMIP/GGCMI)

• Soils in GGCMI.

  • Phase 1 used default vs harmonised management, including a no-N-stress setting; spin-up was not strictly harmonised (Elliott et al. 2015).

  • In Phase 2 (Franke et al. 2020), "soil treatment (which affects soil organic matter and carry-over effects…)" was not standardised. In LPJmL, LPJ-GUESS and APSIM, N mineralisation is part of SOM treatment and cannot be disabled. Carry-over effects, including residue management, differ across models. Soils come largely from HWSD.

  • Several GGCMs do simulate SOM dynamics: EPIC family, LPJmL5 (von Bloh et al. 2018), DSSAT-CENTURY (Basso et al. 2011 initialisation) and APSIM. Verified from Jägermeyr et al. 2021 (Methods) and the GGCMI phase 3 soil-input document:

    • Soils are one static, cropland-weighted dominant HWSD mapping unit per 0.5° cell (topsoil properties). pDSSAT uses GSDE and DSSAT-Pythia uses a global high-resolution profile database.
    • All models are spun up to soil and carbon-pool equilibrium.
    • Soil moisture and temperature are transient in most models.
    • EPIC-IIASA uses dynamic soils only during spin-up, then "static soil handling", i.e. annual re-initialisation of all soil attributes, including SOM fractions.
    • Management (fertiliser, land use, growing seasons, cultivars) is held at 2015 levels after 2015. Residue removal (70%) and tillage are "encouraged" but not followed by all teams, and erosion is excluded.
    • Mineral N comes from LUH2 (manure is grid-specific); all other nutrients are non-limiting.

    Whether SOM evolves in the other models is model-specific and not reported. Even where it does, it starts from equilibrium, not from the observed degraded state, and receives no degrading-management signal.

• Published projections with soil feedbacks.
  • Basso et al. 2018: 8 sites.
  • Balkovič et al. 2018: EPIC, Europe, +2 °C plus soil degradation.
  • Couëdel et al. 2024: 16 models disagree strongly on long-term SOC–yield feedbacks.
  • Couëdel et al. 2026: reset vs continuous runs. Fertility decline gives −24% yield, and the climate × fertility interaction is small.
  • Olin et al. 2015: LPJ-GUESS management.
  • IPCC AR6 WGII Ch. 5 cites only Basso et al. 2018 to quantify how SOC/N loss deepens warming impacts (wheat 9%→13%, maize 14%→19% at +3 °C). This shows the literature base is thin.
• Key risk. Folberth et al. 2016: uncertainty in which soil is assumed within a grid cell can exceed the climate-impact signal, especially in low-input systems. Müller et al. 2024 show large differences in yield sensitivity across GGCMs. A SOC-feedback signal may be smaller than these input and structural uncertainties.

F. Global digital soil mapping and deep learning

• Data and products.
  • WoSIS 2019 and 2023 provide standardised profiles.
  • SoilGrids 2.0 (Poggio et al. 2021): about 240k observations, quantile random forest uncertainty, 250 m, 6 depths.
  • SoilGrids250m (Hengl et al. 2017): random-CV R² 56–83%.
  • iSDAsoil (Hengl et al. 2021): 30 m maps for Africa.
• Deep learning.
  • Padarian et al. 2019: CNN with landscape context, −30% error (Chile).
  • Wadoux 2019: multivariate CNN with uncertainty.
  • GSoilCPM (Zhang et al. 2025): globally pretrained on 106,167 samples and fine-tuned regionally (−10.9% MAE).
  • Fourel et al. 2026 (full text; 16,514 Bavarian samples, longitude-blocked spatial CV): a compact gated CNN-Transformer (≈87k parameters, R² 0.375) matches a 45M-parameter Transformer (0.401). On cropland alone, gradient-boosted trees far outperform the attention models (R² ≈ 0.66 vs ≈ 0.21). The authors recommend land-use-stratified mapping.
  • Demattê et al. 2025 (STOTEN; full text): a global 30 m bare-soil spectral grid (GEOS3, Landsat 1984–2022) covering about 90% of agricultural bare soil. It is a strong candidate covariate for a cropland SOC/clay baseline.
  • Peter-Contesse et al. 2026: Landsat time-series transformer, leave-one-island-out R² 0.57.
  • Shi et al. 2026: discrepancies among global SOC maps.
• Uncertainty and validation. Ploton et al. 2020 (Nat. Commun.) show that random cross-validation can report >50% explained variance where spatial cross-validation shows near-zero skill. Global transformer claims must use spatial or area-of-applicability validation. Meyer & Pebesma 2021 (full text) define the "area of applicability", using a dissimilarity index to the training data in predictor space. Masking global predictions outside it is the recommended way to avoid over-claiming in data-sparse tropical regions.
• Assessment. Fourel et al. 2026 is the most important new finding for step 1: for cropland SOC specifically, tree ensembles may beat transformers under spatial validation. The baseline should therefore benchmark the transformer against gradient-boosted trees and SoilGrids on cropland-only spatial folds before committing to it.
  • A global transformer baseline is feasible and builds on the existing transformer work.
  • Reviewers at Nature-family journals will ask what the transformer adds over SoilGrids for this question. The answer has to be calibrated uncertainty and threshold-exceedance probabilities, not marginal accuracy gains.
  • Clay prediction quality in tropical soils (see I) is likely the binding constraint, not SOC.

G. Process and matrix soil-carbon models under climate projections

• Global offline runs.
  • Gottschalk et al. 2012 (RothC driven by multiple GCMs): global SOC increases by 2100, driven by NPP.
  • Herzfeld et al. 2021 (LPJmL5-tillage): cropland SOC declines in all management scenarios under both RCP2.6 and RCP8.5.
  • Karstens et al. 2022 (IPCC tier-2 type, 0–30 cm): cropland SOC debt grew from 34.6 to 39.6 GtC over 1975–2010. It provides a global management dataset (residues, manure, irrigation, tillage).
  • Morais et al. 2019: RothC across about 17,000 regions. Attainable SOC depends strongly on residues and irrigation.
• Regional runs.
  • Wiesmeier et al. 2016 (Bavaria): −11 to −16% SOC by 2100 with constant inputs, and −3 to −8% even with +20% inputs.
  • Riggers et al. 2021: the carbon inputs needed to hold German cropland SOC.
• Structural uncertainty.
  • Wieder et al. 2017 testbed: models with similar initial stocks diverge in their turnover–temperature responses.
  • Millennial V2 (Abramoff et al. 2022) and MIMICS offer measurable pools.
  • The CLM5 matrix approach (Lu et al. 2020) enables traceability and fast spin-up.
• Key contradiction: the sign of 21st-century cropland SOC change depends mainly on the carbon-input trajectory (yield and residue return) assumed. This creates a circularity for the proposed study: yields drive inputs, and inputs drive SOC. The analysis needs either a coupled iteration or explicit scenarios that bracket inputs.

H. Time of emergence (ToE)

• Methods. Signal-to-noise ToE (Hawkins & Sutton 2012; Hawkins et al. 2020), with low-latitude early emergence (Mahlstein et al. 2011). The agricultural applications found are emergence of precipitation change over crop areas (Rojas et al. 2019) and emergence of crop-yield impacts before 2040 (Jägermeyr et al. 2021).
• Application to soils: none found. The only soil-adjacent ToE work located concerns soil moisture (theses and conference work; not peer-reviewed journal items in this list).
• Conceptual mismatch. Classic ToE measures a forced signal against internal climate variability. SOC has little year-to-year noise, and its "noise" is measurement, spatial and model uncertainty. Recommendation: define "time of threshold crossing" probabilistically, e.g., the year when P(SOC:clay < 1/13 | baseline and trajectory uncertainty) > 0.5 or 0.66. Present it as an adaptation of ToE, not ToE itself.

I. Tropical and highly weathered soils

• Evidence that clay alone is a poor normaliser:

  • Rasmussen et al. 2018 ("Beyond clay"): Fe/Al oxides and exchangeable Ca often predict SOM better than clay.
  • Doetterl et al. 2015: geochemistry × climate.
  • Torn et al. 1997: non-crystalline minerals control SOC in Hawaiian volcanic soils.
  • Huygens et al. 2005: Al-humus-clay complexes in Andisols.
  • von Fromm et al. 2021 (AfSIS, 1,601 samples): oxalate Al/Fe and exchangeable Ca are as important as climate.
  • Barthès et al. 2008 and Feller & Beare 1997: sesquioxide aggregation in tropical soils.

• Partial support: Quesada et al. 2020 find clay explains SOC within kaolinitic Ferralsols and Acrisols, but not in less-weathered soils.

• Measurement problem. In oxide-rich Ferralsols, "pseudo-sand" (stable microaggregates) leads to incomplete dispersion and underestimated clay. That inflates SOC:clay and would misclassify degraded soils as healthy. Primary literature:

  • Bartoli et al. 1991 (Geoderma): a dedicated Na-resin disaggregation protocol is needed for Oxisols (title-level only; verify).
  • Muggler et al. 1997 (Soil Sci.): laser and pipette clay correlate but are "never" 1:1 in deeply weathered Brazilian profiles. There, aggregation was tied to organic matter rather than sesquioxides, which nuances the pseudo-sand mechanism.
  • Neufeldt et al. 2002 (Geoderma): texture and land-use effects on SOM in Cerrado Oxisols.

  Implication: clay values in global databases (WoSIS, SoilGrids) are method-dependent in Oxisols. Clay uncertainty must be propagated into any SOC:clay crossing, and at least one clay-free index should be run in parallel.

• Mineralogy-specific capacity: about 46 vs 82 g C per kg silt+clay for 1:1 vs 2:1 clays (Six et al. 2024). A single 1/13 line therefore cannot be transferred.

• Recommendation: stratify thresholds by clay mineralogy and weathering (e.g., WRB group or a kaolinite/oxide proxy), or use the observed-vs-expected benchmark approach (Poeplau & Don 2023; Feeney et al. 2024), with SOC:clay as one sensitivity.

J. Land degradation and food security

• IPCC SRCCL Ch. 4 (Olsson et al. 2019): degradation affects about a quarter of ice-free land (medium confidence).
• IPCC AR6 WGII Ch. 5:
  • Notes that models increasingly try to include soil and management interactions.
  • Quantifies the SOC/N feedback only via Basso et al. 2018.
  • Cites Iizumi & Wagai 2019 and Kane et al. 2021 for SOC drought buffering.
  • This supports the claim that the global projection literature treats soil degradation qualitatively.
• Global erosion and nutrient depletion:
  • Borrelli et al. 2020 (PNAS, full text): current water erosion about 43 Pg/yr, projected to rise +30 to +66% by 2070 depending on SSP-RCP.
  • Borrelli et al. 2017: land-use-change effects on erosion.
  • Alewell et al. 2020: erosion is more than 50% of phosphorus losses.
  • Montgomery 2007; Evans et al. 2020 (soil lifespans).
• SOC debt: Sanderman et al. 2017 (133 Pg C in the original paper, revised to 116 Pg C in the 2018 correction, doi 10.1073/pnas.1800925115) and Villela et al. 2026 (Brazil).
• Context and framing: Amundson et al. 2015 (Science review); Lal 2004 (Science); Bossio et al. 2020; Amelung et al. 2020.
• Gap: degradation framing is strong, but it is rarely translated into a quantified, spatially explicit term in yield projections.

3. Positioning against Nature/Science literature

3.1 Relevant Nature/Science-family papers

The first table covers papers that overlap or compete directly with the proposed study. The second covers papers relevant to methods or context. Evidence basis: FT = full text read; Abs = abstract or publisher page only.

