{"id":"9c7be333-b0b7-4662-94c7-139e8d81e0b0","arxiv_id":"1909.01244","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"Coordinating telescopes with iterative skymap decrementing and time-delayed overlapping observations improves gravitational-wave counterpart search coverage.","lead":"This paper coordinates multiple telescopes searching for the light from gravitational-wave mergers by extending proven single-telescope scheduling algorithms. Two simple changes, iterative tiling and time-delayed overlap, increase sky coverage and detection probability for real telescope networks.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Coverage gains assume full-night dedication of every telescope; under realistic time limits the iterative advantage may shrink or vanish.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: the analysis assumes full-night dedication of each telescope. This assumption is not a minor technical detail; it is structurally necessary for the iterative method to produce the observed gains. Without it, telescopes cannot be assigned complementary low-probability regions that may only be observable at specific times, and the method reduces to independent scheduling of the highest-priority fields. The paper is honest about this limitation in the conclusion, which supports a conditional verdict rather than an outright rejection. I considered alternative concerns—lack of uncertainties, small sample, and order dependence—but these are secondary to the practical feasibility of the scheduling assumption. The concrete test of varying the time budget directly tests whether the central claim survives contact with realistic telescope operations.","tokens_in":15217,"tokens_out":6792,"duration_ms":69107,"concrete_test":"Modify gwemopt to impose a hard total-observation-time budget per telescope (e.g., 1, 2, 4, and 8 hours) and rerun the S190425z comparison for the GROWTH, PS1/ATLAS, and GRANDMA networks. If the relative gain in integrated probability and sky area between iterative and independent scheduling declines toward zero as the budget shrinks, the reported gains are an artifact of the unlimited-dedication assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that coordinated scheduling yields 'substantial gains,' but the demonstration rests on the assumption, stated in Section 4, that 'taking complete control of each of these systems for the night following the event is appropriate.' In practice, high-value telescopes have limited target-of-opportunity time, so the iterative method's assignment of complementary sky regions may not be executable. The reported gains—e.g., GRANDMA area rising 660→1060 deg², PS1/ATLAS probability 0.51→0.55—come from forcing each telescope to avoid overlap and cover new regions. If each telescope has only a few hours, it will likely spend that time on the highest-probability tiles, which overlap across telescopes, and the incremental coverage from the iterative method will diminish. Additionally, the results are single deterministic runs with no uncertainties or repeated trials, so it is unclear whether even the idealized advantage is statistically robust or particular to the chosen event and network ordering.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents two extensions to the gwemopt scheduling code for coordinating electromagnetic follow-up of gravitational-wave localizations across networks of telescopes. The iterative method sequentially tiles a skymap, decrementing the probability in already-scheduled tiles so that later telescopes avoid regions already covered; the overlapping method imposes a minimum time delay between observations of the same sky field by different telescopes. The iterative method is demonstrated on the event S190425z using three networks (GROWTH, Pan-STARRS/ATLAS, and GRANDMA) and on S190426c using a GRANDMA galaxy-targeted schedule. The paper reports increases in cumulative sky area and integrated probability for the synoptic case, and in the number of galaxies and integrated metric for the galaxy-targeted case, and concludes that network-level coordinated scheduling can yield substantial gains.","tokens_in":15393,"tokens_out":4723,"duration_ms":46477,"significance":"The algorithms are clearly described and implemented in the open-source gwemopt software, which is a concrete contribution for the gravitational-wave follow-up community. The comparisons are fair in the limited sense that the original and iterative runs use the same skymaps and telescope models. If the quantified gains survive more realistic operational constraints, this is a useful first step toward network-level optimization. The evidence base is narrow, however: two gravitational-wave events, deterministic single runs, no uncertainty estimates, and an acknowledged assumption of full-night telescope dedication. Despite these limitations, the paper is a reasonable proof-of-concept for a simple coordination scheme, and the code release makes the results reproducible.","major_comments":[{"comment":"The quantitative headline results depend on the assumption, stated in Section 4, that 'taking complete control of each of these systems for the night following the event is appropriate.' This assumption is load-bearing for the reported gains, such as the GRANDMA area increase from 660 to 1060 