{"id":"9dcb48fc-8974-48dc-b0e2-00b47abbe269","arxiv_id":"2509.10258","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A telluric retrieval algorithm recovers long-term CO2 and CH4 trends from standard star spectra, but with large scatter and an unexplained vertical offset.","lead":"Astroclimes is a new algorithm that measures carbon dioxide and methane in Earth's atmosphere by fitting model spectra to near-infrared observations of standard stars. It could provide nighttime greenhouse gas column measurements from archival astronomical data, but current precision is far below dedicated instruments.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Profile-shape scaling under GGG2020 extrapolation is the key unvalidated assumption; a single layer-independent factor cannot correct shape errors, biasing retrieved abundances and the recovered trend.","rationale":"I read the paper in good faith and agree with the reader's assessment. Astroclimes is a novel, well-structured feasibility study with transparent discussion of limitations: the SKYCALC test validates the forward model and fitting procedure for H2O, the MCMC setup is standard, and the authors explicitly caution that their absolute values require a vertical shift and that seasonal cycles are not recovered. The central claim is not that the method is ready for quantitative climate use, but that archival telluric-standard observations contain recoverable long-term information about CO2 and CH4. For that claim, the most load-bearing assumption is the single-factor scaling of the GGG2020 prior profile. Because GGG2020 profiles after 2018 are extrapolations of flask records with a frozen seasonal cycle, any time-dependent shape error will directly propagate into both the ground-level and column-averaged retrievals and could mimic or suppress the very trends and seasonal cycles the method aims to recover. The paper does not test this sensitivity for CO2 or CH4, only for H2O, where the column average is shown to be insensitive to profile source; that result does not transfer to trace gases with different vertical gradients. I therefore identify the same weakest assumption as the reader, and the recommended test (switching the prior shape to CAMS EGG4 or perturbing the GGG2020 shape) would settle whether the reported long-term trend is robust. This concern does not overturn the paper's modest, well-qualified conclusions; it confirms that the method currently requires prior-shape validation and external calibration, exactly the conditions stated by the reader. Hence the verdict remains CONDITIONAL (UNCHANGED).","tokens_in":27603,"tokens_out":4674,"duration_ms":44909,"concrete_test":"Re-run the MCMC retrieval on the full 436-observation sample with the same settings but replace the GGG2020 CO2 and CH4 prior profiles with time-interpolated CAMS EGG4 profiles from the same grid point (matching the benchmark already used) or with GGG2020 profiles perturbed by height-dependent factors derived from the CAMS-minus-GGG2020 differences. If the retrieved long-term trend, vertical shift, or seasonal-cycle BIC changes by more than the reported scatter (5/9 ppm CO2, 31/42 ppb CH4), the profile-shape assumption is a dominant systematic. A simpler check is to compare retrievals from the pre-2018 subset (genuine flask data) with the post-2018 subset; if trends or offsets differ significantly, extrapolation shape error is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The retrieval (Section 4) fits only a ground-level abundance for each molecule and multiplies the full GGG2020 prior profile by the associated scaling factor. The column-averaged DMF is then computed from this scaled profile via Eq. (22). If the prior vertical shape is wrong, a single scaling factor cannot absorb the error: the fitted ground-level value and the column integral are both biased by an amount that depends on the shape mismatch. Section 3.2.4 acknowledges this directly, noting that MLO/SMO flask records are extrapolated after 2018 (the CARMENES sample spans 2016-2023, so roughly half the data use extrapolated profiles) and that El Nino conditions can corrupt the profile shape; the estimated errors (0.25% for CO2, 0.6% for CH4) are not propagated into the reported uncertainties. No sensitivity test varies the CO2/CH4 profile shape; Section 4.2 only compares different H2O profile sources. Consequently, the central positive claim -- recovery of the long-term trend -- could be an artifact of the smooth extrapolation functions in GGG2020, and the failure to recover seasonal cycles could be caused by the fixed average seasonal cycle imposed by the prior shape.