Direct overlap or competition

Methods and context

3.2 What the proposed study would add beyond the closest papers

3.3 Would the expected result meet the bar?

Stated criteria (current journal pages, accessed 18 Oct 2026):

• Nature Food
  • Article: a substantial novel study. 3,000 words, 150-word abstract, 6 display items, about 50 references.
  • Analysis: a new analysis of existing data leading to "novel and arresting conclusions of importance to a broad audience". Same limits.
  • Scope explicitly includes soil and water management, food security and crop science.
• Nature Climate Change
  • Article and Analysis: same 3,000-word, 6-item format.
  • Aims to publish "the most significant and cutting-edge research" on climate impacts and their implications.
  • Also offers Registered Reports, an option worth considering given the null-result risk.
• Science
  • Source: the live "Information for authors" page, as seen in a browser on 19 Oct 2026. Automated access was blocked.
  • There is no longer a separate "Report" format. The only peer-reviewed format for original research is the Research Article. The old short Reports (about 2,500 words) now fall under it.
  • Research Article (standard):
    • Main text up to 3,000 words, divided into sections with brief subheadings.
    • An abstract, 3–5 display items and about 50 main-text references.
    • Methods go in supplementary materials, in enough detail to replicate the study.
    • All data must be in the paper or in a public repository (Dryad partnership).
  • Research Article (print fit): papers with more than about 4 figures appear in full online. Print carries a one-page structured summary (Introduction, Rationale, Results, Conclusion; about 500 words) and one summary figure.
  • Research Article (extended online format): granted for "a few" papers. It allows up to 6,000 words, 6 display items and 100 references, and the cover letter must justify the extra length.
  • Other peer-reviewed formats:
    • Analytical Reviews: up to 6,000 words and 200 references; may re-analyse published data, but formal meta-analyses count as original research.
    • Policy Articles: 2,000–3,000 words, with new data and policy implications.
  • Stated selection criteria: papers that are "influential in their fields or across fields". They should present "novel and broadly important data, syntheses, or concepts".
    • About 2 weeks to a triage decision. Selected papers get at least two outside reviewers, followed by cross-review.
    • Rejected papers cannot be resubmitted over disagreements about novelty, and declined papers may be transferred to Science Advances.
  • Implication for this study: a Science submission would be a standard 3,000-word Research Article. The extended format is unlikely to be granted for an analysis of this kind. The realistic Science-family fallback is a transfer to Science Advances.

Realistic assessment:

• A Nature Food Analysis is plausible only if the result is spatially explicit, robust across models, and shows a non-trivial omitted term concentrated where food insecurity is highest. A result like "projections underestimate yield losses by X% in low-input tropical systems, negligible in breadbaskets" would fit.
• A uniformly small global term would most likely be redirected to Nature Communications or Communications Earth & Environment.
• Nature Climate Change requires the climate-impact angle to be central (not just soils). Qiao 2022 shows the venue is open to it, so the global version would be judged against that precedent.
• Science is unlikely unless the term is large and robust globally, which current evidence (Couëdel's small interaction; Ma's small marginal effect) makes improbable.

4. Gap analysis

5. Feasibility assessment

5.1 Data availability

5.2 Methodological risks

1. CMIP6 baseline bias. ESMs misrepresent turnover and lack cropland management. Mitigation: use offline models only, with at least 2 structures (RothC + Millennial V2, possibly a matrix/CLM5-type emulator). Constrain them with Ren et al. 2024–style turnover data and bracket carbon-input scenarios.
2. Input–yield circularity. SOC depends on residue inputs, which depend on yield. Mitigation: iterate yield→input→SOC coupling, or run bracketing input scenarios (constant, ±20%, as in Wiesmeier 2016).
3. Scale mismatch. Point SOC is converted to 0.5° yields. Mitigation: aggregate threshold-exceedance probabilities, not mean SOC, to grid cells, and use crop-area weighting.
4. Causality of SOC×climate sensitivity. Mitigation:
   • Use multiple sensitivity sources: observational panels (King 2026 to avoid drought overstatement; Kane 2021; Qiao 2022), within-trial long-term-experiment estimates, and crop-model reset/continuous contrasts.
   • Report the spread across them. Use within-unit panel designs (county or grid fixed effects) rather than cross-sectional partial dependence.
   • Explicitly address the trial-vs-farm gap (Zhao 2026).
5. Threshold transferability. 1/13 lacks yield validation and is biased in high-clay and tropical soils. Mitigation:
   • Use continuous SOC states rather than binary thresholds for yield.
   • Report crossing times for multiple indices: SOC:clay; observed/expected benchmark (Poeplau & Don; Feeney); MAOC-capacity fraction (Georgiou 2022).
   • Stratify by mineralogy.
6. Model-structural uncertainty in crop models. Couëdel 2024 shows 16 models disagree. Mitigation: a multi-model or emulator approach; avoid one crop model.
7. Soil-input uncertainty swamping the signal (Folberth 2016). Mitigation: propagate baseline uncertainty from the transformer or SoilGrids quantiles and test whether the omitted term exceeds it.
8. Circularity in using crop-model-derived sensitivities to critique crop models (He 2025). Mitigation: keep the observational and model pathways separate.

5.3 Likelihood of a null result, and its value

• Probability that the global, production-weighted omitted term is small (less than about 10% of projected climate loss): moderately high (rough estimate 50–70%). This is a judgement from the evidence above, not a calculation.
  • Basso: about 4.5 pp on top of 9–14% losses.
  • Qiao: 14% exacerbation nationally.
  • Couëdel: small interaction.
  • Ma, Hijbeek, Xu: SOC effects are damped by N.
  • Most global production comes from high-input systems.
• Probability that it is material regionally (e.g., low-input SSA, South Asian drylands, degraded tropical croplands): high. Couëdel's 20–50% decline in low-input systems over decades supports this.
• Value of a null result:
  • It sets a quantitative upper bound on a widely asserted but untested claim, which is useful for IPCC AR7 WGII, where the base is a single site study.
  • It shows that GGCMI-based projections are robust to soil omission in breadbaskets.
  • It redirects attention to where the term matters.
  • It provides a reusable framework (decomposition + threshold crossing).
  • Registered Report formats (Nature Climate Change) and Communications Earth & Environment, Earth's Future or ERL value such bounds.

6. Recommended journal targets (ranked)

1. Nature Food (Analysis). Best scope fit: soil management, food security and yield projections. Jägermeyr 2021, Zhao 2026 and Xu 2025 show it publishes both "projections" and "reality-check" soil–yield papers. Requires a clear "where it matters for food security" result.
2. Nature Communications. Realistic primary target for a robust global result of any sign. It has published Ren 2024, Hartley 2021, Georgiou 2022, Ling 2025 and Meng 2026 on soil carbon, and Folberth 2016 and Kornhuber 2023 on crop-projection biases.
3. Nature Climate Change (Article, or Registered Report). If the climate-impact bias is the headline. Must clearly exceed Qiao 2022 in scope. The Registered Report route reduces null-result risk.
4. Communications Earth & Environment / Earth's Future. Strong fallback for a bounded or regional result. Earth's Future published Balkovič 2018 on climate + soil degradation and the 2026 transformer paper.
5. Global Change Biology. Published Couëdel 2026 and King 2026. Good for a mechanistic emphasis.
6. Companion methods papers: SOIL or Geoderma (threshold transferability and tropical normalisation); GMD or Biogeosciences (offline SOC model ensemble under ISIMIP3b).
7. Science (Research Article, 3,000 words): only if the global term is large, robust and surprising. Current evidence makes this unlikely. If submitted, opt in to transfer to Science Advances, which is a realistic home for a robust result of any sign.

7. Design recommendations arising from the literature

1. Adopt Couëdel et al.'s reset vs continuous decomposition globally. Report BL_cont − BL_reset (degradation without climate change), CC_reset − BL_reset (climate only) and the CC×SF interaction. The proposed "underestimation term" is the sum of the last two parts.
2. Benchmark explicitly against GGCMI3 (Jägermeyr 2021). The protocol is verified: equilibrium spin-up, 2015 management fixed, static HWSD soils, annual SOM reset in EPIC-IIASA. The remaining step is to obtain per-model SOM settings from the modelling groups, since Jägermeyr reports them only for EPIC-IIASA.
3. Use ≥2 soil-carbon model structures (RothC + Millennial V2) with bracketed carbon-input scenarios. Use raw CMIP6 cSoil only as a sensitivity check (as planned).
4. Replace a single 1/13 line with continuous states plus multiple normalisations, stratified by mineralogy.
5. Use King 2026 and Kane 2021 together to bracket drought buffering. Do not assume buffering is largest at extremes.
6. Validate the transformer baseline with spatial CV and area-of-applicability analysis (Meyer & Pebesma 2021). Benchmark it against gradient-boosted trees on cropland-only folds (Fourel et al. 2026), consider the GEOS3 bare-soil spectral grid as a covariate (Demattê et al. 2025), and emphasise calibrated probabilities.
7. Pre-register hypotheses and the null-result interpretation (Registered Report option at NCC).

8. Contradictions between studies (explicit)

9. Press versus paper (where checked)

10. Established details and remaining evidence gaps

Established details:

• Hijbeek "farmer-perceived thresholds": this is Hijbeek et al. 2017, Ecological Indicators 83:390–403 (verified via official abstract; §1, §2C, §8).
• GGCMI phase 3 soil treatment: verified from the Jägermeyr et al. 2021 Methods and the GGCMI phase 3 soil-input document (isimip.org/documents/568). Per-model SOM settings beyond EPIC-IIASA are not published and need contact with the modelling groups. That is a design task for the study, not a literature gap.
• No separate GGCMI phase 3 protocol paper exists. The protocol is documented in the Jägermeyr et al. 2021 Methods and on isimip.org.
• Additional sources considered: Meyer & Pebesma 2021, Porwollik et al. 2019, MIRCA2000, SPAM2010, Lugato et al. 2013/2014, and Oxisol clay-dispersion literature (Bartoli 1991, Muggler 1997, Neufeldt 2002).
• AgMIP soil–crop intercomparisons: Couëdel et al. 2024 and 2026 are the AgMIP soil–crop model intercomparison. No global equivalent was found, which is itself part of the gap.
• Source corrections and reading depth: the corrected Sanderman soil carbon debt is 116 Pg C, rather than the original 133 Pg C. Borrelli 2020 is quoted from full text. Bibliographic annotations distinguish full-text, abstract, secondary-source, and metadata-only evidence.

Not examined in full at the time of this assessment (9 papers):

1. Priority 1: Zhao et al. 2026 (Nat. Food) and Hijbeek et al. 2017 (Ecol. Indic.). Annotated from their official abstracts; the key numbers quoted in this report come from those abstracts.
2. Priority 2: Bradford et al. 2019 and Wuepper et al. 2019 (Nat. Sustain.), and Hijbeek 2026 (Nat. Food News & Views).
3. Supporting: Bartoli 1991, Muggler 1997, Neufeldt 2002 (Oxisol clay methods), and Portmann 2010 (MIRCA2000; a dataset description).

These full-text gaps remain qualifications on this assessment. The later historical review states where additional full texts were examined.

Appendix A. Annotated bibliography

1. Lal (2005). Enhancing crop yields in the developing countries through restoration of the soil organic carbon pool in agricultural lands

   Land Degradation & Development · 2005 · DOI · Full text and figures

   Study: Review

   Scale: Developing countries

   Finding: +1 Mg C/ha root-zone SOC → +20–70 kg/ha wheat, 10–50 rice, 30–300 maize.

   Relevance: supports (widely cited coefficients; weak causal basis)

   Evidence: full text

2. Janzen (2006). The soil carbon dilemma: Shall we hoard it or use it?

   Soil Biology and Biochemistry · 2006 · DOI · Full text and figures

   Study: Perspective

   Scale: Global

   Finding: SOC benefits come from decomposition (use) vs storage (hoard) — trade-off.

   Relevance: context

   Evidence: full text

3. Pan et al. (2009). The role of soil organic matter in maintaining the productivity and yield stability of cereals in China

   Agriculture, Ecosystems & Environment · 2009 · DOI · Full text and figures

   Study: Observational (long-term data)

   Scale: China

   Finding: Higher SOM associated with higher productivity and yield stability of cereals.

   Relevance: supports

   Evidence: full text

4. Tittonell & Giller (2013). When yield gaps are poverty traps: The paradigm of ecological intensification in African smallholder agriculture

   Field Crops Research · 2013 · DOI · Full text and figures

   Study: Review/conceptual

   Scale: sub-Saharan Africa

   Finding: Degraded, low-SOC soils become non-responsive to fertilizer (poverty traps).

   Relevance: supports (non-linear degradation in tropics)

   Evidence: full text

5. Hijbeek et al. (2016). Do organic inputs matter – a meta-analysis of additional yield effects for arable crops in Europe

   Plant and Soil · 2016 · DOI · Full text and figures

   Study: Meta-analysis of 20 LTEs

   Scale: Europe

   Finding: Additional yield effect of organic inputs +1.4±1.6% (n.s.) when NPK sufficient; positive for roots/tubers, spring cereals, sandy soils, wet climates.

