{"id":"5c011b43-7279-4433-8fa7-49bcdd8421ed","arxiv_id":"2411.17740","paper_version":3,"verdict":"REJECT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":1,"one_line_summary":"A Strang-type time-split scheme for shallow water flows converges at order four in space and two in time on smooth tests, but its stability restriction is effectively adjustable and its Cameroon flood simulation is unphysical.","lead":"These authors propose a time-split explicit/implicit finite-difference solver for the 2D shallow water equations and report fourth-order spatial, second-order temporal accuracy on analytical tests. The paper also applies the solver to 2024 floods in Cameroon, but the stability condition contains a free parameter and the flood simulation yields depths and velocities that are physically impossible.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.1's stability restriction (47) is vacuous as stated: γ is a free parameter (0<γ≤18) not tied to the matrix A, so any time step can be admitted by choosing γ small; the proof also drops the nonlinear term at (58)/(66) without a quantitative bound.","rationale":"The central claim is that the time-split scheme (38)-(45) is stable under the explicit time step restriction (47). That claim fails for two independent reasons. First, the parameter γ in (47) is declared only to satisfy 0<γ≤18; it is not set equal to ρ_max(A), the spectral radius of the matrix A appearing in the proof. Because γ can be arbitrarily small, condition (47) is vacuously satisfiable for any time step, so it provides no restriction and no stability guarantee. The numerical section compounds this by choosing different γ values (18, 1, 12, 3) across tables, which is only possible if γ is treated as a free tuning parameter rather than a fixed property of the discretization. Second, the proof contains an unjustified approximation at (58) and (66), where a norm involving δ⁴_xE minus a k/4 term is replaced by the norm of δ⁴_xE alone with only an appeal to 'small k'; no bound is given for the discarded term, so the implication is not rigorous. These defects strike at the theorem that constitutes the paper's main theoretical contribution. The numerical convergence tables on smooth Thacker solutions are plausible, but they do not validate a stability condition that is either vacuous or unproven, and the flood simulation yields physically impossible values (velocities ≈ 1.7×10^6 m/s, depths diverging to 10^129 m), which the paper itself attributes to unsuitable initial conditions. The reader's REJECT verdict is therefore appropriate: the central stability claim is not established, and the practical flood-forecasting claim is not supported by the reported results.","tokens_in":28562,"tokens_out":5820,"duration_ms":54312,"concrete_test":"Compute ρ_max(A) numerically for the pentadiagonal matrix A in Eq. (60) using standard linear algebra (e.g., eig in MATLAB). Then re-run Tables 1-4 with γ set to this computed ρ_max(A) in condition (47) instead of the tabulated γ values (1, 3, 12, 18). If the time steps used in the tables violate the resulting bound, the experiments did not test the restriction with the operator-dependent constant derived in the proof; if the scheme remains stable despite violation, the restriction is not necessary.","verdict_should_be":"REJECT","load_bearing_attack":"Theorem 3.1 asserts stability under k ≤ (48/γ) min{ ||β||_0/(√(Mx−3)|||u|||_{0,∞}), |||u|||_{0,∞}/|||u²+½gh|||_{0,∞} } Δx, with 0<γ≤18. The proof derives bounds (64) and (68) with ρ_max(A) in the denominator and only notes 0<ρ_max(A)≤18; γ is never identified with ρ_max(A). Since γ may be any positive number ≤18, for any prescribed k one can choose γ small enough to satisfy (47). Thus (47) imposes no upper bound on k and cannot function as a stability restriction. The numerical tables use γ=18, 1, 12, 3, confirming it is treated as tunable rather than as a property of the discrete operator. Independently, the step from (57) to (59) and (65) to (67) replaces a norm of δ⁴_xE − (k/4)δ²_x(...) by the norm of δ⁴_xE alone, justified only by 'small k'. From k||A−kB||<C one cannot conclude k||A||≤C without control of ||B||; no such bound is supplied. Hence Theorem 3.1 is not established as a sufficient condition for stability.