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents Astroclimes, a synthetic telluric transmission code that generates model spectra from atmospheric profiles and molecular cross-sections, then fits them to high-resolution near-infrared spectra of telluric standard stars using an MCMC likelihood. The method retrieves ground-level and column-averaged dry mole fractions of CO2, CH4, H2O, and O2. The authors apply it to 436 CARMENES observations from 2016 to 2023, using GGG2020 prior profiles scaled by a single ground-level factor per molecule. Benchmarking against the CAMS EGG4 reanalysis, they report long-term scatter of ±5 ppm (ground CO2), ±9 ppm (column CO2), ±31 ppb (ground CH4), and ±42 ppb (column CH4), plus vertical offsets that they correct using the same CAMS comparison. The paper claims recovery of the long-term CO2 and CH4 trends but not their seasonal cycles, and demonstrates short-timescale scatter of order ±1 ppm and ±10 ppb that is comparable to individual MCMC uncertainties.","tokens_in":27929,"tokens_out":6537,"duration_ms":55780,"significance":"If the central claim holds, the method would turn routine telluric-standard observations into a nighttime greenhouse-gas column measurement capability, reusing archival data from multiple observatories. The paper has clear strengths: it uses a large, real dataset; includes a dedicated weather-balloon profile; makes honest and explicit statements about known limitations (e.g., unpropagated GGG2020 extrapolation errors in Section 3.2.4); and provides a SKYCALC benchmark for H2O that quantitatively tests the retrieval of precipitable water vapour. However, the CO2 and CH4 channels are not validated against synthetic spectra or an independent reference free of circularity, and the recovered long-term trend may be inherited from the smooth extrapolation of the GGG2020 priors. These gaps are load-bearing for the paper's main claim, so the contribution is currently more a promising proof-of-concept than a demonstrated measurement technique.","major_comments":[{"comment":"The retrieval in §4 fits only a single ground-level scaling factor that multiplies the full GGG2020 prior profile; if the prior vertical shape is inaccurate, both the fitted ground-level value and the column-averaged DMF from Eq. (22) are biased in a way that a single scale factor cannot correct. Section 3.2.4 acknowledges this directly, stating that the MLO/SMO flask records are extrapolated after 2018, that El Niño conditions can corrupt the profile shape, and that the estimated extrapolation errors (0.25% for CO2, 0.6% for CH4) are not propagated into the reported uncertainties. Since the CARMENES sample spans 2016–2023, roughly half of the data use extrapolated profiles, and the extrapolation functions (19)–(20) are smooth baselines plus a fixed average seasonal cycle. The recovered long-term trend could therefore be an artifact of the smooth extrapolation in the GGG2020 priors rather than a genuine atmospheric signal. The paper provides no sensitivity test that varies the CO2 or CH4 profile shape; Section 4.2 compares only different H2O profile sources. Such a test is necessary before the trend recovery can be attributed to the measurement.","section":"§4, §3.2.4, Eq. (22)"}],"minor_comments":[{"comment":"The data cut from 600 to 511 observations uses an O2 range of 209500±10000 ppm; the threshold is not clearly justified, and since O2 is itself a fitted parameter, the selection could bias the final sample.","section":"§3.1"},{"comment":"The statement that the normalisation parameters 'did not prove to significantly alter the results' is not accompanied by a quantitative check; a brief sensitivity table would support this claim.","section":"§3.1.2"},{"comment":"The sentence reporting the extrapolation errors (0.25% for CO2, 0.6% for CH4) immediately notes that these errors are not propagated; the paper would benefit from a propagation or at least a sensitivity range in the reported uncertainties.","section":"§3.2.4"},{"comment":"There are several typographical issues in the abstract, such as 'Astroclimescan' and 'therently'; these should be corrected throughout the manuscript.","section":"Abstract"},{"comment":"The BIC comparisons in the figure captions are hard to parse; BIC is a relative measure, so a horizontal line 'centred on 0' is an unusual baseline and the caption should state the