   Relevance: undermines (SOC effect largely substitutable by N in well-fertilized systems)

   Evidence: full text

6. Hijbeek (2017). On the role of soil organic matter for crop production in European arable farming

   PhD thesis, Wageningen University · 2017 · DOI · Full text and figures

   Study: PhD thesis

   Scale: Europe

   Finding: Synthesis of SOM role in European arable yields; additional yield effects of SOM small when nutrients sufficient.

   Relevance: undermines / context

   Evidence: full text

7. Hijbeek et al. (2017). Do farmers perceive a deficiency of soil organic matter? A European and farm level analysis

   Ecological Indicators · 2017 · DOI

   Study: Farmer survey (1,452 farmers) + EU risk indicator

   Scale: Europe (5 countries)

   Finding: 18% of farmers perceive high/very high SOM deficiency; farmer-perceived threshold intervals: sand 1.2–4.7%, loam 0.6–2.6%, clay 1.0–2.4% SOM; risk indicator: 7–37% of EU agricultural land at high risk; weak correlation (0.15–0.18) with perceptions.

   Relevance: methods / contradiction (higher thresholds for sand vs SOC:clay logic)

   Evidence: abstract (WUR research portal)

8. Hijbeek et al. (2018). What drives farmers to increase soil organic matter? Insights from the Netherlands

   Soil Use and Management · 2018 · DOI · Full text and figures

   Study: Farmer survey (435 farmers, theory of planned behaviour)

   Scale: Netherlands

   Finding: 90% of Dutch arable farmers intend to raise SOM; organic inputs used more but no perceived SOM increase; Manure Act, input costs, profitable crops impede.

   Relevance: context

   Evidence: full text

9. Poulton et al. (2018). Major limitations to achieving “4 per 1000” increases in soil organic carbon stock in temperate regions: Evidence from long‐term experiments at Rothamsted Research, United Kingdom

   Global Change Biology · 2018 · DOI · Full text and figures

   Study: Long-term experiments (16 LTEs, 114 comparisons)

   Scale: UK (Rothamsted, Woburn, Saxmundham)

   Finding: SOC rises >7‰/yr in 65% of comparisons but mainly with very large FYM inputs; severe limitations to '4 per 1000' in practice.

   Relevance: context (SOC change needs large inputs)

   Evidence: full text

10. Oldfield et al. (2019). Global meta-analysis of the relationship between soil organic matter and crop yields

    SOIL · 2019 · DOI · Full text and figures

    Study: Meta-analysis (observational, cross-study regression)

    Scale: Global

    Finding: Maize/wheat yields rise with SOC and level off near 2% SOC; ~2/3 of maize/wheat land below 2%; potential yield gains +10% (maize), +23% (wheat).

    Relevance: supports (curve shape) / undermines (plateau limits feedback size above 2%)

    Evidence: full text

11. Hijbeek et al. (2019). Use of organic inputs by arable farmers in six agro-ecological zones across Europe: Drivers and barriers

    Agriculture, Ecosystems & Environment · 2019 · DOI · Full text and figures

    Study: Farmer survey (1,263 farmers)

    Scale: Europe, 6 agro-ecological zones

    Finding: Drivers/barriers to organic inputs; perceived soil quality effects are main drivers.

    Relevance: context (farmer-perceived SOM thresholds — verify exact threshold reporting)

    Evidence: full text

12. Bradford et al. (2019). Soil carbon science for policy and practice

    Nature Sustainability · 2019 · DOI

    Study: Perspective

    Scale: Global

    Finding: Soil carbon science for policy: SOC–yield link context-dependent; target setting.

    Relevance: framing

    Evidence: abstract

13. Vonk et al. (2020). European survey shows poor association between soil organic matter and crop yields

    Nutrient Cycling in Agroecosystems · 2020 · DOI · Full text and figures

    Study: Farm survey (observational)

    Scale: Europe, 6 countries (N=1,264)

    Finding: No consistent SOM–yield correlation; wheat positive in Continental, negative in Atlantic climate.

    Relevance: undermines

    Evidence: full text

14. Lal (2020). Soil organic matter content and crop yield

    Journal of Soil and Water Conservation · 2020 · DOI · Full text and figures

    Study: Review

    Scale: Global

    Finding: Argues SOM–yield link; summary of evidence.

    Relevance: supports (narrative)

    Evidence: full text

15. Qiao et al. (2022). Soil quality both increases crop production and improves resilience to climate change

    Nature Climate Change · 2022 · DOI · Full text and figures

    Study: Observational panel + gradient-boosted regression trees + CMIP5 projection

    Scale: National (China, 12,115 site-years)

    Finding: High-quality soils (integrated soil-quality index incl. SOM, Olsen-P etc.) raise mean yield 10.3% and cut variability 15.6%; under RCP8.5 2080–2099 climate-and-soil interactions add 14% to the 11.4 Mt/yr cereal loss when all farmland is uniformly degraded to low quality, and cut it 21% under uniform improvement. Uses CMIP5, stylised (not process-modelled) degradation.

    Relevance: supports — closest precedent (national, stylised degradation)

    Evidence: full text

16. Ma et al. (2023). Global crop production increase by soil organic carbon

    Nature Geoscience · 2023 · DOI · Full text and figures

    Study: Meta-analysis of 13,662 controlled trials; random forest + causal forest

    Scale: Global

    Finding: Yield–SOC plateau optima 43.2–43.9 g/kg maize, 12.7–13.4 wheat, 31.2–32.4 rice; SOC one-fifth as effective as N; optimum SOC +4.3% production, current practices +0.7%; causal-forest adjustment for confounders; single-season trials.

    Relevance: supports (SOC–yield link) / undermines (small marginal effect)

    Evidence: full text

17. Vendig et al. (2023). Quantifying direct yield benefits of soil carbon increases from cover cropping

    Nature Sustainability · 2023 · DOI · Full text and figures

    Study: Meta-analysis (434 paired obs)

    Scale: Global

    Finding: Direct SOC yield benefit only where initial SOC < 9.2 g/kg; not substitutable by N.

    Relevance: supports (threshold behaviour at low SOC)

    Evidence: full text

18. Moinet et al. (2023). Carbon for soils, not soils for carbon

    Global Change Biology · 2023 · DOI · Full text and figures

    Study: Review/critique

    Scale: Global

    Finding: Saturation cuts SOC mitigation 53–81%; yield effects across 21 meta-analyses inconsistent.

    Relevance: undermines

    Evidence: full text

19. Beillouin et al. (2023). A global meta-analysis of soil organic carbon in the Anthropocene

    Nature Communications · 2023 · DOI · Full text and figures

    Study: Second-order meta-analysis

    Scale: Global

    Finding: Cropland conversion causes large SOC loss; indirect climate effects > direct warming.

    Relevance: framing / supports

    Evidence: full text

20. Kätterer & Bolinder (2024). Response of maize yield to changes in soil organic matter in a Swedish long‐term experiment

    European Journal of Soil Science · 2024 · DOI · Full text and figures

    Study: Long-term experiment

    Scale: Sweden (Ultuna)

    Finding: Maize yield +16% per % SOC where not N- or pH-limited.

    Relevance: supports

    Evidence: full text

21. Yang et al. (2024). Diversifying crop rotation increases food production, reduces net greenhouse gas emissions and improves soil health

    Nature Communications · 2024 · DOI · Full text and figures

    Study: Field experiment (6 yr)

    Scale: North China Plain

    Finding: Diversified rotations +38% equivalent yield, SOC +8%.

    Relevance: context

    Evidence: full text

22. Xu et al. (2025). Soil health contributes to variations in crop production and nitrogen use efficiency

    Nature Food · 2025 · DOI · Full text and figures

    Study: Global 0.5° structural equation modelling (19 indicators, cross-sectional)

    Scale: Global

    Finding: Soil health explains ~12% of crop production and 22% of NUE variation; management ~70%; weaker soil influence where fertilizer intensive; authors flag cross-sectional design limits projections.

    Relevance: undermines (soil a modest share of yield variance)

    Evidence: full text

23. Laub et al. (2025). Maize yield responses to soil organic carbon under integrated soil fertility management in tropical environments

    Agronomy for Sustainable Development · 2025 · DOI · Full text and figures

    Study: Long-term experiments (4)

    Scale: Kenya

    Finding: Yield and SOC trends strongly coupled at adequate-rainfall sites (74–84% variance), not at dry sites.

    Relevance: supports (tropics) / nuance (rainfall-conditional)

    Evidence: full text

24. Kan et al. (2025). Soil carbon accrual and crop production enhanced by sustainable subsoil management

    Nature Geoscience · 2025 · DOI · Full text and figures

    Study: 15-yr field experiment + meta-analysis

    Scale: China (rice–wheat)

    Finding: Ditch-buried straw: yield +15%, SOC 0–40 cm +46%.

    Relevance: context (management can reverse trajectory)

    Evidence: full text

25. Zhao et al. (2026). The beneficial relation between soil organic carbon and maize yield in field trials does not translate to real-world farms in Northeast China

    Nature Food · 2026 · DOI · Full text and figures

    Study: Causal inference (trials vs farms)

    Scale: Regional (NE China)

    Finding: SOC–yield benefit in trials (+8–13% to 30–40 g/kg) absent on farms (≤5% to 10 g/kg then decline) due to excess N, knowledge, infrastructure.

    Relevance: undermines (trial-derived sensitivities may not scale to real farms)

    Evidence: abstract (paywalled)

26. Hijbeek (2026). Targeting yield constraints with soil carbon

    Nature Food · 2026 · DOI · Full text and figures

    Study: News & Views

    Scale: —

    Finding: Hijbeek commentary: target yield constraints with soil carbon.

    Relevance: framing (counterpoint)

    Evidence: page

27. Schlenker & Roberts (2009). Nonlinear temperature effects indicate severe damages to U.S. crop yields under climate change

    Proceedings of the National Academy of Sciences · 2009 · DOI · Full text and figures

    Study: Statistical (panel)

    Scale: USA

    Finding: Yields rise with temperature to 29 °C (corn), 30 °C (soy), 32 °C (cotton) and fall steeply above.

    Relevance: methods (breakpoint framework)

    Evidence: full text

28. Lobell et al. (2011). Climate Trends and Global Crop Production Since 1980

    Science · 2011 · DOI · Full text and figures

    Study: Statistical

    Scale: Global

    Finding: Climate trends since 1980 reduced maize and wheat production.

    Relevance: context

    Evidence: full text

29. Ray et al. (2015). Climate variation explains a third of global crop yield variability

    Nature Communications · 2015 · DOI · Full text and figures

    Study: Statistical

    Scale: Global

    Finding: Climate variability explains ~32–39% of yield variability.

    Relevance: context

    Evidence: full text

30. Williams et al. (2016). Soil Water Holding Capacity Mitigates Downside Risk and Volatility in US Rainfed Maize: Time to Invest in Soil Organic Matter?

    PLOS ONE · 2016 · DOI · Full text and figures

    Study: Observational (county 2000–2014)

    Scale: USA (4 states)

    Finding: Soil water-holding capacity lowers maize yield volatility and downside risk.

    Relevance: supports

    Evidence: full text

31. Lesk et al. (2016). Influence of extreme weather disasters on global crop production

    Nature · 2016 · DOI · Full text and figures

    Study: Statistical (1964–2007)

    Scale: Global (national data)

    Finding: Droughts and extreme heat reduced national cereal production 9–10%; no detectable effect of floods or extreme cold.

    Relevance: context

    Evidence: full text

32. Minasny & McBratney (2017). Limited effect of organic matter on soil available water capacity

    European Journal of Soil Science · 2017 · DOI · Full text and figures

    Study: Meta-analysis + 50,000 measurements

    Scale: Global

    Finding: +1% OC raises AWC only ~1.16 mm/100 mm; small effect.

    Relevance: undermines (water mechanism weak)

    Evidence: full text

33. Zhao et al. (2017). Temperature increase reduces global yields of major crops in four independent estimates

    Proceedings of the National Academy of Sciences · 2017 · DOI · Full text and figures

    Study: Multi-method synthesis

    Scale: Global

    Finding: Each °C warming reduces global yields: wheat −6.0%, rice −3.2%, maize −7.4%, soybean −3.1%.

