{"id":"48161a43-11ce-4b45-9876-31dcbe328405","arxiv_id":"2501.09956","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For subcritical fractional dissipation s in (1,2) and any H^sigma initial data with sigma>3, and for critical s=1 with small data and noise, the 3D primitive equations with Stratonovich transport noise have unique local pathwise solutions.","lead":"The paper proves that the three-dimensional primitive equations of ocean-atmosphere dynamics, with fractional dissipation and random transport noise, have unique solutions that exist for a positive (possibly short) time whenever the noise and initial data are regular enough. It is the first local well-posedness result for this fractional stochastic setting and introduces new commutator estimates for the hydrostatic projection.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The main theorem hinges on the new negative-Sobolev commutator estimate in Lemmas B.3/B.4; a half-derivative error there would break the Itô-Stratonovich correction bound (3.6).","rationale":"The reader's weakest_assumption identifies exactly the same load-bearing step: Lemma B.4 combined with Lemma B.3, feeding into Proposition 3.1's estimate (3.6). I agree that this is the most fragile unverified ingredient. I do not find a proven contradiction: the Fourier-splitting argument in Lemma B.3 appears repairable using the lattice lower bound, Lemma B.2 and B.4 are consistent with the stated derivative counts, and the L^p/L^2 gap flagged by the Reader is already addressed by the level-set truncation in the proof of Theorem 2.3. The stopping-time definition in (3.19) may be a typographical issue about W^{1,∞} versus H^σ thresholds, but it is readily repaired and is not the central risk. Since the remaining concern is about an intricate but plausible new estimate rather than an identified error, the CONDITIONAL verdict remains appropriate; I see no basis to move to ACCEPT or REJECT without independent verification of the commutator lemma.","tokens_in":30211,"tokens_out":39949,"duration_ms":376259,"concrete_test":"Re-derive Lemma B.3 on T^3 using a Littlewood-Paley/paraproduct decomposition with explicit Bony estimates, targeting the exact bound ‖Λ^{-1/2}[Λ^σ, b·∇]V‖ ≲ ‖b‖_{σ+3}‖V‖_{σ-1/2} (up to the ∂_z b^h term in Lemma B.4). Then trace the constants through Lemma B.4 into the I1 contribution in (3.4) and check that no term requires more than ∑‖b_k‖^2_{σ+3} or ∑‖b_k‖_{σ+3}‖∂_z b^h_k‖_{σ-3/2}. A useful sub-check is to test the Fourier multiplier inequality numerically for two-mode data b=cos(ℓ·x) and V=cos(m·x) over a range of ℓ,m, including m≈-ℓ; any violation indicates the derivative count in (3.6) is wrong.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 2.3 is built on Proposition 3.1, and the only place where the novel hydrostatic-projection commutator is essential is the estimate of the Itô-Stratonovich corrector at Eq. (3.6). Concretely, the term involving [Λ^σ, B_k][P, B_k]V_n is bounded via ‖Λ^{-1/2}[Λ^σ, B_k][P, B_k]V_n‖ ‖Λ^{σ+1/2}V_n‖, and Lemma B.4 supplies the bound ‖b‖_{σ+3}‖∂_z b^h‖_{σ-3/2}‖V‖_{σ+1/2} + ‖b‖^2_{σ+3}‖V‖_{σ-1/2}. If the true estimate lost even half a derivative, the ‖V‖_{σ+1/2} term would no longer be absorbable by ‖Λ^{σ+s/2}V‖ for s close to 1, and the whole local-well-posedness argument would collapse. The proof of Lemma B.3 is a Fourier-side splitting whose cancellations are delicate; one compressed point is the bound |j|^α/(|k|^α|k-j|^α)≲1, which holds only because the torus frequencies are bounded away from zero. The lemma is not independently verified and is not machine-checked. This is the single most load-bearing unverified ingredient. The Reader's separate L^2-versus-L^p mismatch is not the main risk: the level-set truncation in the proof of Theorem 2.3 already supplies L^p-moment initial data for each piece.