Provisional research output.The underlying studies are being prepared for peer review and values may change. Not for investment or crediting decisions.
AfriERWReadiness Assessment

Methods

Everything on this site is computed from public data by a reproducible build. This page documents each step: the data, the scoring rules, the uncertainty analysis, and how feedstocks, laboratories, Article 6 and field evidence were coded.

Scope and sources

The assessment covers all 54 African countries. It draws on five studies carried out at the Net-Zero Carbon Emission Lab (NCEL), KNUST: three country studies (a readiness assessment, an evidence-graded feedstock register, and an assessment of laboratory, MRV and Article 6 readiness), an integrated study that combines them into the chain of conditions, and a systematic map of ERW evidence across the tropics. The site reads a versioned export of each study (data revisions: readiness 2026-11-04, feedstocks 2026-11-04, laboratories and Article 6 2026-11-04, chain of conditions 2026-11-13, tropical evidence 2026-11-05), and each export is checked against its manifest of row counts and checksums before the site is built. Only public sources were used. Data were captured on 3 November 2026, the South African laboratory census on 4 November 2026 and the social-licence provisions on 13 November 2026. Every dataset is listed on the data and sources page with its licence.

Readiness framework

Readiness describes whether a country's conditions favour starting and verifying enhanced rock weathering (ERW). It is built from 16 performance indicators in 7 dimensions: feedstock, soil and climate, agronomic fit, deployment and net-removal efficiency, MRV capacity, policy and markets, social licence. Literature-based and market indicators are kept out of the readiness index and reported on a separate evidence axis, so that what has been measured is not confused with whether conditions are favourable.

Indicators

Readiness indicators
DimensionIndicatorMeasureScaling to 0 to 5Why it matters
D1 FeedstockF1 Mafic rock near cropland% of cropland within 50 km of mafic rock (GLiM vb, pb, mtam, mtgr)linear 0 to 50%Short haul distances lower cost and transport emissions, which dominate ERW supply-chain emissions after grinding
D1 FeedstockF3 Mining and mine-residue footprintmapped mining area (Maus et al. v2)quintilesExisting extraction, crushing and residue stocks indicate supply capacity; polygons are dominated by metal mines, so this is a proxy
D2 Soil and climateS1 Acidic cropland share% cropland with pH(H2O) < 5.5, 0 to 30 cmlinear 5 to 50%Acidic soils gain most from the liming effect; note that strong acids also consume alkalinity before export (S3)
D2 Soil and climateS2 Weathering climateMAT 10 to 25 C and P/PET 0.2 to 1.0 on croplandcell score, cropland-weightedWarm, moist conditions raise dissolution rates
D2 Soil and climateS4 Drainage and export pathwayP minus AET (Fu-Budyko)linear 50 to 500 mmDrainage is needed to carry bicarbonate out of the root zone
D3 Agronomic fitA1 Cropland areacropland (log)quintilesDeployment scale supports project economics; size bias tested by dropping A1 and by size-controlled validation
D3 Agronomic fitA2 Lime-responsive crop mixshare of harvested area in lime-responsive cropsquintilesAgronomic co-benefits drive farmer uptake
D3 Agronomic fitA3 Fertiliser useN+P2O5+K2O per haquintilesProxy for input-market access and application logistics; direction tested by reversing it
D4 Deployment and net-removal efficiencyL1 Road accessGRIP4 road density on croplandquintilesHaulage of tens of tonnes per hectare needs roads
D4 Deployment and net-removal efficiencyL2 National logistics performanceLPI 2023 (2022 survey)linear 1 to 5Logistics quality affects delivered cost
D4 Deployment and net-removal efficiencyE1 Grid carbon intensityg CO2 per kWhlinear 700 to 100Grinding emissions reduce net removal; merged with logistics so it weighs 1/21 rather than 1/8 of the composite
D5 MRV capacityM1 Accreditation infrastructureILAC MRA statusrubric 0/2/3/5Accredited laboratories are required by crediting protocols
D5 MRV capacityM2 Soil laboratory networklaboratories in AFRILAB or NENALAB meetings (2022)count: 0 to 4 as counted, 5 if 6 or moreLaboratory capacity for soil and geochemical analysis; incomplete self-reported register, hence C
D6 Policy and marketsP1 Article 6 engagementDNA, bilateral agreements, pipeline activity0 to 3 rescaledHost-country authorisation routes
D6 Policy and marketsP4 Governance qualityWGI regulatory quality and rule of lawlinear -2.5 to 2.5Contract enforcement and regulatory reliability
D7 Social licenceT1 Tenure securityshare of adults insecure about main property (Prindex)linear 0 to 50% (inverse)Secure tenure underpins farmer participation and benefit sharing

