A geospatial ex-ante profitability model for Enhanced Rock Weathering (ERW) with basalt across Sub-Saharan African cropland.
🚧 Code release pending. The analysis behind this repository is still being finalised. This repository currently documents what we model, how the pipeline is structured, and preliminary profitability results. The full R pipeline, parameter catalogue, and final outputs will be published here once the working paper is complete. Watch/star the repo to be notified.
We are building a map of where in Sub-Saharan Africa it would make economic sense to spread crushed basalt rock on cropland to (a) reduce soil acidity for farmers and (b) pull carbon dioxide out of the atmosphere at the same time. For every grid cell of cropland across 47 SSA countries and 23 crops, the model answers: if a farmer spread basalt rock dust on this field, would the carbon-removal payments plus the yield improvement outweigh the cost of buying, grinding, transporting, and spreading the rock? The model answers this under different policy assumptions — the carbon price, the cost of measurement, reporting and verification (MRV), the choice and grind size of the rock — and ranks the most promising locations first.
Holding warming below 2 °C requires removing CO₂ from the atmosphere, not just emitting less (IPCC AR6 Synthesis, 2023). Enhanced rock weathering — spreading finely ground silicate rock, almost always basalt, on farmland — is one of the few removal pathways that is:
- Physically scalable — global-cropland deployment estimates reach 0.5–2 Gt CO₂ yr⁻¹ by 2050, with the highest per-hectare potential in the warm-humid tropics (Beerling et al. 2020);
- A co-benefit for farmers — basalt acts like agricultural lime, reducing soil acidity and raising yields. A US Corn-Belt trial reported ~10.5 t CO₂ ha⁻¹ removed alongside a 12–16 % yield uplift over four years; a 2024–25 smallholder trial in Kisumu County, Kenya reported 71 % (year 1) and 79 % (year 2) maize-yield increases from a single 20 t ha⁻¹ basalt application;
- Deployable with existing infrastructure — quarries, lime spreaders, and rural logistics chains already exist;
- Verifiable — outcomes that carbon-credit buyers are willing to pay for.
Sub-Saharan Africa is a uniquely promising setting: its soils are widely acidic and lime-poor (so the agronomic benefit is unusually large), its climate is warm and humid (which makes the rock dissolve fastest and remove the most carbon), and its farmers are largely cut off from carbon markets (so ERW would create a new income stream rather than crowd out an existing one).
But results from field trials are highly heterogeneous — climate, soil pH, and drainage matter enormously. So for a policy-maker or investor, the right question is not "is ERW a good idea in general?" but: at what carbon price, in which country, on which crop, would ERW actually be profitable today? That is the question this model answers, location by location.
For every cropland pixel the model computes a per-hectare gross margin of a basalt-application programme, decomposed into three streams that converge on one accounting identity:
gross margin = CDR credits (t CO₂ removed × carbon price − MRV cost)
+ yield uplift (Δ yield × crop producer price)
− delivered cost (quarry + grinding + transport + spreading)
The model runs three temporal regimes in parallel — year-1, NPV at 10 % over 10 years, and equilibrium (saturating CDR) — with carbon removal phased 30/25/20/15/10 % over five years after application.
| Block | What it does |
|---|---|
| B1 — Working grid | SSA cropland mask (GADM v4) with depth-weighted SoilGrids soil properties (pH, bulk density, exchangeable acidity, exchangeable bases) co-registered with SPAM v2 harvested area, production, and yield for 23 crops. |
| B2 — Basalt rate | How much basalt does each pixel need? Lime-equivalent CaCO₃ requirements (Kamprath, Cochrane, and LiTAS-targeted rules) converted to t basalt ha⁻¹ using feedstock CaO/MgO chemistry, a trial-calibrated reactive fraction, and a grain-size lever (10–500 µm). Uniform 10/20/50 t ha⁻¹ counterfactuals run in parallel. |
| B3 — CDR surface | How much CO₂ does a tonne of basalt remove here? Per-tonne CDR potential from CaO/MgO stoichiometry, scaled by a temperature × moisture climate factor and a soil-pH term, with an aridity-based deduction for alkalinity that precipitates locally as pedogenic carbonate instead of reaching the ocean. |
| B4 — Logistics | Where does the rock come from and what does hauling it cost? Basalt and gabbro outcrops from the GLiM global lithological map, cost-distance routing over the Malaria Atlas Project motorised friction surface, monetised per tonne-minute, producing a delivered-price surface at the farm gate. |
| B5 — Energy | Per-country industrial electricity prices and grid carbon intensities, driving both the grinding cost and the grinding-emissions (LCA) penalty on net CDR. |
| B6 — Agronomic response | Per-crop yield response to acidity correction: EcoCrop pH and acidity-saturation tolerance curves for all 23 crops, plus a Bayesian hierarchical model fit to a curated ERW field-trial dataset that produces posterior yield-uplift surfaces where trial evidence supports them. |
| B7 — Profitability | The accounting core: decomposed costs, carbon credits net of MRV, yield revenue at FAOSTAT 2016–20 producer prices, across three temporal regimes and four basalt-allocation rules. |
