the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Jan 2026
ARTEMIS version 1.0: A Reactive Transport Enhanced Rock Weathering Model with Coupled Soil Carbon and Nutrient Dynamics
Abstract. Enhanced rock weathering (ERW) is increasingly considered to be a promising carbon dioxide reduction (CDR) strategy, but carbon removal can be is difficult to verify with field measurements. Reactive transport models (RTMs) have the potential to shed light on the soil dynamics affecting CDR, and to quantify the timescales involved. Here, we present a new 1-D RTM representing all major processes affecting the chemistry of soils. These processes include nitrogen cycling kinetics, sorption and the choice of open or closed systems with respect to gas diffusion. We demonstrate this model’s utility with a detailed investigation examining the impact of those key ERW and soil processes on CDR and topsoil pH at a site in the United States Corn Belt. Given continued annual applications of a metabasalt for 55 years, results indicate a 20-year lag time to achieve 10 tCO2 ha−1 for CDR based on solute export in drainage water, with long-term topsoil pH (7.5–8.0) maintained by sorption. Topsoil pH would stabilise below the maximum recommended limit of 7.4 with triennial metabasalt treatments, but the lag time would double. Five-year model runs with four annual metabasalt treatments suggest doubled bicarbonate export in the absence of nitrogen kinetics due to reduced strong acid weathering. Calcite deposition in the upper soil occurs if the metabasalt is replaced with a pure CaSiO3 feedstock, reducing CDR efficiency. For a pure Mg2SiO4 feedstock, calcite deposition limits Mg export because Mg replaces exchangeable Ca on soil clay surfaces. Without sorption, calcite saturation maintains topsoil pH near 8 for all feedstocks under open system conditions. However, topsoil pH was unrealistically high (∼ 10) for the CaSiO3 feedstock with a closed system. With these model runs, we illustrate the process representation useful for predicting solute export through soils at individual field sites. Critically, we also discuss the limitations of this model and possibilities for development of the next generation of ERW models.
Competing interests: DJB has minority equity stakes in two companies aiming to help mitigate climate change (Future Forest and Undo), and is an advisory board member of The Carbon Community, a UK carbon removal charity.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(3251 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 05 Apr 2026)
- CC1: 'Comment on egusphere-2025-5823', Carl Steefel, 31 Jan 2026 reply
-
RC1: 'Comment on egusphere-2025-5823', Anonymous Referee #1, 17 Feb 2026
reply
The manuscript presents a comprehensive technical description of ARTEMIS and an application example. However, in its current form, it does not yet demonstrate the verification, benchmarking, and diagnostic analyses typically expected for publication in GMD. Several key areas require additional clarification and testing to meet what I would consider the standards of appropriate scientific model development. The manuscript is also incredibly long and particular sections could be more suitable for a technical manual. Here are the high-level concerns wrt to appropriateness in this journal.
- Model couplings: The manuscript describes multiple coupled process modules (hydrology, PHREEQC reactions, plant uptake, gas boundary conditions), but does not systematically assess how numerical coupling choices influence model outcomes. Given the operator-splitting structure, a quantitative analysis of coupling sensitivity or timestep dependence would strengthen confidence in the model predictions.
- Benchmarking against analytical solutions or established open-source RTMs for limiting cases (e.g., 1D advection–reaction, equilibrium carbonate system, or simple infiltration scenarios) would allow readers to distinguish structural model behavior from implementation-specific artifacts, especially those emerging at the end. More specifics and suggestions are provided below.
- Section 2 provides detailed implementation descriptions typical of a user manual. While valuable for documentation, the manuscript would benefit from clearer separation between (i) implementation details appropriate for a code manual and (ii) scientific diagnostics demonstrating what new understanding is gained through this model relative to existing frameworks
- Minimal validation. The comparison to drainage chemistry provides a useful consistency check. However, it does not constitute a full validation in the GMD sense, because flow path depth is treated diagnostically rather than tested independently. Additional constraints and testing on a well-characterized system would help evaluate whether the discrepancies arise from site characteristics or model structure.
- ARTEMIS occupies an intermediate position between fully coupled reactive transport solvers and lumped parameter soil biogeochemistry models. The manuscript would benefit from explicitly positioning ARTEMIS within this modeling landscape and clarifying which classes of problems it is uniquely suited to address.