    Relevance: baseline sensitivities

    Evidence: full text

34. Iizumi & Wagai (2019). Leveraging drought risk reduction for sustainable food, soil and climate via soil organic carbon sequestration

    Scientific Reports · 2019 · DOI · Full text and figures

    Study: Statistical on global gridded data

    Scale: Global

    Finding: Small topsoil OC increase raises drought tolerance on 70% of harvested area; +16% output in drought years.

    Relevance: supports

    Evidence: full text

35. Deng et al. (2020). Soil indigenous nutrients increase the resilience of maize yield to climatic warming in China

    Environmental Research Letters · 2020 · DOI · Full text and figures

    Study: Observational (county 1980–2010)

    Scale: China

    Finding: Maize −2.6%/°C; soil indigenous N/P/K explain regional warming response.

    Relevance: supports (soil modulates warming response)

    Evidence: full text

36. Rigden et al. (2020). Combined influence of soil moisture and atmospheric evaporative demand is important for accurately predicting US maize yields

    Nature Food · 2020 · DOI · Full text and figures

    Study: Observational (SMAP)

    Scale: US Midwest

    Finding: Soil moisture × VPD interaction predicts maize damage; ignoring soil moisture doubles projected damage.

    Relevance: methods (water-supply pathway)

    Evidence: full text

37. Kane et al. (2021). Soil organic matter protects US maize yields and lowers crop insurance payouts under drought

    Environmental Research Letters · 2021 · DOI · Full text and figures

    Study: Observational (12,376 county-years)

    Scale: USA

    Finding: Under severe drought +1% SOM → +2.2 Mg/ha maize and −36% insurance payouts; effect weaker in clay soils.

    Relevance: supports (SOC × drought)

    Evidence: full text

38. Renwick et al. (2021). Long-term crop rotation diversification enhances maize drought resistance through soil organic matter

    Environmental Research Letters · 2021 · DOI · Full text and figures

    Study: Long-term experiment

    Scale: USA

    Finding: Rotation diversification improves maize drought resistance partly via SOM.

    Relevance: supports (mechanism)

    Evidence: full text

39. Nouri et al. (2021). Conservation agriculture increases the soil resilience and cotton yield stability in climate extremes of the southeast US

    Communications Earth & Environment · 2021 · DOI · Full text and figures

    Study: Long-term experiment (29 yr)

    Scale: Tennessee, USA

    Finding: No-till improves cotton yield stability under climate extremes via structure and SOC.

    Relevance: supports

    Evidence: full text

40. Ortiz-Bobea et al. (2021). Anthropogenic climate change has slowed global agricultural productivity growth

    Nature Climate Change · 2021 · DOI · Full text and figures

    Study: Econometric

    Scale: Global

    Finding: Anthropogenic climate change reduced agricultural TFP ~21% since 1961.

    Relevance: framing

    Evidence: full text

41. Feng et al. (2022). Soil properties resulting in superior maize yields upon climate warming

    Agronomy for Sustainable Development · 2022 · DOI · Full text and figures

    Study: Observational

    Scale: China

    Finding: Soil properties explain ~72% of spatial variation in maize T-sensitivity; SOC largest contributor (28.9%).

    Relevance: supports

    Evidence: full text

42. Deng et al. (2023). Building soil to reduce climate change impacts on global crop yield

    Science of The Total Environment · 2023 · DOI · Full text and figures

    Study: Random forest on global gridded data

    Scale: Global

    Finding: Per °C warming: maize −3.4%, wheat −2.4%, rice −0.3%, soy −5.0%; SOC dominant explainer of warming-response heterogeneity; medium SOC sequestration avoids ~3–5%/°C; irrigation not modelled (data limitation).

    Relevance: supports / overlaps strongly (improvement direction)

    Evidence: full text

43. Hendrawan et al. (2023). Possible factors determining global-scale patterns of crop yield sensitivity to drought

    PLOS ONE · 2023 · DOI · Full text and figures

    Study: Machine learning on global gridded data

    Scale: Global (0.5°)

    Finding: Crop drought sensitivity explained by precipitation, soil WHC, acidity, PET, fertilizer, season — SOC not top driver.

    Relevance: mixed

    Evidence: full text

44. Mahmood et al. (2023). Soil health explains the yield-stabilizing effects of soil organic matter under drought

    Soil & Environmental Health · 2023 · DOI · Full text and figures

    Study: Observational (county-level rainfed corn 2000–2016; SHAPE soil health score)

    Scale: USA

    Finding: Site-specific soil health score (from SOM, adjusted for climate, texture, suborder): +0.5 SHS ≈ +1.15±0.18 Mg/ha corn under the most severe drought; predictive power rises with drought severity; effect largely independent of texture and suborder.

    Relevance: supports (SOC×drought) / contradicts King 2026 at extremes

    Evidence: full text

45. Teng et al. (2024). Conservation agriculture improves soil health and sustains crop yields after long-term warming

    Nature Communications · 2024 · DOI · Full text and figures

    Study: Long-term warming experiment

    Scale: North China Plain

    Finding: Conservation agriculture improves soil health +21%, sustains yields under 8-yr warming.

    Relevance: supports (management × warming)

    Evidence: full text

46. He et al. (2025). Soil organic carbon improvement for mitigating crop yield losses under global warming

    European Journal of Agronomy · 2025 · DOI · Full text and figures

    Study: Multi-model (9 GGCMs × 37 CMIP6 GCMs × 3 SSPs)

    Scale: Global

    Finding: 1980–2010 per °C: maize −2.3%, soybean +3.0%; 2050–2080: maize −6.6 to −7.5%, soybean −8.9 to −10.7% per °C; soil properties explain 51%/59% of spatial T-sensitivity, SOC strongest; SOC improvement reduces losses by 0.5–1.1 %/°C (maize), 1.3–2.5 %/°C (soy), less under SSP585.

    Relevance: overlaps strongly (step 4) / indicates modest buffering magnitude

    Evidence: full text

47. Feng et al. (2025). Globally increased cropland soil exposure to climate extremes in recent decades

    Nature Communications · 2025 · DOI · Full text and figures

    Study: Satellite time series

    Scale: Global croplands

    Finding: 86% of croplands increasingly exposed to climate extremes; higher-SOC areas more exposed.

    Relevance: supports (degradation risk)

    Evidence: full text

48. King et al. (2026). Soil Organic Carbon Improves Crop Yield and Yield Resilience

    Global Change Biology · 2026 · DOI · Full text and figures

    Study: Observational (67,000 county-years)

    Scale: USA

    Finding: SOC benefits largest at moderate water supply, not drought; cannot compensate extreme scarcity.

    Relevance: undermines (SOC×drought breakpoint assumption)

    Evidence: full text

49. Hassink (1997). The capacity of soils to preserve organic C and N by their association with clay and silt particles

    Plant and Soil · 1997 · DOI · Full text and figures

    Study: Data synthesis

    Scale: Temperate/tropical

    Finding: Capacity to preserve C relates to clay+silt content (Hassink equation).

    Relevance: methods

    Evidence: full text

50. Six et al. (2002). Stabilization mechanisms of soil organic matter: Implications for C-saturation of soils

    Plant and Soil · 2002 · DOI · Full text and figures

    Study: Conceptual review

    Scale: Global

    Finding: Stabilization mechanisms; C saturation framework.

    Relevance: methods

    Evidence: full text

51. Loveland (2003). Is there a critical level of organic matter in the agricultural soils of temperate regions: a review

    Soil and Tillage Research · 2003 · DOI · Full text and figures

    Study: Literature review

    Scale: Temperate (UK focus)

    Finding: Widely held 2% SOC (≈3.4% SOM) critical threshold: quantitative evidence 'slight'; possibly a desirable range, but little quantitative evidence that SOC decline reduces yields.

    Relevance: undermines (no evidence for critical threshold)

    Evidence: full text

52. Stewart et al. (2007). Soil carbon saturation: concept, evidence and evaluation

    Biogeochemistry · 2007 · DOI · Full text and figures

    Study: Concept/evidence review

    Scale: Global

    Finding: Soil C saturation concept and evidence.

    Relevance: methods

    Evidence: full text

53. Dexter et al. (2008). Complexed organic matter controls soil physical properties

    Geoderma · 2008 · DOI · Full text and figures

    Study: Theory + 4 databases (France, Poland)

    Scale: Europe

    Finding: Complexed organic carbon (COC): 1 g OC complexes with n=10 g clay; non-complexed OC drives physical behaviour; origin of 1/10 ratio.

    Relevance: methods (theoretical basis of ratio)

    Evidence: full text

54. Schjønning et al. (2012). Clay Dispersibility and Soil Friability—Testing the Soil Clay‐to‐Carbon Saturation Concept

    Vadose Zone Journal · 2012 · DOI · Full text and figures

    Study: Field test (clay 100–220 g/kg)

    Scale: Denmark

    Finding: No universal lower OC threshold for structure; clay-to-carbon saturation concept tested as predictor of dispersibility/friability; rotation effects minor.

    Relevance: methods / undermines universal thresholds

    Evidence: full text

55. Patrick et al. (2013). Soil Organic Carbon Thresholds and Nitrogen Management in Tropical Agroecosystems: Concepts and Prospects

    Journal of Sustainable Development · 2013 · DOI · Full text and figures

    Study: Review

    Scale: Tropics

    Finding: ~2% SOC often cited as quality threshold; no universal threshold for N response.

    Relevance: undermines universal thresholds

    Evidence: full text

56. Johannes et al. (2017). Optimal organic carbon values for soil structure quality of arable soils. Does clay content matter?

    Geoderma · 2017 · DOI · Full text and figures

    Study: Field study (161 Swiss Cambi-Luvisol samples, visual structure + bulk volumes)

    Scale: Switzerland

    Finding: SOC–pore volume relation linear at all clay contents, no optimum (contradicts Dexter's complexed-C theory); SOC:clay 1:8 = average very good structure, 1:10 = good/medium limit, 1:13 = poor-structure boundary (abstract misprints '1:8 or smaller'; corrigendum 10.1016/j.geoderma.2017.05.009).

    Relevance: methods / undermines non-linear 'collapse' (relationship is linear)

    Evidence: full text

57. Bünemann et al. (2018). Soil quality – A critical review

    Soil Biology and Biochemistry · 2018 · DOI · Full text and figures

    Study: Review

    Scale: Global

    Finding: Soil quality indicators critically reviewed; thresholds rarely validated.

    Relevance: undermines

    Evidence: full text

58. Wiesmeier et al. (2019). Soil organic carbon storage as a key function of soils - A review of drivers and indicators at various scales

    Geoderma · 2019 · DOI · Full text and figures

    Study: Review

    Scale: Global

    Finding: Drivers and indicators of SOC storage across scales.

    Relevance: context

    Evidence: full text

59. Prout et al. (2020). What is a good level of soil organic matter? An index based on organic carbon to clay ratio

    European Journal of Soil Science · 2020 · DOI · Full text and figures

    Study: Indicator development (national inventory)

    Scale: National (England & Wales, 3,809 sites)

    Finding: SOC:clay 1/8, 1/10, 1/13 boundaries; 38.2% arable, 6.6% grassland, 5.6% woodland sites 'degraded'; land use, soil type, rainfall, pH explain only 21% of ratio variance.

    Relevance: methods (threshold definition)

    Evidence: full text

60. Hartley et al. (2021). Temperature effects on carbon storage are controlled by soil stabilisation capacities

    Nature Communications · 2021 · DOI · Full text and figures

    Study: Observational (WoSIS >9,000 profiles)

    Scale: Global (non-cropland)

    Finding: Soil C declines ~1.4× per 10 °C; decline >3× stronger in coarse soils; ESM fails to reproduce.

    Relevance: supports texture-normalized framing

    Evidence: full text

61. Prout et al. (2022). Changes in organic carbon to clay ratios in different soils and land uses in England and Wales over time

    Scientific Reports · 2022 · DOI · Full text and figures

    Study: Resampled national inventory + LTEs

    Scale: National (England & Wales)

    Finding: ~Half of arable soils <1/13 vs 5% grass/woodland; soils ≥1/8 lost more SOC; proposed land-use targets; warns sesquioxidic tropical soils may differ.