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a time-split linearized explicit/implicit finite-difference scheme for the two-dimensional shallow-water equations, with a stability restriction claimed in Theorem 3.1 (Eq. (47)). It reports fourth-order spatial and second-order temporal convergence on two Thacker analytical solutions, and applies the scheme to a 2024 flood event in the Far North region of Cameroon.","tokens_in":28858,"tokens_out":7509,"duration_ms":63597,"significance":"If the stability theorem and convergence rates were fully established, the scheme would offer a useful high-order time-split solver for shallow-water flows. The numerical convergence tables (Tables 1-4) are coherent and benchmark against external analytical solutions, which gives credible evidence of accuracy on smooth problems. The Cameroon flood application, however, produces nonphysical values, and the stability theorem as stated is vacuous because of the free parameter γ.","major_comments":[{"comment":"The stability restriction (47) involves a free parameter γ with 0 < γ ≤ 18 that is never tied to the discrete operator. The proof itself derives the bounds (64) and (68) with ρ_max(A) — the spectral radius of the pentadiagonal matrix A — in the denominator, but the theorem substitutes an arbitrary γ. Consequently, for any positive time step k and any data, one can choose γ small enough (while keeping γ > 0) so that (47) holds; the condition therefore imposes no upper bound on k and cannot function as a stability restriction. The numerical tables use γ = 18, 1, 12, 3, confirming that γ is treated as tunable. The theorem should be restated with γ identified as ρ_max(A) or with an explicit upper bound derived from A.","section":"Theorem 3.1, Eq. (47)"},{"comment":"The step from (57) to (59) and from (65) to (67) replaces a norm of δ⁴_x E − (k/4)δ²_x(...) by the norm of δ⁴_x E alone, justified only by 'for small values of the time step k' and an unquantified assertion that dropping the term 'does not compromise the result on stability.' No bound on the dropped term is proved, and the direction of the inequality is not preserved by this replacement. Since Theorem 3.1 claims a sufficient condition for stability, the missing control of the k² term is a load-bearing gap.","section":"Theorem 3.1 proof, Eqs. (57)-(59) and (65)-(67)"},{"comment":"The flood simulation results are nonphysical: reported peak values include water depth 2×10² m, water depth up to 3.3×10¹²⁹ m, and velocities up to 1.7×10⁶ m/s and 10¹⁵³ m/s. The text itself notes that small initial depths or large initial velocities destroy the computed solutions, yet the section still presents the peak date of 21 September 2024 as a forecast. As reported, the application does not support the claim that the scheme is a reliable flood-prediction tool.","section":"Section 4, flood application, Figure 5"},{"comment":"The paper asserts that the implicit second stage (39) is unconditionally stable and that the scheme is temporally second-order accurate and spatially fourth-order convergent, but no proof is given for either statement: stability of the nonlinear implicit stage is stated without argument, and the convergence order is only inferred from the local truncation error estimates (28) and (36) together with the numerical tables. If the stability theorem is repaired, the unconditional stability of stage (39) should also be justified.","section":"Sections 2 and 3, stability of stage (39) and convergence order"}],"minor_comments":[{"comment":"Equation (15) uses the same notation (·,·)_∼ and ∥·∥∼ for both R^{Mx−3} and R^{My−3} without distinguishing the dimensions; this is confusing for the reader.","section":"Eq. (15)"},{"comment":"The symbol ≈ in equations (58) and (66) is informal; if the dropped terms are truly negligible, the proof should state the precise inequality that justifies the replacement.","section":"Eqs. (58) and (66)"},{"comment":"The Manning friction formulas in (4) have unusual powers of u and v; a reader cannot verify dimensional consistency from the text, and the relation to the standard Manning formula should be clarified.","section":"Eq. (4)"},{"comment":"In the caption of Figure 5, the mesh size in the y-direction is given as both 12.36 and 111; the discrepancy should be resolved.","section":"Figure 5 caption"},{"comment":"There are numerous typos, including 'T ables' and 'Abowitz' in reference [47]; the manuscript needs a careful proofreading before resubmission.