{"id":"9f06610e-14a1-4d12-9f55-8423065ee135","arxiv_id":"2412.05473","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":10,"one_line_summary":"A six-night, high-cadence SPIRou dataset yields the first Maunakea NIR Moon-continuum measurement, 126 resolved OH doublets, and about 30-minute OH variability timescales that set practical sky-sampling requirements.","lead":"This paper uses 1075 high-resolution infrared sky spectra from SPIRou on Maunakea to measure how hydroxyl airglow lines and moonlight vary over minutes to years. It provides concrete sky-subtraction guidelines for infrared surveys, including when the Moon must be included in sky models.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 10-degree Moon-separation guideline is extrapolated from fits that reach only ~6 degrees, with no reported uncertainties and no lunar-phase or aerosol corrections; this is the least secure link in the central claim.","rationale":"The strongest claim combines two recommendations: the 10° Moon-separation rule and the ~10-minute ABBA cadence for 1% sky subtraction. The GPR-based cadence is derived from a well-defined time-series analysis and the reported histograms include dispersion, so it is reasonably supported. The Moon rule, however, is based on a serendipitous passage on three nights, with data only to 6° and no uncertainties or atmospheric covariates. The paper itself notes the aerosol caveat (§4.2). Since this rule is in the abstract and would directly affect survey design, it is the most load-bearing assumption. Independent validation or refit with proper uncertainties is needed. The reader's weakest assumption aligns with this; I agree. No reason to change the conditional verdict.","tokens_in":114,"tokens_out":4191,"duration_ms":52580,"concrete_test":"Recompute the Moon background fits separately for each of the three nights and also with lunar illumination as a covariate; propagate the fit covariance to the separation at which the Moon continuum equals 1% of the band-integrated sky background. If the resulting threshold is outside, say, 6–14°, the 10° guideline needs a caveat. Cross-check with an independent model (e.g., ESO SkyCalc or Jones et al. 2013) at 6–10° separation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline survey guideline—that the Moon contribution must be accounted for at separations ≥10° for PCA sky reconstruction, targets fainter than ~15 mag, or sky subtraction better than 1%—depends on an exponential fit to the Moon continuum in Fig. 6/Table 1. The text states the data only constrain the continuum \"to about 6° in all bands\" (§4.2), yet the 10° threshold is presented without any uncertainty or extrapolation warning. The fit itself pools three nights with lunar illumination varying from 95% to 80% and no phase or aerosol correction, and the flux calibration relies on telluric standards. If the true scattering profile is flatter, or if phase change or molecular scattering introduces a systematic gradient, the 10° number could shift enough to change which observations need Moon accounting. This is not a fatal flaw in the raw spectra, but it is the load-bearing part of the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes 1075 high-resolution NIR sky spectra from SPIRou on Maunakea, including two three-night runs with 5.5-minute cadence, to characterize OH line variability and the Moon's continuum contribution. The authors fit OH lines with one or two Gaussian profiles on a high-pass-filtered continuum, identify 126 resolved doublets, and study temporal variability with Gaussian process regression (GPR) and Lomb-Scargle periodograms. Two survey guidelines are presented: (1) the Moon contribution must be accounted for at separations within 10 degrees when using PCA sky reconstruction, observing targets fainter than about 15 mag, or aiming for sky subtraction better than 1%; and (2) OH lines have a GPR correlation length of about 30 minutes, implying that an ABBA sky-subtraction pattern needs roughly 10 minutes between sky observations for a 1% residual. The data are released on Zenodo (doi:10.5281/zenodo.13363061) and the analysis code is public on GitHub.","tokens_in":18956,"tokens_out":7754,"duration_ms":75156,"significance":"If the guidelines are robust, this would be a direct and useful input for NIR survey scheduling and sky-subtraction strategies, complementing the ESO SkyCalc modeling efforts. The dataset itself is a strong asset: six nights of high-cadence, high-resolution sky spectra with serendipitous Moon passages are rare, and the public data/code release is exemplary. The doublet catalog and the explicit GPR correlation timescales are concrete products that can inform future observing programs. However, the two headline guidelines rest on an extrapolated, uncertainty-free Moon fit and on an incompletely described ABBA residual calculation; as it stands, the paper is more secure as a characterization dataset than as a quantitative survey rule.","major_comments":[{"comment":"The 10-degree Moon-separation guideline is extrapolated from data that the text itself says constrain the continuum only 'to about 6 degrees in all bands' (Section 4.2). Table 1 reports no uncertainties on the exponential fit slopes or intercepts, and the Figure 6 caption states that the three nights