How surface roughness scaling can mislead enhanced weathering predictions

Not all modeling choices in enhanced weathering (EW) have clear right answers, but some choices are clearly wrong. Over the last six years, enhanced weathering modeling has widely adopted a misinterpreted parameterization for rock surface area that can drastically inflate modeled carbon removal outcomes.

Parameterizing the surface area of crushed rock is a consequential EW modeling decision. Higher surface areas tend to mean faster weathering and more carbon removal — the same principle that explains why crushed ice melts faster than a single block. In this case, a minor misinterpretation of a paper from 2007¹1A Navarre-Sitchler and S Brantley (2007) Basalt weathering across scales Earth and Planetary Science Letters (written by one of the co-authors of this article) has led to many EW modeling studies using an equation that we believe inflates surface areas by ~10-100x, potentially producing similarly inflated carbon removal outcomes.

The goal of this piece is to correct the record and propose an alternative approach for the EW community. We explain the error, show the magnitude of its effect in a set of example simulations, and discuss how to interpret the affected papers.

Correcting the misused parameterization

Most EW models calculate initial rock surface area by treating rock grains as perfect spheres, computing their surface area from the radius, and then multiplying by a “roughness factor,” as grains aren’t perfectly round. The key question here is what roughness factor models should use.²2Since spheres have the lowest surface area to volume ratio of any 3D shape, valid roughness factors are greater than or equal to one.

Today, perhaps the most common approach in EW modeling is to calculate the roughness factor from the grain radius using an equation put forth by Navarre-Sitchler and Brantley, 2007 (NSB07).¹1A Navarre-Sitchler and S Brantley (2007) Basalt weathering across scales Earth and Planetary Science Letters Its use in EW models can be traced back to Beerling et al. (2020) — a highly influential paper that presented pathways for scaling EW to two billion tonnes of carbon dioxide removal per year.³3D J Beerling et al. (2020) Potential for large-scale CO₂ removal via enhanced rock weathering with croplands Nature Since then, at least 14 other studies across six different bespoke EW models have used the same surface roughness parameterization.⁴4S K Anand et al. (2026) Soil structure and mixing controls on water‐rock contact: Implications for enhanced weathering Water Resources Research ^,55S H Baek et al. (2023) Impact of climate on the global capacity for enhanced rock weathering on croplands Earth's Future ^,66D J Beerling et al. (2025) Transforming US agriculture for carbon removal with enhanced weathering Nature ^,77M B Bertagni et al. (2025) Advancing enhanced weathering modeling in soils: Critical comparison with experimental data JAMES ^,88R M Eufrasio et al. (2022) Environmental and health impacts of atmospheric CO₂ removal by enhanced rock weathering depend on nations’ energy mix Communications Earth & Environment ^,99J Jerden et al. (2024) The impact of geochemical and life-cycle variables on carbon dioxide removal by enhanced rock weathering: Development and application of the Stella ERW model Applied Geochemistry ^,1010E P Kantzas et al. (2022) Substantial carbon drawdown potential from enhanced rock weathering in the United Kingdom Nature Geoscience ^,1111Y Kanzaki et al. (2022) Soil Cycles of Elements simulator for Predicting TERrestrial regulation of greenhouse gases: SCEPTER v0.9 Geoscientific Model Development ^,1212Y Kanzaki et al. (2024) In silico calculation of soil pH by SCEPTER v1.0 Geoscientific Model Development ^,1313T Kukla et al. (2025) Swapping carbonate for silicate in agricultural enhanced rock weathering CDRXIV (preprint) ^,1414L Taylor et al. (2026) ARTEMIS version 1.0: A reactive transport enhanced rock weathering model with coupled soil carbon and nutrient dynamics EGUsphere (preprint) ^,1515M Val Martin et al. (2023) Improving nitrogen cycling in a land surface model (CLM5) to quantify soil N₂O, NO, and NH₃ emissions from enhanced rock weathering with croplands Geoscientific Model Development ^,1616N Vakilifard et al. (2021) The role of enhanced rock weathering deployment with agriculture in limiting future warming and protecting coral reefs Environmental Research Letters ^,-1717Z Zhang et al. (2025) An integrated modelling framework to determine terrestrial carbon dioxide removal via enhanced rock weathering Global Change Biology The problem is the parameterization was incorrectly applied from the start.

