{"id":"92d6840f-56e5-41b8-92f6-7c7ed0f585c1","arxiv_id":"2603.00371","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Using existing FRB host magnitude and redshift-dispersion models, LSST will identify most coherent-mode FRB hosts, with photometric redshift errors degrading H0 sensitivity by only ~7%.","lead":"This paper forecasts how many fast radio burst (FRB) host galaxies the Rubin Observatory's LSST survey will spot, for FRBs found by ASKAP and MeerKAT. It predicts that single-visit images will catch 65% of ASKAP's CRACO hosts, and that photo-z errors only slightly weaken Hubble constant measurements.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table 1 counts galaxies brighter than a 5σ limit as 'identified', but the cited B. Anderson et al. (in prep) shows that host association (PATH posterior ≥ 0.9) is biased against hosts with m_r > 22; the 65%/81% fractions are upper limits on detection, not identification.","rationale":"The reader identified the Marnoch+2023 host-magnitude distribution as the weakest assumption; that is certainly a concern and I partially agree with it. However, the most load-bearing issue for the headline claim is that the paper's 'identify' percentages are actually detection percentages. The calculation in §2.4 uses only a photometric magnitude limit, not a host-association criterion, yet the abstract and conclusion use the word 'identify'. The paper itself cites B. Anderson et al. (in prep) showing that host association at a 90% posterior level is biased against hosts with m_r > 22, which implies that the detectable fraction and the identifiable fraction are not the same. This is an internal mismatch rather than an external extrapolation, and it directly affects the quantitative claims in Table 1 and the abstract. The concrete test—running PATH on simulated LSST catalogs with realistic astrometric uncertainties—would settle the size of the overestimate. The reader's verdict of CONDITIONAL remains appropriate; my concern reinforces the need for conditioning, so the verdict does not change.","tokens_in":16301,"tokens_out":9121,"duration_ms":105604,"concrete_test":"Simulate CRACO and MeerTRAP coherent FRB localizations with realistic astrometric errors (e.g., 1″, 3″, 5″) and inject host galaxies drawn from the paper's Marnoch+2023 mr(z) model into LSST-like single-visit and 10-year co-add catalogs (with field galaxies from observed number counts). Run PATH (Aggarwal et al. 2021) requiring a 90% posterior probability of association, and compute the identifiability fraction as a function of magnitude limit. Compare these fractions to f24.7 and f27.5 in Table 1. If the PATH-based fractions are >10 percentage points lower, the abstract and Table 1 overstate host identification.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (§2.4, Table 1) equates host identification with m_r < m_lim (24.7 single visit, 27.5 co-add). This ignores the association step: a detected galaxy is not necessarily the FRB host if chance coincidences are likely. The paper itself notes in §2.2 that 'requiring a posterior likelihood of at least 90% results in significant bias against hosts with m_r > 22' (B. Anderson et al., in prep). This bias applies a fortiori to the LSST co-add limit (27.5), where the field-galaxy surface density is much higher. The model in §2.4 therefore computes the maximum fraction of hosts that are photometrically detectable, not the fraction that would be uniquely identified. This is a metric mismatch with the abstract's 'identify'. The size of the overestimate depends on the adopted radio localization uncertainty (not specified in the paper) and on the host-galaxy prior; even a 1″ positional error at m_r=27.5 yields a non-negligible chance-coincidence probability, and errors of a few arcseconds make it dominant. The 81% co-add figure is especially exposed. The host-magnitude distribution (Marnoch+2023) is a separate uncertainty, but the detection/identification gap is a direct, internal mismatch.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper forecasts how many FRB host galaxies detected by ASKAP's CRACO and MeerKAT's coherent mode will be observable in LSST, using a model of the r-band host magnitude distribution from Marnoch et al. (2023) convolved with redshift distributions from the ZDM code. It predicts that 65% of CRACO hosts and 81% of MeerKAT coherent hosts will be detectable in single-visit and co-added LSST images respectively. The second part simulates a simplified H0 analysis with photometric redshifts, finding that the adopted photo-z error (sigma_z=0.035) degrades H0 precision by ~7% for CRACO, while missing faint hosts degrades it by ~47%, or ~62% combined.","tokens_in":16723,"tokens_out":3855,"duration_ms":44731,"significance":"If the headline numbers hold, the paper provides a strong practical argument that LSST can relieve the optical follow-up bottleneck