two competing models explicitly.","section":"Figures 6–9"},{"comment":"The Rayleigh scattering expression mixes a wavelength in micrometres with an exponent involving λ; please specify the units of λ in this equation.","section":"Eq. (6)"},{"comment":"Using the 800 hPa pressure level for surface pressures of 770–790 hPa introduces a small systematic; please state the approximate magnitude of this effect.","section":"§3.3"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal and the idea is worthwhile. My main concern is that the central claim — recovery of the long-term CO2 and CH4 trends — may be driven by the GGG2020 prior extrapolation rather than by the astronomical data. The authors should add the sensitivity tests and independent validation described in the major comments before the paper can be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this is a serious feasibility study for a genuinely new idea — measuring nighttime column-averaged CO2 and CH4 from archival telluric standard spectra. The method, Astroclimes, fits synthetic telluric transmission with an MCMC and reports uncertainties, which the usual telluric-correction tools (Molecfit, TelFit, TAPAS) don't do. If the approach matures, it would fill a real gap: nighttime column measurements are scarce. The paper's honesty is a plus; it explicitly says the precision is not yet state of the art.\n\nWhat it does well: the machinery is careful and transparent. The SKYCALC PWV retrieval validates the H2O modeling. The H2O results on the balloon night track the weather station, which is a good sanity check. The nightly scatter (~1 ppm CO2, ~10 ppb CH4) is on par with the MCMC uncertainties, suggesting the measurement precision itself is reasonable. The data reduction is documented in enough detail to reproduce.\n\nThe soft spots are real and mostly in the validation chain. The biggest issue is the profile-shape assumption. The retrieval scales the entire GGG2020 prior profile by a single ground-level factor. If the prior vertical shape is wrong, the fitted ground value and the column average are both biased, and no amplitude factor can fix that. Section 3.2.4 acknowledges the MLO/SMO extrapolation error (~0.25% CO2, 0.6% CH4) but doesn't propagate it. There's no sensitivity test varying the CO2/CH4 profile shape. That means the recovered long-term trend could be inherited from the smooth extrapolation functions in the priors rather than from the data. The failure to recover seasonal cycles might also be a side effect of the fixed seasonal shape in the prior.\n\nSecond, the CAMS comparison is a circle. The vertical shift (14 ppm CO2, 42 ppb CH4) is derived from the same comparison it's used to correct, and CAMS reports no uncertainties. That's okay for a qualitative benchmark, but it can't establish absolute accuracy. Also, the sample is cut from 600 to 436 observations after O2-outlier and wavelength-shift cuts; the cuts are described, but that's a quarter of the data.\n\nOverall: the claims are appropriately modest — the paper says \"promise\" and means it. The core idea is worth supporting, and the limitations are identified, if not solved. A serious referee could ask for a profile-shape sensitivity test and an independent validation (TCCON or a satellite overpass, even if approximate) before the CO2/CH4 numbers are used quantitatively.\n\nI'd send this out. It deserves referee time.","headline":"A well-built feasibility study for a genuinely new way to get nighttime greenhouse-gas columns from archival telluric spectra, but the profile-shape scaling and CAMS-based shift keep the CO2/CH4 numbers from being trusted yet.","tokens_in":28421,"tokens_out":3800,"would_cite":false,"duration_ms":32824,"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":"Astroclimes shows that ground-based astronomical spectra of telluric standard stars can recover the long-term rise of atmospheric CO2 and CH4.","keywords":["greenhouse gases","telluric spectroscopy","CO2 retrieval","CH4 retrieval","MCMC","CARMENES","atmospheric column","nighttime measurements"],"falsifier":"A direct test would be to generate synthetic telluric spectra with the ESO Sky Model but with deliberately altered CO2 and CH4 vertical profile shapes (same ground value, different distribution with height), run the Astroclimes retrieval on them, and check whether the retrieved column-averaged dry mole fractions recover the known inputs or instead show a shape-dependent bias. Independently, comparing Astroclimes column-averaged