were fit without correcting for lunar phase change (illumination varying from 95% to 80%). Because this number appears in the abstract as a survey rule, the authors should either restrict the guideline to the measured range with an explicit extrapolation warning, or quantify the extrapolation, for example by fitting a physical scattering model (e.g., Jones et al. 2013) or by reporting Monte Carlo uncertainties that include phase and aerosol systematic terms.","section":"Section 4.2, Table 1, Figure 6"},{"comment":"The criterion that the Moon must be accounted for when 'observing targets at YJHK mags fainter than ~15' is not derived anywhere in the paper. No calculation is shown that connects the measured Moon surface brightness (mag arcsec^-2) to a target magnitude threshold for a given sky-subtraction accuracy. Please provide the underlying contrast/signal-to-noise estimate, or explicitly label the 15-mag threshold as an illustrative rule of thumb rather than a quantitative result of this analysis.","section":"Abstract and Section 4.2"},{"comment":"The method for computing the ABBA sky-subtraction timescales is not described. The text states that 'From the GPR fit of each line, we determined the timescale corresponding to a 1%, 2% and 5% sky-subtraction error in an ABBA scenario' but gives no equation or algorithm for converting the GP covariance into an ABBA residual as a function of delay (for example, the residual after subtracting a linear trend over a sliding window). Without this, the headline result of about 10 minutes for a 1% ABBA residual is not reproducible; please add the explicit formula or a clear pseudo-code description.","section":"Sections 3.3 and 4.5, Figure 9"},{"comment":"The GPR prior on the correlation length is a log-uniform distribution bounded between 10 minutes and 1 day. The derived ABBA 1% timescale is 10.4 +/- 1.6 min, close to the lower prior boundary, and the simple-subtraction 1% timescale is about 2 minutes, which is below the prior range. This raises the possibility that the prior's lower bound affects the short-lag behavior of the squared-exponential kernel and biases the quoted timescales. Please report a sensitivity test with a lower bound below the sampling cadence (e.g., 2 minutes) or otherwise demonstrate that the 10-minute prior does not drive the ABBA result.","section":"Section 3.3 and Section 4.5, Figure 9"}],"minor_comments":[{"comment":"The sentence 'subtracted a linear fit of the persistence contribution for angles ranging from 50 to 20 degrees' is unclear: it should specify what is being fit (e.g., background versus time or versus Moon separation) and the intended direction of the angular range (50 to 20 degrees is descending).","section":"Section 4.2"},{"comment":"Table 1 should report uncertainties on the intercept and slope values and should state the exact functional form used (the text says 'exponential decay' but the table columns are labeled 'Slope (degree^-1)', which suggests a different parametrization).","section":"Table 1 and Figure 6"},{"comment":"The phrasing 'at separation distances less than 10 degrees' (abstract) and 'at Moon separation distances of at least 10 degrees' (Section 4.2) is contradictory; use a consistent phrase such as 'within 10 degrees of the Moon' throughout.","section":"Abstract and Section 4.2"},{"comment":"The doublet identification thresholds (separation within +/-0.05 Angstrom and flux ratio between 0.78 and 1.3) are stated without an estimate of the false-positive rate; a comparison with HITRAN line separations or a bootstrap-style test would strengthen confidence in the 126-doublet catalog.","section":"Section 4.4 and Table A.1"},{"comment":"There are several typographical issues: 'Gaussian Proccess' in the Section 3.3 heading, 'a a unique set' in the Conclusions, and 'minuutes' in the Figure 12 caption; the reference list also contains duplicated entries for Noll et al. (2023) and Oliva et al. (2015a).","section":"Various"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the journal's astro-im instrumentation and techniques scope. The Zenodo data release and public code are clear strengths. My main concern is that the headline Moon guideline is presented more confidently than the data support: the 10-degree number is an extrapolation beyond the fitted range with no reported uncertainty, and the magnitude threshold is not derived. The ABBA timescale result also needs a full methodological description. These issues are fixable within the manuscript's scope, so major revision is appropriate rather than rejection. I saw no concerns about citation ethics or novelty disclosure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nShort take: this is a solid empirical paper with a publicly released dataset, and the OH variability numbers are probably the most durable part. The 30-minute GPR correlation length and the ~10-minute ABBA cadence for 1% sky residuals are concrete and useful for survey scheduling. The 10-degree Moon guideline, by contrast, is the weakest link—the underlying exponential fit only reaches about 6 degrees of separation, has no reported uncertainties in Table 1, and pools three nights with Moon illumination changing from 95% to 80% without a phase or aerosol correction. The paper itself admits the data only track to ~6 degrees, and that aerosol can change the numbers. So the guideline is a reasonable extrapolation, not a measurement, and should be flagged as such when cited.