{"id":"fd407f82-97bf-4a07-b5c3-81904fe4d10c","arxiv_id":"2412.14666","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A fast semi-analytical covariance for the galaxy-PSF tau-statistics is derived and validated against simulations and jackknife on UNIONS data, including a redefined tau-statistic that breaks the alpha-eta degeneracy.","lead":"Weak lensing surveys correct galaxy shapes for the telescope's point spread function, and this paper introduces a fast way to compute the uncertainty on the standard PSF systematics diagnostics. The method, tested on UNIONS survey data, gives results consistent with slower simulation-based and jackknife approaches, which will speed up systematics checks for current and future cosmic shear surveys.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (21) mislabels the ξγγ ξbc cosmic-variance term, making the published semi-analytical covariance derivation internally inconsistent.","rationale":"The reader's weakest assumption (Gaussianity of the fields) is a legitimate approximation, but the paper tests it in Appendix B and finds no significant non-Gaussianity; moreover, the PSF fields are modelled as deterministic in the covariance derivation, so the test with fixed stars is less decisive than it appears. The more load-bearing issue is the algebraic consistency of the central covariance formula. The index mismatch between Eqs. (20) and (21) means the published derivation is internally inconsistent, and the missing code prevents checking which expression was used. If the published formula is the one implemented, the covariance is biased at the separation argument level, undermining the claim of comparability with jackknife and simulation covariances. I recommend CONDITIONAL: the authors must fix the derivation or provide code to verify the implementation, and rerun the validation.","tokens_in":24025,"tokens_out":19112,"duration_ms":142311,"concrete_test":"Re-derive Eq. (21) from Eq. (20) by explicit index summation, or obtain the code and inspect whether the covariance uses ξγγ(ϑik)ξbc(ϑjl) or ξγγ(ϑil)ξbc(ϑjk). Then recompute the τ-statistics covariance with the corrected expression and compare the resulting α, β, η contours and ξ_PSF,sys to the jackknife and GLASS estimates. If the corrected version removes the χ2_red discrepancy and the β shift, the published Eq. (21) is the source of the error.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central new result is the semi-analytical covariance of the τ-statistics, Eq. (21). Comparing Eq. (21) to its own derivation, Eq. (20), reveals an index inconsistency. Eq. (20) gives the Wick-expanded cosmic-variance contribution as ξγγ(ϑik)ξbc(ϑjl) + ξγc(ϑil)ξγb(ϑjk) (plus the corresponding '−' terms with cos(4(φik−φjl)) and cos(4(φil−φjk))). Eq. (21) instead writes all four terms with the same arguments (ϑil, ϑjk), e.g. ξγγ(ϑil)ξbc(ϑjk) and cos(4(φil−φjk)). This cannot be obtained by exchanging k and l: in the weighted estimator (4), k labels a galaxy with weight w^e_k and l labels a star with weight w^c_l, so such an exchange changes the weight structure. The ensemble-average derivation in Sect. 2.5.2 (Eqs. 27–31) also treats only terms of the form F1(θ23)F2(θ14), which correspond to the ξγc ξγb term; the ξγγ ξbc term has arguments θ13 and θ24 and is not covered by that substitution. If the published formula is implemented as written, the shear-shear×PSF-PSF contribution to the covariance is evaluated at the wrong angular separations (galaxy-star instead of galaxy-galaxy and star-star), biasing the covariance. The elevated reduced chi-square (1.63 vs 0.86 for jackknife) and the systematic shifts in β may be consequences. The absence of the code link prevents checking which expression was actually used, so the central validation cannot be reproduced as published.