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The stability theorem is the central theoretical result, and the free-parameter issue is a genuine correctness problem; however, the numerical accuracy evidence is sound. I believe a major revision addressing the theorem statement, the missing estimates, and the unphysical flood outputs could make the paper acceptable. I also note that the manuscript cites 21 previous works by the author among 60 references; this is not by itself improper, but the novelty relative to those papers should be made explicit."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper has one credible numerical result and two serious problems. On Thacker's analytical solutions, the split scheme shows clean fourth-order spatial and second-order temporal convergence. That part is reproducible and suggests the method does work on smooth problems.\n\nBut the stability theorem is not what it claims to be. Restriction (47) contains a free parameter gamma in (0,18] that never gets tied to the discrete operator. In the proof, gamma would have to be rho_max(A), but since it is not, you can satisfy (47) for any time step by taking gamma small enough. The numerical section uses gamma=18, 1, 12, 3, which confirms it is treated as a tuning knob. On top of that, the step from (58) to (59) and (65) to (67) drops the k/4 terms with only \"for small k\" as justification, and no bound on the neglected product is supplied. So Theorem 3.1 is not a proven sufficient condition; it is a heuristic dressed as a theorem.\n\nThe Cameroon flood case is worse. Reported velocities around 1.7e6 m/s and depths blowing up to 1e129 m are not flood forecasts; they are signs that the simulation has broken down. The paper itself admits small initial depth or large velocities destroy the solutions, but then still presents those numbers as part of the flood analysis. That section does not support the applied claim.\n\nWhat is genuinely new is the specific combination of Strang splitting with a fourth-order explicit x-step and an implicit y-step. That is a reasonable component to exist, and the convergence tables give some evidence it performs as advertised on smooth test problems. The writing is clear and the scheme is described in enough detail to reproduce.\n\nFor a serious referee, I would want the stability theorem fixed or honestly reframed as a CFL heuristic, and the flood section either validated against observations or removed. In its current form, the paper's central theoretical claim and its practical application both fail. I would not cite it this year, and I would not send it to review as is.\n\nThat's my read.","headline":"The convergence tests on Thacker solutions look real, but the stability theorem is vacuous as stated and the Cameroon flood application is unphysical; the scheme might be salvageable, but the paper in its current form should not be published.","tokens_in":29363,"tokens_out":3892,"would_cite":false,"duration_ms":36802,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M12","65M06"],"pacs":[],"model":"deepseek-v4-flash","headline":"A symmetric time-split explicit/implicit algorithm for the two-dimensional shallow water equations is stable under a computable $\\Delta x$-scaled time-step restriction and, on test problems, converges at fourth order in space and second…","keywords":["two-dimensional shallow water equations","time-split explicit/implicit scheme","symmetric operator splitting","fourth-order spatial accuracy","second-order temporal accuracy","stability time-step restriction","Cameroon far north floods","hydrodynamic flood forecasting"],"falsifier":"Run the full scheme (38)-(45) on a smooth periodic test and compare its largest stable time step with restriction (47) across a range of $\\gamma$: a run that satisfies (47) with $\\gamma$ near 18 and nevertheless grows in the $L^2$ norm would falsify the claimed sufficient condition. A more direct check is to evaluate the dropped term $\\frac{k}{4}\\delta^2_x(u\\delta^{3\\mp}_xE+h\\delta^{3\\mp}_xE)$ alongside $\\delta^4_xE$ over many steps; if its norm is not small compared with the retained term uniformly as $k\\to 0$, the approximation behind (58) and (66) fails.","tokens_in":28336,"feed_emoji":"🌊","tokens_out":12305,"duration_ms":108191,"temperature":0.7,"pith_summary":"This