Indicators reported but not scored:

Context indicators
IndicatorRoleMeasureRule
P3 Market activityvalidation and contextregistry listing, credit sale, offtake or company-reported operationissued credits 5; registry listing verified 4; reported sale or offtake 3; company-reported operations 2; none 0
T3 Population near minescontextGHS-POP within 5 km of mine polygonsinverse quintiles (sensitivity only)
PV Political stabilitysensitivityWGI PVlinear -2.5 to 2.5
S3 Soil-barrier flagcontextshare of acidic cropland with beta at or below the boundary, doses 10 to 50 t/hacontext
P2 Policy recognition of CDR or ERWstatusNDC textconstant: no NDC qualifies
M4 Water monitoring baselinestatusopen river alkalinity stations (GFQA)missing in all states

Scoring and aggregation

Each indicator is scaled to 0 to 5 by a rule fixed before scoring: linear between stated bounds, quintiles across countries, or a rubric. For a linear rule with bounds ai and bi (with bi<ai where lower values are better),

si=5min(1,max(0,xi−aibi−ai))(1)

A dimension score is the mean of its nj indicator scores:

Dj=1nj∑i∈jsi(2)

The readiness index is the weighted geometric mean of the dimension scores, each floored at 0.1 so that a zero does not collapse the product:

R=exp(∑jwjlnmax(Dj,0.1)∑jwj)(3)

The published weights are equal (wj=1/7). A geometric mean is only partly compensatory: a strong dimension cannot fully offset a weak one. Arithmetic and minimum aggregation are offered as alternatives on the readiness page. Alternative weight sets used for sensitivity:

Weight sets
DimensionEqualPrincipal componentsEntropyBiophysical emphasisInstitutional emphasis
D1 Feedstock0.1430.0820.1750.2000.120
D2 Soil and climate0.1430.0640.2660.2000.120
D3 Agronomic fit0.1430.2690.0510.2000.120
D4 Deployment and net-removal efficiency0.1430.0380.0490.1000.120
D5 MRV capacity0.1430.2790.3090.1000.200
D6 Policy and markets0.1430.2670.1150.1000.200
D7 Social licence0.1430.0010.0360.1000.120

Missing values. A missing indicator score is filled with the cross-country median of observed scores for the index only; it is shown as missing and marked imputed everywhere on the site. 16 countries have at least one imputed value. Each affected country profile reports the rank it would have if the imputed values were low or high.

Evidence grades and data coverage

Every indicator value carries an evidence grade:

  • Evidence grade A Measured national data from an authoritative open source
  • Evidence grade B Modelled or derived from open global data, or a verified register search
  • Evidence grade C Proxy, partial coverage, or a literature search that found nothing
  • Evidence grade D Weak proxy or grey evidence only
  • Evidence no data No data

A dimension's grade is the lower median of its indicator grades. The data-coverage check requires at least five of the 7 dimensions to be graded A or B; 53 countries pass. A country below the threshold is still shown, with a note.

Uncertainty and tiers

Rank uncertainty is estimated with 10,000 Monte Carlo draws (seed 20261103). In each draw:

  1. every observed indicator score is perturbed by a beta distribution that keeps its mean,
    si′=5X,X∼Beta(μκ,(1−μ)κ),μ=si5,κ=μ(1−μ)v−1(4)
    where the variance v on the unit scale follows from the standard deviation σg of the evidence grade (A 0.25, B 0.5, C 0.75, D 1.0 on the 0 to 5 scale), capped so that both beta parameters stay positive (μ is held between 0.005 and 0.995):
    v=min((σg5)2,0.95μ(1−μ))(5)
  2. each missing score is replaced by a randomly chosen observed score of the same indicator;
  3. the J dimension weights are drawn from a Dirichlet distribution centred on the chosen weights w¯j, with concentration α=5.0 per dimension (equal weights give Dir(α,…,α)):
    (w1,…,wJ)∼Dir(αJw¯1,…,αJw¯J)(6)
  4. the readiness index and the ranks of all 54 countries are recomputed.