square degrees and the Pan-STARRS/ATLAS integrated probability increase from 0.51 to 0.55. The authors acknowledge the limitation but do not test how the iterative advantage degrades when telescopes have limited target-of-opportunity time. A simple experiment that caps each telescope's total available observing time (or number of tiles) would establish whether the gains survive realistic scheduling constraints and would materially strengthen the central claim.","section":"Section 4 (Conclusion)"},{"comment":"The improvements are reported as single deterministic realizations without uncertainties, repeated trials, or exploration of the telescope-ordering choice. The text notes that the first telescope 'should likely be the best telescope' and that ordering can depend on several event- and network-specific factors, but no alternative orderings are tested. The reader therefore cannot assess the robustness of claims such as the GRANDMA area improvement or the galaxy-targeted improvement from 1303 to 1929 galaxies. A sensitivity test over telescope orderings, or at least a statement of the variation in the metrics, is needed before the gains can be regarded as more than anecdotal.","section":"Section 2 (Figures 3-5)"},{"comment":"The galaxy-targeting demonstration relies on an 'integrated metric' whose improvement from 85% to 99% is quoted as a primary result, but the metric is never defined in the text. The description says only that galaxy weights include a proxy for location within the localization and galaxy mass or star-formation rate, with possible sensitivity-based corrections. Without the actual weighting formula or a precise reference to the implementation in gwemopt, the 85% to 99% claim is not quantitatively assessable, and the claim that the total number of galaxies imaged improves from 1303 to 1929 cannot be evaluated independently.","section":"Section 2 (galaxy-targeting paragraph and Figure 5 caption)"}],"minor_comments":[{"comment":"There is a typo in the sentence 'we use use this method to schedule eleven telescopes' — 'use' is repeated.","section":"Section 2 (galaxy-targeting paragraph)"},{"comment":"The one-hour minimum time delay is introduced as an example, but the dependence of the resulting schedules on this user-chosen free parameter is not explored; a sentence on why one hour is sufficient for the stated asteroid/differentiation goal would help.","section":"Section 3 (overlapping algorithm)"},{"comment":"The caption reads 'On the left is the original algorithm where the telescopes are scheduled separately, and on the right, where they are scheduled iteratively.' The 'where' after 'the right' is grammatically awkward and should be rephrased, for example, 'on the right, the iterative algorithm is used.'","section":"Figure 5 caption"},{"comment":"Several entries are cited as 'Coughlin et al. 2019' with different arXiv numbers; these should be disambiguated with letters (e.g., 2019a, 2019b, 2019c) in both the text and the reference list to avoid ambiguity.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"This is a useful methods paper for the gravitational-wave follow-up community, and the open-source implementation is a genuine strength. My recommendation for major revision is driven by the gap between the strength of the quantitative claims and the narrowness of the demonstrations: two events, deterministic runs, no ordering sensitivity, and an idealized full-night dedication assumption. These are fixable by adding sensitivity tests and clearly framing the results as proof-of-concept, rather than as robust operational predictions. The paper fits the journal's scope well; I do not see grounds for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does one useful thing: it extends gwemopt's single-telescope tiling to networks by decrementing the skymap after each telescope (the iterative method) and adds a time-delay constraint for repeated coverage (the overlapping method). The ideas are straightforward extensions of existing greedy/hierarchical scheduling, but the network-level application is new, the implementation is in open-source code, and the demonstrations on real events (S190425z and S190426c) show tangible improvements in unique sky area, integrated probability, and galaxy-targeted metric coverage. The gains are modest but real—GRANDMA's area goes from 660 to 1060 square degrees, Pan-STARRS/ATLAS probability from 0.51 to 0.55, and the galaxy-targeted metric from 85% to 99%—and they come from actual skymaps, not synthetic tests. That is credit where it is earned.\n\nThe soft spots are also clear. Only two events are shown, with no uncertainties or repeated trials, so the numbers may be particular to these localizations and network configurations. Telescope ordering matters and is basically chosen by hand; the paper mentions a flag for picking the first telescope based on observability but does not explore how sensitive the results are to ordering. The 'golden tiles' parameter lets users preserve overlap in the core, which changes the balance between redundancy and unique coverage. The biggest practical caveat is the full-night dedication assumption: the authors state in Section 4 that they assume complete control of each system for the night following the event, and stress-testing that assumption shows the iterative advantage may shrink if telescopes have only a few hours of target-of-opportunity time, since they would then likely spend it on the highest-probability tiles that overlap across telescopes. This is a real limitation, but it is not hidden—the authors explicitly flag it as a focus for future work.