{"id":"1b2b1fde-fa4c-480d-966c-30c1a65c2f83","arxiv_id":"1908.08137","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"CO2 can build enough capillary pressure to invade thin seals when water films in the gas plume block water flow, as a new layered-aquifer model shows.","lead":"This paper models how injected CO2 migrates and leaks across thin sealing layers in layered aquifers. It finds that the water relative permeability inside the gas plume strongly controls capillary pressure, making seal invasion plausible at Sleipner even without a conduit.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sleipner conclusion rests on unconstrained k*_rw; if the true value is not far below 10^-4, the predicted capillary pressure drops below seal entry pressure and the leakage claim fails.","rationale":"The reader's weakest-assumption analysis identifies exactly the load-bearing point: the Sleipner conclusion switches on the value of k*_rw, the water relative permeability in the gas plume. The paper itself acknowledges that this parameter is 'very poorly constrained', and the quantitative threshold for the high-capillary-pressure regime is k*_rw << Lambda_s_w h_n, which is roughly 10^-4 for the reference parameters. The paper's k*_rw ≈ 0 end-member produces the 3.2 MPa capillary pressure quoted in Section IV, but a moderate value such as 10^-2 would reduce the maximum capillary pressure by orders of magnitude and eliminate the leakage prediction. This is not an internal inconsistency or a mathematical error; it is a parameter-identification problem in the field application. The model and the qualitative mechanism are valuable, and the sensitivity sweeps in Figures 3 and 9 are appropriate. Because the reader already assigned a CONDITIONAL verdict on precisely this basis, the stress-test pass does not change the verdict. The concrete test—directly measuring k*_rw at representative conditions—would determine whether the Sleipner claim is a plausible prediction or only one possible scenario.","tokens_in":21302,"tokens_out":5749,"duration_ms":62872,"concrete_test":"Perform a steady-state CO2-brine relative permeability measurement on Utsira Formation core material at reservoir pressure and temperature, targeting the residual water saturation regime (S_w ≈ 0.2) to obtain a direct value for k*_rw. Then rerun the Sleipner calculation in Section IV with that measured value rather than the k*_rw ≈ 0 end-member. If the measured k*_rw exceeds approximately 10^-4, the Eq. (47) limit is not active and the predicted maximum capillary pressure falls below the 2-5 MPa seal entry pressure range, so the distributed-leakage claim for Sleipner would not be supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central field claim—that CO2 at Sleipner can invade the intermediate mudstone seals without a focused conduit—depends on the high-capillary-pressure limit of Eq. (47), which requires Lambda_s_w h_n / k*_rw >> 1 (Eq. 44). With the reference Lambda_s_w = 10^-4 and h_n ~ O(1), this requires k*_rw << 10^-4. The authors state that k*_rw is 'very poorly constrained' and that 'the most likely scenario is k*_rw << 1', but they provide no measurement, core-flood data, pore-network calculation, or field calibration to support a value below this threshold. Their Sleipner estimate p_c ≈ 3.2 MPa is computed in the k*_rw ≈ 0 end-member. If the true value is at the higher end of plausibility, say k*_rw = 10^-2, Figure 3 shows that max p_c drops by two to three orders of magnitude, returning to the buoyant-overpressure limit of Eq. (46), which cannot exceed a 2 MPa seal entry pressure with realistic plume thicknesses. The model derivation is internally consistent and the mechanism is plausible, but the headline 'contrary to conventional wisdom' conclusion is a scenario conditioned on an unconstrained parameter, not a robust prediction.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper extends the vertically integrated two-phase model of Jenkins et al. (2019) to include vertical gas leakage across thin seals with a capillary entry threshold. The authors derive coupled PDEs for water and gas in layered aquifers, coupled through vertical water fluxes and through gas fluxes controlled by a smoothed step function of capillary pressure. They analyze a two-aquifer reference case and show that the maximum capillary pressure at the base of the seal is extremely sensitive to the relative permeability of water in the gas region, k*_rw. Two limiting approximations are derived: buoyant overpressure (Eq. 46) and hydraulically disconnected water column (Eq. 47). They then apply