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a semi-analytical estimator for the covariance matrix of the τ-statistics, i.e., cross-correlations of galaxy ellipticities with PSF ellipticity, ellipticity-residual, and size-residual fields, and applies it to the UNIONS v1.3 shear catalogue. The covariance is constructed from measured ρ-, τ-, and shear correlation functions plus a shot-noise term, using a Gaussian-field assumption to reduce four-point correlators via Wick's theorem. The authors compare constraints on the PSF error parameters (α, β, η) obtained with this covariance against jackknife and GLASS-simulation covariances, report broad agreement, and propose a redefinition of τ5 that breaks the α–η degeneracy. The paper claims a factor of about 8 speed-up relative to jackknife and positions the method as a fast PSF diagnostic for current and Stage IV weak-lensing surveys.","tokens_in":24436,"tokens_out":14653,"duration_ms":95556,"significance":"The methodological goal is timely and the application is nontrivial: the paper contains a detailed derivation, a real-data demonstration, a comparison against two independent covariance estimators, and a practical discussion of degeneracies. If the covariance formula is correct, the semi-analytical approach is genuinely useful for fast PSF-systematics checks in current and future surveys, and the proposed τ~-statistics are a reasonable way to separate leakage from size-residual contamination. However, the central claim of 'comparable results' is currently weakened by an apparent index inconsistency in the main formula and by the incomplete Gaussianity validation.","major_comments":[{"comment":"The published cosmic-variance term in Eq. (21) is inconsistent with its own derivation in Eq. (20). Eq. (20) gives the first Wick term as ξγγ(ϑik)ξbc(ϑjl) along with cos(4(φik−φjl)), while Eq. (21) writes this term, and the second term, with the arguments (ϑil, ϑjk) and cos(4(φil−φjk)). The sentence after Eq. (21) that the result follows from invariance under exchange of the indices k and l is not valid, because k labels a galaxy with weight w^e_k and l labels the PSF-related field c with weight w^c_l; exchanging them changes the weight structure and the field identities. Correspondingly, the ensemble-average treatment in Sect. 2.5.2 (Eq. 27) uses F1(θ23)F2(θ14), which covers only the ξγc ξγb term; the ξγγ ξbc term depends on θ13 and θ24 and is not evaluated there. If the code implements Eq. (21) as written, the covariance is biased at the level of the main new result, and this may explain the elevated reduced chi-square (1.63 vs 0.86) and the visible shift in β. The authors should correct Eq. (21), update the ensemble-average derivation accordingly, re-run the full comparison, and provide the actual code link (the text currently gives only '/gtb').","section":"Sect. 2.5.1, Eqs. (20)–(21)"},{"comment":"The Gaussianity test used to justify Wick's theorem in Eq. (19) is based on 300 GLASS realizations in which only the galaxy catalog is simulated; the star and PSF catalogs are the observed data, repeated identically in every realization. The test therefore cannot detect non-Gaussianity in the star and PSF fields, even though those fields enter the covariance through the ξbc and ξγc factors. The Gaussian assumption for the PSF-related fields thus remains unvalidated. Please either extend the test to simulated star/PSF fields, or provide a specific physical and statistical justification for treating those sparse, selection-affected fields as Gaussian.","section":"App. B"},{"comment":"The semi-analytical covariance yields χ²_red = 1.63 for 57 degrees of freedom, compared with 0.86 for jackknife and 1.18 for simulations. This is a statistically poor fit, and the text's statement that the fit is 'similar' is not supported by any quantitative comparison. The discrepancy is a red flag that the published semi-analytical covariance does not fully describe the noise. The authors should either demonstrate that the corrected Eq. (21) removes the excess chi-square, or explicitly qualify the claim of comparable results by reporting this goodness-of-fit difference and its consequences for parameter inference.","section":"Sect. 4.1, Table 1"}],"minor_comments":[{"comment":"The text refers to 'Table 3' when summarizing parameter constraints; the table summarizing these results is Table 1.","section":"Sect. 4.1"},{"comment":"The repository link is given as '/gtb' without a resolvable URL; a complete link to the public repository must be provided for reproducibility.","section":"Sect. 6"},{"comment":"The sentence 'The first two terms depend on the scalars θ23 and θ14' is unclear, given that the ξγγ ξbc term depends on θ13 and θ24; please clarify the indexing used in the ensemble-average derivation.","section":"Sect. 2.5.2"},{"comment":"The left panel is described as 'analytical expressions from Eq. (18)', but Eq. (18) is the general covariance definition; the actual semi-analytical expression is Eq. (21).","section":"Fig. 1 caption"},{"comment":"The