    Relevance: methods / supports threshold use in temperate soils

    Evidence: full text

62. Georgiou et al. (2022). Global stocks and capacity of mineral-associated soil organic carbon

    Nature Communications · 2022 · DOI · Full text and figures

    Study: Observational (1,144 profiles)

    Scale: Global

    Finding: MAOC 899 PgC; soils at 42%/21% of mineral capacity; agricultural soils most undersaturated.

    Relevance: methods (capacity-normalized thresholds)

    Evidence: full text

63. Poeplau & Don (2023). A simple soil organic carbon level metric beyond the organic carbon‐to‐clay ratio

    Soil Use and Management · 2023 · DOI · Full text and figures

    Study: National inventory analysis

    Scale: Germany (2,958 topsoils)

    Finding: SOC:clay strongly clay-biased; 94% of Chernozems classed degraded; propose benchmark vs expected SOC.

    Relevance: undermines (1/13 threshold transferability)

    Evidence: full text

64. Soinne et al. (2023). Effects of organic carbon and clay contents on structure‐related properties of arable soils with high clay content

    European Journal of Soil Science · 2023 · DOI · Full text and figures

    Study: Field study

    Scale: Finland (high-clay soils)

    Finding: SOC and clay effects on structure in clay-rich arable soils; ratio behaviour in high clay.

    Relevance: methods / undermines

    Evidence: full text

65. Begill et al. (2023). No detectable upper limit of mineral‐associated organic carbon in temperate agricultural soils

    Global Change Biology · 2023 · DOI · Full text and figures

    Study: Observational (189 samples)

    Scale: Germany

    Finding: No detectable upper limit of MAOC.

    Relevance: undermines saturation-based thresholds

    Evidence: full text

66. Cotrufo et al. (2023). The robust concept of mineral‐associated organic matter saturation: A letter to Begill et al., 2023

    Global Change Biology · 2023 · DOI · Full text and figures

    Study: Letter

    Scale: —

    Finding: Defends MAOC saturation concept against Begill et al.

    Relevance: contradiction flagged

    Evidence: full text

67. Feeney et al. (2024). Benchmarking soil organic carbon (SOC) concentration provides more robust soil health assessment than the SOC/clay ratio at European scale

    Science of The Total Environment · 2024 · DOI · Full text and figures

    Study: Indicator development (LUCAS n=19,855)

    Scale: Europe

    Finding: Observed/typical SOC by pedo-climate zone outperforms SOC:clay; SOC:clay misleading.

    Relevance: undermines / methods (alternative index)

    Evidence: full text

68. Rabot et al. (2024). Relevance of the organic carbon to clay ratio as a national soil health indicator

    Geoderma · 2024 · DOI · Full text and figures

    Study: National indicator evaluation (RMQS)

    Scale: France

    Finding: SOC/clay and SOC/(fine silt+clay) poor indicators of bulk density and aggregate stability; raw SOC better; pH confounds; 1/13 would class 63% cropland, 81% permanent crop, 23% grassland as unhealthy.

    Relevance: undermines

    Evidence: full text

69. Mäkipää et al. (2024). Is the organic carbon-to-clay ratio a reliable indicator of soil health?

    Geoderma · 2024 · DOI · Full text and figures

    Study: Indicator evaluation (LUCAS 2009)

    Scale: Europe (EU-wide)

    Finding: SOC:clay 1:13 threshold inadequate outside the conditions used to derive it; SOC variation driven also by climate and land use; degraded-soil shares inconsistent with national GHG-inventory stock changes.

    Relevance: undermines (1/13 transferability)

    Evidence: full text

70. Sauzet et al. (2024). The organic carbon‐to‐clay ratio as an indicator of soil structure vulnerability, a metric focused on the condition of soil structure

    Soil Use and Management · 2024 · DOI · Full text and figures

    Study: Commentary

    Scale: France/Switzerland

    Finding: Defends SOC:clay as a soil-structure vulnerability index; argues Poeplau & Don's SOC:SOCexp can classify highly depleted soils as 'good'.

    Relevance: methods (contradiction: structure-index vs benchmark camps)

    Evidence: full text

71. Georgiou et al. (2024). Emergent temperature sensitivity of soil organic carbon driven by mineral associations

    Nature Geoscience · 2024 · DOI · Full text and figures

    Study: Observational + model benchmarking

    Scale: Global

    Finding: POC T-sensitivity 28% higher than MAOC; ESM protected-pool fraction 16–85%.

    Relevance: methods (pool-resolved models e.g. Millennial)

    Evidence: full text

72. Six et al. (2024). The six rights of how and when to test for soil C saturation

    SOIL · 2024 · DOI · Full text and figures

    Study: Opinion/synthesis

    Scale: Global

    Finding: Saturation testing rules; max MAOC ~82 g C/kg silt+clay (2:1) vs ~46 (1:1 clays).

    Relevance: methods (mineralogy-specific capacity; key for tropics)

    Evidence: full text

73. Zhou et al. (2024). Global turnover of soil mineral-associated and particulate organic carbon

    Nature Communications · 2024 · DOI · Full text and figures

    Study: Global mapping + data–model

    Scale: Global

    Finding: MAOC ~975 PgC, POC ~330 PgC; turnover ~129 vs ~23 yr.

    Relevance: methods / data

    Evidence: full text

74. Heinemann et al. (2024). No saturation of soil carbon under long-term extreme manure additions

    Plant and Soil · 2024 · DOI · Full text and figures

    Study: Long-term experiment

    Scale: Germany

    Finding: No SOC saturation under extreme long-term manure additions.

    Relevance: contradiction flagged

    Evidence: full text

75. Ling et al. (2025). Soil organic carbon thresholds control fertilizer effects on carbon accrual in croplands worldwide

    Nature Communications · 2025 · DOI · Full text and figures

    Study: Meta-analysis + continental field experiments

    Scale: Global

    Finding: Initial SOC threshold of 15 g C/kg switches N-fertilizer effect from POC accrual (below) to MAOC accrual (above).

    Relevance: methods (threshold behaviour of C pools, N interaction)

    Evidence: full text

76. King & Sokol (2025). Soil carbon formation is promoted by saturation deficit and existing mineral-associated carbon, not by microbial carbon-use efficiency

    Science Advances · 2025 · DOI · Full text and figures

    Study: Manipulative experiment (118 soils, 15 sites)

    Scale: USA

    Finding: New MAOC formation promoted by existing MAOC and saturation deficit, not microbial CUE.

    Relevance: methods

    Evidence: full text

77. Todd-Brown et al. (2013). Causes of variation in soil carbon simulations from CMIP5 Earth system models and comparison with observations

    Biogeosciences · 2013 · DOI · Full text and figures

    Study: Model evaluation

    Scale: Global (CMIP5)

    Finding: ESM SOC varies ~5-fold; NPP+T explains most ESM spatial SOC but ~10% of observed.

    Relevance: methods / undermines raw ESM SOC

    Evidence: full text

78. Wieder et al. (2013). Global soil carbon projections are improved by modelling microbial processes

    Nature Climate Change · 2013 · DOI · Full text and figures

    Study: Model comparison

    Scale: Global

    Finding: Microbial model better matches present SOC; projections range from gains to large losses.

    Relevance: methods

    Evidence: full text

79. Todd-Brown et al. (2014). Changes in soil organic carbon storage predicted by Earth system models during the 21st century

    Biogeosciences · 2014 · DOI · Full text and figures

    Study: Model analysis

    Scale: Global (CMIP5)

    Finding: 21st-century ΔSOC −72 to +253 PgC; all ESMs reduce turnover time 15–28%; NPP gains offset.

    Relevance: methods

    Evidence: full text

80. Olin et al. (2015). Soil carbon management in large-scale Earth system modelling: implications for crop yields and nitrogen leaching

    Earth System Dynamics · 2015 · DOI · Full text and figures

    Study: Global DVM with cropland management

    Scale: Global

    Finding: LPJ-GUESS links residue/tillage/cover-crop/manure management to SOC, yields and N leaching.

    Relevance: methods / overlaps (yield–SOC coupling in a global model)

    Evidence: full text

81. Chappell et al. (2015). The global significance of omitting soil erosion from soil organic carbon cycling schemes

    Nature Climate Change · 2015 · DOI · Full text and figures

    Study: Model modification (RothC+erosion)

    Scale: Australia/global

    Finding: Omitting erosion misattributes SOC loss; global SOC erosion 0.3–1.0 PgC/yr.

    Relevance: methods (erosion term)

    Evidence: full text

82. Luo et al. (2016). Toward more realistic projections of soil carbon dynamics by Earth system models

    Global Biogeochemical Cycles · 2016 · DOI

    Study: Review/perspective

    Scale: Global

    Finding: Roadmap for more realistic ESM soil carbon (data assimilation, matrix approach).

    Relevance: methods

    Evidence: metadata only; the available file contains a different article, so the intended full text was unavailable

83. Bradford et al. (2016). Managing uncertainty in soil carbon feedbacks to climate change

    Nature Climate Change · 2016 · DOI · Full text and figures

    Study: Perspective

    Scale: Global

    Finding: Limited direct evidence of warming-driven soil C loss; models encode outdated turnover knowledge.

    Relevance: undermines (uncertainty in SOC decline)

    Evidence: full text

84. Crowther et al. (2016). Quantifying global soil carbon losses in response to warming

    Nature · 2016 · DOI · Full text and figures

    Study: Field warming synthesis

    Scale: Global

    Finding: Warming-driven soil C losses concentrated in high-C cold soils; loss scales with initial stocks.

    Relevance: context (croplands not focus)

    Evidence: full text

85. Koven et al. (2017). Higher climatological temperature sensitivity of soil carbon in cold than warm climates

    Nature Climate Change · 2017 · DOI · Full text and figures

    Study: Observational constraint

    Scale: Global

    Finding: Climatological T-sensitivity of turnover higher in cold than warm climates; ESMs fail pattern.

    Relevance: methods

    Evidence: full text

86. Shi et al. (2018). Model structures amplify uncertainty in predicted soil carbon responses to climate change

    Nature Communications · 2018 · DOI · Full text and figures

    Study: Model structural uncertainty

    Scale: Global

    Finding: Vertically resolved and microbial models widen projection uncertainty, even sign of feedback.

    Relevance: methods (structural ensemble needed)

    Evidence: full text

87. Varney et al. (2020). A spatial emergent constraint on the sensitivity of soil carbon turnover to global warming

    Nature Communications · 2020 · DOI · Full text and figures

    Study: Emergent constraint

    Scale: Global

    Finding: Spatial emergent constraint on sensitivity of soil C turnover to warming; constrains ESM soil C loss.

    Relevance: methods

    Evidence: full text

88. Arora et al. (2020). Carbon–concentration and carbon–climate feedbacks in CMIP6 models and their comparison to CMIP5 models

    Biogeosciences · 2020 · DOI · Full text and figures

    Study: Model analysis (C4MIP)

    Scale: Global

    Finding: CMIP6 vs CMIP5 carbon–concentration and carbon–climate feedbacks.

    Relevance: methods

    Evidence: full text

89. Varney et al. (2022). Evaluation of soil carbon simulation in CMIP6 Earth system models

    Biogeosciences · 2022 · DOI · Full text and figures

    Study: Model evaluation

    Scale: Global (CMIP6)

    Finding: CMIP6 soil C still inconsistent; NPP improved but τs not; unrealistically high NPP–Cs correlation.

    Relevance: methods / undermines raw cSoil use

    Evidence: full text

90. Varney et al. (2023). Simulated responses of soil carbon to climate change in CMIP6 Earth system models: the role of false priming

    Biogeosciences · 2023 · DOI · Full text and figures

    Study: Model analysis

    Scale: Global (CMIP6)

    Finding: Reduced CMIP6 ΔCs spread partly due to 'false priming' in multi-pool models (NPP rise lowers effective τ).

    Relevance: methods (do not trust raw cSoil deltas)

    Evidence: full text

91. Wang et al. (2023). Responses of soil organic carbon to climate extremes under warming across global biomes

    Nature Climate Change · 2023 · DOI · Full text and figures

    Study: Space-for-time + meta-analysis

    Scale: Global

    Finding: SOC −10.5% at 1.5 °C; extremes worsen losses in 22/33 indices; temperate croplands sensitive.

