Enhanced rock weathering (ERW) false-positive explorer

Interactive tool for exploring how different ERW data characteristics affect the detectability of weathering under the TiCAT mass-balance framework. Adjust tracer concentrations and noise to see how the implied rock fraction and weathering extent responds, then run tests to check whether the inferred signal is statistically significant. Both forward and inverse (Bayesian) tests are included.

Note: some of the mobile tracer sliders can move with the feedstock sliders, to keep things physically realistic. If you see strange behavior, check to make sure your inputs are as expected!

Product of Mati Carbon PBC, based in part on the approach of Jordan et al. (2026). Data distributions are idealized and not intended to support data validation.

Study parameters

182

Immobile tracer data

Change in concentration (ppm)
4,583 ppm
17,292 ppm
713 ppm
1,384 ppm
194% of mean Δ
Is ΔI significantly different from the soil baseline?
Paired t-test
p-value—
Statistical power—
Population (Welch)
p-value—
Statistical power—
n for 80% power (paired)—
Paired test assumes Pearson's r = 0.75, which sets the level of information gained by pairing.
⚠ Feedstock concentration must exceed soil concentration for a valid immobile tracer.
⚠ Inferred rock fraction is at the prior upper bound — results may be unreliable.

Mobile tracer data

Change in concentration (ppm)
3,340 ppm
68,440 ppm
3,063 ppm
4,955 ppm
162% of mean Δ
Is ΔX significantly different from the soil baseline?
Paired t-test
p-value—
Statistical power—
Population (Welch)
p-value—
Statistical power—
n for 80% statistical power (paired)—
Paired test assumes Pearson's r = 0.75, which sets the level of information gained by pairing.
⚠ Observed ΔX exceeds the mixing line (τ = 0). At the inferred rock fraction, even zero weathering cannot produce this much mobile element change — the signal is inconsistent with the dilution model.

Bayesian posterior

Posterior density MAP Median ΔX insignificant 68% HPD 95% HPD
Rock fraction in soil
MAP—
Median—
68% CI—
95% CI—
Weathering extent
MAP—
Median—
68% CI—
95% CI—
Initial CDR (posterior mean)
r × τ—
per hectare—
Calculations use ΔX = r·[(1−τ)·Xb − Xs] to account for soil cation dilution [Jordan et al. (2026)]. The cross-hatched area marks where ΔX becomes statistically insignificant — here, the mobile tracer cannot distinguish the observed dissolution from τ = 1. Gridding artifacts may be present at low σ.

For CDR calculations

Rock fraction in soil (r) —
CDR potential 0.300 t CO₂/t feedstock
Sampling depth 6.0 cm
Effective application rate —
—
CDR potential and sampling depth are illustrative. Calculation assumes a soil bulk density of 1.5 g/cm³.

False-positive rate tests

5,000 null draws per scenario. Each draw samples ΔI and ΔX from a bivariate Gaussian under the null condition (no rock addition), then runs the Bayesian inversion. Null draws that violate the mixing line [i.e., ΔX > r·(Xb − Xs), implying τ < 0] are discarded as physically implausible.
Click "Run FPR tests" to compute null probabilities.

Summary

Run the FPR tests to see whether the simulated CDR claim is statistically significant.