| B8 — Deployment frontier | Pixels ranked by gross margin per tonne of basalt, producing supply curves and deployment maps under annual basalt-supply caps from 1 to 500 Mt yr⁻¹. |
| B9 — Sensitivity | Five-axis robustness scan: carbon price grid, yield-uplift multipliers, MRV cost (0–80 $/t CO₂), grain size (10–500 µm), and allocation rule — all crossed with the three temporal regimes and 23 crops. |
The policy and engineering levers exposed by the model:
| Variable | Default |
|---|---|
| Feedstock chemistry (CaO / MgO mass fractions) | 10 % / 7 % |
| Grain size | 100 µm |
| Basalt allocation rule | soil-targeted (vs uniform 10/20/50 t ha⁻¹) |
| Carbon price | $150 / t CO₂ |
| MRV cost | $30 / t CO₂ |
| Discount rate (NPV regime) | 10 % |
All inputs are public datasets:
| Dataset | Role |
|---|---|
| GADM v4 boundaries | 47-country SSA mask |
| SoilGrids Africa (30 arc-sec) | soil pH, bulk density, exchangeable acidity and bases |
| SPAM v2 | harvested area, production, yield for 23 crops |
EcoCrop (via Recocrop) |
crop pH / acidity tolerance curves |
| Curated ERW field-trial dataset | Bayesian yield-uplift model |
| GLiM v1.0 lithology (Hartmann & Moosdorf 2012) | basalt and gabbro source outcrops |
| Malaria Atlas Project friction surface (2019) | motorised travel-time routing |
| WorldClim v2 | temperature and precipitation for weathering kinetics |
| IEA / Ember / GET.invest (2022–24) | country electricity prices and grid carbon intensity |
| FAOSTAT producer prices (2016–20) | crop revenue valuation |
Preliminary outputs from the current model run (NPV regime, soil-targeted allocation, carbon price $150 per t CO₂, MRV $30 per t CO₂ unless noted). Maps and numbers may change before the final release.
Where ERW pays. The flagship output: the crop-weighted gross margin of a basalt programme on every cropland pixel, combining carbon credits, yield uplift, and delivered costs.
Profitable cropland — pixels where the combined gross margin is positive at current assumptions. Profitable clusters concentrate where acidic soils, humid climate, and nearby basalt coincide: the Guinea coast, the Ethiopian and East African highlands, the Congo-basin margins, Madagascar, and parts of southern Africa.
Breakeven carbon price — the carbon price at which each pixel turns profitable. The best locations break even well below $100 per t CO₂; remote or weakly responding areas need several hundred dollars.
Per-crop gross margins — the same accounting crop by crop. Producer prices and acidity response shift the profitable frontier from one crop to the next; maize (SSA's most widely grown staple) and groundnut shown here.
The full preliminary set — three temporal regimes (year-1 / NPV / equilibrium), four allocation rules, per-crop gross margins, profitable-area masks, and breakeven carbon prices — is in docs/maps/profitability/. The twelve input-indicator maps behind these results (soil pH, basalt requirement, CDR surfaces, logistics, energy) are in docs/maps/.
| Output | Description |
|---|---|
| ERW profitability maps | Per-pixel gross margin, decomposed into agronomic-only, CDR-only, and combined terms — 4 allocation rules × 3 temporal regimes × 23 crops. |
| Supply curves & deployment frontiers | Marginal cost per t CO₂ and total removal as a function of the annual basalt-supply cap (1–500 Mt yr⁻¹), with deployment masks. |
| Sensitivity tables | Profitable area and gross margin under sweeps of carbon price, MRV cost, yield uplift, grain size, and allocation rule. |
| Full R pipeline | The reproducible geospatial pipeline (all model code), with a parameter catalogue citing every constant and assumption. |
| Documentation | A plain-language tutorial, living model documentation, and an evidence review of the ERW literature. |
- The Bayesian yield-response model is fit on a still-small trial dataset (~14 trials) and is prior-dominated; it grows as new ERW field trials are published. The EcoCrop tolerance curves serve as the fallback response.
- Soil-moisture is proxied by mean annual precipitation rather than a full aridity index (P/PET).
- Compaction emissions from basalt traffic and transient N₂O effects of tilling-in fines are not modelled.
- Country-level carbon-credit eligibility (Article 6 / corresponding adjustments) is not modelled.
- Land tenure and aggregator-finance structure (who pays the upfront basalt cost, who owns the credit) are outside the model.
- Industrial by-product feedstocks (steel slag, cement-kiln dust, mine tailings) are not represented and would materially shift the supply geography.
ex-ante-erw/
├── README.md ← you are here
├── LICENSE ← MIT
├── CITATION.cff
├── erw/ ← R pipeline scripts (placeholder — released with the working paper)
├── data/ ← input & intermediate rasters (placeholder — fetch scripts to follow)
└── docs/ ← figures and documentation
├── erw-conceptual-framework.{png,svg}
├── erw-pipeline-flow.{png,svg}
└── maps/ ← 12-panel indicator-map gallery
└── profitability/ ← preliminary gross-margin, breakeven, and profitable-area maps
If you reference this work before the working paper is out, please cite the repository (see CITATION.cff).
Bisrat Gebrekidan — b.gebrekidan@cgiar.org
This repository is a public placeholder for an analysis in progress. Figures and numbers shown here are preliminary and may change before the final release.