    Relevance: supports (climate-driven SOC loss)

    Evidence: full text

92. Varney et al. (2024). Soil carbon-concentration and carbon-climate feedbacks in CMIP6 Earth system models

    Biogeosciences · 2024 · DOI · Full text and figures

    Study: Model analysis

    Scale: Global (CMIP6)

    Finding: βs (CO2) dominates soil C change to 2×CO2; γs (climate) increasingly uncertain at high CO2.

    Relevance: methods

    Evidence: full text

93. Ren et al. (2024). Projected soil carbon loss with warming in constrained Earth system models

    Nature Communications · 2024 · DOI · Full text and figures

    Study: Data–model constraint

    Scale: Global

    Finding: ESMs overestimate intrinsic turnover ~30%; constrained soils lose 19–45 PgC by 2100 vs +19–30 PgC gain in raw ESMs.

    Relevance: supports (SOC decline trajectories) / methods

    Evidence: full text

94. Portmann et al. (2010). MIRCA2000—Global monthly irrigated and rainfed crop areas around the year 2000: A new high‐resolution data set for agricultural and hydrological modeling

    Global Biogeochemical Cycles · 2010 · DOI

    Study: Dataset

    Scale: Global (5 arcmin, ~2000)

    Finding: Monthly irrigated/rainfed areas for 26 crop classes; 25% of harvested area irrigated (rice ~60%, wheat ~30%, maize ~20%).

    Relevance: data (irrigation stratification)

    Evidence: abstract

95. Basso et al. (2011). Procedures for Initializing Soil Organic Carbon Pools in the DSSAT‐CENTURY Model for Agricultural Systems

    Soil Science Society of America Journal · 2011 · DOI · Full text and figures

    Study: Methods

    Scale: Field (DSSAT-CENTURY)

    Finding: SOC pool initialization strongly affects simulated SOC dynamics.

    Relevance: methods risk (initialization)

    Evidence: full text

96. Rosenzweig et al. (2013). Assessing agricultural risks of climate change in the 21st century in a global gridded crop model intercomparison

    Proceedings of the National Academy of Sciences · 2013 · DOI · Full text and figures

    Study: Multi-model projection (GGCMI phase 1)

    Scale: Global

    Finding: Strong negative impacts at high warming and low latitudes; simulations with explicit N stress give much more severe climate impacts.

    Relevance: supports (soil-nutrient representation changes impacts)

    Evidence: full text

97. Elliott et al. (2015). The Global Gridded Crop Model Intercomparison: data and modeling protocols for Phase 1 (v1.0)

    Geoscientific Model Development · 2015 · DOI · Full text and figures

    Study: Protocol

    Scale: Global

    Finding: GGCMI Phase 1 protocol; harmonized vs default management, no-N-stress setting.

    Relevance: data / methods

    Evidence: full text

98. Folberth et al. (2016). Uncertainty in soil data can outweigh climate impact signals in global crop yield simulations

    Nature Communications · 2016 · DOI · Full text and figures

    Study: Model sensitivity

    Scale: Global (EPIC)

    Finding: Within-grid soil choice changes yields more than climate impact signals in places, esp. low-input systems.

    Relevance: methods risk (soil-input uncertainty may swamp SOC feedback)

    Evidence: full text

99. Müller et al. (2017). Global gridded crop model evaluation: benchmarking, skills, deficiencies and implications

    Geoscientific Model Development · 2017 · DOI · Full text and figures

    Study: Model evaluation

    Scale: Global

    Finding: GGCM skill varies by crop, region and harmonization setting.

    Relevance: methods

    Evidence: full text

100. Balkovič et al. (2018). Impacts and Uncertainties of +2°C of Climate Change and Soil Degradation on European Crop Calorie Supply

     Earth's Future · 2018 · DOI · Full text and figures

     Study: Model projection (EPIC + EURO-CORDEX)

     Scale: Europe

     Finding: +2 °C mostly raises calorie yields (CO2 offsets warming) but soil degradation (nutrient depletion) creates calorie vulnerability of 0–50 Gcal/ha.

     Relevance: supports / overlaps (regional degradation term)

     Evidence: full text

101. Basso et al. (2018). Soil Organic Carbon and Nitrogen Feedbacks on Crop Yields under Climate Change

     Agricultural & Environmental Letters · 2018 · DOI · Full text and figures

     Study: Multi-model ensemble

     Scale: 8 sites worldwide

     Finding: Including SOC and N losses reduces yields 13% (wheat) and 19% (maize) at +3 °C, ~4.5 pp beyond temperature-only; CO2 compensation insufficient.

     Relevance: supports — direct precedent (site scale)

     Evidence: full text

102. von Bloh et al. (2018). Implementing the nitrogen cycle into the dynamic global vegetation, hydrology, and crop growth model LPJmL (version 5.0)

     Geoscientific Model Development · 2018 · DOI · Full text and figures

     Study: Model description

     Scale: Global (LPJmL5)

     Finding: Coupled N cycle in LPJmL with SOM mineralization.

     Relevance: methods (GGCM with SOM)

     Evidence: full text

103. Folberth et al. (2019). Parameterization-induced uncertainties and impacts of crop management harmonization in a global gridded crop model ensemble

     PLOS ONE · 2019 · DOI · Full text and figures

     Study: Model uncertainty

     Scale: Global

     Finding: Parameterization and management harmonization drive GGCM spread.

     Relevance: methods

     Evidence: full text

104. Lange (2019). Trend-preserving bias adjustment and statistical downscaling with ISIMIP3BASD (v1.0)

     Geoscientific Model Development · 2019 · DOI · Full text and figures

     Study: Method

     Scale: Global

     Finding: ISIMIP3BASD trend-preserving bias adjustment.

     Relevance: methods / data

     Evidence: full text

105. Porwollik et al. (2019). Generating a rule-based global gridded tillage dataset

     Earth System Science Data · 2019 · DOI · Full text and figures

     Study: Dataset (rule-based allocation)

     Scale: Global (5 arcmin, ~2005)

     Finding: Six crop-specific tillage classes (incl. conservation-agriculture no-till) mapped globally; CA areas downscaled for 54 countries.

     Relevance: data (management layer)

     Evidence: full text

106. Franke et al. (2020). The GGCMI Phase 2 experiment: global gridded crop model simulations under uniform changes in CO ₂ , temperature, water, and nitrogen levels (protocol version 1.0)

     Geoscientific Model Development · 2020 · DOI · Full text and figures

     Study: Protocol

     Scale: Global

     Finding: GGCMI Phase 2 CTWN; soil treatment/SOM carry-over not standardized; N mineralization tied to SOM in some models.

     Relevance: methods (evidence GGCMs treat SOM inconsistently)

     Evidence: full text

107. Iizumi & Sakai (2020). The global dataset of historical yields for major crops 1981–2016

     Scientific Data · 2020 · DOI · Full text and figures

     Study: Dataset

     Scale: Global (0.5°)

     Finding: GDHY gridded yields 1981–2016.

     Relevance: data

     Evidence: full text

108. Yu et al. (2020). A cultivated planet in 2010 – Part 2: The global gridded agricultural-production maps

     Earth System Science Data · 2020 · DOI · Full text and figures

     Study: Dataset

     Scale: Global (5 arcmin, ~2010)

     Finding: SPAM2010 area, yield and production for 42 crops under four farming systems.

     Relevance: data (crop area weights, farming-system stratification)

     Evidence: full text

109. Jägermeyr et al. (2021). Climate impacts on global agriculture emerge earlier in new generation of climate and crop models

     Nature Food · 2021 · DOI · Full text and figures

     Study: Multi-model projection (GGCMI3 × CMIP6)

     Scale: Global

     Finding: Maize end-century −6% (SSP126) to −24% (SSP585); emergence before 2040. Methods: management fixed at 2015, all models spun up to soil/C-pool equilibrium, EPIC-IIASA re-initializes soil attributes incl. SOM annually, erosion excluded, 70% residue removal encouraged, other nutrients non-limiting.

     Relevance: baseline; verified evidence that projections omit degradation trajectories

     Evidence: full text

110. Minoli et al. (2022). Global crop yields can be lifted by timely adaptation of growing periods to climate change

     Nature Communications · 2022 · DOI · Full text and figures

     Study: Model projection

     Scale: Global

     Finding: Adapting growing periods lifts projected yields.

     Relevance: context (adaptation counterfactual)

     Evidence: full text

111. Kornhuber et al. (2023). Risks of synchronized low yields are underestimated in climate and crop model projections

     Nature Communications · 2023 · DOI · Full text and figures

     Study: Multi-model analysis

     Scale: Global

     Finding: Synchronized low-yield risks underestimated in climate–crop projections.

     Relevance: context (framing of underestimation)

     Evidence: full text

112. Adalibieke et al. (2023). Global crop-specific nitrogen fertilization dataset in 1961–2020

     Scientific Data · 2023 · DOI · Full text and figures

     Study: Dataset

     Scale: Global

     Finding: Crop-specific N fertilization 1961–2020.

     Relevance: data

     Evidence: full text

113. Couëdel et al. (2024). Long-term soil organic carbon and crop yield feedbacks differ between 16 soil-crop models in sub-Saharan Africa

     European Journal of Agronomy · 2024 · DOI · Full text and figures

     Study: Multi-model comparison (16 soil–crop models; blind vs full calibration)

     Scale: 4 LTEs, sub-Saharan Africa

     Finding: First multi-model comparison of long-term yield–SOC feedbacks in SSA; ensemble accuracy improves with full calibration; models differ strongly in simulated long-term SOC and yield feedbacks.

     Relevance: methods (model-structural uncertainty in feedback)

     Evidence: full text

114. Müller et al. (2024). Substantial Differences in Crop Yield Sensitivities Between Models Call for Functionality‐Based Model Evaluation

     Earth's Future · 2024 · DOI · Full text and figures

     Study: Model evaluation

     Scale: Global (GGCMI)

     Finding: Large differences in crop yield sensitivities among GGCMs; call for functionality-based evaluation.

     Relevance: methods

     Evidence: full text

115. Couëdel et al. (2026). Beyond Climate Change: The Role of Integrated Soil Fertility Management for Sustaining Future Maize Yield in Sub‐Saharan Africa

     Global Change Biology · 2026 · DOI · Full text and figures

     Study: Multi-model ensemble (15 models)

     Scale: 4 LTEs, sub-Saharan Africa

     Finding: No-input soil fertility decline −24% yield (−20 to −50% over decades) exceeds single climate-factor effects; fertility×climate interaction small; ISFM +182%.

     Relevance: supports (degradation term large) / undermines (interaction small → term mostly additive)

     Evidence: full text

116. Hengl et al. (2017). SoilGrids250m: Global gridded soil information based on machine learning

     PLOS ONE · 2017 · DOI · Full text and figures

     Study: Global DSM product

     Scale: Global (250 m)

     Finding: SoilGrids250m ensemble ML; CV R² 56–83% (random CV).

     Relevance: data

     Evidence: full text

117. Padarian et al. (2019). Using deep learning for digital soil mapping

     SOIL · 2019 · DOI · Full text and figures

     Study: Deep learning DSM

     Scale: National (Chile)

     Finding: CNN with contextual covariates cut error 30% vs point models.

     Relevance: methods

     Evidence: full text

118. Wadoux (2019). Using deep learning for multivariate mapping of soil with quantified uncertainty

     Geoderma · 2019 · DOI · Full text and figures

     Study: Deep learning DSM

     Scale: National (France, LUCAS)

     Finding: Multivariate CNN with quantified uncertainty.

     Relevance: methods

     Evidence: full text

119. Batjes et al. (2020). Standardised soil profile data to support global mapping and modelling (WoSIS snapshot 2019)

     Earth System Science Data · 2020 · DOI · Full text and figures

     Study: Database

     Scale: Global

     Finding: WoSIS 2019 standardized profile data (OC, texture, BD, etc.).

     Relevance: data

     Evidence: full text

120. Wadoux et al. (2020). Machine learning for digital soil mapping: Applications, challenges and suggested solutions

     Earth-Science Reviews · 2020 · DOI · Full text and figures

     Study: Review

     Scale: Global

     Finding: ML for DSM: challenges incl. extrapolation, uncertainty, interpretation.

     Relevance: methods

     Evidence: full text

121. Ploton et al. (2020). Spatial validation reveals poor predictive performance of large-scale ecological mapping models

     Nature Communications · 2020 · DOI · Full text and figures

     Study: Methods critique

     Scale: Central Africa

     Finding: Spatial CV reveals near-zero predictive power where random CV suggested >50%.