retrievals to collocated aircraft flask profiles or TCCON measurements over the same nights at a site such as Calar Alto would reveal whether profile-shape errors produce the observed vertical offset and missing seasonal cycle.","tokens_in":27434,"feed_emoji":"🌍","tokens_out":3029,"duration_ms":28116,"temperature":0.7,"pith_summary":"This paper presents a new algorithm, Astroclimes, that measures the abundance of CO2 and CH4 in Earth's atmosphere by fitting model telluric transmission spectra to near-infrared observations of telluric standard stars taken by ground-based telescopes. Because these stars are routinely observed for calibration purposes, the method could turn archival astronomical data into a nighttime greenhouse-gas monitoring network, complementing sunlight-dependent satellites and ground networks. Running the algorithm on 436 CARMENES spectra from 2016 to 2023, the paper finds that the retrievals reproduce the known long-term increase in both CO2 and CH4 but do not recover their seasonal cycles. Benchmarked against the CAMS EGG4 reanalysis, the retrievals show an overall vertical offset and long-term scatter of about ±5 ppm (ground CO2), ±9 ppm (column CO2), ±31 ppb (ground CH4), and ±42 ppb (column CH4), while scatter within a single night is much smaller, around ±1 ppm and ±10 ppb. The paper argues the method is not yet competitive with dedicated state-of-the-art instruments but is a promising proof of concept.","feed_headline":"Starlight reveals Earth's long-term CO2 and methane rise","feed_subtitle":"A new algorithm turns routine calibration star spectra into nighttime greenhouse-gas column measurements, with nightly precision near 1 ppm.","key_machinery":"The central object is the synthetic telluric transmission spectrum, computed from molecular cross-sections (a newly calculated low-temperature grid based on HITRAN), an atmospheric profile of pressure, temperature, and molecular abundances, and the Beer-Lambert law with additional terms for Rayleigh scattering, aerosol extinction, and collision-induced absorption by O2. The free parameters are the ground-level abundances of CO2, CH4, H2O, and O2; each molecular abundance profile from the GGG2020 pipeline is multiplied by a single scaling factor. The model is compared to observations through a cross-correlation-based log-likelihood, and the parameters are estimated with an MCMC using the emcee sampler. Column-averaged dry mole fractions are derived by integrating the retrieved number-density profiles relative to O2, whose atmospheric fraction is assumed constant.","core_discovery":"The paper's central claim is that astronomical observations of telluric standard stars, which are routinely taken for calibration and contain strong telluric absorption lines, can be used to estimate the column-averaged and ground-level dry mole fractions of CO2 and CH4 in Earth's atmosphere. Using a Markov-Chain Monte Carlo fit of synthetic transmission spectra to CARMENES near-infrared data, the authors recover the long-term upward trend of both gases over 2016 to 2023, though the seasonal cycle is not detected at a statistically significant level. The paper also quantifies the current precision limits: on hourly timescales the scatter is comparable to the individual measurement uncertainties, indicating that the residual long-term scatter is dominated by systematic effects rather than random noise.","pith_inferences":["If the nightly precision holds after correcting the systematic offset, the method could turn existing exoplanet-atmosphere telluric calibration data into a climate dataset spanning years, enabling retrospective studies of how CO2 and CH4 columns evolved at dozens of observatory sites.","The assumed constancy of the vertical profile shape is likely the main source of the seasonal-cycle failure; a retrieval that fits profile shape parameters (e.g., a scale height or a stratification parameter) rather than a single ground-level scaling factor might recover the seasonal signal.","Combining the near-infrared CO2 and CH4 lines with the visible O2 A-band in a joint retrieval would anchor the dry-air column directly from the same stellar spectrum, reducing the reliance on external O2 assumptions and potentially lowering the column-averaged scatter.","A direct test of the method's sensitivity to profile shape could be made by applying Astroclimes to spectra taken during a known El Niño event, when the extrapolated MLO and SMO flask records are most likely to be in error."],"forward_implications":["Archival