A country's tier is set by its median rank: Tier 1 for ranks 1 to 10, Tier 2 for 11 to 25, Tier 3 for 26 to 40 and Tier 4 for 41 to 54. Each profile also shows the probability of being in that tier, the 5 to 95% rank range and the probability of being in the top ten. A country is robustly in the top ten if that probability is at least 0.8; 5 countries are. Wider weight uncertainty (α = 1 with random aggregation), noise only, narrower weights (α = 20) and uniform imputation were run as sensitivity cases.

Analysis in your browser

With the published settings, the readiness page shows the published results exactly. When you change the weights or the aggregation rule, the same procedure is repeated in your browser: weights are drawn from a Dirichlet distribution centred on your weights (concentration 5.0 per dimension; a weight of zero removes the dimension), beta noise and imputation are applied as above, and 4,000 seeded draws are run in a background worker. Beta and Dirichlet variates are generated from gamma variates (Marsaglia and Tsang method), so individual draws differ from the published run while the distributions are the same; at equal weights the browser results agree with the published tiers within Monte Carlo error.

Evidence axis

Five evidence indicators describe what has been measured on ERW in each country:

Evidence indicators
DomainIndicatorMeasureScoring
feedstockF2 Feedstock characterisationin-country characterisation of candidate rockcatalogue or >=3 peer-reviewed 5; 2: 4; 1: 3; grey only 2; none 0
feedstockF4 Contaminant screeningNi, Cr, asbestos or radionuclide screening of candidate rock>=2 peer-reviewed 5; 1: 3; grey only 2; none 0
fieldM3A Field evidence with ERW-relevant rockin-country field or pot trials with mafic, ultramafic or alkaline silicate rock>=3 peer-reviewed field 5; 2: 4; 1: 3; pot only 2; grey only 1; none 0
fieldM3B Measured CO2 removalin-country peer-reviewed measurement of CO2 removal (cropland or mine residue)>=2 peer-reviewed 5; 1: 4; grey only 2; none 0
socialT4 Social evidencepublished perception or acceptance study of ERW>=2 peer-reviewed 5; 1: 4; grey only 2; none 0

The evidence score is the mean of the five evidence indicator scores ek,

E=15∑k=15ek(7)

The evidence class counts the credibility domains (feedstock, field, social) with peer-reviewed evidence (grade B). The credibility gate requires at least one such domain (8 countries) and the strict gate two (3 countries). Evidence records come from a systematic literature search and the feedstock register; rock that is not an ERW feedstock (granite, micaschist, unidentified quarry dust) is recorded but not counted.

Feedstock register

Candidate rocks and mine residues were identified by a systematic search of peer-reviewed and grey literature, theses, survey reports and company material, and each record was screened and graded. Deposits are placed on a cumulative evidence ladder:

  • Tier 0: Mapped or proxy occurrence
  • Tier 1: Described rock or residue with identity and location
  • Tier 2: Plus measured bulk major oxides (at least CaO and MgO)
  • Tier 3: Plus quantitative mineralogy (normally qXRD)
  • Tier 4: Plus measured particle size and screening for all four contaminant groups

Evidence records for deposits use their own grades:

  • Evidence grade A Peer-reviewed article read in full
  • Evidence grade B Peer-reviewed article assessed from the abstract or partial text
  • Evidence grade C Grey literature, theses, conference abstracts, preprints or company material
  • Evidence grade D Proxy geochemistry (tailings by commodity), not deposit-specific

Theoretical potential. For samples with measured major oxides, the oxide-based potential follows Renforth (2019):

Epot=MCO2100(CaOMCaO+MgOMMgO+Na2OMNa2O+K2OMK2O−SO3MSO3−P2O5MP2O5)×103η(8)

in kg CO2 per tonne of rock, with each oxide in wt% and M its molar mass in g/mol (MCO2=44.009); sulphate and phosphate are subtracted because they consume alkalinity, and missing oxides count as zero. The factor η accounts for carbonate-system buffering in seawater: 1.5 by default (range 1.4 to 1.7); η = 2 is the stoichiometric maximum before buffering. Analyses are used only if they pass quality checks: a stated oxide basis and method, and plausible values. 10 of 13 samples pass; the excluded analyses and the reasons are listed on the feedstocks page. For deposits with quantitative mineralogy, a reactive-phase potential counts only cations held in reactive minerals. The realised range applies an efficiency distribution for the rock type.