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a local well-posedness theorem for the three-dimensional primitive equations with fractional dissipation ( -Δ)^{s/2}, s∈[1,2), driven by Stratonovich transport noise. The main result, Theorem 2.3, states that for σ>3 and noise coefficients in ℓ²(N,H^{σ+3}), every initial datum V0∈L²(Ω,H^σ) admits a maximal pathwise solution when s∈(1,2); in the critical case s=1 the same conclusion is claimed for small initial data and small vertical noise variation as quantified by (2.3). The proof is organized around a Galerkin approximation of a cut-off system, uniform energy estimates (Proposition 3.1), a compactness argument yielding martingale solutions (Proposition 3.2), and pathwise uniqueness via a double cut-off (Proposition 3.3). The central new ingredient is a collection of commutator estimates involving the hydrostatic Leray projection, especially Lemma B.4, whose proof depends on the negative-Sobolev commutator estimate Lemma B.3. The paper also contains a short discussion of the supercritical case s<1, explaining why the method would require analytic initial data and why the Stratonovich-corrector cancellation fails for spatially dependent noise in that setting.","tokens_in":30533,"tokens_out":36443,"duration_ms":340202,"significance":"If the main theorem is correct, this is the first local well-posedness result for 3D fractionally dissipated primitive equations with transport noise, and it gives a natural interpolation between the fully viscous and inviscid stochastic theories. The paper is unusually concrete: the Galerkin estimates are written out in detail, the compactness passage is standard but complete, and the new commutator lemmas in Appendix B are proved in the paper rather than cited. The treatment of the critical case with an explicit smallness condition on ∂_z b^h is a genuine contribution, and the supercritical discussion is honest and includes a worked example showing why analytic weights break the key cancellation. The main weaknesses are localized to the final localization step and to the statement of the critical smallness threshold; these are repairable without changing the strategy.","major_comments":[{"comment":"The stopping time τ = inf{t≥0: ||V||_σ > ρ} does not ensure that the cut-off function θρ(||V||_{W^{1,∞}}) is identically 1 on [0,τ). The cut-off is activated when ||V||_{W^{1,∞}} > ρ/2, while the Sobolev embedding only gives ||V||_{W^{1,∞}} ≤ C0||V||_σ; hence ||V||_σ can be strictly below ρ at the first time the cut-off drops below 1. Consequently the stopped cut-off solution is not in general a solution of the original equation (2.4) on the whole interval [0,τ). The argument should use τ = inf{t≥0: ||V||_{W^{1,∞}} > ρ/2}, or otherwise prove that the cut-off remains equal to 1 up to the stopping time under the stated choice of ρ.","section":"Proof of Theorem 2.3, after (3.19)"},{"comment":"The displayed condition 2C0M < ρ < 1/(2Cσ) with M := 1/(4C0Cσ) is inconsistent, because 2C0M = 1/(2Cσ). The intended smallness condition is 2C0||V0||_σ < ρ, which is satisfiable only if ||V0||_σ < 1/(4C0Cσ) up to the exact universal constants. The theorem statement's threshold ||V0||_σ < 1/C0 is therefore not justified by the proof as written. Please correct the constant in both the statement and the proof; the qualitative claim that sufficiently small initial data are allowed in the critical case remains plausible.","section":"Proof of Theorem 2.3, critical case s=1"},{"comment":"The stochastic integral appearing after the definition of Y_t is, under the stated L²-type hypotheses, only a continuous local martingale: the integrand has a.s. finite ∫|·|²dt, but the expectation of the square root need not be finite. Before concluding E[Y_t||V(t)||²_{σ−1/2}] ≤ 0, one should apply a localization argument, for instance stopping at τ_N = inf{t : ∫_0^t (sum_k |⟨Λ^{σ−1/2}PB_kV, Λ^{σ−1/2}V⟩|²) dr ≥ N}, and then pass N→∞. This is a standard but necessary step.","section":"Proposition 3.3, final step"}],"minor_comments":[{"comment":"In the displayed chain for I21, the first line drops the factor |j| and the second line introduces |k−j|^s|j|^{s−α}; as written this is not a valid algebraic inequality. The desired bound still follows by using |j| ≤ |k−j| on the region |j| ≤ 1/2|k−j|, but the display should be rewritten for clarity.","section":"Lemma B.3, estimate of I21"},{"comment":"In the bound for the Itô-Stratonovich corrector, the norms of ~V_{n_j}−~V and ~V appear to be L² norms. Since (PB_k)^2 is a second-order operator, the estimate needs H^σ norms of the difference (or a test function with two derivatives) to be valid as written.","section":"Proposition 3.2, linear term convergence"},{"comment":"The Wiener process components are denoted (~W^k)_{k≥0}, while all sums in the paper start at k=1; the indexing should be made consistent.","section":"Definition 2.2"},{"comment":"The linear operator P defined by P e_k = ~p_k and the hydrostatic Leray projection P defined in (2.1) share the same symbol; this can be confusing and the notation should be changed.","section":"Section 2.2"},{"comment":"In the line \"V ∈ L²(Ω; C([0,T;H^σ]) ∩ L²(0,T;H^{σ+s/2}))\", the bracket in C([0,T;H^σ]) should read C([0,T];H^σ).","section":"Proof of Theorem 2.3"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious contribution and I do not see a fatal obstruction in the central commutator estimates. The two main issues — the stopping time/cut-off mismatch and the inconsistent critical smallness threshold — are localized in the final step of the proof and in the statement of Theorem 2.3, and both are repairable. The local martingale point in Proposition 3.3 is also standard to fix. I recommend major revision rather than rejection, and I expect the corrected version to be a strong paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The result is new: first local well-posedness for 3D fractionally dissipated primitive equations with transport noise, subcritical s in (1,2) for arbitrary H^sigma data, critical s=1 for small data and small vertical noise variation. The paper is honest about what it does not do: supercritical s<1 with general noise remains open, and the analytic-setting cancellation failure is stated plainly.\n\nThe proof architecture is standard but well executed: Galerkin uniform estimates, compactness to martingale solutions, pathwise uniqueness via a double cutoff, then Yamada–Watanabe. The double cutoff for uniqueness is a nice touch. The genuinely original piece is the commutator machinery in Appendix B, especially Lemma B.4.\n\nThat is also the soft spot. Lemma B.4 is load-bearing: it alone controls the Itô–Stratonovich corrector through estimate (3.6), which the fractional dissipation must absorb. If that bound lost even half a derivative, the whole local-well-posedness argument collapses near s=1. The proof is intricate and not machine-checked. The stress-test flags the Fourier splitting in Lemma B.3, and that is the right place to look. I do not see an obvious error—the frequency-splitting cases cover the regimes, and the torus bound |j|^alpha/(|k|^alpha|k-j|^alpha) ≲ 1 is legitimate because frequencies are bounded away from zero—but the derivative counts are delicate enough that a referee should verify them line by line.\n\nThe L^2-versus-L^p mismatch the reader flagged is minor. The proof of Theorem 2.3 explicitly localizes by level sets of ||V0||_sigma, which gives L^p bounds for each piece; it is a missing sentence, not a gap.\n\nI saw no evidence of curve-fitting or circularity. Self-citations are contextual and do not carry the argument. The paper is clearly written for a specialist audience, and the new commutator estimates will likely be reused elsewhere.\n\nRecommendation: send to peer review. A serious referee should check Lemmas B.3 and B.4; if they hold, this is a solid contribution to the stochastic primitive-equation literature.","headline":"Credible local well-posedness for a new stochastic 3D primitive-equation regime, resting on a single novel commutator estimate that deserves referee scrutiny.","tokens_in":31054,"tokens_out":1593,"would_cite":true,"duration_ms":17347,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q86","60H15","76M35","35Q35","86A10"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves local existence and pathwise uniqueness for the three-dimensional primitive equations with fractional dissipation and Stratonovich transport noise.","keywords":["stochastic primitive equations","transport noise","fractional dissipation","hydrostatic Leray projection","pathwise uniqueness","local well-posedness","critical dissipation","Sobolev spaces"],"falsifier":"Do the explicit Fourier calculation that Lemma B.4 is designed to control: take $b=(0,0,\\sin z)$ and $\\phi=(e^{{\\rm i}N x_1}\\cos(2\\pi z),0,0)$ on the three-torus, with the symmetries of Section 2, and compare both sides of the second bound in Lemma B.4 as $N\\to\\infty$; if the left side grows faster than $O(N^{s+1/2})$, the estimate and the main theorem collapse.","tokens_in":29996,"feed_emoji":"🌊","tokens_out":9974,"duration_ms":98257,"temperature":0.7,"pith_summary":"This paper proves that the three-dimensional primitive equations of large-scale ocean and atmosphere dynamics, damped by the fractional Laplacian $(-\\Delta)^{s/2}$ for $s\\in(1,2)$, have unique local pathwise solutions in $H^\\sigma$ ($\\sigma>3$) for arbitrary initial data, when the equations are driven by Stratonovich transport noise. At the critical dissipation index $s=1$, the same conclusion holds provided the initial data are small and the vertical shear of the noise coefficients is small. The significance is