the model to Sleipner and argue that if k*_rw is small, capillary pressures may reach ~3.2 MPa and invade interbedded mudstone seals without a focused conduit.","tokens_in":21542,"tokens_out":4588,"duration_ms":42852,"significance":"If the model and its Sleipner interpretation hold, the paper provides a physically plausible mechanism for distributed CO2 leakage across thin intermediate seals, which would challenge the widespread inference of a focused conduit at Sleipner. The derivation is internally consistent, and the two limiting closures (Eqs. 46 and 47) are tested against the full numerical model in Figs. 2 and 3. The paper also gives a clear, falsifiable prediction: the leakage fraction should be strongly controlled by k*_rw. Its main limitation is that the field-scale conclusion is conditioned on a parameter, k*_rw, that the authors admit is very poorly constrained, and the model closures have not been validated against higher-fidelity simulations or experimental data.","major_comments":[{"comment":"The Sleipner estimate pc ≈ 3.2 MPa is computed in the k*_rw ≈ 0 end-member. The high-capillary limit in Eq. (47) requires k*_rw << Λ_s_w h_n, which with the reference values (Λ_s_w = 10^-4 and h_n ~ O(1)) means k*_rw << 10^-4. The paper provides no measurement, core-flood data, pore-network calculation, or field calibration to support values below this threshold. If k*_rw is near the high end of plausibility, say 10^-2, Fig. 3 shows that the maximum capillary pressure drops by two to three orders of magnitude, returning to the buoyant-overpressure limit of Eq. (46), which would not exceed the ~2 MPa seal entry pressure. The statement in the Abstract and Section IV that CO2 at Sleipner 'may be able to leak across the intermediate seals in the absence of a focused conduit' is therefore a conditional scenario rather than a robust prediction; the required constraint on k*_rw should be stated explicitly and the conclusions reframed accordingly.","section":"Section IV, Eq. (47), Fig. 3"},{"comment":"The vertical-flux ansatz (piecewise-linear q_w,z and vertically uniform q_g,z) and the assumption that leaked gas appears instantaneously in the overlying aquifer are central to the model, but they are not validated against full 2D two-phase simulations or laboratory experiments. Since the paper uses the model to make quantitative leakage predictions (e.g., M2_g fractions in Figs. 4–9), a comparison with a high-resolution numerical model for a few representative parameter sets would materially strengthen the central claims. At present, the predictive accuracy of the upscaled closure is an assumption rather than a demonstrated property.","section":"Section II.A, Eqs. (5), (10), (15)"}],"minor_comments":[{"comment":"The dimensionless notation is inconsistent: Eq. (34) defines tilde quantities, but the tildes are dropped immediately afterward, and some source terms in Eqs. (36)–(37) still appear with tildes. Please make the notation uniform throughout.","section":"Section II.D/E"},{"comment":"The sentence 'M 1 g (t = 1) = 2−M 1 g (t = 1)< 2' appears to contain an indexing error; the second term should presumably refer to M 2 g (t = 1).","section":"Section III.B"},{"comment":"The text 'see §33' when referring to the characteristic pressure should refer to the appropriate section or equation (§II.D or Eq. (33)).","section":"Section IV"},{"comment":"The entry-pressure transition function R in Eq. (23) depends on the smoothing parameter ϑ, but no numerical test of independence with respect to ϑ is reported. Please include such a test or state the range of ϑ for which the results are verified.","section":"Section II.B.2"},{"comment":"The curve labels in Fig. 3 are difficult to distinguish at printed size; using distinct line styles or markers in addition to color would improve readability.","section":"Figure 3"},{"comment":"The neglect of gas transit time through the seals is a reasonable approximation for thin seals, but its effect on the time at which gas arrives in the upper aquifer should be mentioned in the Sleipner discussion, where the seismic images show vertically stacked plumes.","section":"Section II.A.2"}],"recommendation":"major_revision","confidential_remarks":"For the editor: this is a solid modeling contribution with an internally consistent derivation, but the Sleipner conclusion rests on an unconstrained parameter, k*_rw, and the upscaled closures lack direct validation. I recommend major revision so that the authors either constrain k*_rw more strongly or clearly present the Sleipner interpretation as a conditional scenario, and so that the model is tested against a higher-fidelity reference."