for FRB cosmology. The approach is transparent: it uses publicly available code (ZDM), observational inputs, and explicitly states its model assumptions. However, the central claim rests on a magnitude-limit calculation presented as a host-identification forecast, while the association step is not modeled. The host-magnitude distribution is extrapolated from 23 low-redshift hosts with no evolution, and the photo-z simulation uses one algorithm's error excluding outliers. The paper is therefore a useful planning estimate, but the quantitative headline figures should be treated as upper limits until the association step and systematic uncertainties are incorporated.","major_comments":[{"comment":"The abstract and §2.4 state that LSST will 'identify' 65% of ASKAP/CRACO hosts and 81% of MeerKAT hosts, but the calculation in §2.4 counts only hosts with m_r below the LSST 5σ limits (24.7 and 27.5). The association step is not modeled. The paper itself notes in §2.2 (citing B. Anderson et al., in prep) that requiring a 90% posterior association probability is significantly biased against hosts with m_r > 22. At the co-add limit of 27.5, the field-galaxy surface density is far higher than at 22, so a non-negligible fraction of detected galaxies will be chance coincidences, especially for MeerKAT's larger localization uncertainties. Thus Table 1 gives upper limits on photometric detection, not the fraction of uniquely identified hosts. The 81% co-add figure is especially exposed. Please either revise the claims to 'photometrically detectable' or include a host-association model (e.g., P","section":"§2.4 / Abstract / Table 1"},{"comment":"The host magnitude model is built from 23 FRB hosts with z ≲ 0.5, extrapolated unchanged to z ~ 2 with no galaxy evolution, and assumes a normal distribution in m_r at each z. The paper acknowledges these limitations, but the headline fractions in Table 1 and the abstract are point estimates with no systematic uncertainty. Since the 65% and 81% figures are direct outputs of this assumed distribution, the manuscript should quantify how these fractions vary under plausible perturbations: e.g., adopting a different scatter, including a simple luminosity-evolution term, or using the empirical step-function counting used in earlier work (Marnoch et al.; Caleb et al.). Without such a robustness check, the abstract's precise percentages overstate the confidence of the forecast.","section":"§2.2 / Figure 1 / Table 1"},{"comment":"The photo-z simulation adopts sigma_z = 0.035 from the kNN algorithm in the DP1 study (T. Zhang et al. 2025) after excluding outliers, while the same study reports 10–20% outliers with large redshift errors. The 6.8% (CRACO) and 3.3% (MeerKAT) precision losses in Table 2 therefore assume that outliers are perfectly removed by quality cuts. Since the abstract states 'only 7%' without this caveat, the simulation should either include a realistic outlier fraction (with a statement of how quality cuts would remove them) or the abstract should be qualified. The simplified likelihood also omits selection biases and host-DM systematics noted in §3.1; this is acceptable for an order-of-magnitude estimate, but the abstract's precision claim should reflect that caveat.","section":"§3.1 / Table 2"}],"minor_comments":[{"comment":"The text uses 'dim' in the abstract but 'faint' in the body; please be consistent. Also, 'MeerKA T' is split across line breaks in several places (e.g., Section 2.3, 2.4, Table 2), which should be fixed.","section":"General"},{"comment":"The column header 'ASKAPO/CRACO' contains a typo ('ASKAPO'). Also, the abstract rounds 46.8% to 47% and 62.4% to 62%, which is fine, but the table values should be consistent in the text.","section":"Table 2"},{"comment":"Table 1 gives no uncertainties, and Figure 2 shows only the model curve without indicating the 23-host sample scatter. Adding error bands or a shaded region would help readers assess the robustness of the 65% and 81% numbers.","section":"Table 1 / Figure 2"},{"comment":"The caption says 'solid curves' for the redshift distributions and 'dashed'/'dotted' for the LSST-accessible fractions, but the figure legend is not reproduced here. Please ensure the line styles are clearly labeled and the dashed/dotted description matches the visible plot.","section":"Figure 3"},{"comment":"The paper relies on several 'in prep' or 'in prep., 2026' works for key assumptions (B. Anderson et al.; Yuanming Wang et al.; L. Spitler, E. Keane et al.). While this is acceptable for a forward-looking planning paper, please check that the most recent public versions are cited where possible, and consider adding a short description of the B. Anderson et al. association-bias result in the text, since it directly motivates the major concern in §2.4.