{"id":"b7f51315-c0ab-4497-be06-52f484cf07e6","arxiv_id":"2607.13236","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Visibility-graph Forman-Ricci curvature collapses at the onset of the two largest solar-cycle-25 Forbush decreases and tracks their rigidity spectra.","lead":"This paper applies a network-geometry tool—visibility-graph curvature—to the two largest cosmic-ray Forbush decrease events of solar cycle 25. It reports that the curvature dips sharply at event onset across six neutron monitors, and uses the dip to estimate each event's rigidity spectrum.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Whole-series NVG curvature is acausal: long-range visibility edges can shift the detected curvature minimum earlier than the physical onset, as seen in Gannon pre-shock minima (14:30–16:30 UT vs 17:00 UT shock); the time-local sliding-window check is relegated to robustness.","rationale":"The reader's weakest_assumption correctly identifies that whole-series NVG curvature is not time-local, but I sharpen this into a concrete timing artifact: the acausal visibility can move the curvature minimum earlier, as evidenced by Gannon pre-shock minima. This is load-bearing because the paper's primary result is the temporal location of the minimum at onset. However, the paper's own sliding-window analysis provides a plausible escape: if the local construction still shows a dip near onset, the qualitative conclusion survives, though the whole-series timing claim must be demoted. Thus the verdict remains CONDITIONAL, not ACCEPT, and not REJECT absent the test. I partially agree with the reader because they focused on baseline contamination rather than the direct shift in timing, which is the sharper threat.","tokens_in":11514,"tokens_out":10882,"duration_ms":123347,"concrete_test":"Recompute the nodal FR curvature for the Gannon event using the 6-h sliding-window NVG (step 30 min, as in §4.3) and record the time of each station's minimum curvature. Compare these times to the whole-series minima reported in §4.2. Specifically, check whether the pre-shock minima (14:30–16:30 UT) shift to after the 17:00 UT shock. If any station's minimum moves later, the whole-series timing is contaminated by non-local edges; if all minima remain pre-shock, the timing may be a genuine precursor, but a further causal (past-only) graph test would be needed to distinguish.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central timing claim—that nodal FR curvature exhibits a sharp, coherent minimum at FD onset—depends critically on the whole-series NVG being a time-local descriptor. But Eq. (1) links any mutually visible pair across the entire record, so F(t) at time t encodes information about future samples. This is not merely baseline contamination; it can move the location of the detected minimum. In §4.2, five of six Gannon stations have curvature minima between 14:30 and 20:30 UT on May 10, bracketing the 17:00 UT shock. Some of these minima occur hours before the shock. Under a time-local construction, a node at 14:30 UT should not respond to a Forbush decrease whose shock arrives at 17:00 UT. The most natural explanation is that the pre-shock node is connected by long visibility edges to the future FD hub, pulling its curvature down. The sliding-window graphs in §4.3 are immune to non-stationarity outside the window, but the paper presents them only as robustness checks; the primary whole-series results (Figs. 2–3, §4.2) drive the 'onset geometry' claim. If those pre-shock minima vanish under a local sliding-window construction, the claimed onset timing is an artifact of acausal visibility, and the 'onset-to-quiet' phase contrasts are not purely local. The placement test does not resolve this, because its null windows are drawn from the same globally constructed graph and are themselves contaminated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper applies nodal Forman–Ricci (FR) curvature of natural visibility graphs (NVG) to the two largest Forbush decreases of solar cycle 25, using 30-min NMDB data from six NM64 stations spanning cutoff rigidities 0–7 GV. It reports a sharp, coherent curvature minimum at FD onset in both events, with onset-to-quiet median curvature ratios of 2.5–5.2 (Gannon) and 3.3–8.2 (June 2025), supported by a block-permutation placement test at p below the attainable floor for all 24 station–phase combinations. The paper also fits FD rigidity spectra (γ = 0.80 and 0.50), reports an amplitude–curvature scaling |F_min| ∝ A^0.57, and includes robustness checks using edge-weighted curvature, sliding-window graphs, and an amplitude-blind horizontal-visibility control.","tokens_in":11889,"tokens_out":3772,"duration_ms":43486,"significance":"If the central timing claim survives scrutiny, the paper introduces a novel and computationally cheap geometric descriptor for FD morphology, with the attractive properties of single-station operation and sensitivity beyond the degree sequence. The study is commendably transparent: the placement test preserves autocorrelation, the HVG control is a sensible specificity check, the sliding-window experiments directly address non-stationarity, and the authors list several limitations explicitly. However, the main claim that nodal FR curvature marks FD onset in a time-local manner rests on whole-series NVG construction, whose acausal long-range edges can shift or create pre-onset minima. This issue is load-bearing for the paper's primary conclusion and must be resolved before the results can be accepted as stated.","major_comments":[{"comment":"The whole-series NVG makes F(t) depend on the entire record: Eq. (1) links any mutually visible pair across all n samples, so a node at time t can be connected to a future FD hub. In §4.2 five of six Gannon stations have their curvature minima between 14:30 and 20:30 UT on May 10, several hours before the 17:00 UT shock. Under a truly time-local construction, a 14:30 UT node should not respond to a shock