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is a solid upscaled model paper that makes a genuine new point—water-film relative permeability in the gas plume can dramatically change capillary pressure at the base of a seal and therefore the amount of distributed gas leakage. The Sleipner implication is real but conditional: it holds in the end-member where k*_rw is effectively zero, a parameter the authors admit is very poorly constrained. If k*_rw is not orders of magnitude below 10^-4, the leakage prediction reverts to the old buoyant-overpressure limit.\n\nThe derivation is careful and internally consistent. The model extends Jenkins et al. [25] by adding gas leakage with a capillary entry threshold, and the two limiting approximations, Eqs. (46) and (47), cleanly explain the numerical behavior in Figures 2 and 3. The parameter study is thorough, and the identification of lambda_s_w h_n / k*_rw as the controlling group is a real contribution. No quantity is fitted to the target leakage result, so there is no circularity problem.\n\nSoft spots: the Sleipner conclusion is an end-member scenario. The paper says the most likely scenario is k*_rw << 1 but gives no core-flood data, pore-network calc, or field calibration to support a value below the threshold needed for their ~3 MPa estimate. If k*_rw is only 10^-2, Figure 3 shows the maximum capillary pressure drops by two to three orders of magnitude, back to the buoyant limit, which cannot exceed a 2 MPa entry pressure with realistic plume thicknesses. That is a load-bearing uncertainty. Also, the model closures are not validated against full 2D simulations or experiments; the piecewise-linear vertical flux assumption is plausible but unverified. These are not fatal for a theory paper, but they should be acknowledged more prominently and the Sleipner application should be framed as scenario exploration, not a prediction.\n\nBottom line: the paper is a good contribution for researchers working on CO2 storage, stratified flows, and seal integrity. It deserves a serious referee. A revision should add a sensitivity analysis over k*_rw and soften the 'contrary to conventional wisdom' framing. I would cite it for the mechanism even if I wouldn't rely on the Sleipner number.","headline":"A well-derived upscaled model showing that water-film relative permeability can control seal leakage, but the Sleipner claim is an end-member scenario that needs better parameter constraints.","tokens_in":22084,"tokens_out":2858,"would_cite":true,"duration_ms":25729,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["47.56.+r"],"model":"deepseek-v4-flash","headline":"CO2 injected into layered aquifers can build enough capillary pressure to leak through thin seals without a fault or fracture conduit, and this may explain the stratified leakage at Sleipner.","keywords":["CO2 storage","layered aquifers","capillary entry pressure","gravity current","relative permeability of water in gas region","Sleipner","distributed leakage","pressure dissipation"],"falsifier":"Measure the relative permeability of water in CO2-brine systems at residual water saturation under reservoir conditions, for example with core floods at $s_{wr}\\approx0.2$: if $k^\\star_{rw}$ is of order $10^{-2}$ or larger, then $\\Lambda^s_w h_n/k^\\star_{rw}\\ll 1$ in the Sleipner scenario, the capillary pressure falls to the roughly 11 kPa buoyant-overpressure value, and the predicted distributed leakage across the mudstone seals does not occur.","tokens_in":21069,"feed_emoji":"💨","tokens_out":12775,"duration_ms":103315,"temperature":0.7,"pith_summary":"Using a vertically integrated two-phase gravity-current model, this paper argues that the capillary pressure capable of driving CO2 through a thin sealing layer is not controlled by buoyancy alone: it depends strongly on whether the gas plume blocks vertical water flow. When water can flow freely through the gas region, the capillary pressure at the base of the seal is just the buoyant overpressure, which is far too small to breach typical seals. When the water relative permeability inside the gas region is small, the water column is hydraulically disconnected and the capillary pressure becomes larger by more than two orders of magnitude in the reference case. Applying this to the Sleipner site, the model finds the capillary pressure can reach about 3.2 MPa, enough to invade mudstone seals with entry pressures around 0.5-5 MPa. This suggests that the distributed CO2 leakage observed at Sleipner could occur through the pore space of