section heading contains a typo: 'τ-statisics' should be 'τ-statistics'.","section":"Sect. 2.4"}],"recommendation":"major_revision","confidential_remarks":"The index inconsistency in Eq. (21) is load-bearing and, combined with the missing code link, makes it impossible to determine from the manuscript which expression was actually implemented. This is fixable within the manuscript's scope, so I recommend major revision rather than rejection. The paper otherwise fits the journal's scope and, once the covariance formula is corrected and revalidated, could be a useful contribution. I also note that the Gaussianity test in Appendix B does not cover the star/PSF fields, so the validation of Wick's theorem remains incomplete."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a useful paper with a genuine new method, but the central derivation has an index error that needs fixing before the formula can be trusted as published.\n\nWhat's new: the semi-analytical covariance for the tau-statistics. That's a real extension of Schneider et al. (2002), and the speedup (about 8x vs jackknife) makes it attractive for quick PSF diagnostics. The validation on UNIONS data is mostly convincing: parameter contours from the semi-analytical covariance, jackknife, and GLASS simulations agree, and the final systematic level is consistent. The redefined tau-statistics in Section 5.2 break the alpha-eta degeneracy and are a nice practical contribution. The paper is clearly written and the authors are honest about limitations, including the missing shot-noise treatment and the fourth-moment extension.\n\nSoft spots. The main one is the index inconsistency between Eq (20) and Eq (21). In Eq (20) the cosmic-variance term has two distinct structures: ξγγ(ik)ξbc(jl) and ξγc(il)ξγb(jk). Eq (21) writes all four terms with arguments (il) and (jk), claiming invariance under exchange of k and l. But k runs over galaxies and l runs over stars; those are different object sets, so swapping them is not a symmetry of the sum. The first term should stay at galaxy-galaxy and star-star separations. The same issue appears in the ensemble-average derivation in Section 2.5.2, which writes all terms as F1(θ23)F2(θ14). This is a load-bearing error in the derivation, even if the numerical code happens to be correct. The code link is a placeholder (\"GitHub/gtb\"), so there is no way to check which formula was actually used. The elevated reduced chi-square (1.63 vs 0.86 for jackknife) and the visible beta shift could be consequences.\n\nTwo other points, both minor. The Gaussianity test in Appendix B uses GLASS simulations that only simulate galaxies, so it does not test non-Gaussianity in the star/PSF fields that enter the covariance. And the covariance is built from the same measured rho- and tau-statistics that are then fit; the paper notes this, and the validation against jackknife mitigates the concern, but it is worth stating clearly.\n\nBottom line: the method is promising and the paper deserves a serious referee, but it should not be accepted as-is. The authors need to correct the index structure in Eq (21) and the corresponding derivation, provide the code, and re-validate. The redefined tau-statistics and the general approach are still worth publishing after that.","headline":"Useful and fast PSF systematics diagnostic, but the central covariance derivation has an index error that must be fixed before the formula is usable.","tokens_in":25005,"tokens_out":11806,"would_cite":false,"duration_ms":76358,"reading_group":"yes","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives a semi-analytical covariance matrix for the galaxy--PSF cross-correlations ($\\tau$-statistics) that reproduces the parameter uncertainties from jackknife and simulation-based estimates at about one-eighth the…","keywords":["weak gravitational lensing","point spread function","PSF systematics","tau-statistics","rho-statistics","semi-analytical covariance","cosmic shear","UNIONS"],"falsifier":"Run the semi-analytical covariance on a suite of simulated galaxy catalogs with strongly non-Gaussian fields and with known injected PSF leakage parameters, and show that the least-squares posteriors cover the injected values; failure would demonstrate bias from the Gaussian truncation.","tokens_in":23860,"feed_emoji":"🔭","tokens_out":10964,"duration_ms":82901,"temperature":0.7,"pith_summary":"This paper tries to establish