The science, including the parts that are hard for us.

Silicate minerals react with carbonic acid — rainwater that has dissolved CO₂ from the air. The reaction breaks the mineral down, releasing calcium and magnesium and converting the dissolved CO₂ into bicarbonate. This is the process that has regulated atmospheric carbon over geological time.

Its limitation is rate. Weathering is governed by the surface area of rock exposed to water, and intact bedrock exposes very little. Crushing basalt to a fine particle size raises that surface area by orders of magnitude; spreading it across cultivated soil places it where water, root exudates and biological acids are already concentrated. The chemistry is not modified. Only the speed is.

What makes the removal durable is where the carbon ends up. Bicarbonate is stable in solution and, on reaching the ocean, remains part of the marine inorganic carbon pool on a timescale of tens of thousands of years. That is a different class of claim from carbon held in biomass or topsoil, which can be released by a fire, a drought, or a change in how the land is farmed.

Dissolution rate, agronomic benefit, and how hard the removal is to detect all change with the ground beneath the rock. We hold a separate interpretation model per agro-climatic zone, written and owned by the science team. Two of the five are hard cases for us, and they are listed as such.

Why publish the difficult ones. A methodology that works everywhere is a methodology that has not been tested anywhere. The Deccan Vertisols are the cheapest soils for us to reach and among the hardest to measure against — both facts are true at once, and a buyer evaluating this work needs the second one as much as the first.

Every sample runs the same panel. The farmer-facing soil card and the carbon accounting are two readings of one physical sample, not two separate exercises.

Baseline before anything is applied

A georeferenced, multi-depth baseline is taken on every plot before a single tonne of rock is spread — at minimum 0–15 cm and 15–30 cm. This is the one step that cannot be recovered later. A plot sampled after application has no defensible before-state, and no amount of subsequent work repairs that.

Mass balance against an immobile tracer

Establishing how much of the applied rock has actually weathered requires a reference that does not weather. Titanium, zirconium and niobium are effectively immobile in these systems, so the change in the ratio of a mobile element to an immobile one gives depletion directly, without depending on knowing precisely how much rock reached a given square metre.

Concentration is not a quantity

Laboratory results arrive as concentrations. Converting them to tonnes per hectare requires bulk density and texture for the specific horizon sampled. Assuming a standard value here is a common and quiet source of error, so we measure it.

Uncertainty is an output, not a caveat

Every removal estimate is produced with its uncertainty propagated through the calculation rather than appended to it. A figure without an interval is not a measurement, and in carbon removal a confidently wrong number is worse than an honestly wide one.

Every figure carries its category

Measured, modelled and projected numbers answer different questions, and they are never merged into one. When a figure is published it will carry one of these labels.