Contaminant screening. For rocks with measured trace elements, the soil concentration after n applications is estimated by mass balance,

Csoil(n)=b+(Crock−b)nAnA+Msoil(9)

with A=20 t/ha per application mixed into the top 20 cm (bulk density 1.3 t/m3, so Msoil=2600 t/ha), no leaching or crop uptake, and a background b of zero (half the limit as a variant). Concentrations are compared with the soil limits of Puro.earth (Table 4.3, mg/kg: As 5, Cd 1, Cr 100, Co 20, Cu 100, Pb 60, Ni 150, V 100, Zn 200). An element passes if the soil stays below its limit after ten applications (200 t/ha in total). This is a descriptive screen, not a protocol test.

Priority deposits are described deposits (Tier 1 or 2, grades A to C) of mafic, ultramafic or alkaline rock with a known location and at least 1 km2 of acidic cropland within 50 km: the candidates where further characterisation would add most. Deposits known only from an agronomic trial site are listed separately. Deposits within 50 km of each other are grouped into clusters, and cropland areas are counted once per cluster. The geochemical knowledge layer counts published whole-rock analyses from GEOROC near cropland.

Laboratories and measurement

Accreditation coverage (M1) is 5 for a national body that is a full ILAC signatory, 3 for coverage by a regional signatory body (SADCAS or SOAC-WAAS), 2 for an associate member and 0 otherwise. Laboratories were taken from the published registers and schedules of accreditation bodies; the completeness of each body's census is listed on the laboratories page. Each entry is assigned a scope tier: soil chemistry (soil matrix with pH, nutrients, extractable or total major cations, or carbon); major-element geochemistry (rocks, ores or soils analysed for CaO and MgO, or Na and K, by fusion XRF, ICP or AAS); assay (precious or base metals only); environmental soil (contaminants only); or outside ERW scope. The status (current, expired, suspended) is read from the certificate or register.

Domestic measurement capacity requires accreditation coverage (M1 > 0) and a current accredited soil-chemistry laboratory in the country. Regional measurement access requires accreditation coverage and such a laboratory in the country or in another member state of a regional economic community it belongs to (SADC, EAC, ECOWAS, UEMOA, ECCAS, AMU or IGAD). Shares carry Clopper-Pearson 95% intervals: for k countries of n, with B(q;a,b) the q quantile of a beta distribution and α=0.05,

[B(α2;k,n−k+1),B(1−α2;k+1,n−k)](10)

with the lower bound 0 when k=0 and the upper bound 1 when k=n. Protocol routes count countries where a protocol's laboratory requirement could be met in country or in the region; the Puro.earth and Rainbow routes are upper bounds because they accept evidenced quality management.

Article 6 indicators

Article 6 engagement (P1) combines the designated Article 6.4 authority, bilateral agreements and pipeline activities. The national carbon-market framework is coded 0 (none found), 1 (draft or strategy), 2 (policy or guidelines) or 3 (act or regulation) from a dated search for each country. Institutional arrangements (P6) add one point each for a listed authority, an authorising body in the 2026 implementation report, published host-Party participation requirements and a CARP initial report, plus one point for a registry in place (half for one in progress). Correlations with biophysical suitability are Spearman coefficients with 5,000 bootstrap resamples.

Field studies

The register of African ERW studies combines a search of OpenAlex, Crossref and Semantic Scholar in English and French with the records of the readiness and feedstock studies. Whether a study quantifies CO2 removal was first coded from titles and abstracts and then checked against the full text where one could be obtained (15 of 34 counted studies). Removal is coded Yes if a removal value is reported, Partial if removal-relevant measurements are reported without a removal value, and No otherwise.

Chain of conditions

The integrated study follows each country along seven links from rock to credit to consent. Each link is coded met or not met under a main threshold, and each value carries an evidence grade. On this site the links are coded C1 to C7; in the study they are L1 to L7 (the readiness index already uses L1 and L2 for logistics indicators).

Links of the chain
LinkMain conditionCode in the study
C1 RockMafic rock mapped within 50 km of cropland, or a documented feedstock depositL1
C2 Soil and climateAt least 10% of cropland with topsoil pH below 5.5L2
C3 Characterised feedstockA deposit with measured CaO and MgO (characterisation Tier 2 or above)L3
C4 Field evidenceA peer-reviewed field or pot trial with ERW rock, or a measured removal, in the countryL4
C5 MeasurementAccreditation coverage and an accredited soil-chemistry laboratory at home or in a regional partner stateL5
C6 Crediting and authorisationAn Article 6.4 national authority and a carbon-market policy or legal instrument in forceL6
C7 Social licenceBenefit-sharing and community consent or participation provisions in forceL7