\n\nWhat's new: first Maunakea NIR spectroscopic measurement of the Moon's continuum contribution (serendipitous), 126 resolved OH doublets at R~75,000, and a high-cadence six-night dataset with 1075 SPIRou spectra. The data and code are on Zenodo/GitHub, which is the kind of thing that makes this worth referee time. The methods are standard—GPR, LSP, Gaussian fitting—but applied to a unique dataset.\n\nSoft spots, in order: (1) the Moon fit as above; (2) the text says \"remove fits with η <15% to ensure only good measurements were kept\" but η is relative error and smaller is better, so that cut reads inverted—likely a typo, but it makes the LSP results hard to reproduce; (3) the two-Gaussian flux model after high-pass filtering is simplistic, acknowledged by the authors, and could bias doublet detection counts; (4) the LSP analysis is limited by the 5.5-minute sampling, also acknowledged. None of these sink the central OH variability claim, which is based on a large number of lines and a GPR with explicit white-noise separation.\n\nWho should read it: people planning NIR multi-object or fiber surveys on Maunakea, anyone building an ESO-SkyCalc-like site model, and instrumentalists who care about ABBA/nodding cadence. The survey guidelines are actionable if you treat the Moon number as provisional.\n\nVerdict: send to peer review. The data release and the variability timescales are worth publishing, and the Moon measurement is a start even if the 10-degree threshold needs to become a range with error bars. A careful referee should ask for the Table 1 uncertainties, a clarification of the η cut, and a check of the doublet selection thresholds against HITRAN. This is not a reject; it's a revise-and-resubmit.","headline":"The OH variability timescales and the public dataset are the real contributions; the 10-degree Moon rule is an uncalibrated extrapolation and should be treated as provisional.","tokens_in":19564,"tokens_out":4073,"would_cite":false,"duration_ms":37663,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper establishes that the Moon's continuum contribution must be included in near-infrared sky subtraction at Moon separations of at least 10 degrees for faint targets and sub-1% residuals, and that OH airglow lines vary on roughly…","keywords":["night sky brightness","OH airglow","near-infrared spectroscopy","sky subtraction","Gaussian process regression","Lomb-Scargle periodogram","Maunakea observing site","SPIRou spectrograph"],"falsifier":"Measure the NIR continuum background as a function of Moon separation on a photometric night with known aerosol optical depth, extending to separations of 10-15 degrees; if the Moon's contribution at 10 degrees is below 1% of the band-integrated sky background, or if the exponential decay slope differs materially from the values in Table 1, the recommended 10-degree threshold would need revision.","tokens_in":18492,"feed_emoji":"🌙","tokens_out":6473,"duration_ms":59852,"temperature":0.7,"pith_summary":"This paper aims to give survey planners concrete numbers for near-infrared (NIR) sky subtraction. Analyzing 1075 high-resolution spectra of the Maunakea night sky, it finds that the Moon's continuum, usually ignored at NIR wavelengths, must be accounted for at Moon separation distances of at least 10 degrees when reconstructing the sky with principal component analysis, when targets are fainter than about 15 mag, or when a sky subtraction better than 1% is attempted. The same data show that 481 OH airglow lines vary on a common timescale of about 30 minutes, so a 1% residual in an ABBA sky-science-sky pattern requires the two sky frames to be within about 10 minutes. The paper also provides a catalog of 126 resolved OH doublets and makes the spectra and fitting code public.","feed_headline":"Moon matters for faint NIR targets out to 10 degrees","feed_subtitle":"OH airglow lines shift on ~30-minute timescales, telling surveys when to take sky frames.","key_machinery":"The argument is carried by a two-Gaussian flux model for each of 481 hydroxyl lines, fitted to continuum-filtered spectra, and by two time-series tools: Gaussian process regression with a squared-exponential plus white-noise kernel, which yields the correlation length, and Lomb-Scargle periodograms, which reveal dominant variability periods. The Moon's contribution is separated from the band background by subtracting a linear persistence-decay fit and then fitting the residual flux as an exponential function of angular separation to the Moon. The correlation length is the central quantity because it directly converts the time delay between sky and science frames into a fractional sky-subtraction error.","core_discovery":"The central discovery is that the NIR night sky on Maunakea has two previously underappreciated variability drivers with directly actionable timescales. The Moon's scattered continuum, measured serendipitously as it passed within about a degree of the spectrograph fiber, decays exponentially with angular separation; the fit indicates the Moon dominates the band-integrated background within roughly 4 degrees in J and 8 degrees