the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Reconciling the role of clay mineral formation on net alkalinity generation during rock weathering: Insights from reactive transport models of shale, granite, and basalt
Abstract. Clay minerals, while ubiquitous in soil, have wide-ranging and countervailing impacts on global biogeochemical cycles. Their compositional diversity makes predicting their formation conditions and impacts on soil solution chemistry a grand challenge. In this proof of concept study, we unpack the influence of clay mineral formation on alkalinity generation during rock weathering: a key process regulating atmospheric CO2 levels over geologic time. Because they remove aqueous Si and Al from solution during formation, clay minerals may enhance silicate mineral dissolution and thereby enhance alkalinity generation. However, clays can also offset the alkalinity produced during silicate dissolution by generating acidity when they form. To identify whether, or to what extent, clay minerals enhance alkalinity generation, we employ reactive transport models of idealized granite, basalt, and shale weathering over a range of environmental conditions. Clay mineral formation typically increases net alkalinity generation by enabling greater silicate dissolution. This enhancement is most pronounced at low fluid flow rates and low soil CO2 concentrations when thermodynamic conditions require clays for continued silicate dissolution and sustained alkalinity generation. However, clay-enhanced alkalinity generation diminishes at low erosion rates when primary mineral supply decreases; in less common instances where chlorite forms, clay formation can lead to a net alkalinity consumption. Overall, clay minerals promote CO2 drawdown and modify the silicate weathering feedback.
- Preprint
(1393 KB) - Metadata XML
-
Supplement
(845 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-5323', Benjamin Tutolo, 30 Oct 2026
-
RC2: 'Comment on egusphere-2026-5323', Anonymous Referee #2, 02 Nov 2026
This study uses batch and one-dimensional reactive-transport models to investigate whether secondary clay formation increases net alkalinity generation by removing dissolved Si and Al and thereby sustaining primary-silicate dissolution. The authors first examine a simplified plagioclase-kaolinite system in batch and 1-D simulations. They vary the initial plagioclase abundance and compare cases with and without kaolinite formation. They then simulate weathering of idealized shale, granite, and basalt assemblages containing additional primary and secondary minerals. In both the simplified plagioclase-kaolinite models and the 1-D rock-weathering models, the authors compare simulations with and without clay formation and vary CO2 concentration. In the rock-weathering models, the authors additionally test the effects of water flow and physical erosion rates. They conclude that clay formation generally increases alkalinity by keeping primary silicates undersaturated, except where chlorite forms at high pH.
The question of how clay formation affects net alkalinity is worthwhile. However, the manuscript does not clearly establish what its results add to previous reactive-transport studies of secondary-mineral formation and silicate dissolution. The central result largely follows from the model design, and key choices governing secondary-mineral formation and the magnitude of the modeled effect are insufficiently justified or tested. The presentation is difficult to follow. I recommend rejection of the manuscript.
Major comments
Novelty and knowledge gaps. That secondary-mineral precipitation maintains primary-mineral undersaturation and raises weathering rates is established (e.g., Maher et al., 2006, 2009), and the authors cite a number of studies in lines 60–66. The gap offered on lines 76–78 (clays are numerous, so more study is needed) is weak. Line 121 calls previous reactive transport studies “inconsistent” without saying how, and the title promises to “reconcile” something, but the Introduction never identifies which discrepancies are being reconciled (lines 120–124 only suggest that they may stem from mineralogy or modeling strategy). Some justification for the work appears in Methods (lines 261–265) and should be in the Introduction. The Introduction also alternates between generic statements on clays and rock-type descriptions without explaining what differs in the process of silicate weathering among shale, granite and basalt relative to state of the art modeling practices. The Introduction frames the question around the diversity of clay minerals and their different effects on alkalinity (lines 69–78), yet the simulations mostly compare “clay” with “no clay” scenarios. The simplified models contain only kaolinite, which by construction does not consume alkalinity, and the main text presents no simulation that isolates the effect of individual clay types in the rock models. The study therefore does not directly address the gap it states.
The clay-free baseline determines the headline results. The central result (clay formation enhances silicate weathering) largely follows from the model design. The clay-free baseline excludes the Al-bearing sinks that would normally limit Al accumulation. In the batch no-clay run, total Al is 125 µM at pH 6.15 (line 303). This is orders of magnitude above gibbsite solubility, so the baseline apparently suppresses all Al-bearing phases, including gibbsite. Primary silicates therefore approach equilibrium quickly by construction. A more informative comparison would allow an Al hydroxide or amorphous phase only, versus a full clay assemblage, and compare different clay assemblages. Similarly, no Si sink other than clays and quartz is allowed. In the kaolinite batch run, Si reaches 1.58 mM (line 311), close to the solubility of amorphous silica, so the Si accumulation that drives the plagioclase affinity term may be unrealistic.