With respect to model development, some aspects of the ARTEMIS wrapper are very useful. For example, the fertilizer parameterization. It also tackles some complex couplings and introduces new frameworks for nutrient and plant cycling from other contexts. However, the following areas in the ARTEMIS formulation are potentially problematic or lack the rigorous testing needed to have confidence in the model predictions. Philosophically, I am not sure it makes sense to call this a reactive transport model, it sits in between more traditional lumped parameter black box models and true RTMs. As such, it needs more careful attention than is currently provided in the manuscript.
- Hydrological formulation: ARTEMIS uses a "simplistic" bucket model for percolation and soil moisture, which moves water only when it exceeds field capacity. It does not solve the Richards equation or calculate unsaturated hydraulic conductivity, which are standard for variably saturated flow in codes like TOUGHREACT, MIN3P, SCEPTER, CrunchFlow and PFLOTRAN Given the simplified hydrologic representation, it would be helpful to demonstrate under what conditions the bucket formulation reproduces solute flushing dynamics comparable to a Richards-based model, particularly during transient rewetting events.
- Gas diffusion dynamics: A key limitation of ARTEMIS is that gas diffusion between soil layers and the soil surface is not modeled. Instead, it uses an analytical solution for pCO2 profiles or simple daily replenishment via the ideal gas law. Most models explicitly model gas-phase advection and diffusion as dynamic transport continua. Because gas-phase diffusion strongly controls soil CO₂ gradients, especially under wet conditions, sensitivity analyses exploring how the analytical pCO₂ profile assumption influences alkalinity generation would strengthen the manuscript.
- Numerical coupling and operator splitting: ARTEMIS relies on a MATLAB preprocessor to calculate mixing fractions for PHREEQC's MIX function. This is a decoupled, sequential approach that can may introduce operator splitting errors; discussion of timestep constraints and numerical stability criteria would improve confidence transparency.
- Sorption and secondary phase coupling: ARTEMIS seems to lacks dynamic coupling of sorption to secondary phases that dissolve or precipitate during the simulation (e.g., ferrihydrite or gibbsite) and it is confusing that sorption cannot occur on neo-formed phases. State-of-the-art models typically update surface properties and porosity as these phases evolve to reflect feedbacks on chemistry and transport.
- Surface area updating: While surface area treatment remains a confounding issue, and the various options for user-defined exploration could prove useful, ARTEMIS updates Reactive Surface Area (RSA) using a "shrinking sphere" approach that assumes smooth, monomineralogic particles. The shrinking-sphere assumption may not fully capture surface roughening and secondary phase coating typical of basalt weathering. Clarifying how sensitive results are to RSA parameterization would help contextualize uncertainty.
- Plant module: The inclusion of a plant nutrient uptake module is potentially one of the most novel and valuable aspects of ARTEMIS, as relatively few reactive transport models explicitly integrate vegetation processes in this way. However, the manuscript would benefit from clearer conceptual positioning of this component. At present, the interleaving of SWAT-derived phenology, biomass trajectories, and michaelis–menten uptake kinetics could be misread as a mechanistic plant growth model. Clarifying this structural modeling choice would significantly improve interpretability. A brief comparison with alternative approaches (e.g., vegetation implementations in MIN3P or other RTMs) would also help contextualize the contribution and clarify how ARTEMIS differs conceptually from existing frameworks.
Line-by-line comments
Line 35: The stated aim of the paper combines a technical description of ARTEMIS with an example calibration/validation case. However, the manuscript does not clearly articulate what conceptual, numerical, or process-level advance ARTEMIS provides relative to existing models. It would strengthen the paper to explicitly identify the unique scientific contribution—e.g., which modeling gap ARTEMIS fills, what new capability it enables, or what limitation in existing frameworks it overcomes.
Line 4: “all major processes affecting chemistry of soils” This statement appears overly broad. ARTEMIS includes many important processes, but not all major soil chemical processes (e.g., full variably saturated flow, dynamic gas transport, emergent plant growth, or certain microbial feedbacks). I recommend qualifying this statement to more accurately reflect the scope of the implemented processes.