that the covariance of the $\\tau$-statistics---the cross-correlations between galaxy shapes and PSF model fields that quantify weak-lensing PSF contamination---can be computed semi-analytically rather than from simulations or jackknife resampling. The authors derive the covariance from a Gaussian four-point expansion, using the measured $\\rho$- and $\\tau$-statistics as inputs, and show on UNIONS data that the resulting parameter constraints on the PSF error amplitudes $\\alpha$, $\\beta$, $\\eta$ and the total systematic level $\\xi_{\\mathrm{PSF,sys}}$ agree with the two standard approaches. If the method holds, PSF systematics checks become cheap enough to run on many catalogs, patches, or tomographic bins. The paper also introduces a redefinition of the $\\tau$-statistics that breaks a degeneracy between the leakage parameter and the size-residual parameter.","feed_headline":"New formula makes telescope-blur systematics checks 8x faster","feed_subtitle":"Semi-analytical covariance matches jackknife and simulation estimates on UNIONS data, cutting runtime from hours to minutes.","key_machinery":"The $\\tau$-statistics $\\tau_0=\\langle e\\,e_p\\rangle$, $\\tau_2=\\langle e\\,\\delta e_p\\rangle$, $\\tau_5=\\langle e\\,\\delta T_p\\rangle$ are linear functions of the $\\rho$-statistics through parameters $\\Omega=(\\alpha,\\beta,\\eta)$, and the paper's new machinery is an analytic covariance for those $\\tau$-values. It reduces the four-point correlation of the estimators to a mixed term (galaxy shape noise times a $\\rho$-statistics) and cosmic-variance products of shear and PSF correlation functions, with the sums over galaxy positions replaced by ensemble averages over survey area and number density. The connected four-point term is dropped via Wick's theorem, and the shot-noise diagonal is taken from a tree-code estimator of the observed catalog; the resulting matrix is then used in a least-squares solution or a Gaussian likelihood. The redefinition $\\tilde{\\tau}_5$ replaces the spin-2 star ellipticity factor in the size residual by a tangential unit vector, turning the size-residual contribution into a scalar correlation that no longer mixes with PSF leakage.","core_discovery":"The central claim is that the covariance matrix of the $\\tau$-statistics can be assembled from analytic integrals over the measured two-point functions: a mixed shot-noise term involving the galaxy shape noise $\\sigma_e^2$ and the relevant $\\rho$-statistics, plus cosmic-variance terms built from products of the shear correlation function and the PSF-field correlations, with the connected four-point term set to zero under the Gaussian assumption. Applied to the UNIONS v1.3 shape catalogue with a data-driven multi-CCD PSF model, this semi-analytical covariance yields contours for $(\\alpha,\\beta,\\eta)$ and a systematic level $\\xi_{\\mathrm{PSF,sys}}$ that agree at the 68% confidence level with jackknife and simulation-based covariances. The computation takes 27 minutes on 48 cores versus 227 minutes for jackknife, and unlike the simulations it retains galaxy--star cross terms. A reformulated $\\tilde{\\tau}$-statistics, treating the PSF size residual as a scalar field through a tangential unit-vector factor, removes the $\\alpha$--$\\eta$ degeneracy while reproducing the same systematic level.","pith_inferences":["A natural next test is to validate the semi-analytical covariance on fields with strongly non-Gaussian small-scale clustering or with injected star-position correlations, since the Gaussian truncation is only tested on simulated galaxy fields in the paper.","The $\\tilde{\\tau}$ redefinition could be paired with the same semi-analytical covariance derivation, an extension the paper leaves to future work that would give an end-to-end fast diagnostic with both degeneracy removal and low runtime.","If the runtime advantage holds at future survey sizes, PSF systematics could be tracked as a function of survey depth, observing conditions, or mask geometry in near real time, informing calibration before the final cosmology run.","The covariance formalism should transfer to fourth-order PSF-moment systematics with an analogous structure, as the paper itself notes."],"forward_implications":["PSF systematic parameters and their error bars can be obtained from a single galaxy-plus-star catalog, with no simulation campaign or patch resampling needed.","The semi-analytical