Rate laws and kinetic parameterization. Eq. 3 is a generic linear TST law, (1 − Q/K), used for both plagioclase dissolution and kaolinite precipitation. The exponents, the exact form of the Al dependence, the surface areas, and the kaolinite rate constant are not given in the main text. The authors cite non-linear alternatives (Alekseyev et al., 1997; Hellmann and Tisserand, 2006) and the Al-inhibition model of Oelkers et al. (1994) (lines157-162). They describe the rate law as “pH- and Al-dependent” (lines 160–162), but Eq. 3 is a generic product of activities rather than the Oelkers model, and the affinity term remains linear. This matters because the clay effect arises through two distinct routes. Through the affinity term, clays increase dissolution only where plagioclase approaches equilibrium. Through Al inhibition, the rate depends on aqueous Al concentration at any saturation state. The manuscript does not separate the two.
In Maher et al. (2009), a linear TST law required feldspar rate constants about two orders of magnitude below laboratory values to reproduce weathered silicate profiles, whereas the Al-inhibition and parallel laws gave laboratory-consistent constants. Maher and co-workers concluded that inferred rates and mass depletion depend strongly on the form of the rate law near equilibrium. The magnitude of the clay effect here (e.g., the factor of 8.5 in batch runs, 180–370% in rock models) may therefore be sensitive to the rate-law choice, yet no sensitivity test is shown.
The same applies to kaolinite precipitation, which also uses a linear TST law. The authors argue that results are insensitive to this choice because kaolinite is highly oversaturated (lines 600–602), but Fig. 2A shows that precipitation kinetics set the time to equilibrium, and Maher et al. (2009) found the kaolinite rate constant, solubility and initial abundance to control the extent of primary-mineral dissolution. No sensitivity test of these choices is shown.Representation of clay precipitation and mineral thermodynamics. In the rock models the precipitating phases (illite, smectite, chlorite, kaolinite, gibbsite) are imposed a priori and seeded at 0.1% in granite and basalt (lines 221–224). The clay types that form are determined by the phase list, the precipitation rate laws, and the database solubilities chosen by the authors, so they cannot be interpreted as a prediction of which clays form during weathering. Clay formation at low temperature is poorly represented by TST rate laws combined with database solubilities, which is why nucleation-based and precursor-phase models have been developed (e.g., Steefel and Van Cappellen, 1990). The authors justify the TST choice by the unknown parameters of these models (lines 595–599), but this avoids rather than answers the problem that motivates those models. No solid-phase data (mineral or clay abundance profiles) are used to test the model in the main text, so the interpretation of clay-type results (lines 425–439) and the chlorite exception rest on the chosen phase list and database. Mg-chlorite precipitation is not typically observed in low-temperature weathering profiles and probably reflects the database. The chlorite exception in the abstract, the conclusion that clays strengthen the feedback in basalt (lines 587–590), and the conclusions on enhanced rock weathering depend on it.
Missing alkalinity-relevant processes. Carbonate precipitation. Carbonate precipitation appears absent from the shale, granite and basalt assemblages (Table 2), although Fig. 1 itself shows calcite precipitation. The models therefore lack the main sink that buffers pH and removes alkalinity. This may explain why basalt pH reaches about 10 (Fig. 5H) while the authors cite about 9 as observed (line 437). The text should state whether dolomite is dissolution-only (line 203 says “Ca carbonate”, Table 2 says dolomite). Other phases and exchange. Fe-bearing smectites, allophane or other amorphous phases are not included. Cation exchange is not explicitly included, although the Introduction stresses the alkalinity consumption of high-CEC clays (lines 72–75). These omissions bear directly on the alkalinity budget. The authors state that the models are not meant to replicate realistic kinetics and flow (line 594). All models are a simplified representation of reality, but the processes that control the quantity being predicted must be represented. The omissions above are of that kind.
Hydrology, transport, and boundary conditions• Saturation and gas transport. Profiles are fully saturated (line 251) but the text states gases diffuse at 0.01 cm2 s-1 (line 252), a value typical of soil air, roughly three orders of magnitude faster than dissolved aqueous species. In a fully saturated domain a gas phase is not physically meaningful, and the description is inconsistent. Partial saturation controls CO2 and O2 supply and residence-time distributions, which are central to the approach-to-equilibrium argument.
• CO2. The “Initial CO2 concentration of 15,000 ppm” (line 252) in the 1D model is unclear. Is it a top boundary condition or a finite amount defined in the domain? What about the batch simulations? Is the batch system open or closed to CO2?