in Y, and the paper recommends treating separation distances under 10 degrees as contaminated for demanding observations. Independently, Gaussian process regression on the line fluxes gives a correlation length of about 30 minutes, with a median of 39 minutes for high signal-to-noise lines, implying that simple on/off sky subtraction needs a 3-minute cadence for 1% accuracy, while ABBA subtraction extends that to about 10 minutes. These two results are framed as the first step toward a data-driven sky model for Maunakea.","pith_inferences":["If aerosol optical depth varies significantly, the fixed 10-degree threshold may shift; combining these spectra with simultaneous aerosol measurements could replace the threshold with a separation-versus-contamination curve.","The homogeneous ~30-minute OH timescale suggests that low-dimensional models (PCA or a trained autoencoder) could predict the full 481-line sky from a few monitored lines, potentially relaxing the required sky cadence.","The Moon's relative contribution is band-dependent, so a target's science band should set the Moon-separation constraint rather than a single value for all NIR observations.","Extending this analysis to a second site or to visible airglow would test whether the correlation length and the Moon scaling are transferable or site-specific."],"forward_implications":["Survey schedulers should treat Moon separation under 10 degrees as a constraint for faint NIR targets, especially with PCA-based sky reconstruction or when trying to reach sub-1% residuals.","ABBA-style observations should keep the two sky frames within about 10 minutes to hold 1% sky-subtraction error, while simple on/off subtraction would require about 3 minutes of sampling.","The 126 resolved OH doublets form a wavelength-dependent catalog that can be used to predict sky-line contamination at high spectral resolution and below.","The public spectra and fitting tools are a reusable input for building a Maunakea-specific sky model."],"supporting_citations":[{"why":"Supplies the 481 OH line wavelengths and identifications used as the fitting list.","marker":"Rousselot et al. (2000)"},{"why":"Provides the Gaussian process regression framework used to derive correlation lengths.","marker":"Rasmussen et al. (2006)"},{"why":"Earlier study of scattered moonlight with the VLT that this work complements and extends to NIR wavelengths.","marker":"Jones et al. (2019)"},{"why":"Model of lunar phase dependence of albedo that contextualizes the Moon's NIR continuum.","marker":"Kieffer & Stone (2005)"},{"why":"Establishes the 0.5-1% sky subtraction accuracy goal for PCA-based methods that motivates the 1% criterion.","marker":"Sharp & Parkinson (2010)"},{"why":"Provides OH doublet intensity-ratio behavior that informs the doublet classification and variability discussion.","marker":"Noll et al. (2023)"}],"fun_headline_variants":["Moon's glow contaminates NIR sky out to 10 degrees","OH airglow varies on 30-minute timescales, setting sky cadence","NIR sky subtraction: Moon out to 10°, OH lines ~30 min","Maunakea NIR: Moon and OH variability drive sky subtraction","Sky cadence for NIR surveys: Moon out to 10°, OH ~30 min"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 10-degree Moon guideline rests on an exponential fit to only three nights of serendipitous Moon passages, observed out to about 6 degrees, with no correction for lunar phase changes or aerosol scattering, so the threshold is an extrapolation beyond the measured range.","fun_headline_variants_meta":{"raw":{"variants":["Moon's glow contaminates NIR sky out to 10 degrees","OH airglow varies on 30-minute timescales, setting sky cadence","NIR sky subtraction: Moon out to 10°, OH lines ~30 min","Maunakea NIR: Moon and OH variability drive sky subtraction","Sky cadence for NIR surveys: Moon out to 10°, OH ~30 min"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000721,"raw_usage":{"total_tokens":3295,"prompt_tokens":1066,"completion_tokens":2229,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":682,"completion_tokens_details":{"reasoning_tokens":2125}},"tokens_in":682,"tokens_out":2229,"duration_ms":14033,"temperature":1.0,"reasoning_tokens":2125,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T20:40:57.306834+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the NIR continuum background as a function of Moon separation on a photometric night with known aerosol optical depth, extending to separations of 10-15 degrees; if the Moon's contribution at 10 degrees is below 1% of the band-integrated sky background, or if the exponential decay slope differs materially from the values in Table 1, the recommended 10-degree threshold would need revision.","supporting_citations":[{"cited_title":"E., Williams, C","cited_arxiv_id":null,"evidence_quote":"Provides the Gaussian process regression framework used to derive correlation lengths."},{"cited_title":"2019, A&A, 624, A39, doi: 10.1051/0004-6361/201833759","cited_arxiv_id":null,"evidence_quote":"Earlier study of scattered moonlight with the VLT that this work complements and extends to NIR wavelengths."},{"cited_title":"H., & Stone, T","cited_arxiv_id":null,"evidence_quote":"Model of lunar phase dependence of albedo that contextualizes the Moon's NIR continuum."}],"review_version":1}