the seals without any focused conduit.","feed_headline":"Thin seals at Sleipner may leak CO2 without a conduit","feed_subtitle":"Small water permeability in the gas plume lifts capillary pressure to ~3 MPa, breaching mudstone seals.","key_machinery":"The machinery is the dimensionless ratio $\\Lambda^s_w h_n/k^\\star_{rw}$, which compares the vertical resistance to water flow through the gas-saturated region with that through the seal. It controls the hydraulic connectivity of the water column across the plume: when the ratio is small the capillary pressure follows the buoyant-overpressure limit, and when it is large the water pressure is disconnected and the capillary pressure is instead set by the water pressure in the aquifer above. The model also couples the layers through a smoothed capillary entry threshold for gas flow across seals, with dimensionless groups $\\Lambda^s_w$, $M^s_z$, and $\\tilde{p}^E_c$ governing the strength and extent of leakage.","core_discovery":"The paper's central claim is that distributed gas leakage in layered aquifers can be switched on, or dramatically amplified, by a single material property: the relative permeability of water inside the gas plume, $k^\\star_{rw}$. In the conventional limit where $k^\\star_{rw}$ is large, the water pressure remains nearly hydrostatic across the plume and the capillary pressure on the underside of the seal equals the buoyant overpressure $(\\rho_w-\\rho_g)gh_n$; at the reference parameter set this is about 11 kPa. If $k^\\star_{rw}$ is small enough that $\\Lambda^s_w h_n/k^\\star_{rw}\\gg 1$, the plume blocks vertical water flow, the water pressure is set by the overlying aquifer, and the capillary pressure rises to about 3.2 MPa. The latter exceeds the estimated entry pressure of the intermediate mudstone seals at Sleipner, so the authors conclude that CO2 may be able to leak across those seals by distributed flow even with no fault or fracture conduit.","pith_inferences":["If this mechanism operates, seismic images showing multiple stratified CO2 plumes at Sleipner may be reinterpreted as evidence of distributed seal breakthrough rather than hidden faults; a testable check is whether the predicted leakage onset matches the timing of plume appearance in each layer.","The same ratio-based mechanism should apply to other buoyant gases stored underground, such as methane or hydrogen, and to hydrocarbon migration, so containment assessments for any layered storage system should include water-relative-permeability effects in the gas-saturated zone.","Because $k^\\star_{rw}$ depends on the conductivity and connectivity of residual water films, direct core-flood measurements of CO2-brine relative permeability at representative saturations could calibrate the model and narrow the factor-of-300 uncertainty in predicted capillary pressure.","If film connectivity changes with geochemistry over time, for example through drying or mineral reactions near the injection well, leakage risk could evolve during a storage project, a coupling the present model does not include."],"forward_implications":["If $k^\\star_{rw}$ is small, the maximum capillary pressure beneath a seal can exceed the buoyant overpressure by two to three orders of magnitude, making gas leakage possible where a buoyancy-only estimate would rule it out.","Decreasing $k^\\star_{rw}$ by one to two orders of magnitude can initiate gas leakage or substantially increase the leaked mass, so uncertainty in this parameter translates directly into uncertainty in storage security.","At Sleipner parameter values, the disconnected-water limit gives a capillary pressure near 3.2 MPa, exceeding entry-pressure estimates of 0.5-5 MPa and implying that distributed leakage across the intermediate mudstone seals is plausible without a focused conduit.","The amount of gas that leaks grows roughly linearly with the seal-to-aquifer conductance ratio $\\Lambda^s_w$ and the seal gas-to-water mobility ratio $M^s_z$ when leakage is weak, and this growth saturates once leakage itself modifies the pressure field.","Distributed gas leakage can also alter plume geometry, narrowing the plume in the injection aquifer by up to about 25 percent relative to pressure dissipation alone."],"supporting_citations":[{"why":"Supplies the layered-aquifer brine-leakage model that this paper extends to include gas leakage, including the water-pressure coupling equations.","marker":"[25]"},{"why":"Introduces capillary entry pressure into gravity-current leakage and defines the buoyant-overpressure limit that the new model recovers when water is