Line 30: The discussion appears to underrepresent the extensive literature on biological controls on soil biogeochemistry. Even if certain processes are not specific to enhanced weathering, they may still influence mineral dissolution, nutrient cycling, and carbon dynamics. Incorporating a broader biological context would improve the framing.
Line 42: “all the key processes” – Similar to line 4, this claim should be moderated. While ARTEMIS integrates multiple important processes, it does not include all key processes typically associated with fully coupled reactive transport models. A more precise statement would improve clarity.
Line 50: Beginning in Section 2.1, the manuscript shifts toward detailed implementation descriptions. While documentation is valuable, the paper would benefit from distinguishing between (i) code-level implementation details appropriate for a user manual and (ii) scientific diagnostics demonstrating how these implementations improve or extend current modeling approaches. Consider condensing descriptive content and emphasizing conceptual or methodological novelty.
Line 82: It would be helpful to clarify why ARTEMIS does not permit alternative surface representations (e.g., bulk surface proxies) when mineralogical distributions are poorly constrained. Many field applications lack detailed mineralogy; flexibility in surface representation could enhance usability.
Line 84: The text suggests that direct solid-to-solid transformations are not allowed. Because most geochemical models treat such transformations through dissolution–precipitation pathways, it would be helpful to clarify whether this statement simply reflects that mathematical structure, or whether certain transformations are restricted in ARTEMIS..
Line 95: This section largely reiterates established descriptions of TST-based mineral kinetics common to RTMs (e.g., Steefel et al., 2015). Rather than restating standard formulations, the manuscript would be strengthened by clearly identifying what differs in ARTEMIS (e.g., surface area evolution, coupling strategy, parameterization flexibility) relative to existing implementations.
Line 120: The discussion of non-linear TST rate laws would benefit from clearer positioning relative to prior work. Many RTMs (e.g., GWB, CrunchFlow, PFLOTRAN) already implement flexible non-linear rate formulations. If ARTEMIS offers a distinct implementation or addresses a specific limitation, that should be made explicit. Otherwise, the paragraph may appear redundant. The key issue seems to be the degree of user discretion in choosing rate forms (noted later around line 140), which may merit more focused discussion.
Line 175: This section again reads primarily as implementation documentation. Equation (4) and particle size distribution approaches are standard in many RTMs, including SCEPTER and others. The manuscript would benefit from explicitly stating what is novel in ARTEMIS’s treatment—whether numerical flexibility, coupling structure, or uncertainty exploration.
Line 260: The plant uptake section provides a helpful explanation of the implemented processes. However, its placement and narrative structure make it difficult to distinguish between descriptive documentation and scientific contribution. It would improve clarity to explicitly state at the beginning of Section 2.6 what is conceptually new in ARTEMIS relative to existing RTM approaches with vegetation components.
Line 280: The repeated use of “redox decoupling” between ammonium and nitrate may be confusing, as it suggests a structural independence rather than a kinetic or microbially mediated separation of reaction pathways to avoid the PHREEQC treatment of kinetic species as equilibrium controlled. Consider rephrasing to clarify that nitrification and denitrification are treated as kinetically controlled microbial reactions rather than dynamically coupled redox equilibria.
Line 385: Please clarify how volatilized nitrogen species are treated. Are they represented as diffusing gas species within the soil column, or are they removed through a prescribed flux? A brief explanation of the gas-phase handling would improve transparency.
Line 455: The coupling between prescribed demand and Michaelis–Menten limitation via Km could be clarified. A schematic illustrating the three key processes—(i) prescribed demand, (ii) uptake limitation, and (iii) charge-balancing proton flux—would help readers understand which terms influence soil chemistry directly.
Section 2.6
The plant uptake module is potentially one of the more interesting components of the manuscript, as relatively few reactive transport models incorporate vegetation processes in this manner. However, the conceptual structure of this module would benefit from clearer positioning. Early in Section 2.6, it should be stated explicitly that plant growth and nutrient demand are prescribed closures rather than emergent outcomes of carbon assimilation or resource limitation. In the current formulation, vegetation functions as an externally imposed, time-varying demand for nutrients, with an optional prescribed partitioning of evapotranspiration, rather than as a mechanistic carbon-assimilating organism governed by photosynthesis and allocation dynamics.