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a useful planning forecast, but the central claim conflates photometric detectability with host identification. The authors explicitly recognize the association bias in §2.2, so the fix is within scope: either soften the abstract/Table 1 claims or add a straightforward association-probability calculation. I also recommend requesting a robustness test of the host-magnitude model, as the 23-host, no-evolution extrapolation is the backbone of the quantitative forecasts. The photo-z section is honestly framed as an order-of-magnitude estimate, but the abstract's 'only 7%' should carry a caveat about outlier removal. These are fixable with moderate additional work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Good planning paper, not a discovery. The genuinely new thing is the convolution of Marnoch et al.'s host magnitude model with z-DM predictions to get mr(z) for specific radio surveys, plus the photo-z H0 sensitivity estimates. The zdm code is public and the assumptions are stated clearly, so the numbers can be traced and re-run. The authors also check their host model against independent DSA and MeerTRAP hosts. That's real work and useful.\n\nThe qualitative conclusions are probably right: LSST will catch most bright FRB hosts, and photo-z errors are not the bottleneck; losing faint high-z hosts hurts H0 more. The paper is honest about the 23-host, no-evolution host model, and the simplified H0 likelihood.\n\nThe soft spots are real. The stress-test point lands: Table 1 and the abstract say \"identify,\" but the calculation only counts hosts brighter than a 5σ limit. The paper itself cites B. Anderson (in prep) showing that a 90% posterior host-association cut biases against hosts with mr > 22. At the co-add depth of 27.5, chance coincidences with unrelated galaxies are not negligible, and the radio localization uncertainty is not specified. So 65% and 81% are upper limits on photometric detectability, not on unique identification. The 81% MeerTRAP number is the most exposed because it relies on the deep co-add. The fix is not hard: state the localization error, apply a chance-coincidence calculation, or at least re-label the numbers and stop calling it identification.\n\nThe host magnitude model is a separate uncertainty: 23 low-z hosts, assumed Gaussian, extrapolated to z~2 with no evolution. The fractions have no error bars, so they should be quoted as rough point estimates. Several inputs are still in prep, which makes the forecast harder to audit. The H0 numbers are explicitly simplified and should be treated as order-of-magnitude.\n\nBottom line: this deserves peer review. An editor should send it to a referee, with the request that the association bias be acknowledged or modeled and the headline fractions re-labeled. It is a useful planning paper for the FRB and Rubin communities, not a breakthrough, and worth citing once the numbers are tightened.","headline":"Useful planning forecast with reproducible code; treat 65%/81% as upper limits on host identification until association bias is folded in.","tokens_in":17207,"tokens_out":3464,"would_cite":true,"duration_ms":39524,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"If this paper is right, the Rubin Observatory's LSST alone will identify most fast radio burst host galaxies, and photometric redshifts will barely hurt Hubble-constant measurements.","keywords":["fast radio bursts","FRB host galaxies","LSST","Rubin Observatory","photometric redshifts","Hubble constant","dispersion measure","ASKAP / MeerKAT"],"falsifier":"Take FRBs localized by ASKAP/CRACO over the first few years of LSST, count the fraction with a host galaxy brighter than r = 24.7 within the localization region, and compare to the predicted 65%; a similar check at r = 27.5 for MeerKAT hosts, or a measurement of the m_r distribution at z > 0.5, would confirm or refute the model.","tokens_in":16174,"feed_emoji":"🔭","tokens_out":5432,"duration_ms":52449,"temperature":0.7,"pith_summary":"This paper tries to predict how often the Vera C. Rubin Observatory's LSST will identify the host galaxies of fast radio bursts (FRBs) detected in coherent modes by ASKAP and MeerKAT, and whether LSST photometric redshifts will suffice for cosmology. The authors combine a model of FRB host galaxy r-band brightness as a function of redshift with simulated redshift–dispersion measure distributions for the two radio surveys. They find that a single LSST visit should reveal 65% of ASKAP/CRACO hosts, and the full 10-year co-added images 81% of MeerKAT coherent hosts. They also argue that photometric redshift errors cost only about 7% in H0 precision for CRACO, while missing faint high-redshift hosts is the larger problem, degrading precision by roughly 47% (and 62% when combined). The point is to show that dedicated optical follow-up may become largely unnecessary for FRB cosmology.","feed_headline":"One Rubin visit will catch 65% of ASKAP FRB