connected.","marker":"[20]"},{"why":"Provides the measured CO2-water interfacial tension used to estimate the seal entry pressure at reservoir conditions.","marker":"[34]"},{"why":"Provides pore-throat radii for North Sea shales, giving the 10-100 nm range from which entry pressures of 0.5-5 MPa are estimated.","marker":"[35]"},{"why":"Gives an independent Sleipner seal entry pressure estimate of 2-5 MPa and the geological context that the model's 3.2 MPa result is compared against.","marker":"[36]"},{"why":"Seismic characterization showing injected CO2 ponding against thin intermediate seals at Sleipner, the field observation motivating distributed leakage.","marker":"[9]"},{"why":"Documents the spatial and temporal evolution of the stratified CO2 plumes at Sleipner that the model seeks to explain.","marker":"[10]"}],"fun_headline_variants":["Water blockage in CO2 plume may breach Sleipner seals","CO2 pressure buildup may breach thin seals without faults","Sleipner study: low water flow in plume can trigger CO2 leak","Capillary pressure rise from CO2 injection may cross seals","Model shows CO2 can leak across Sleipner seals without conduit"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that water is nearly immobile inside the CO2 plume ($k^\\star_{rw}$ very small); the paper says the value is very poorly constrained and likely small, but no direct measurement is given, and if water can flow through the plume easily the capillary pressure reverts to the buoyant-overpressure limit and the Sleipner conclusion fails.","fun_headline_variants_meta":{"raw":{"variants":["Water blockage in CO2 plume may breach Sleipner seals","CO2 pressure buildup may breach thin seals without faults","Sleipner study: low water flow in plume can trigger CO2 leak","Capillary pressure rise from CO2 injection may cross seals","Model shows CO2 can leak across Sleipner seals without conduit"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000718,"raw_usage":{"total_tokens":3264,"prompt_tokens":1024,"completion_tokens":2240,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":640,"completion_tokens_details":{"reasoning_tokens":2152}},"tokens_in":640,"tokens_out":2240,"duration_ms":14062,"temperature":1.0,"reasoning_tokens":2152,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:47:54.979902+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the relative permeability of water in CO2-brine systems at residual water saturation under reservoir conditions, for example with core floods at $s_{wr}\\approx0.2$: if $k^\\star_{rw}$ is of order $10^{-2}$ or larger, then $\\Lambda^s_w h_n/k^\\star_{rw}\\ll 1$ in the Sleipner scenario, the capillary pressure falls to the roughly 11 kPa buoyant-overpressure value, and the predicted distributed leakage across the mudstone seals does not occur.","supporting_citations":[{"cited_title":"Here, we extend the work of Jenkins et al","cited_arxiv_id":null,"evidence_quote":"Supplies the layered-aquifer brine-leakage model that this paper extends to include gas leakage, including the water-pressure coupling equations."},{"cited_title":"Buoyant dispersal of CO 2 during geological storage,","cited_arxiv_id":null,"evidence_quote":"Introduces capillary entry pressure into gravity-current leakage and defines the buoyant-overpressure limit that the new model recovers when water is connected."},{"cited_title":"Impact of pressure dissipation on fluid injection into layered aquifers,","cited_arxiv_id":null,"evidence_quote":"Provides the measured CO2-water interfacial tension used to estimate the seal entry pressure at reservoir conditions."},{"cited_title":"Large-scale impact of CO2 storage in deep saline aquifers: A sensitivity study on pressure response in stratified systems,","cited_arxiv_id":null,"evidence_quote":"Provides pore-throat radii for North Sea shales, giving the 10-100 nm range from which entry pressures of 0.5-5 MPa are estimated."},{"cited_title":"Evaluation of large-scale CO 2 storage on fresh-water sections of aquifers: An exam- ple from the Texas Gulf Coast Basin,","cited_arxiv_id":null,"evidence_quote":"Gives an independent Sleipner seal entry pressure estimate of 2-5 MPa and the geological context that the model's 3.2 MPa result is compared against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Seismic characterization showing injected CO2 ponding against thin intermediate seals at Sleipner, the field observation motivating distributed leakage."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the spatial and temporal evolution of the stratified CO2 plumes at Sleipner that the model seeks to explain."}],"review_version":1}