Saibro plans the soil sampling rounds of enhanced rock weathering projects: MRV software for the lead who turns lab sheets into a claim
From your baseline fingerprint, rock analysis, spreading log, rainfall and lab sheets, Saibro's weathering model forecasts every field with a range. It sends the next round's cores where it is least sure, as many as each Unit needs to reach the protocol's power. The claim is then computed from the cores alone.
A noisy round one today means re-coring at several times the density.
Book a call before your next round
Below ten samples a hectare, one field's weathering is hard to tell from noise
10per ha
samples, with over 100 t/ha of rock, for one field's mass balance error to average under 15% (five US fields, 2025)
Noisy fields stay in the claim, since a protocol cannot drop a field for showing no signal. Credits issue at the 30th percentile, so noise costs tonnes.
Example: ten a hectare on 3,000 ha is 30,000 cores a round.
Round two's cores go where the model is least sure, and stop at the protocol's power
Example project, Unit 4 (olive groves, 290 ha): 470 cores for round two, most in the North and East fields. That is under two a hectare, because the protocol sets power per Unit, not per field. A seeded random draw places each core so your verifier can redraw it, and some land on ground the model already knows. Each round's cores refit the model.
The plan stops at 80% power: if the rock weathered as much as round one measured, a round this size detects it four times in five.
Terradot's platform draws on 75,000+ soil samples gathered over five years. Saibro starts from published trials and your cores, so first-season ranges run wide and plans larger.
- Land-use historyten years a field, read with soil maps; names and parcel numbers are stripped first
- 90% rangeeach field's weathering estimate, the band nine cores in ten should land in
- Stays in the claimevery noisy field and every outlier; no plan is sized on the forecast
- Approve plannothing reaches the field app until you press it
$6 a hectare spread, per reporting period
One round alone is $2 a hectare in the Units it covers; from 10,000 ha a period, $4.50. You pay on hectares, known before the round: the model planning your evidence never earns more from a bigger claim, and a Unit whose cores find little costs the same.
Up to 9,999 ha a period · $6 a hectare
From 10,000 ha a period · $4.50 a hectare
$6 per hectare / reporting period | One round $2 per ha in the round's Units One sampling round for the Units you choose, usually the ones round one left too noisy to call. Round one came back noisy and round two is soon Book a call | Early access Reporting period $6 per ha spread / reporting period Every round of a reporting period, and the claim at its end with the verifier's files. We want every round and the claim in one place Book a call | Portfolio $4.50 per ha / period, from 10,000 ha Several projects from 10,000 hectares a period, with the Puro export and our scientist on every plan. We run more than one project Book a call |
|---|---|---|---|
| Sampling plan per Unit | Units you choose | Every Unit | Every Unit, every project |
| Weathering forecast per field | Yes | Yes | Yes |
| Claim at the protocol's percentile | No | Yes | Yes |
| Verifier's file pack | No | Isometric v1.2 | Isometric v1.2 and Puro export |
| Held-out model check | Before the plan | Every round and period | Every round and period |
| Seats for field team and lab | Unlimited | Unlimited | Unlimited |
| Rounds | One | Up to four a period | Up to four a period, per project |
| Plan review with our scientist | No | On request | Every plan |
| In the plan |
|
|
|
A first plan for up to three of your noisiest Units, on the call, before any contract
No charge for
- Seats for your field team, lab and verifier
- Lab sheets read, every round
- Claims re-run after a protocol update
- Exports for the registry and the ERW Data Quarry
Example: 2,980 ha is $17,880 for the period, or $1,860 for one round on three noisy Units (930 ha).
For scale, MRV averaged GBP 148 a tonne in Frontier's applicant data. Divide $6 by your own tonnes a hectare.
Round planner · Sample project
| Unit | Signal after round one | Cores in round two | Status |
|---|---|---|---|
| Unit 1, maize, 520 ha | Clear | 0 | No round two needed |
| Unit 2, maize, 470 ha | Clear | 0 | No round two needed |
| Unit 3, maize, 380 ha | Too noisy to call | 340 | Plan waiting for you |
| Unit 4, olives, 290 ha | Too noisy to call | 470 | Plan waiting for you |
| Unit 5, olives, 610 ha | Clear | 0 | No round two needed |
| Unit 6, olives, 260 ha | Too noisy to call | 360 | Approved |
| Unit 7, maize, 450 ha | Clear | 0 | No round two needed |
Sample project after round one: every Unit, its signal and its cores for round two
In the example project, four Units came back clear and three need round two
2,980 ha of maize and olive groves in the Ribatejo and the Alentejo, seven Units. Each with its signal after round one, the cores the model proposes for round two and where its plan stands. On a clear Unit, Saibro shows what ten more cores would likely add.
Unit 6 is approved. Units 3 and 4 wait for you.
A plan, not a claim: the cores are still the evidence, and nothing here is credited until they are measured.
Every core in the verifier's files traces back to the point it was drawn at
From the example project: controls matched on the baseline fingerprint, round two at the lab, the check the model must pass each period on fields it never saw, and the files themselves.
Controls matched on the baseline fingerprint: pH, inorganic carbon, texture, CEC, base saturation, cations and tracers.
The lab round as it comes back: each Unit's forecast and power update as its sheets arrive.
The model checked against fields it never saw, every reporting period; a failed check switches it off for planning.
The files a verifier asks for: sampling plan and power analysis, rarefaction check, lab sheets, and the model's code and theory manual.
Beta since August 2026
Planner and offline field app since August, claim page since October, Puro export due February 2027. Founded in April 2026; taking developers whose next round is still ahead.
Your claim counts soil cores, one of the protocol's two measured routes. The porewater validation check is optional now: porewater sampling cannot yet reliably catch the signal.
A Unit the model was sure of claims what its cores show, even when that is less than it expected.
The claim comes from the cores, at the 30th percentile, with the verifier's files
The example claim page, period one: 1,186 tonnes at the 30th percentile, meaning seven in ten Monte Carlo draws came out higher. Beside it, outside the sum, the model's forecast of 950 to 1,700 t. The verifier's files carry the planning model's code and held-out results, as the draft model module would require.
- Surfacethe basalt as spread, tonnes a hectare from your log
- Top 20 cmevery core is cut at least this deep
- Tillage depth plus 5 to 10 cmwhere tilled-field cores stop
- Belowdissolved products move on; the registry's losses module counts them
The case against Saibro
Core every Unit again at several times the density and you need no model: the protocol credits only measured removal.
It works, and the protocol rewards it. It also costs up to $200 a tonne in field measurement, a lab queue once samples run to hundreds of thousands a year, and a first delivery pushed later, on removal that already lags spreading by years.
A first core count for your noisiest Units, before round two
Tell us the hectares spread and their crops. They decide which plan fits, and show whether the model has seen ground like yours: maize, olives, vines, cereals and pasture yes, flooded rice no.