telluric standard star spectra from high-resolution spectrographs at multiple observatories could become a new data source for nighttime greenhouse-gas column monitoring.","The short-timescale scatter (about ±1 ppm for CO2 and ±10 ppb for CH4 within a night) suggests that once systematic offsets are understood, the method could approach the precision needed to detect regional carbon-cycle signals.","The inability to recover seasonal cycles with the current dataset indicates that longer time baselines or better control of systematic biases are required before the method can be used for trend attribution studies.","Applying the same algorithm to other near-infrared instruments, especially those covering the stronger O2 A-band in the visible, could improve the column-averaged retrievals and reduce scatter.","The overall vertical shift of the retrieved abundances relative to the CAMS reanalysis points to a correctable bias that, if removed, would make the method directly comparable to established column networks."],"supporting_citations":[{"why":"Provides the cross-correlation-based log-likelihood function that drives the MCMC fit of model transmission spectra to observations.","marker":"Brogi & Line (2019)"},{"why":"Supplies the emcee MCMC sampler used to estimate posterior distributions of the retrieved abundances.","marker":"Foreman-Mackey et al. (2013)"},{"why":"Defines the GGG2020 atmospheric prior profiles (via ginput) that the retrieval scales, and the equations converting between dry mole fraction and number density.","marker":"Laughner et al. (2023a)"},{"why":"Provides the HITRAN line lists from which the low-temperature molecular cross-sections used in the model are computed.","marker":"Gordon et al. (2022)"},{"why":"Supplies the formulation for Rayleigh and aerosol scattering as well as the ESO Sky Model Calculator used to generate reference spectra and emission-line masks.","marker":"Noll et al. (2012)"},{"why":"Describes the CAMS EGG4 reanalysis used as the benchmark for evaluating the retrieved ground-level and column-averaged abundances.","marker":"Agustí-Panareda et al. (2023)"},{"why":"Documents the CAMS global reanalysis framework that EGG4 is part of, providing the context for the benchmark data.","marker":"Inness et al. (2019)"}],"fun_headline_variants":["Starlight measures CO2 and methane from Earth's atmosphere","Calibration star spectra gauge CO2 and methane levels","Nighttime CO2 and methane measured via starlight","Star spectra yield long-term CO2 and methane trends"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The retrieval assumes the vertical shape of each molecular abundance profile from the GGG2020 pipeline is correct and only scales that entire profile by a single constant; if the true shape differs, for example because the underlying flask records are extrapolated through an El Niño year, the column-averaged values will be biased even when the ground-level value fits well.","fun_headline_variants_meta":{"raw":{"variants":["Starlight measures CO2 and methane from Earth's atmosphere","Calibration star spectra gauge CO2 and methane levels","Nighttime CO2 and methane measured via starlight","Star spectra yield long-term CO2 and methane trends"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000795,"raw_usage":{"total_tokens":3520,"prompt_tokens":987,"completion_tokens":2533,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":2467}},"tokens_in":603,"tokens_out":2533,"duration_ms":17490,"temperature":1.0,"reasoning_tokens":2467,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:55:40.431992+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct test would be to generate synthetic telluric spectra with the ESO Sky Model but with deliberately altered CO2 and CH4 vertical profile shapes (same ground value, different distribution with height), run the Astroclimes retrieval on them, and check whether the retrieved column-averaged dry mole fractions recover the known inputs or instead show a shape-dependent bias. Independently, comparing Astroclimes column-averaged retrievals to collocated aircraft flask profiles or TCCON measurements over the same nights at a site such as Calar Alto would reveal whether profile-shape errors produce the observed vertical offset and missing seasonal cycle.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Describes the CAMS EGG4 reanalysis used as the benchmark for evaluating the retrieved ground-level and column-averaged abundances."}],"review_version":2}