• Passivation. Surface passivation is omitted. “CrunchTope does not have a method” (line 621) is a software argument, not a scientific one. A sensitivity test with scaled reactive surface area is feasible.
• Permeability. The discussion of permeability (lines 604-608) does not apply to a 1-D column with a prescribed Darcy flux, since permeability changes cannot alter the flow. The limitation that matters is the prescribed flux itself, because clogging would reduce flow in a head-controlled system.
• Erosion. Erosion is introduced only as an option the software allows (lines 234–242), and the values are not justified.Model validation. Given the departures from the field-constrained models, the authors should show that the model reproduces observations (e.g., the Santa Cruz chronosequence, Shale Hills, or a basaltic profile). Fig. 7 is not that test. It compares stream Si and alkalinity from mixed-lithology catchments with model averages, the clay and no-clay averages are nearly indistinguishable, and shale alkalinity is underpredicted. The comparison cannot support or reject the clay effect.
Results are trivial or confounded. That dissolution is faster with more plagioclase (line 324) and that more CO2 gives more alkalinity (lines 337–338) are not new, and Section 3.1 reads as a parameter sweep without a clear purpose. The last paragraph of Section 3.1 (lines 356–362) describes spatial variation without showing a figure. Lithologies differ in many properties at once (plagioclase composition and content, carbonate, pyrite, clay seeds, porosity, and hence residence time: about 1.35 yr for 13.5% porosity versus 0.44 yr for 4.4% at the same flux). Differences cannot be attributed to “lithology” or “clay type” without a factorial design.
Overall quality of the writing and figures. The manuscript needs a thorough language edit, and the reasoning behind the modeling exercise is hard to follow. Many statements, particularly those describing the models, are vague or inaccurate. The central message (clays consume alkalinity but also enhance silicate dissolution) is repeated in nearly every section (abstract, Introduction, Sections 2.2, 3.4, 4.2 and Conclusions), while the design of the simulations and the purpose of each set are never stated clearly. The examples below illustrate the problem and are not exhaustive.
• “The dynamics of clay mineral formation along weathering profiles may shift dramatically amongst shale, granite, and basalt” (line 100). “Shift” is not the right term, and the sentence states a hypothesis without saying what is expected to differ or why.
• “An addition of 10% kaolinite to each clay model promotes robust clay formation” (line 176). What does “robust” mean here?
• “The chemical evolution of silicate dissolution and clay formation is roughly equivalent spatially along a weathering profile versus temporally in a batch reaction” (lines 185–186). This is unclear. The analogy holds only under specific conditions, which are not stated. In a batch system products accumulate and nothing is transported, whereas in the 1-D column solutes are advected and dispersed, gases diffuse, and the solid evolves differently at each depth.
• “However, in the 1-D models, since mineralogy slowly evolves and a gradational weathering front appears, reaction rates and aqueous chemistry are slightly affected” (lines 187–188). It is unclear what is affected relative to what, and “slightly” is not quantified. "Gradational weathering front" is also an unusual expression, and “appears” is the wrong verb.
• “Weathering profiles are simulated over 250,000 years” (line 191). Do the authors mean that the development of the profile is simulated for 250,000 years?
• “The rainwater contains low concentrations of primary solutes” (line 246). Primary and secondary species are a numerical distinction in the speciation model, not a chemical property of the water.
• “Our simple batch models provide direct evidence for clay mineral formation engendering additional plagioclase dissolution” (line 295). Model output is not direct evidence. The simulations show the consequence of the authors' assumptions, and evidence requires data.
• “Al concentrations initially increase with depth as silicate dissolves then decrease and remain depleted as clay minerals become saturated and form down-profile” (lines 407–409). What is meant by “down-profile”? Also, minerals precipitate when the solution is supersaturated with respect to them, not when they “become saturated”.
• “pH only increases from 4.97 to 6.15 with Al3+ still representing a significant proportion of the total Al concentration (125 µM) at 0.41 µM” (lines 302–304). The sentence is garbled, and 0.41 µM is about 0.3% of 125 µM.
• “Since clays promote additional plagioclase dissolution, plagioclase weathers away faster” (lines 417–418). What does "weathers away" mean?
• “At higher erosion rates (0.2 mm/year), weathering profiles near steady state with primary silicates still available” (line 420). This is not a sentence.
Figures should be substantially improved. For instance, for figure 1:
• The arrows are unclear and precipitation cannot be distinguished from dissolution.
• Aqueous speciation is shown only for Al, although other species also form complexes (line 248).
• The caption says “carbonate dissolution generates alkalinity”, but this is not shown in the figure.
• The dashed arrows (Na+/Ca2+ to clays, clays to H+) are not explained.