paper aims to replace expensive two-dimensional shallow-water solves with a three-stage split: an explicit sweep in $x$, an implicit sweep in $y$, and another explicit sweep in $x$. The paper claims a stability theorem (Theorem 3.1) giving an explicit, $\\Delta x$-scaled restriction on the time step, and numerical tests on exact solutions suggest the scheme is fourth-order accurate in space and second-order in time. If the scheme behaves as claimed, it offers a fast high-order alternative for dam-break and overland flood simulation, and the paper applies it to the 2024 floods in Cameroon's far north region, locating the peak of inundation around 21 September. The claim matters because flood forecasting needs solvers that are fast enough to run in real time and accurate enough to resolve sharp moving water fronts.","feed_headline":"Fourth-order time-split scheme maps Cameroon flood peaks","feed_subtitle":"The split solver runs 1D steps, reaches second order in time, and places the 2024 Cameroon flood peak on September 21.","key_machinery":"The central object is the symmetric operator composition $P(k)=P_1(k/2)P_2(k)P_1(k/2)$: the two outer explicit steps use the same operator, which makes the composite sequence symmetric and hence second-order in time, while the inner implicit step is unconditionally stable. The stability proof is carried by the matrix form of the fourth-order difference operator, $\\delta^4_x=\\frac{1}{12\\Delta x}A$, where $A$ is a normal pentadiagonal matrix. A standard eigenvalue localization argument bounds the norm of $A$ by 18, and that single bound is what converts the stability inequality into the user-visible restriction (47) involving the computed solution's norms $\\|u\\|_{0,\\infty}$ and $\\|u^2+\\frac12 gh\\|_{0,\\infty}$.","core_discovery":"The paper's central claim is that the time-split linearized explicit/implicit scheme (38)-(45) is stable under the time-step restriction (47) and numerically achieves fourth-order spatial accuracy with second-order temporal convergence. The scheme is written as the symmetric composition $P(k)=P_1(k/2)P_2(k)P_1(k/2)$, where $P_1$ is an explicit one-dimensional operator built from fourth-order spatial differences and a quadratic linearization of the $x$-flux, and $P_2$ is an implicit operator obtained by averaging the explicit operator at the two intermediate levels for the $y$-flux and source term. The proof of Theorem 3.1 treats the implicit middle step as unconditionally stable and reduces the stability of the whole scheme to the two explicit steps; the fourth-order difference $\\delta^4_x$ is represented as $\\frac{1}{12\\Delta x}A$ with a normal pentadiagonal matrix $A$ whose spectral radius is bounded by 18. From that bound the paper derives restriction (47), and the reported experiments on the two exact shallow-water solutions show $L^\\infty$ errors decreasing by factors close to $3^4$ under mesh refinement and close to $3^2$ under time-step refinement. The same algorithm is then applied to the Logone basin to simulate the July-October 2024 flood season.","pith_inferences":["The proof of Theorem 3.1 discards the $\\frac{k}{4}\\delta^2_x(\\cdot)$ correction term inside the explicit operator without a bound on that term, so a careful reader should treat restriction (47) as a validated guideline with a tuning parameter $\\gamma$ rather than as a fully proven sufficient condition.","The same matrix-bounding argument does not use the detailed form of the shallow-water source terms, so the split-template should transfer to other hyperbolic balance laws; testing it on the full Saint-Venant system with wetting and drying would be a natural next step.","The Cameroon application is driven by 2013 discharge climatology, so a true operational forecast would require comparing the computed inundation extents against observed gauge or satellite data for 2024."],"forward_implications":["Each time level costs one explicit sweep in $x$, one implicit sweep in $y$, and one explicit sweep in $x$, so the solver avoids inverting two-dimensional block systems while keeping fourth-order spatial accuracy.","On the exact shallow-water tests, the $L^\\infty$ errors for water depth and both velocity components shrink by factors close to $3^4$ as the mesh is refined by 3 and by factors close to $3^2$ as the time step is refined by 3, matching