that fractional dissipation interpolates between the fully viscous case, where global well-posedness is known, and the inviscid case, which is ill-posed and can blow up; the paper shows the subcritical and critical regimes remain well-posed despite the loss of horizontal derivatives and the singular nature of the hydrostatic Leray projection. The argument hinges on new commutator estimates that make the Itô–Stratonovich correction controllable by the fractional dissipation.","feed_headline":"Fractionally dissipated ocean equations get unique local solutions","feed_subtitle":"Subcritical dissipation handles any data; the critical case needs small data and small noise.","key_machinery":"The engine is the hydrostatic Leray projection $P\\phi=\\phi-\\nabla_h\\Delta_h^{-1}\\nabla_h\\cdot\\phi$, which removes the barotropic component and eliminates pressure; unlike the usual Leray projection, its symbol is singular along the entire vertical-frequency axis, so standard commutator cancellations fail. To compensate, the paper proves commutator estimates in negative Sobolev norms (Lemma B.3) and, as the central new input, the two bounds of Lemma B.4 for $[P,b\\cdot\\nabla]\\phi$ and for $\\Lambda^{-1/2}[\\Lambda^s,b\\cdot\\nabla][P,b\\cdot\\nabla]\\phi$. These estimates decompose the Itô–Stratonovich corrector into terms whose regularity loss is at most half a derivative, which the fractional Laplacian can dominate; they are what makes the energy estimates of Proposition 3.1 close.","core_discovery":"The central claim is Theorem 2.3: for $\\sigma>3$, $s\\in(1,2)$, and noise coefficients in $\\ell^2(\\mathbb{N},H^{\\sigma+3})$, every $H^\\sigma$ initial datum gives a maximal pathwise solution of the fractionally dissipated primitive equations; for $s=1$ the same holds under the smallness conditions (2.3) and $\\|V_0\\|_\\sigma<1/C_0$. The mechanism is a cancellation between the Itô–Stratonovich corrector and the noise energy input: the paper proves that $\\langle\\Lambda^\\sigma P(b_k\\cdot\\nabla P(b_k\\cdot\\nabla))V,\\Lambda^\\sigma V\\rangle + \\|\\Lambda^\\sigma P(b_k\\cdot\\nabla V)\\|^2$ is bounded by order $\\|b_k\\|^2\\|V\\|_{\\sigma+1/2}^2$ plus lower-order terms. Because this is exactly the order of the quadratic nonlinearity, subcritical dissipation $s>1$ absorbs it by interpolation, while critical dissipation $s=1$ absorbs it only when the coefficient in front of the $\\sigma+1/2$ norm is small, which is what the noise and data smallness conditions enforce.","pith_inferences":["If Lemmas B.3–B.4 are correct, the same corrector-cancellation template should apply to other anisotropic fluid models whose projection symbol has a singular set of positive dimension, giving a general route for transport noise under weak dissipation.","The smallness condition singles out $\\partial_z b^h$ as the quantity controlling the critical case; the paper does not explore it, but a natural conjecture is that large vertical shear of the noise causes finite-time loss of regularity at $s=1$ even from small data.","The analytic-class failure in the supercritical case suggests that any future well-posedness result for $s<1$ will need either noise independent of the vertical variable, or a weakening of the solution concept.","A sharpened version of (2.3) that tracks only the highest-order symbol of $\\partial_z b^h$ might relax the critical smallness assumption; testing this would require only revisiting the estimates in Proposition 3.1."],"forward_implications":["For any $\\sigma>3$ and $s\\in(1,2)$, local well-posedness holds for every initial datum of finite $H^\\sigma$ norm, with a maximal existence time characterized by the norm exceeding any prescribed level.","For $s=1$, local well-posedness holds for small initial data, and the required noise smallness (2.3) is automatically satisfied when the horizontal noise components are independent of $z$.","Pathwise uniqueness holds, so the martingale solutions constructed by compactness are actually strong (pathwise) solutions on the original system up to the explosion time.","The double cutoff technique used for uniqueness improves the earlier analytic-class argument for stochastic inviscid primitive equations, replacing analytic regularity with Sobolev regularity in the subcritical regime.","For $s<1$ the argument cannot work: Sobolev well-posedness is impossible, and the cancellation behind the corrector bound fails in the analytic class unless the noise coefficient is spatially constant, leaving the supercritical