arriving at 17:00 UT; the most natural explanation is long visibility edges to the future depression. The paper's primary 'onset geometry' claim therefore rests on an acausal descriptor. The sliding-window analysis in §4.3 mitigates this, but it is presented only as a robustness check, and the placement test does not resolve the issue because its null windows are drawn from the same globally constructed graph and are equally contaminated. The primary analysis should be repeated with a ca","section":"§3.1, Eq. (1); §4.2, Fig. 2"},{"comment":"The amplitude–curvature relation |F_min| ≈ 26.6 A^0.57 is fit to 12 station–event pairs, but both A and |F_min| are derived from the same count-rate series and the same NVG; the fit reports no uncertainties on the prefactor or exponent, no scatter plot residuals, and no test of sensitivity to excluding either event or individual stations. With 12 points clustered into two events, Spearman ρ = 0.75 (p = 5×10^-3) is suggestive but not a strong scaling law. The paper should provide confidence intervals on the power-law fit, show per-event fits, and discuss the effective number of independent samples.","section":"§4.4, Fig. 4b"},{"comment":"Part of the headline 'curvature collapse' is a mathematical consequence of the definition: Eq. (3) gives F(u) = 4 − k_u − ⟨k⟩_{N(u)}, and a deep negative excursion becomes a visibility hub by the NVG rule, so any node connected to that hub acquires strongly negative curvature. The paper acknowledges this mechanism in §5, but the framing in the abstract and conclusions presents the collapse as an empirical discovery rather than as an expected property. The empirical content lies in the timing, coherence, rigidity ordering, and amplitude scaling of the collapse. Once the acausality of the whole-series graph is addressed, the paper should explicitly separate the tautological component from the genuinely data-driven component.","section":"Eq. (3); §5"}],"minor_comments":[{"comment":"The title contains a typo: 'CUR V ATURE' should be 'CURVATURE'.","section":"Title"},{"comment":"The phrase 'locality in time (a curvature value per sample)' is misleading because the whole-series NVG curvature at a sample depends on the full record unless a sliding-window or causal construction is used. Please rephrase to avoid contradicting §3.1.","section":"§5"},{"comment":"The text says 'Figure 5 compares...' but the figure appears after Fig. 4; please check the numbering and in-text references.","section":"§4.3"},{"comment":"The statement 'Analysis code and derived products are available from the authors on request' is weaker than the reproducibility standards of most journals. Please consider depositing the code in a permanent repository, especially since the graph construction and permutation test are central to the paper.","section":"Data and code availability"},{"comment":"The GLE74 masking by linear interpolation over 8 h at polar stations is acknowledged as a limitation, but the possible effect on curvature in the recovery phase is not quantified. A brief sensitivity test (e.g., masking a few hours more or less) would be useful.","section":"§3.2 and §5"}],"recommendation":"major_revision","confidential_remarks":"The core statistical machinery is appropriate and the paper is readable, but the central timing claim is not yet established because the primary whole-series NVG is acausal. I would like to see a revised version in which the main analysis uses a causal or sliding-window construction, or otherwise demonstrates that the pre-shock curvature minima survive without long-range future visibility. The amplitude–curvature power law also needs proper uncertainty quantification. This is well within the scope of the journal and the approach is promising, but the current framing overstates the time-locality of the descriptor."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does two things well. It is the first application of Forman–Ricci curvature on visibility graphs to Forbush decreases, and it documents the effect carefully: six stations, two large events, a placement test that preserves autocorrelation, an amplitude-blind HVG control, and a decomposition showing that the two-hop neighbor-degree term carries most of the signal. The qualitative result—curvature collapses at onset, with ratios of 2.5 to 8.2 and large effect sizes—is credible and consistent with a deep negative excursion creating visibility hubs. The spectral indices (γ = 0.80 and 0.50) and the amplitude–curvature scaling (A^0.57) are new numbers, even if the sample is small. I also credit the authors for being explicit about limitations: they say detection is not an open problem, they mask the GLE74 interval, and they acknowledge the placement test floor.\n\nThe soft spot is not the statistics or the graph construction per se; it is the claim that nodal curvature is time-local. Equation (1) connects mutually visible pairs across the entire record, so F(t) at a quiet-time node can be pulled down by a future FD hub. The paper's own Figure 2 shows Gannon curvature minima at 14:30–16:30 UT, hours before the 17:00 UT shock. Under a local construction, that should not happen. The authors mention this indirectly when they say curvature responds to the gradient rather than the floor, but the pre-shock minima are not explained by that. The sliding-window NVGs in §4.3 are exactly the right fix, but they are presented only as robustness checks, and the timing claim in §4.2 is driven by the whole-series graphs. The placement test does not resolve this, because the null windows are drawn from the same globally constructed graph and are themselves contaminated. The amplitude–curvature power law is fit to the same 12 points that define the amplitudes, so it is suggestive, not established. And no code is shipped.