• Calcite precipitation is shown, but calcite is not in the modeled assemblages (Table 2).Table 1 is the only summary of the simulation design, yet it falls well below what is expected for a modeling study. It conveys almost no information and is in places inconsistent with the text.
• “Simple 1-D reactions” and “1-D rock reactions” are undefined.
• “(and kaolinite)” in brackets is unclear
• “Aqueous reactions: None” for the batch and simple 1-D models appears to contradict the Al speciation reported on lines 303-304.
• The parameters tested are unclear and incomplete (e.g., the table lists oxalate concentration, which appears nowhere in the main-text results).Citation: https://doi.org/10.5194/egusphere-2026-5323-RC2
Data sets
Reconciling the role of clay mineral formation on net alkalinity: Insights from reactive transport models of shale, granite, and basalt Noah Jemison https://doi.org/10.5281/zenodo.17064490
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 737 | 259 | 36 | 1,032 | 95 | 47 | 48 |
- HTML: 737
- PDF: 259
- XML: 36
- Total: 1,032
- Supplement: 95
- BibTeX: 47
- EndNote: 48
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
In this manuscript, Jemison and co-authors prevent a series of batch and reactive transport models designed to test the effect of clay mineral precipitation on alkalinity generation during the weathering of single minerals and multi-mineralic rocks. Their overall conclusion is that clay mineral precipitation tends to yield enhanced primary silicate mineral dissolution, which, in turn yields higher net alkalinity generation. Overall, the manuscript reads well, is very thorough, and would have a substantial impact on the weathering/alkalinity generation community’s understanding of the feedbacks between clay mineral precipitation and primary mineral dissolution. My comments, listed below, are generally minor, and I think the manuscript would be acceptable for publication after some minor revisions.
Firstly, I think the manuscript would benefit from the inclusion of an activity diagram, or series of activity diagrams, to demonstrate the controls on clay mineral precipitation in this system. The ancient diagrams included in Lasaga et al. (1994) provide some level of guidance for why this exercise would be useful. We included one that is perhaps more relevant in the supplement to Tutolo et al. (2021) – Fig. S3 – which shows how pH constrains whether chlorite, smectite, or kaolinite would precipitate from alteration of basalt. Again, as a guide, including something like this could be very helpful here as a way of outlining the regimes that are encountered in the model results.
Secondly, I appreciate the authors’ discussion of the uncertainties surrounding precipitation rates and hence timescales, but the simulations do not really explore these uncertainties per se. While this is OK, because the manuscript is already very thorough, I would recommend digging a bit more into the (very limited) literature on this subject. In Tutolo et al. (2015), which involves experiments conducted at much higher temperatures and is perhaps less relevant, we explored how reactive surface area and reaction rates impact outlet fluid chemistry from core experiments conducted on K-feldspar rich sandstone. These results indicated that if kaolinite precipitates too quickly, K-feldspar dissolves too quickly, and hence the models overshoot the K concentrations in outlet fluids. This study is not mechanistic by any means, but at least gets at the processes under exploration here. I’m sure there are others out there in the literature worth digging into, but probably not many. In other words, the fact that we don’t know how fast clay minerals grow is an important gap in our knowledge and I think it is worth directly pointing that out here.
Other, more minor concerns:
The manuscript indicates that the simulations achieved equilibrium (e.g., line 179), but I’m not entirely clear what that means in this context. Does it mean steady state, or does it mean that the fluids are equilibrating with the primary plagioclase, or does it mean that all of the primary plagioclase has dissolved (which would only be possible in non-erosion simulations). Equilibrium with primary plagioclase at low-temperatures seems unlikely, which is worth pointing out as a potential discrepancy between the model and reality.
Line 220: only biotite is redox-active, so this should be reworded
Line 226 and elsewhere: I’m not really familiar with feldspars being called “more mafic” or “more felsic” – maybe it’s worth just saying these are more similar to those that would be encountered in basaltic or granitic rocks. Also it’s worth pointing out that pure anorthite seems unlikely in basalt.
Fig. 5 (and elsewhere) – it took me a while to realize which lines represented “no clay” versus which ones represented “with clay” – can you explain this in the caption as well as including it in panel B?
Line 468: here is an example of where an activity diagram might be useful.
Fig. 7 is really cool, but the outliers tend to extend the y-axis in a way that obscures the variation across rock types. Could the figure be made taller or could a broken axis be added to show this more clearly?
Line 612: Unless I’m misunderstanding something, this is only true of the permeability doesn’t change and the hydraulic head gradient remains constant. While that may be the case in your model, it is unlikely in the real world. Could this be revised to say “does not influence Darcy velocity in the model?