Countries meeting C1 and C2 are called biophysically favourable (17 countries). The six core links are C1 to C6; C7 is reported separately because its evidence base is thinnest. Counts and shares carry Clopper-Pearson 95% intervals. The binding link of a country is the failed link with the lowest score on a common 0 to 5 scale, which does not assume an order. The link scores are

S1=D1,S2=D2,S3=1.25Tmax,S4=max(eM3A,eM3B)(11)
S5=12(M1+M2b),S6=13(P1+P5+P6),S7=mean(52b,52c,5g,D7)(12)

where D1, D2 and D7 are readiness dimension scores, Tmax is the highest characterisation tier of the country's deposits (0 to 4), eM3A and eM3B are the field-trial and measured-removal evidence scores, M1 and M2b are accreditation coverage and accredited laboratory capacity, P1, P5 and P6 are Article 6 engagement, the national framework and Article 6 arrangements, and b, c and g are the coded benefit-sharing (0 to 2), consent (0 to 2) and grievance (0 or 1) provisions; missing terms are left out of the mean. With F the set of failed links,

binding link=argmink∈FSk(13)

and ties are broken in the order C1 to C7. The first failing link in that fixed order is reported as a check.

Sensitivity. 864 combinations of thresholds were tested: acidic cropland share (5, 10 or 25%), characterisation (Tier 2, Tier 3, or Tier 3 with contaminants within limits), field evidence (any or trials only), laboratory route (regional, domestic, any ERW-relevant laboratory, or the Puro.earth equivalence route), crediting framework (authority only, any policy or legal instrument, or a binding legal instrument), social provisions (both or either) and the treatment of provisions that could not be determined. The threshold explorer on the chain page re-evaluates the links in your browser and reproduces every one of these variants.

Verifiable cropland and soil barrier

Favourable cropland is cropland (Copernicus land cover fraction, 2019) with topsoil pH(H2O) below 5.5 (0 to 30 cm, SoilGrids 2.0) and a warm-humid weathering climate (CHELSA V2.1):

Sclim=12[ℓ(T;10,25)+ℓ(AI;0.2,1.0)]≥2.5(14)

where ℓ(x;a,b) is the linear rule of equation 1, T is the mean annual temperature in °C and AI the aridity index. Verifiable cropland is favourable cropland within 50 km (great-circle distance) of a deposit characterised to Tier 2 or above with no known contaminant exceedance, in a country with domestic or regional access to an accredited soil laboratory; a strict variant requires Tier 3 and a domestic laboratory. Areas are computed on a 1/24° grid from cropland fractions and cell areas.

The soil-barrier ratio compares the exchangeable acidity and exchange pool of the soil with the charge supplied by a basalt dose d (t/ha; 7.5 kmol of charge per tonne):

β=kCEC7.5d,k=46.75,d=20tha−1(15)

with the cation exchange capacity CEC from SoilGrids 2.0 (0 to 100 cm) and k in the range 36.8 to 56.7. Where β>0.865 at 20 t/ha, alkalinity did not break through in the diagnostic, so early monitoring must rely on soil analyses. The cropland maps on the chain of conditions page show the classes of the 1/24° grid reprojected to the map frame (nearest cell, no smoothing); verifiable cells are enlarged for visibility.

For the 22 countries with a carbon-market policy or legal instrument in force, provisions were coded from the instruments and two comparative legal reviews: benefit sharing 2 (binding rule, defined share or mandatory agreement), 1 (general provision) or 0 (none identified); consent 2 (explicit consent), 1 (consultation or participation) or 0; grievance or dispute mechanism 1 or 0. Where the sources did not establish a provision it is coded as not determinable, treated as absent in the main analysis and as present in a sensitivity variant. Link C7 requires benefit sharing and consent or participation (both at least 1). Coded on 13 November 2026.

Markets and validation

ERW market activity is a registry listing or a company or press report of operations, credit sales or offtakes (graded C unless confirmed on a registry); it was reported in 7 countries. Necessary condition analysis tests whether a minimum level of each link score is required for market activity. With a binary outcome yi (activity reported or not) and link score xi, the ceiling-envelopment effect size is the share of the range of the score that lies below the lowest score among countries with activity,

d=mini:yi=1xi−miniximaxixi−minixi(16)

and its P value is the share of 5,000 permutations of y that give an effect size at least as large. Discrimination of market activity by the chain score without field evidence (to avoid circularity), the verification index, the readiness index and the biophysical sub-index is summarised by the area under the curve (AUC) with 2,000 bootstrap resamples. The verification index is the mean of the measurement and crediting link scores defined above,

V=12(S5+S6)(17)

Pathways

For each biophysically favourable country, each failed core link is mapped to an action: characterising a documented deposit to Tier 2 with contaminant screening, or sampling a mapped mafic unit where none is documented (C3); a public reference field site (C4); accreditation coverage or a new accredited soil laboratory (C5); a carbon-market policy or law (C6); and benefit-sharing and consent provisions (C7). The smallest number of new laboratories is found by exact set cover over regional economic communities, counting only countries with no accredited soil laboratory in any partner state. Costs are not attached, because comparable cost data were not available.