We refer to the mistaken parameterization as the B20 equation, since it comes from Beerling et al. (2020). The B20 equation and the NSB07 equation are written identically:

Both equations solve for the roughness factor, λ, and rely on constants, a and d, which are discussed below. The key difference — and the mistake in the B20 equation — is the interpretation of β. B20 defines β as the grain radius, while NSB07 defines it as the smallest feature resolvable by the “ruler” you use to measure the surface area.

To understand the correct interpretation of β, it helps to first understand the problem NSB07 was trying to solve. Rather than provide a roughness factor for EW models, the NSB07 equation was part of a framework for comparing basalt weathering rates measured at different scales. Researchers need to measure surface area to get a weathering rate in terms of the mass of rock weathered per area per time, but the ruler they use affects the result. This is analogous to the “coastline paradox,” where coastline perimeters appear shorter when measured with larger rulers. Like coastlines, rock surfaces are fractal, so larger rulers smooth out more detail. NSB07 presented a framework to correct for that effect.

The smallest commonly used ruler for rock surface areas is a gas molecule, typically N₂, and its size sets the “target” resolution for surface area in EW models. Using gas phase adsorption, this ruler measures the “BET” (Brunauer, Emmett, and Teller) surface area, which approximates the surface area that is accessible to water and, therefore, weatherable. Coarser rulers need a factor, λ, to scale the surface area they measure up to the BET surface area — with larger λ values needed at coarser scales (Figure 1). For example, at the grain scale, if you measure the spherical surface area, you’d have to apply a scaling factor to account for the actual roughness of the grain surface. At the far larger watershed scale, if you measure the area of the basaltic bedrock, you would smooth out even more features — such as larger cracks and fracture networks — and would have to apply a much larger scaling factor to get back to BET surface area.

In the NSB07 formulation, β is the size of the smallest feature your ruler can resolve. It gets normalized to the denominator, a — the resolution of the BET ruler — and raised to the power of d, the empirically determined fractal dimension. Critically, this means that when applying NSB07 correctly, each ruler has a single smallest resolvable feature, and therefore a single corresponding λ value.

Since the EW models are typically trying to estimate surface area starting from a spherical grain size ruler, they should use the β value from NSB07 that is associated with that ruler — yielding a λ of ~20. By instead letting λ vary with grain radius, the B20 equation leads to higher λ values, and therefore higher surface areas, for nearly all EW grain sizes — ~2-10x higher than the correct interpretation of NSB07 (Figure 2).

That said, even the correct interpretation of NSB07, which results in substantially smaller values of λ, is likely still inappropriate for EW applications. That’s because a λ of 20 is about twice as high as the λ values typically measured for ground, fresh mineral surfaces.¹⁸18C Anbeek (1992) Surface roughness of minerals and implications for dissolution studies Geochimica et Cosmochimica Acta ^,1919L E Beckingham et al. (2016) Evaluation of mineral reactive surface area estimates for prediction of reactivity of a multi-mineral sediment Geochimica et Cosmochimica Acta ^,2020J D Rimstidt et al. (2012) Systematic review of forsterite dissolution rate data Geochimica et Cosmochimica Acta ^,2121A F White and S L Brantley (2003) The effect of time on the weathering of silicate minerals: Why do weathering rates differ in the laboratory and field? Chemical Geology ^,-2222A F White and M L Peterson (1990) Role of reactive-surface-area characterization in geochemical kinetic models ACS Publications That discrepancy is acceptable in the context of NSB07, which derived λ values from a post-mortem analysis of published data spanning ~13 orders of magnitude. For EW, however, the relevant question is what roughness factors actually look like for the freshly crushed rock that gets applied to fields — and the observational data is likely to be a more appropriate constraint.

Inflated carbon removal estimates

Artificially high roughness factors can lead to artificially high carbon dioxide removal (CDR) estimates, and that’s what we found when we used a model to compare the B20 equation to roughness factors more consistent with the observational data. For the two basalt feedstocks we tested at a range of grain sizes, using the B20 equation inflated CDR estimates by 2x at the low end, and ~250x at the high end.