     Relevance: methods risk (validation of global DSM)

     Evidence: full text

122. Poggio et al. (2021). SoilGrids 2.0: producing soil information for the globe with quantified spatial uncertainty

     SOIL · 2021 · DOI · Full text and figures

     Study: Global DSM product

     Scale: Global (250 m)

     Finding: SoilGrids 2.0: ~240k observations, quantile RF uncertainty, 6 depths.

     Relevance: data / baseline comparison

     Evidence: full text

123. Hengl et al. (2021). African soil properties and nutrients mapped at 30 m spatial resolution using two-scale ensemble machine learning

     Scientific Reports · 2021 · DOI · Full text and figures

     Study: Continental DSM

     Scale: Africa (30 m)

     Finding: iSDAsoil maps of soil properties and nutrients.

     Relevance: data

     Evidence: full text

124. Meyer & Pebesma (2021). Predicting into unknown space? Estimating the area of applicability of spatial prediction models

     Methods in Ecology and Evolution · 2021 · DOI · Full text and figures

     Study: Method (dissimilarity index / area of applicability)

     Scale: Generic (spatial ML)

     Finding: Defines the 'area of applicability' where a spatial ML model's cross-validated performance holds, via a weighted dissimilarity index to training data.

     Relevance: methods (essential for global transformer DSM validation)

     Evidence: full text

125. Padarian et al. (2022). Soil carbon sequestration potential in global croplands

     PeerJ · 2022 · DOI · Full text and figures

     Study: Global modelling

     Scale: Global croplands

     Finding: Cropland SOC sequestration potential estimated by ML.

     Relevance: data / methods

     Evidence: full text

126. Batjes et al. (2024). Providing quality-assessed and standardised soil data to support global mapping and modelling (WoSIS snapshot 2023)

     Earth System Science Data (discussion/preprint) · 2024 · DOI · Full text and figures

     Study: Database

     Scale: Global

     Finding: WoSIS 2023 snapshot; new ETL, ISO 28258 model.

     Relevance: data

     Evidence: full text

127. Zhang et al. (2025). Regional-scale soil carbon predictions can be enhanced by transferring global-scale soil–environment relationships

     Geoderma · 2025 · DOI · Full text and figures

     Study: Transfer learning DSM

     Scale: Global pretrain → 4 regions

     Finding: GSoilCPM: pretrained on 106,167 samples; MAE −10.9%, CCC +29% regionally.

     Relevance: methods / competitor (global pre-trained SOC model)

     Evidence: full text

128. Demattê et al. (2025). A global soil spectral grid based on space sensing

     Science of The Total Environment · 2025 · DOI · Full text and figures

     Study: Earth-observation data product (GEOS3, Landsat 1984–2022 median bare-soil composite)

     Scale: Global (30 m)

     Finding: First global fine-resolution bare-soil spectral grid; bare soil detected for ~90% of agricultural areas; spectral patterns consistent with terrestrial spectral libraries.

     Relevance: data (candidate covariate for transformer/DSM baseline)

     Evidence: full text

129. Fourel et al. (2026). Compact CNN-Transformers match larger models for spatially-robust soil organic carbon mapping at regional scale

     Geoderma · 2026 · DOI · Full text and figures

     Study: Deep learning vs tree ensembles, spatially blocked CV (16,514 samples, 2007–2023)

     Scale: Regional (Bavaria)

     Finding: Compact gated CNN-Transformer (≈87k params, R² 0.375) ≈ 45M-param Transformer (0.401); RF 0.342, XGBoost 0.317 overall; but ON CROPLAND gradient-boosted trees far outperform attention models (R² ≈0.66 vs ≈0.21); recommends land-use-stratified mapping.

     Relevance: undermines a transformer-only cropland baseline / methods

     Evidence: full text

130. Peter‐Contesse et al. (2026). Transformer‐Based Digital Soil Mapping Reveals Land‐Use Legacy Effects on (Sub)Tropical Soil Carbon

     Earth's Future · 2026 · DOI · Full text and figures

     Study: Transformer DSM (Landsat time series)

     Scale: Regional (Hawaii)

     Finding: R² 0.57 leave-one-island-out; inferred SOC decline 2017–2024, former sugarcane −0.96 pp.

     Relevance: methods (author's own prior work)

     Evidence: full text

131. Shi et al. (2026). Diagnosing global soil organic carbon mapping discrepancies with interpretable machine learning: A comparison of soil type linkage and digital soil mapping

     Geoderma · 2026 · DOI · Full text and figures

     Study: Interpretable ML comparison

     Scale: Global

     Finding: Diagnoses discrepancies among global SOC maps.

     Relevance: methods (baseline uncertainty)

     Evidence: full text

132. Gottschalk et al. (2012). How will organic carbon stocks in mineral soils evolve under future climate? Global projections using RothC for a range of climate change scenarios

     Biogeosciences · 2012 · DOI · Full text and figures

     Study: Process model projection (RothC × GCMs)

     Scale: Global

     Finding: Global SOC increases by 2100 (NPP-driven) with losses at high latitudes.

     Relevance: undermines (decline not robust; input assumption decisive)

     Evidence: full text

133. Lugato et al. (2013). A new baseline of organic carbon stock in European agricultural soils using a modelling approach

     Global Change Biology · 2013 · DOI · Full text and figures

     Study: Process model platform (CENTURY; ~164,000 soil–climate–land-use combinations)

     Scale: Europe

     Finding: Pan-European agricultural SOC baseline with management (irrigation, fertilization, tillage) from official statistics; validated against EIONET inventories and ~20,000 LUCAS 2009 samples.

     Relevance: methods (regional precedent for management-explicit SOC modelling)

     Evidence: full text

134. Sommer & Bossio (2014). Dynamics and climate change mitigation potential of soil organic carbon sequestration

     Journal of Environmental Management · 2014 · DOI · Full text and figures

     Study: Global RothC

     Scale: Global

     Finding: SOC sequestration dynamics and limits.

     Relevance: methods

     Evidence: full text

135. Wieder et al. (2015). Representing life in the Earth system with soil microbial functional traits in the MIMICS model

     Geoscientific Model Development · 2015 · DOI · Full text and figures

     Study: Model description

     Scale: Cross-site

     Finding: MIMICS microbial-mineral model.

     Relevance: methods

     Evidence: full text

136. Wiesmeier et al. (2016). Projected loss of soil organic carbon in temperate agricultural soils in the 21st century: effects of climate change and carbon input trends

     Scientific Reports · 2016 · DOI · Full text and figures

     Study: Process model projection (RothC)

     Scale: Bavaria

     Finding: SOC −11 to −16% by 2100 with constant inputs; −19 to −24% with −20% inputs.

     Relevance: supports

     Evidence: full text

137. Abramoff et al. (2017). The Millennial model: in search of measurable pools and transformations for modeling soil carbon in the new century

     Biogeochemistry · 2017 · DOI · Full text and figures

     Study: Model description

     Scale: Conceptual/global

     Finding: Millennial model with measurable pools (POM, MAOM, aggregates, microbes).

     Relevance: methods (candidate offline model)

     Evidence: full text

138. Wieder et al. (2017). Carbon cycle confidence and uncertainty: Exploring variation among soil biogeochemical models

     Global Change Biology · 2017 · DOI · Full text and figures

     Study: Model testbed

     Scale: Global

     Finding: Three soil models with common forcing diverge in T–turnover relationships.

     Relevance: methods

     Evidence: full text

139. Zomer et al. (2017). Global Sequestration Potential of Increased Organic Carbon in Cropland Soils

     Scientific Reports · 2017 · DOI · Full text and figures

     Study: Global estimate

     Scale: Global croplands

     Finding: Croplands could sequester 0.90–1.85 PgC/yr.

     Relevance: context

     Evidence: full text

140. Morais et al. (2019). Detailed global modelling of soil organic carbon in cropland, grassland and forest soils

     PLOS ONE · 2019 · DOI · Full text and figures

     Study: Process model (RothC)

     Scale: Global (~17,000 regions)

     Finding: Attainable SOC strongly depends on residue cover and irrigation.

     Relevance: methods

     Evidence: full text

141. Lu et al. (2020). Full Implementation of Matrix Approach to Biogeochemistry Module of CLM5

     Journal of Advances in Modeling Earth Systems · 2020 · DOI · Full text and figures

     Study: Model development

     Scale: Global (CLM5)

     Finding: Matrix approach for CLM5 biogeochemistry; fast spin-up/traceability.

     Relevance: methods (matrix model option)

     Evidence: full text

142. Herzfeld et al. (2021). Soil organic carbon dynamics from agricultural management practices under climate change

     Earth System Dynamics · 2021 · DOI · Full text and figures

     Study: Process model projection (LPJmL)

     Scale: Global croplands

     Finding: Cropland SOC declines under all management scenarios and both RCP2.6/8.5.

     Relevance: supports (SOC decline trajectory)

     Evidence: full text

143. Riggers et al. (2021). How much carbon input is required to preserve or increase projected soil organic carbon stocks in German croplands under climate change?

     Plant and Soil · 2021 · DOI · Full text and figures

     Study: Process model projection (RothC)

     Scale: Germany

     Finding: C inputs needed to maintain cropland SOC under climate change.

     Relevance: methods / supports

     Evidence: full text

144. Lessmann et al. (2021). Global variation in soil carbon sequestration potential through improved cropland management

     Global Change Biology · 2021 · DOI · Full text and figures

     Study: Meta-analysis × spatial management data

     Scale: Global croplands

     Finding: Improved management sequestration potential 0.44–0.68 GtC/yr (max complementarity), realistically 0.28–0.43 GtC/yr; manure availability limits.

     Relevance: methods / context

     Evidence: full text

145. Abramoff et al. (2022). Improved global-scale predictions of soil carbon stocks with Millennial Version 2

     Soil Biology and Biochemistry · 2022 · DOI · Full text and figures

     Study: Model evaluation

     Scale: Global

     Finding: Millennial V2 improves global-scale predictions of SOC stocks.

     Relevance: methods

     Evidence: full text

146. Karstens et al. (2022). Management-induced changes in soil organic carbon on global croplands

     Biogeosciences · 2022 · DOI · Full text and figures

     Study: Global reduced-complexity model (IPCC tier 2)

     Scale: Global croplands

     Finding: Cropland SOC debt 34.6 GtC (1975) growing to 39.6 GtC (2010); management dataset.

     Relevance: data / methods (management layers)

     Evidence: full text

147. Gaudaré et al. (2023). Soil organic carbon stocks potentially at risk of decline with organic farming expansion

     Nature Climate Change · 2023 · DOI · Full text and figures

     Study: Modelling

     Scale: Global

     Finding: 100% organic conversion lowers C inputs 29–39% and SOC.

     Relevance: supports (input-driven SOC decline)

     Evidence: full text

148. Meng et al. (2026). Nonlinear temperature change responses shape soil organic carbon loss-gain transitions in global Mollisol croplands

     Nature Communications · 2026 · DOI · Full text and figures

     Study: Satellite + DL time series

     Scale: Global Mollisol croplands

     Finding: Mollisol cropland SOC −5.59% 1990–2020; nonlinear threshold in warming rate.

     Relevance: supports (observed cropland decline) / methods (threshold claims)

     Evidence: full text

149. Mahlstein et al. (2011). Early onset of significant local warming in low latitude countries

     Environmental Research Letters · 2011 · DOI · Full text and figures

     Study: Method application

     Scale: Global

     Finding: Local warming emerges earliest in low-latitude countries.

     Relevance: methods (ToE)

     Evidence: full text

150. Hawkins & Sutton (2012). Time of emergence of climate signals

     Geophysical Research Letters · 2012 · DOI · Full text and figures

     Study: Method

     Scale: Global

     Finding: ToE methodology for individual models; median ToE earlier at low latitudes; ToE uncertainty ≥30–60 yr from signal and noise differences.

     Relevance: methods (ToE)

     Evidence: full text

151. Rojas et al. (2019). Emergence of robust precipitation changes across crop production areas in the 21st century

     Proceedings of the National Academy of Sciences · 2019 · DOI · Full text and figures

     Study: ToE application

     Scale: Global crop areas

     Finding: Precipitation-trend ToE <2040 for all four crops; reduced (increased) precipitation trends cover 1–14% (3–31%) of production.