case open."],"supporting_citations":[{"why":"Frames the fractional-dissipation threshold via the deterministic 2D primitive equations and supplies the ill-posedness results for supercritical and critical large data.","marker":"[2]"},{"why":"Supplies the Stratonovich-to-Itô conversion and the turbulent-pressure transport-noise formulation used to write the equation.","marker":"[6]"},{"why":"Provides the pseudodifferential commutator estimate (B.1) on which Lemma B.2 and the double-commutator bounds rest.","marker":"[16]"},{"why":"Supplies the fractional Leibniz/product estimates of Lemma A.1 used throughout the nonlinear and noise estimates.","marker":"[19]"},{"why":"Provides the analytic-class existence framework for stochastic inviscid primitive equations that the supercritical discussion compares against and whose uniqueness argument is improved here.","marker":"[35]"},{"why":"Gives the Aubin–Lions–Simon compactness lemma used to prove tightness and the martingale-solution step.","marker":"[54]"}],"fun_headline_variants":["Unique local ocean solutions: subcritical any data, critical small","Fractional dissipation + transport noise: local well-posedness","Ocean equations: novel commutator estimates yield local uniqueness","Subcritical all data, critical small: pathwise primitive equation solutions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof stands on the new commutator estimates in Lemmas B.3 and B.4; if the double commutator $\\Lambda^{-1/2}[\\Lambda^s,b\\cdot\\nabla][P,b\\cdot\\nabla]\\phi$ actually loses half a derivative more than the paper claims, the fractional dissipation cannot absorb the noise, and the energy estimates fail.","fun_headline_variants_meta":{"raw":{"variants":["Unique local ocean solutions: subcritical any data, critical small","Fractional dissipation + transport noise: local well-posedness","Ocean equations: novel commutator estimates yield local uniqueness","Subcritical all data, critical small: pathwise primitive equation solutions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000273,"raw_usage":{"total_tokens":1628,"prompt_tokens":930,"completion_tokens":698,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":546,"completion_tokens_details":{"reasoning_tokens":628}},"tokens_in":546,"tokens_out":698,"duration_ms":7715,"temperature":1.0,"reasoning_tokens":628,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:31:24.494368+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Do the explicit Fourier calculation that Lemma B.4 is designed to control: take $b=(0,0,\\sin z)$ and $\\phi=(e^{{\\rm i}N x_1}\\cos(2\\pi z),0,0)$ on the three-torus, with the symmetries of Section 2, and compare both sides of the second bound in Lemma B.4 as $N\\to\\infty$; if the left side grows faster than $O(N^{s+1/2})$, the estimate and the main theorem collapse.","supporting_citations":[{"cited_title":"W ell-posedness and ill-posedness of the primitive equations with fraction al horizontal dissipation","cited_arxiv_id":null,"evidence_quote":"Frames the fractional-dissipation threshold via the deterministic 2D primitive equations and supplies the ill-posedness results for supercritical and critical large data."},{"cited_title":"The stochastic primitive equations with non-isothermal turbulent pressure","cited_arxiv_id":"2210.05973","evidence_quote":"Supplies the Stratonovich-to-Itô conversion and the turbulent-pressure transport-noise formulation used to write the equation."},{"cited_title":"Generalized surface quasi- geostrophic equations with singular velocities","cited_arxiv_id":null,"evidence_quote":"Provides the pseudodifferential commutator estimate (B.1) on which Lemma B.2 and the double-commutator bounds rest."},{"cited_title":"Lon g time dynamics of forced critical SQG","cited_arxiv_id":null,"evidence_quote":"Supplies the fractional Leibniz/product estimates of Lemma A.1 used throughout the nonlinear and noise estimates."},{"cited_title":"Local martingale solutions a nd pathwise uniqueness for the three-dimensional stochast ic inviscid primitive equations","cited_arxiv_id":null,"evidence_quote":"Provides the analytic-class existence framework for stochastic inviscid primitive equations that the supercritical discussion compares against and whose uniqueness argument is improved here."},{"cited_title":"Compact sets in the space Lp(0, T ; B)","cited_arxiv_id":null,"evidence_quote":"Gives the Aubin–Lions–Simon compactness lemma used to prove tightness and the martingale-solution step."}],"review_version":1}