Our analysis used the reactive transport model SCEPTER (v1.0.2), where the B20 equation is the default roughness factor treatment.¹¹11Y Kanzaki et al. (2022) Soil Cycles of Elements simulator for Predicting TERrestrial regulation of greenhouse gases: SCEPTER v0.9 Geoscientific Model Development Our EW simulations spread 10 tonnes of rock per hectare annually for 50 years, testing different roughness factor treatments, feedstocks, and grain sizes. You can find our model results here, and code for our analysis here.

Figure 3 shows how much less CDR is achieved using roughness factors of 1-10, or the NSB07 equation, compared to the B20 equation. For both basalts — the glassy basalt¹²12Y Kanzaki et al. (2024) In silico calculation of soil pH by SCEPTER v1.0 Geoscientific Model Development and Blue Ridge basalt²³23A L Lewis et al. (2021) Effects of mineralogy, chemistry and physical properties of basalts on carbon capture potential and plant-nutrient element release via enhanced weathering Applied Geochemistry — larger grains lead to more inflated CDR with the B20 equation, consistent with a larger discrepancy in λ values.²⁴24In all roughness factor treatments, coarser grains lead to less surface area per unit mass and less CDR (not shown). Since the B20 equation assumes coarser grains are also rougher, the decrease in surface area (and thus CDR) is smaller. We focus on the results where λ is less than 10, consistent with data. In those simulations, at a grain radius of 100 μm, CDR reaches just 1-30 percent of what the B20 equation achieves, depending on the roughness factor and the feedstock. At a radius of 600 μm, that range drops to ~0.4-8 percent.

The CDR response to a lower roughness factor depends in part on the composition of the feedstock.²³23A L Lewis et al. (2021) Effects of mineralogy, chemistry and physical properties of basalts on carbon capture potential and plant-nutrient element release via enhanced weathering Applied Geochemistry ^,2525F Qin and L E Beckingham (2021) The impact of mineral reactive surface area variation on simulated mineral reactions and reaction rates Applied Geochemistry In our simulations, the glassy basalt is slightly more sensitive to the roughness factor treatment than the Blue Ridge basalt. The difference is likely due to the presence of reactive minerals in the Blue Ridge basalt that approach chemical equilibrium under typical soil conditions, such that reaction progress is governed more by the renewal of dilute soil water than by reactive surface area.²⁶26Surface roughness is not as important a factor for all feedstocks. As an example, we repeated these simulations with calcite and, consistent with previous work, found the surface roughness parameterization had a negligible effect on CDR outcomes. These calcite simulations are just for illustrative purposes. The B20 equation has only been applied to basalt, consistent with the intent of NSB07.

Moving away from the B20 equation

Models don’t require a roughness factor — EW practitioners could prescribe the specific surface area directly, and some have²⁷27H Deng et al. (2023) The environmental controls on efficiency of enhanced rock weathering in soils Scientific Reports ^,2828H Green et al. (2024) Carbon dioxide removal via weathering of sugarcane mill ash under different soil conditions Applied Geochemistry — but the concept remains useful in EW. Grain size is one of the most commonly measured feedstock characteristics. It relates to the emissions required to crush the rock, and the mesh sizes of the sieves used to sort it. The roughness factor lets modelers connect their downstream weathering estimates to these upstream processes.