     Relevance: methods (ToE applied to agriculture)

     Evidence: full text

152. Hawkins et al. (2020). Observed Emergence of the Climate Change Signal: From the Familiar to the Unknown

     Geophysical Research Letters · 2020 · DOI · Full text and figures

     Study: Method

     Scale: Global

     Finding: Observed emergence framework (signal-to-noise relative to variability).

     Relevance: methods (ToE)

     Evidence: full text

153. Bartoli et al. (1991). Disaggregation and clay dispersion of Oxisols: Na resin, a recommended methodology

     Geoderma · 1991 · DOI

     Study: Laboratory method

     Scale: Oxisols

     Finding: Na-resin disaggregation recommended for clay dispersion of Oxisols (title-level; abstract not retrievable).

     Relevance: methods (I: clay measurement in oxide-rich soils)

     Evidence: title/metadata only — verify

154. Feller & Beare (1997). Physical control of soil organic matter dynamics in the tropics

     Geoderma · 1997 · DOI · Full text and figures

     Study: Review

     Scale: Tropics

     Finding: Physical control of SOM in tropical soils; texture/mineralogy effects differ from temperate.

     Relevance: methods (I)

     Evidence: full text

155. Torn et al. (1997). Mineral control of soil organic carbon storage and turnover

     Nature · 1997 · DOI · Full text and figures

     Study: Observational (chronosequences)

     Scale: Hawaii volcanic soils

     Finding: Non-crystalline minerals control SOC storage and turnover.

     Relevance: undermines clay-only normalization (Andisols)

     Evidence: full text

156. Muggler et al. (1997). LASER GRAIN-SIZE DETERMINATION IN SOIL GENETIC STUDIES 2. CLAY CONTENT, CLAY FORMATION, AND AGGREGATION IN SOME BRAZILIAN OXISOLS

     Soil Science · 1997 · DOI

     Study: Laboratory (laser diffraction vs pipette)

     Scale: Brazil (3 deeply weathered profiles)

     Finding: Laser and pipette clay correlate well but never 1:1; aggregation linked to organic matter rather than sesquioxides in these profiles.

     Relevance: methods (I: method-dependent clay values; nuance on pseudo-sand mechanism)

     Evidence: abstract (WUR/search record)

157. Neufeldt et al. (2002). Texture and land-use effects on soil organic matter in Cerrado Oxisols, Central Brazil

     Geoderma · 2002 · DOI

     Study: Field fractionation

     Scale: Brazil (Cerrado Oxisols)

     Finding: Texture and land-use effects on SOM in clayey vs loamy Oxisols; management effects most visible in sand-size (POM) fractions (per related 1999 report).

     Relevance: data / methods (I)

     Evidence: partial abstract (Embrapa record) — verify

158. Huygens et al. (2005). Aggregate and soil organic carbon dynamics in South Chilean Andisols

     Biogeosciences · 2005 · DOI · Full text and figures

     Study: Field fractionation

     Scale: Chile (Andisols)

     Finding: Al-humus-clay complexes stabilize aggregates; amorphous Al dominates.

     Relevance: undermines clay-only ratio in Andisols

     Evidence: full text

159. Barthès et al. (2008). Texture and sesquioxide effects on water-stable aggregates and organic matter in some tropical soils

     Geoderma · 2008 · DOI · Full text and figures

     Study: Field study

     Scale: Tropics

     Finding: Sesquioxides drive water-stable aggregation and OM protection in tropical soils.

     Relevance: undermines clay-only normalization (I)

     Evidence: full text

160. Doetterl et al. (2015). Soil carbon storage controlled by interactions between geochemistry and climate

     Nature Geoscience · 2015 · DOI · Full text and figures

     Study: Observational

     Scale: Global

     Finding: SOC storage controlled by geochemistry–climate interactions (Fe/Al, pH).

     Relevance: undermines clay-only normalization

     Evidence: full text

161. Rasmussen et al. (2018). Beyond clay: towards an improved set of variables for predicting soil organic matter content

     Biogeochemistry · 2018 · DOI · Full text and figures

     Study: Data synthesis

     Scale: Global

     Finding: Clay is often a weak SOM predictor; Fe/Al oxides and exchangeable Ca matter more in many soils.

     Relevance: undermines clay-only normalization

     Evidence: full text

162. Quesada et al. (2020). Variations in soil chemical and physical properties explain basin-wide Amazon forest soil carbon concentrations

     SOIL · 2020 · DOI · Full text and figures

     Study: Observational (147 forest soils)

     Scale: Amazon Basin

     Finding: Clay explains SOC in kaolinitic Ferralsols/Acrisols but not less-weathered soils.

     Relevance: mixed (clay normalization may work within kaolinitic soils)

     Evidence: full text

163. von Fromm et al. (2021). Continental-scale controls on soil organic carbon across sub-Saharan Africa

     SOIL · 2021 · DOI · Full text and figures

     Study: Observational (AfSIS 1,601 samples)

     Scale: sub-Saharan Africa

     Finding: Oxalate Al/Fe and exchangeable Ca as important as climate for SOC.

     Relevance: undermines clay-only index in SSA

     Evidence: full text

164. Lal (2001). Soil degradation by erosion

     Land Degradation & Development · 2001 · DOI · Full text and figures

     Study: Review

     Scale: Global

     Finding: Erosion-driven soil degradation and productivity loss.

     Relevance: framing

     Evidence: full text

165. Lal (2004). Soil Carbon Sequestration Impacts on Global Climate Change and Food Security

     Science · 2004 · DOI · Full text and figures

     Study: Perspective

     Scale: Global

     Finding: Agricultural/degraded soil sink capacity 50–66% of historic loss (42–78 GtC); +1 t C/ha in degraded cropland → +20–40 kg/ha wheat, 10–20 rice, 30–300 maize (approx.).

     Relevance: framing (classic claim; coefficients weakly evidenced)

     Evidence: full text

166. Montgomery (2007). Soil erosion and agricultural sustainability

     Proceedings of the National Academy of Sciences · 2007 · DOI · Full text and figures

     Study: Synthesis

     Scale: Global

     Finding: Conventional agriculture erosion exceeds soil production by 1–2 orders of magnitude.

     Relevance: framing

     Evidence: full text

167. Amundson et al. (2015). Soil and human security in the 21st century

     Science · 2015 · DOI · Full text and figures

     Study: Review

     Scale: Global

     Finding: Soil degradation as a threat to human security this century.

     Relevance: framing

     Evidence: full text

168. Smith et al. (2015). Global change pressures on soils from land use and management

     Global Change Biology · 2015 · DOI · Full text and figures

     Study: Review

     Scale: Global

     Finding: Global change pressures on soils from land use and management.

     Relevance: framing

     Evidence: full text

169. Borrelli et al. (2017). An assessment of the global impact of 21st century land use change on soil erosion

     Nature Communications · 2017 · DOI · Full text and figures

     Study: Global modelling (RUSLE)

     Scale: Global

     Finding: Land-use change effects on soil erosion.

     Relevance: framing / data

     Evidence: full text

170. Sanderman et al. (2017). Soil carbon debt of 12,000 years of human land use

     Proceedings of the National Academy of Sciences · 2017 · DOI · Full text and figures

     Study: Data-driven global model + HYDE

     Scale: Global

     Finding: Agricultural land use caused soil carbon loss of 133 Pg C (original); corrected to 116 Pg C in 2018 correction (10.1073/pnas.1800925115); hotspots in major cropping regions.

     Relevance: framing / data

     Evidence: full text

171. Lugato et al. (2018). Soil erosion is unlikely to drive a future carbon sink in Europe

     Science Advances · 2018 · DOI · Full text and figures

     Study: Biogeochemistry–erosion model (CENTURY + erosion) under climate projections

     Scale: EU agricultural soils

     Finding: C input gains offset by higher heterotrophic respiration; with erosion feedbacks EU agricultural soils are a net C source of 0.92–10.1 Tg C/yr over 2016–2100 (vs 1.8 Tg C/yr without erosion).

     Relevance: supports (projected cropland SOC loss) / methods (erosion term)

     Evidence: full text

172. Wuepper et al. (2019). Countries and the global rate of soil erosion

     Nature Sustainability · 2019 · DOI

     Study: Econometric (discontinuity)

     Scale: Global

     Finding: Country policies/institutions shape erosion rates.

     Relevance: framing

     Evidence: metadata verified; annotation from abstract/prior knowledge - verify

173. Olsson et al. (2019). Land degradation

     IPCC Special Report on Climate Change and Land (Cambridge Univ. Press 2022) · 2019 · DOI · Full text and figures

     Study: Assessment chapter

     Scale: Global

     Finding: IPCC SRCCL Ch.4: land degradation status, drivers, climate interactions.

     Relevance: framing (authoritative)

     Evidence: full text

174. Alewell et al. (2020). Global phosphorus shortage will be aggravated by soil erosion

     Nature Communications · 2020 · DOI · Full text and figures

     Study: Global modelling

     Scale: Global croplands

     Finding: Erosion contributes >50% of P losses; P depletion 4–19 kg/ha/yr under shortage.

     Relevance: framing

     Evidence: full text

175. Borrelli et al. (2020). Land use and climate change impacts on global soil erosion by water (2015-2070)

     Proceedings of the National Academy of Sciences · 2020 · DOI · Full text and figures

     Study: Global projection (RUSLE-based × SSP-RCP)

     Scale: Global

     Finding: Current water erosion ~43 Pg/yr; erosion projected to increase +30 to +66% by 2070 depending on scenario; conservation agriculture mitigates.

     Relevance: framing / methods (erosion trajectory)

     Evidence: full text

176. Evans et al. (2020). Soil lifespans and how they can be extended by land use and management change

     Environmental Research Letters · 2020 · DOI · Full text and figures

     Study: Global analysis

     Scale: Global

     Finding: Soil lifespans; many eroding soils <100 years.

     Relevance: framing

     Evidence: full text

177. Amelung et al. (2020). Towards a global-scale soil climate mitigation strategy

     Nature Communications · 2020 · DOI · Full text and figures

     Study: Perspective

     Scale: Global

     Finding: Cropland soils with large yield gaps/SOC losses have the greatest sequestration potential.

     Relevance: framing

     Evidence: full text

178. Bossio et al. (2020). The role of soil carbon in natural climate solutions

     Nature Sustainability · 2020 · DOI · Full text and figures

     Study: Global analysis

     Scale: Global

     Finding: Soil carbon in natural climate solutions (~25% of NCS potential).

     Relevance: framing

     Evidence: full text

179. Prăvălie (2021). Exploring the multiple land degradation pathways across the planet

     Earth-Science Reviews · 2021 · DOI · Full text and figures

     Study: Review/global assessment

     Scale: Global

     Finding: Multiple land degradation pathways mapped globally.

     Relevance: framing

     Evidence: full text

180. Bezner Kerr et al. (2022). Food, Fibre and Other Ecosystem Products

     Climate Change 2022: Impacts, Adaptation and Vulnerability (IPCC AR6 WGII; CUP 2023) · 2022 · DOI · Full text and figures

     Study: Assessment chapter

     Scale: Global

     Finding: IPCC AR6 WG2 Ch.5: food, fibre, ecosystem products; soil degradation treated qualitatively in yield projections.

     Relevance: framing (gap statement)

     Evidence: full text

181. Wang et al. (2023). Temperature effect on erosion-induced disturbances to soil organic carbon cycling

     Nature Climate Change · 2023 · DOI · Full text and figures

     Study: Model/observational

     Scale: Global

     Finding: Temperature modulates erosion-induced disturbance to SOC cycling.

     Relevance: methods

     Evidence: full text

182. Villela et al. (2026). Soil carbon debt from land use change in Brazil

     Nature Communications · 2026 · DOI · Full text and figures

     Study: Inventory/modelling

     Scale: Brazil

     Finding: Soil carbon debt from land-use change in Brazil.

     Relevance: data (tropical cropland trajectories)

     Evidence: full text

183. Liao et al. (2026). Governing black soils for food and climate security

     Nature Communications · 2026 · DOI · Full text and figures

     Study: Perspective/analysis

     Scale: Global black soils

     Finding: Governance of black soils for food and climate security.

     Relevance: framing

     Evidence: full text