the advertised orders.","In the Logone basin application the computed water depth and velocities reach their peak around 21 September 2024, which the paper identifies with the observed flood peak in the far north region.","Because the restriction (47) depends on norms of the computed solution, a user must monitor the admissible time step during a run; the numerical section reports that too-small initial depths or too-large initial velocities can let oscillations destroy the solution."],"supporting_citations":[{"why":"Supplies the symmetric operator-splitting construction used to make the composite scheme second-order in time.","marker":"[46]"},{"why":"Provides the exact shallow-water solutions with moving shorelines used to verify the convergence orders.","marker":"[48]"},{"why":"Supplies the eigenvalue localization and normality results used to bound the norm of the pentadiagonal difference matrix.","marker":"[45]"},{"why":"Provides the fourth-order difference approximations on which the spatial accuracy of the scheme rests.","marker":"[28]"},{"why":"Defines the two-dimensional hydrodynamic overland-flow model and the closed-boundary treatment used in the simulations.","marker":"[9]"},{"why":"Introduces the predictor-corrector idea from which the time-split rapid solver is built.","marker":"[25]"},{"why":"Gives the finite-difference scheme and the CFL condition used as the guideline for the allowable time step.","marker":"[10]"}],"fun_headline_variants":["4th-order split solver maps Cameroon floods","Cameroon flood peaks from a fourth-order split scheme","Fourth-order split scheme tames 2D Cameroon floods","2024 Cameroon flood simulated by fourth-order split solver"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that, for small time steps, the correction term $\\frac{k}{4}\\delta^2_x(\\cdots)$ inside the explicit operator can be dropped from the stability estimate without hiding an instability; the proof relies on this approximation but does not bound the discarded term.","fun_headline_variants_meta":{"raw":{"variants":["4th-order split solver maps Cameroon floods","Cameroon flood peaks from a fourth-order split scheme","Fourth-order split scheme tames 2D Cameroon floods","2024 Cameroon flood simulated by fourth-order split solver"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001007,"raw_usage":{"total_tokens":4259,"prompt_tokens":950,"completion_tokens":3309,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":566,"completion_tokens_details":{"reasoning_tokens":3247}},"tokens_in":566,"tokens_out":3309,"duration_ms":21506,"temperature":1.0,"reasoning_tokens":3247,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:57:40.200205+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the full scheme (38)-(45) on a smooth periodic test and compare its largest stable time step with restriction (47) across a range of $\\gamma$: a run that satisfies (47) with $\\gamma$ near 18 and nevertheless grows in the $L^2$ norm would falsify the claimed sufficient condition. A more direct check is to evaluate the dropped term $\\frac{k}{4}\\delta^2_x(u\\delta^{3\\mp}_xE+h\\delta^{3\\mp}_xE)$ alongside $\\delta^4_xE$ over many steps; if its norm is not small compared with the retained term uniformly as $k\\to 0$, the approximation behind (58) and (66) fails.","supporting_citations":[{"cited_title":"Stoer, R","cited_arxiv_id":null,"evidence_quote":"Supplies the symmetric operator-splitting construction used to make the composite scheme second-order in time."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the exact shallow-water solutions with moving shorelines used to verify the convergence orders."},{"cited_title":"Sportisse","cited_arxiv_id":null,"evidence_quote":"Supplies the eigenvalue localization and normality results used to bound the norm of the pentadiagonal difference matrix."},{"cited_title":"Fiedler, J.A","cited_arxiv_id":null,"evidence_quote":"Defines the two-dimensional hydrodynamic overland-flow model and the closed-boundary treatment used in the simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the predictor-corrector idea from which the time-split rapid solver is built."},{"cited_title":"Garcia, R","cited_arxiv_id":null,"evidence_quote":"Gives the finite-difference scheme and the CFL condition used as the guideline for the allowable time step."}],"review_version":1}