\n\nOne might worry that the iterative gain is partly by construction, since decrementing the skymap forces unique area. But integrated probability also improves, which is not purely tautological; by assigning complementary regions, the network covers more of the localization. The effect is modest but genuine.\n\nThe paper is a small, solid contribution to the multi-messenger follow-up toolbox. The code is open and usable, the algorithms are clearly described, and the limitations are stated honestly. I would accept this for peer review. A referee could ask for more events and uncertainty estimates, and a deeper exploration of telescope ordering and time constraints, but the core idea is worth publishing. For someone working on gravitational-wave counterpart follow-up, this is worth reading; for a general astrophysicist, it is not essential.","headline":"A simple, open-source extension of gwemopt to multi-telescope scheduling that shows real but conditional gains, with the main caveat being the full-night dedication assumption the authors themselves acknowledge.","tokens_in":15916,"tokens_out":2990,"would_cite":true,"duration_ms":28750,"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":"Coordinated telescope networks can cover substantially more sky than the same telescopes acting independently in gravitational-wave counterpart searches.","keywords":["gravitational waves","electromagnetic counterparts","telescope scheduling","observation optimization","multi-messenger astronomy","kilonova","skymap tiling","gwemopt"],"falsifier":"Run both the independent and iterative schedulers on a set of real O3 alerts while capping each telescope's available time at, say, 50% of the night and letting the first telescope be chosen per alert rather than fixed; if the iterative method no longer beats independent scheduling in cumulative probability and area, the paper's central claim would fail under realistic operating conditions.","tokens_in":15060,"feed_emoji":"🔭","tokens_out":5799,"duration_ms":50724,"temperature":0.7,"pith_summary":"The paper argues that gravitational-wave counterpart searches should be scheduled as a coordinated network rather than as a set of independent telescopes. It introduces two small modifications to existing single-telescope scheduling: an iterative tiling rule that removes already-covered sky before the next telescope chooses its tiles, and an overlapping rule that forbids different telescopes from imaging the same field within a set time window. Tested on the neutron-star merger alert S190425z and the black-hole–neutron-star alert S190426c, these rules raise the covered sky area and the cumulative probability of finding the counterpart over what the same telescopes would achieve alone. This matters because most gravitational-wave alerts have huge, patchy sky localizations, and the growing number of follow-up telescopes makes network-level coordination a practical lever for finding kilonovae.","feed_headline":"Team scheduling nearly doubles telescope coverage of GW alerts","feed_subtitle":"A simple deconfliction rule lifts the GRANDMA network's covered sky from 660 to 1060 square degrees.","key_machinery":"The central object is the gravitational-wave sky localization map, whose pixels carry the probability that the source lies there. The load-bearing mechanism is the 'iterative' tiling algorithm: after each telescope's schedule is fixed, every pixel covered by that schedule is set to zero in the map, so the next telescope automatically targets the highest-probability sky that remains uncovered. A second mechanism, 'overlapping' scheduling, treats the time since a field was last observed as a scheduling constraint—like airmass or moon distance—so a different telescope will not revisit the same field within a user-chosen delay (one hour in the examples). Together they turn a network of independent telescopes into a single deconflicted instrument, and the paper implements them in the open-source scheduling code gwemopt.","core_discovery":"On the paper's own terms, extending two proven single-telescope scheduling techniques to a telescope network yields substantially better electromagnetic follow-up of gravitational-wave alerts. In the iterative method, the first telescope tiles the skymap normally; its covered pixels are then set to zero in the skymap, and the next telescope tiles the remaining probability, and so on. In the overlapping method, a minimum time delay is imposed between different telescopes' observations of the same field, so duplicate coverage is spread out in time rather than wasted. Applied to the real alert S190425z, iterative scheduling increased the GRANDMA network's covered area from 660 to 1060 square degrees and raised the Pan-STARRS/ATLAS cumulative probability from 0.51 to 0.55; applied to S190426c with galaxy targeting, it raised the number of galaxies imaged by eleven GRANDMA telescopes from 1303 to 1929 and the cumulative metric from 85% to 99%.","pith_inferences":["If the gains hold