hosts","feed_subtitle":"Ten-year co-adds reach 81% of MeerKAT hosts; photometric redshifts barely dent H0 precision.","key_machinery":"The central object is a model of the FRB host galaxy r-band magnitude distribution m_r(z), taken to be Gaussian with redshift-dependent mean and scatter derived from 23 hosts with optical spectra, extrapolated without evolution out to z ~ 2. This is convolved with the FRB redshift–dispersion measure distribution predicted for each radio survey (the paper's simulation of the z–DM relation), and the resulting m_r distribution is compared against LSST r-band magnitude limits of 24.7 (single visit) and 27.5 (10-year co-add). For the H0 analysis, the same machinery is extended with a model of photometric redshift errors (Gaussian, sigma_z = 0.035) and the single-visit magnitude limit.","core_discovery":"The paper claims that LSST images alone will identify most FRB host galaxies detected by coherent radio surveys, without dedicated optical follow-up. Combining a model of host galaxy r-band magnitudes with predicted redshift–dispersion measure distributions for ASKAP's CRACO and MeerKAT's coherent mode, the authors estimate that a single LSST visit (r-band limit 24.7) will reveal 65% of CRACO hosts, and that 10-year co-added images (r-band limit 27.5) will reveal 81% of MeerKAT coherent hosts. They further claim that using LSST photometric redshifts instead of spectroscopic ones inflates H0 uncertainty by only about 7% for CRACO (3% for MeerKAT), whereas missing faint, high-redshift hosts is","pith_inferences":["If the host magnitude model is even approximately correct, the LSST–CRACO combination will yield a nearly complete low-redshift host sample, allowing the community to directly measure the m_r(z) distribution and test the no-evolution assumption within a few years.","Because missing high-redshift hosts cost far more than photo-z noise, a mixed strategy — LSST identification for most hosts plus targeted deep imaging of the faintest candidates — may be the most efficient path to FRB cosmology.","The photo-z treatment ignores catastrophic outliers; if LSST photo-zs produce a significant tail of large errors, the small H0 degradation quoted here would be an underestimate.","The same LSST maps could power field-level baryon inference along FRB sight lines, but only if photo-z accuracy at z < 1.5 is sufficient; that application is flagged by the paper as needing further work."],"forward_implications":["A single LSST visit should remove the need for dedicated optical follow-up for 65% of ASKAP/CRACO FRB hosts, freeing telescope time.","The 10-year LSST co-adds should identify 81% of MeerKAT coherent hosts, with similar expectations for SKA-Mid.","Photometric redshifts from LSST are nearly as good as spectroscopic redshifts for H0 from FRBs: only ~7% worse precision for CRACO.","The main enemy is not photo-z noise but missing faint hosts: H0 precision degrades by ~47% for CRACO when faint hosts are lost, and ~62% with photo-z errors combined.","The prediction method is general and can be applied to any optical follow-up survey."],"fun_headline_variants":["Rubin's single shot reveals 65% of ASKAP FRB hosts","LSST alone IDs most FRB host galaxies, no follow-up needed","Photo-z errors only add 7% to H0 uncertainty","Missing faint FRB hosts cuts H0 sensitivity by 47%","LSST co-adds net 81% of MeerKAT FRB hosts"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The visibility fractions rest on the assumption that the r-band magnitudes of FRB host galaxies follow a normal distribution with the mean and scatter measured from 23 mostly low-redshift hosts, unchanged out to z ~ 2; if real hosts are fainter or evolve with redshift, the 65% and 81% figures are too optimistic.","fun_headline_variants_meta":{"raw":{"variants":["Rubin's single shot reveals 65% of ASKAP FRB hosts","LSST alone IDs most FRB host galaxies, no follow-up needed","Photo-z errors only add 7% to H0 uncertainty","Missing faint FRB hosts cuts H0 sensitivity by 47%","LSST co-adds net 81% of MeerKAT FRB hosts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000788,"raw_usage":{"total_tokens":3403,"prompt_tokens":925,"completion_tokens":2478,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":669,"completion_tokens_details":{"reasoning_tokens":2381}},"tokens_in":669,"tokens_out":2478,"duration_ms":16928,"temperature":1.0,"reasoning_tokens":2381,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T19:55:07.769876+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take FRBs localized by ASKAP/CRACO over the first few years of LSST, count the fraction with a host galaxy brighter than r = 24.7 within the localization region, and compare to the predicted 65%; a similar check at r = 27.5 for MeerKAT hosts, or a measurement of the m_r distribution at z > 0.5, would confirm or refute the model.","supporting_citations":[],"review_version":1}