- Because the text interleaves SWAT-derived phenology, biomass trajectories, and Michaelis–Menten uptake kinetics, it may be misread as representing mechanistic growth modeling. Clarifying this modeling choice would improve transparency and prevent confusion about the degree of biological realism represented.
- A brief comparison with alternative vegetation implementations in other reactive transport frameworks (e.g., earlier approaches in MIN3P or related models) would help contextualize the contribution of ARTEMIS and clarify what is novel in this implementation relative to prior work.
- The structure adopted in this module has important implications for interpreting soil chemistry results. Because biomass trajectories are prescribed, nutrient limitation does not dynamically reduce plant growth except through the uptake cap. As a result, soil chemistry responds to unmet nutrient demand, and proton fluxes are generated through charge-balancing rules to maintain electroneutrality. Under sustained nutrient depletion, plants may continue to “attempt” uptake while soil pools absorb the imbalance? Is this correct?
- This modeling choice is not inherently incorrect, but it has consequences that should be discussed explicitly, particularly in the context of alkalinity and weathering studies. The explicit representation of H⁺ or OH⁻ extrusion and tunable proton release during nitrogen fixation means that substantial alkalinity signals may arise independently of mineral dissolution. Without careful framing, these plant-driven ion exchange processes could be misinterpreted as enhanced weathering signals, especially in a manuscript framed around carbon dioxide removal. The relative magnitude of plant-driven proton fluxes compared to mineral buffering would therefore be important to discuss or quantify.
It would also be useful for the authors to propose criteria by which the plant module can be evaluated or validated. Because growth trajectories are prescribed rather than predicted, traditional biomass validation in a predictive sense is not possible. Instead, evaluation might focus on nutrient uptake fluxes, seasonal soil depletion patterns, or charge-balance-induced proton fluxes. Clarifying how this module can be tested against observations would strengthen confidence in its implementation.
Finally, the plant module introduces a substantial number of parameters, including phenological timing, biomass partitioning, stoichiometric uptake curves, root depth evolution, depth-weighted uptake distributions, Michaelis–Menten kinetics, and charge-balance coefficients. Many of these parameters are only weakly constrained by soil chemistry data alone. Without independent plant or nutrient flux measurements, there is a risk of equifinality, whereby different parameter combinations produce similar soil chemical outcomes. A discussion of parameter identifiability and calibration constraints would therefore improve the robustness and interpretability of this component.
Section 3.3. Calibration of the model The calibration procedure would be strengthened by including a spin-up phase under steady forcing to ensure that internal pools and fluxes are equilibrated prior to feedstock introduction. There appears to be no spin-up phase, which is usually necessary to ensure the underlying fluxes and processes are established prior to introducing a perturbation such as a feedstock. Without spin-up, internal pools and fluxes may not be equilibrated, and early-year dynamics (e.g., first 5–10 years) may reflect transient adjustment rather than system response to perturbation. Clarifying whether such a procedure was applied would improve confidence in the calibration.
Section 3.4: This is not validation of a model in a standard sense. The authors compared model chemistry from several soil layers to the drainage water. They found that the best-matching model layer (56–76 cm) was significantly shallower than the physical drain depth (130 cm). From this discrepancy, the authors characterize the site as having macropore flow, where water bypasses deeper matrix layers without questioning whether issues in the model result in this finding.
Figure 4 and general presentation of model results. The presentation of derived quantities such as “Mg CDR” is difficult to interpret without comparison to observed concentrations or fluxes. Where possible, presenting standard state variables (e.g., Ca²⁺, Mg²⁺, alkalinity, pH, DIC) alongside derived metrics would improve transparency and allow broader comparison with expected field ranges.
Final section. The paper is incredibly long. Aspects of the Appendix contain useful discussion and presentation. Thus, the paper may be inverted from what is the optimal presentation: could much of the text on formulation and user guidance by relegated to appendix, to focus on presentation of model insights.
Citation: https://doi.org/10.5194/egusphere-2025-5823-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 204 | 110 | 18 | 332 | 15 | 25 |
- HTML: 204
- PDF: 110
- XML: 18
- Total: 332
- BibTeX: 15
- EndNote: 25
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
See attached PDF