Unified evidence register

The three source studies searched OpenAlex, Crossref, Semantic Scholar and repositories (19,245 records identified, 15,250 unique after de-duplication by DOI and normalised title). Records that any study included as ERW evidence for an African country were merged into one register of 106 records (64 peer-reviewed), coded by evidence domain. Country-level evidence flags recomputed from the register reproduce those of the readiness study exactly.

Tropical evidence map

The tropical evidence map is a systematic map of ERW evidence in tropical environments. A Boolean search of OpenAlex, a semantic search, targeted website searches and a reference library yielded 862 coded records; the Boolean search retrieved 80.4% of 143 independently identified benchmark studies. 440 records are direct studies (390 at core tropical sites or under simulated tropical forcing, 50 tropical-adjacent) in 26 countries. Evidence is organised in 12 value-chain stages and 20 strata in five dimensions (climate, soil, crop system, farm scale and region), giving 240 stage-stratum cells.

Each cell takes the highest evidence-maturity level supported by a qualifying study: U, the stratum was not reported; 0, no eligible evidence; 1, inferred only; 2, controlled or early evidence; 3, early field evidence; 4, established field evidence. Levels 2 to 4 require a primary study at a core tropical site that measured the stage-defining outcome and reported the stratum; studies under simulated tropical forcing are capped at level 2, and modelling, reviews and standards support level 1 only. Sensitivity runs:

  • Main assessment: 113 of 240 cells at level 2 or above
  • Study design only (no outcome anchor): 129 of 240 cells at level 2 or above
  • Tropical-adjacent studies counted as direct: 118 of 240 cells at level 2 or above
  • Two concordant studies required for level 3: 86 of 240 cells at level 2 or above
  • Without the database-search records: 81 of 240 cells at level 2 or above
  • Unclassified cells treated as inferred: 113 of 240 cells at level 2 or above

Countries are normalised as on the study map: multi-country studies are placed at the first-named country, and studies located only by author affiliation are flagged. 39 direct studies are African by region; one of them is a multi-country study placed at Brazil, so African country profiles list 38. Research gaps are classified by the study needed to close them (tier 1 to 3).

Use of AI and verification

Literature screening and the coding of the crosswalk, laboratory register, policy indicators and field evidence were carried out with AI assistance and are recorded in the study code. A verification sheet covering every coded value that drives a headline result has been prepared for second coding by a person; until it is complete, coded values should be read as provisional. The status of each study is: readiness, provisional; manuscript in preparation for peer review; feedstocks, provisional; manuscript in preparation for peer review; laboratories and Article 6, provisional; manuscript in preparation for peer review; chain of conditions, provisional; manuscript in preparation for peer review; tropical evidence map, provisional; manuscript in preparation for peer review.

Limitations

  • National values hide large differences within countries; a low national score does not rule out good local sites.
  • Indicators describe readiness to start and verify ERW, not realised carbon removal.
  • Public registers and literature are incomplete; absence of evidence is graded, not assumed to be absence. The laboratory census is incomplete where accreditation bodies do not publish their scopes.
  • Theoretical potentials are upper bounds from rock chemistry. Contaminant screening uses a simple mass balance and published limits.
  • Article 6 data change monthly and protocols are revised often; results describe a snapshot.
  • The chain treats each condition as pass or fail at a threshold; the threshold explorer shows how results change. Market activity rests on company and press reports, and the necessary condition analysis on only 7 countries with activity.
  • Many legal provisions could not be determined from public sources and need checking against gazetted texts. Legal provisions are not evidence of community acceptance; the searches found no study of how African communities view ERW.
  • The soil-barrier diagnostic is a screening model parameterised from gridded soil data, not a substitute for site measurement.
  • The tropical evidence map is limited by what is reported: cells are unclassified where studies did not report the stratum.
  • Boundaries are for orientation only and imply no position on any territorial claim.

Versions and citation

The release notes and the recommended citation are on the data and sources page.