But until more empirically grounded formulations are developed, we suggest modelers should use roughness factors within the 1-10 range, independent of the grain size. This range is supported by data, including mechanically ground rock, and appears in a range of minerals including feldspars, pyroxenes, and olivines.¹⁸18C Anbeek (1992) Surface roughness of minerals and implications for dissolution studies Geochimica et Cosmochimica Acta ^,1919L E Beckingham et al. (2016) Evaluation of mineral reactive surface area estimates for prediction of reactivity of a multi-mineral sediment Geochimica et Cosmochimica Acta ^,-2020J D Rimstidt et al. (2012) Systematic review of forsterite dissolution rate data Geochimica et Cosmochimica Acta ^,2222A F White and M L Peterson (1990) Role of reactive-surface-area characterization in geochemical kinetic models ACS Publications It’s also consistent with suggested modeling practices in marine EW²⁹29L J J Geerts et al. (2025) Review and syntheses: Ocean alkalinity enhancement and carbon dioxide removal through marine enhanced rock weathering using olivine Biogeosciences and with a number of terrestrial EW papers that used a roughness factor of one.³⁰30A Chen et al. (2023) Experimentally-calibrated estimation of CO₂ removal potentials of enhanced weathering Science of the Total Environment ^,3131G Cipolla et al. (2021) The role of hydrology on enhanced weathering for carbon sequestration I. Modeling rock-dissolution reactions coupled to plant, soil moisture, and carbon dynamics Advances in Water Resources ^,3232G Cipolla et al. (2021) The role of hydrology on enhanced weathering for carbon sequestration II. From hydroclimatic scenarios to carbon-sequestration efficiencies Advances in Water Resources ^,3333G Cipolla et al. (2022) Effects of precipitation seasonality, irrigation, vegetation cycle and soil type on enhanced weathering – modeling of cropland case studies across four sites Biogeosciences ^,-3434J P M Vink and P Knops (2023) Size-fractionated weathering of olivine, its CO₂-sequestration rate, and ecotoxicological risk assessment of nickel release Minerals The two other parameterizations we’ve seen in the terrestrial EW literature yield λ values in the ~5-30 range, with higher values at smaller grain sizes,³⁵35S L Brantley and N P Mellott (2000) Surface area and porosity of primary silicate minerals American Mineralogist ^,3636J Strefler et al. (2018) Potential and costs of carbon dioxide removal by enhanced weathering of rocks Environmental Research Letters though these equations rely on a limited set of data. In aggregate, a systematic relationship between λ and grain size may only emerge when we account for additional parameters such as mineralogy and grinding method.³⁷37M E Hodson et al. (1998) Determination of mineral surface area in relation to the calculation of weathering rates Geoderma

The roughness factor can also be measured directly as the ratio of the BET and spherical surface areas. But this approach is complicated by secondary minerals that can have extremely high surface areas without contributing to CDR.³⁸38U Kuila and M Prasad (2013) Specific surface area and pore‐size distribution in clays and shales Geophysical Prospecting ^,3939F Macht et al. (2011) Specific surface area of clay minerals: Comparison between atomic force microscopy measurements and bulk-gas (N₂) and -liquid (EGME) adsorption methods Applied Clay Science ^,4040B K G Theng et al., (1982) Surface properties of allophane, halloysite, and imogolite Clays and Clay Minerals One recent study found that even trace amounts of these secondary phases (<0.1 weight percent) can account for most of the BET surface area.⁴¹41B A Fisher et al. (2023) Mineral surface area in deep weathering profiles reveals the interrelationship of iron oxidation and silicate weathering Earth Surface Dynamics Without careful screening, these secondary phases could increase the roughness factor without increasing the unweathered surface area that matters for CDR.⁴²42L E Beckingham et al. (2017) Evaluation of accessible mineral surface areas for improved prediction of mineral reaction rates in porous media Geochimica et Cosmochimica Acta

With more roughness factor data being collected across grain sizes and feedstocks, better parameterizations may emerge down the road. But in this early phase of EW modeling, restricting λ to less than 10 is a responsible choice because it is less likely to inflate modeled CDR outcomes in ways inconsistent with the available evidence.

Recalibrating expectations

The mistaken B20 equation has been widely used in the EW modeling community, and the CDR estimates reported in affected papers should be interpreted accordingly.⁴³43Some of the authors of this piece (Tyler and Freya) have published a preprint that uses the mistaken parameterization. Some affected papers make points that likely hold true regardless of the roughness factor parameterization. But papers that make claims about the amount or scale of CDR that EW can achieve — including the foundational Beerling et al. (2020) paper — risk miscalibrating our expectations. Those results should be understood as likely inflating CDR outcomes, perhaps by as much as two orders of magnitude.

The broader lesson here is that we are still in the learning phase of EW. That mistakes exist in the EW modeling literature is not itself a problem — in these early stages, mistakes are all but guaranteed. Problems arise, however, when we place high confidence in published results whose methods are still actively being developed. The academic literature can help stakeholders calibrate expectations for EW as a CDR pathway, but those expectations should be treated as provisional until the underlying methods mature.