across the full O3/O4 alert population, the biggest payoff may come from networks of small telescopes: the iterative method disproportionately increases the contribution of the smallest-field-of-view members by assigning them yet-unexplored sky.","A natural extension the authors leave implicit is jointly optimizing the telescope ordering instead of taking 'best first'; the gains could be larger still if the order is chosen per event from alert properties such as localization area and site weather.","The golden-tile idea suggests a testable hybrid: reserving the inner 50% probability region for redundant, multi-filter follow-up while iteratively covering the outer region would trade a little raw coverage for robustness against weather, a trade-off that could be measured on historical alerts."],"forward_implications":["Any telescope network that adopts iterative tiling can expect its combined coverage and cumulative probability to at least match, and generally exceed, the sum of independently scheduled observations, with the largest gains when field-of-view sizes and site locations differ.","Galaxy-targeted follow-up, not just wide-field tiling, benefits the same way: decrementing the weights of already-scheduled galaxies lets a network of small-aperture telescopes image roughly 50% more galaxies inside the 90% credible contour.","Because the overlapping rule spreads duplicate observations across time, a network can use multiple telescopes to measure kilonova color evolution and reject asteroids without sacrificing coverage of high-probability fields.","The open-source implementation means these coordinated strategies can be adopted by any follow-up team that already uses gwemopt, with no new telescope infrastructure required."],"supporting_citations":[{"why":"The gwemopt codebase and single-telescope scheduling methods that the iterative and overlapping algorithms extend.","marker":"Coughlin et al. 2018b"},{"why":"The hierarchical and greedy tiling schemes whose decrementing of map probability is reused in the network method.","marker":"Ghosh et al. 2017"},{"why":"BAYESTAR, which produces the gravitational-wave skymaps whose pixel probabilities drive the tiling.","marker":"Singer & Price 2016"},{"why":"The S190425z alert and initial localization used for the main demonstration.","marker":"Singer et al. 2019b"},{"why":"The GRANDMA network, the real telescope network whose coverage improves in the examples.","marker":"Antier et al. 2019"},{"why":"The earlier hand-built network coordination that this paper formalizes into an algorithm.","marker":"Waratkar et al. 2019"},{"why":"Probability-weighted exposure time allocation that the scheduling step relies on.","marker":"Coughlin & Stubbs 2016"}],"fun_headline_variants":["Smarter telescope teamwork widens coverage of GW alerts","Network scheduling boosts gravitational-wave counterpart coverage","Iterative tiling lifts telescope network coverage by 400 sq deg","Teamwork: simple scheduling rules improve GW follow-up","Coordinated telescopes cover more sky for GW alerts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that each telescope in the network can be fully dedicated to the gravitational-wave follow-up for the entire night; if real telescopes keep most of their scheduled science and take part in only a fraction of alerts, the demonstrated coverage gains would shrink.","fun_headline_variants_meta":{"raw":{"variants":["Smarter telescope teamwork widens coverage of GW alerts","Network scheduling boosts gravitational-wave counterpart coverage","Iterative tiling lifts telescope network coverage by 400 sq deg","Teamwork: simple scheduling rules improve GW follow-up","Coordinated telescopes cover more sky for GW alerts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000908,"raw_usage":{"total_tokens":3911,"prompt_tokens":959,"completion_tokens":2952,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":575,"completion_tokens_details":{"reasoning_tokens":2877}},"tokens_in":575,"tokens_out":2952,"duration_ms":20094,"temperature":1.0,"reasoning_tokens":2877,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:23:04.549503+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run both the independent and iterative schedulers on a set of real O3 alerts while capping each telescope's available time at, say, 50% of the night and letting the first telescope be chosen per alert rather than fixed; if the iterative method no longer beats independent scheduling in cumulative probability and area, the paper's central claim would fail under realistic operating conditions.","supporting_citations":[{"cited_title":"L., Brady, P","cited_arxiv_id":null,"evidence_quote":"The hierarchical and greedy tiling schemes whose decrementing of map probability is reused in the network method."},{"cited_title":"P., & Price, L","cited_arxiv_id":null,"evidence_quote":"BAYESTAR, which produces the gravitational-wave skymaps whose pixel probabilities drive the tiling."},{"cited_title":"2019, GCN, 24316","cited_arxiv_id":null,"evidence_quote":"The earlier hand-built network coordination that this paper formalizes into an algorithm."},{"cited_title":"2016, Experimental Astronomy, 1","cited_arxiv_id":null,"evidence_quote":"Probability-weighted exposure time allocation that the scheduling step relies on."}],"review_version":1}