covariance includes galaxy--PSF cross terms that the simulation-based covariance omits, so it is a more complete comparison target.","Runtime drops from about 227 to 27 minutes for the UNIONS setup, and using the least-squares solver instead of MCMC compounds the gain to a factor of about 8.2.","The redefined $\\tilde{\\tau}$-statistics separate the leakage contribution from the size-residual contribution, giving a cleaner null test of PSF size modelling.","Large tomographic and future survey analyses can run repeated PSF diagnostics across redshift bins or catalog cuts without rebuilding covariances each time."],"supporting_citations":[{"why":"Provides the two-point correlation estimators, the pairwise weighting, and the ensemble-average technique that the semi-analytical covariance derivation adapts.","marker":"Schneider et al. (2002)"},{"why":"Introduces the $\\tau$-statistics and the jackknife-based parameter estimation that this paper's semi-analytical covariance is compared against.","marker":"Gatti et al. (2021)"},{"why":"Defines the PSF size-residual term and the additive systematics model used in Eq. (8).","marker":"Paulin-Henriksson et al. (2008)"},{"why":"Defines the $\\rho$-statistics used to build the PSF systematic level.","marker":"Jarvis et al. (2016)"},{"why":"Provides the simulation-based covariance approach for $\\tau$-statistics that serves as one of the two comparison methods.","marker":"Zhang et al. (2023)"},{"why":"Supplies the lognormal simulations used to produce the comparison covariance from 300 realisations.","marker":"Tessore et al. (2023)"},{"why":"Implements the tree-code correlation-function estimates used to supply the measured statistics and the shot-noise diagonal to the semi-analytical covariance.","marker":"Jarvis et al. (2004)"}],"fun_headline_variants":["Analytic covariance makes PSF systematics checks 8x faster","Semi-analytic covariance accelerates PSF systematics checks","New covariance formula for PSF systematics: 8x faster","Telescope blur checks get analytic boost on UNIONS data","Semi-analytic covariance speeds PSF systematics checks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation assumes the galaxy and PSF shape fields are close enough to Gaussian that a four-point correlation can be split into products of two-point correlations; the paper's Gaussianity test uses simulated galaxies only, not simulated stars.","fun_headline_variants_meta":{"raw":{"variants":["Analytic covariance makes PSF systematics checks 8x faster","Semi-analytic covariance accelerates PSF systematics checks","New covariance formula for PSF systematics: 8x faster","Telescope blur checks get analytic boost on UNIONS data","Semi-analytic covariance speeds PSF systematics checks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000638,"raw_usage":{"total_tokens":3023,"prompt_tokens":1116,"completion_tokens":1907,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":732,"completion_tokens_details":{"reasoning_tokens":1821}},"tokens_in":732,"tokens_out":1907,"duration_ms":11473,"temperature":1.0,"reasoning_tokens":1821,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:01:25.647605+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the semi-analytical covariance on a suite of simulated galaxy catalogs with strongly non-Gaussian fields and with known injected PSF leakage parameters, and show that the least-squares posteriors cover the injected values; failure would demonstrate bias from the Gaussian truncation.","supporting_citations":[{"cited_title":"2021, Monthly Notices of the Royal Astronomical Society, 504, 4312","cited_arxiv_id":null,"evidence_quote":"Introduces the $\\tau$-statistics and the jackknife-based parameter estimation that this paper's semi-analytical covariance is compared against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the PSF size-residual term and the additive systematics model used in Eq. (8)."},{"cited_title":"2023, Monthly Notices of the Royal Astronomical Society, 525, 2441","cited_arxiv_id":null,"evidence_quote":"Provides the simulation-based covariance approach for $\\tau$-statistics that serves as one of the two comparison methods."},{"cited_title":"2004, , 352, 338","cited_arxiv_id":null,"evidence_quote":"Implements the tree-code correlation-function estimates used to supply the measured statistics and the shot-noise diagonal to the semi-analytical covariance."}],"review_version":1}