\n\nThis is not a fatal flaw. The central result—that the curvature collapses at onset and is statistically distinguishable from quiet-time fluctuations—likely survives a local reanalysis; the June 2025 event shows the effect in sliding windows too. But the paper overstates time-locality and the onset geometry claim needs rework. A serious referee should ask the authors to present the sliding-window analysis as primary, report where the minima fall under local construction, and ideally release the code. It is worth engaging with, and it should be sent to peer review rather than desk-rejected.","headline":"Whole-series visibility-graph curvature collapses at FD onset and survives permutation testing, but the primary analysis is not time-local and some 'onset' minima occur before the shock; the sliding-window version that would fix this is only a robustness check.","tokens_in":12384,"tokens_out":2574,"would_cite":false,"duration_ms":28940,"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":"Curvature of the visibility graph built from neutron-monitor count rates collapses sharply and coherently at the onset of the two deepest Forbush decreases of solar cycle 25, and the signature survives a structure-preserving permutation tes","keywords":["Forbush decrease","visibility graph","Forman-Ricci curvature","neutron monitor","rigidity spectrum","cosmic-ray modulation","geomagnetic storm","time-series network"],"falsifier":"Compute nodal curvature on a sliding window shorter than the main-phase gradient (e.g., 3 hours) for the May 2024 event; if the sharp onset minimum persists at the shock time, the signal is local, whereas if it shifts or disappears the whole-series result is contaminated by long-range visibility links. Alternatively, replace the FD interval in the record with quiet-time fluctuations, rebuild the visibility graph from the remaining samples, and check whether a curvature minimum still appears at the same time.","tokens_in":11371,"feed_emoji":"🌌","tokens_out":7365,"duration_ms":78986,"temperature":0.7,"pith_summary":"The paper is trying to establish that a geometric network measure—the nodal Forman–Ricci curvature of a natural visibility graph—turns the onset of a Forbush decrease into a sharp, statistically significant dip in a time series, and that this dip carries information about the event's amplitude and rigidity spectrum. The two test events are the largest of solar cycle 25: the fast, sheath-dominated May 2024 storm and the deeper, compound June 2025 event, observed by the same six neutron monitors. The claim matters because it offers a compact, single-station, real-time-computable descriptor of Forbush-decrease morphology—onset timing, depth, and spectral hardness—rather than just a detection of the decrease itself. The analysis anchors every graph-derived quantity to shock-based event phases and checks specificity with an amplitude-blind control.","feed_headline":"Graph curvature pinpoints cosmic-ray storm onset","feed_subtitle":"A geometric network measure turns the cycle's two biggest neutron-monitor dips into sharp, statistically verified onset signals.","key_machinery":"The machinery is the natural visibility graph (NVG)—nodes are 30-minute count-rate samples, and two samples are linked if every intermediate sample lies below the straight line joining them—combined with the nodal Forman–Ricci curvature F(u) = 4 − k_u − ⟨k⟩_N(u), where k_u is the node's degree and ⟨k⟩_N(u) is the mean degree of its neighbors. This quantity is a two-hop local measure: a node adjacent to a hub inherits strongly negative curvature even if its own degree is modest. In a Forbush decrease, the few extreme samples of the steep gradient become hubs visible to large stretches of the record, so the curvature collapses exactly where the count-rate gradient is steepest. Unit-weight, tem","core_discovery":"The central discovery is that nodal Forman–Ricci curvature F(u) of the natural visibility graph built from 30-minute neutron-monitor records drops from quiet-time medians of roughly −10 to −14 to onset medians of −20 to −134, giving onset-to-quiet median ratios of 2.5–5.2 for the May 2024 storm and 3.3–8.2 for the June 2025 event, with the curvature minimum reached within hours of the shock arrival and before the count-rate minimum. A placement test that compares each observed phase-window mean with all contiguous windows of equal length in the pre-event interval finds the observed mean below every null window for all 24 station–phase combinations, so the attained p-value sits at the test fl","pith_inferences":["If the curvature response is read as a geometric marker of the transition from fluctuation-dominated to gradient-dominated transport, the onset dip should also appear in other impulsive cosmic-ray modulations (such as solar energetic particle events, after masking) whenever a steep unidirectional gradient is present.","The authors' explicit caveat that the placement-test p-value is floored at about 2×10⁻³ by the length of the quiet interval suggests that longer pre-event baselines, or surrogate-based tests, could sharpen significance and allow phase-resolved comparisons between events of different morphology.","The sublinear exponent 0.57 hints that curvature saturates for very deep decreases; a calibration with synthetic profiles injected into real quiet-time records—which the authors propose as next work—would determine whether the relation is universal across station sizes, counting rates, and cutoff rigidities.","A catalog-scale application would require hourly or finer data and short event-matching windows, as the paper itself states; this is an extension, not an established result."],"forward_implications":["If the curvature collapse tracks onset, a single neutron monitor can timestamp the arrival of the driving interplanetary shock or sheath without waiting for the count-rate minimum.","The rigidity-ordered amplitudes yield FD spectral indices (0.80 and 0.50) that place both events in the canonical range and show the June 2025 event is both deeper and spectrally harder.","The |F_min| ∝ A^0.57 relation means curvature can serve as a proxy for FD amplitude once calibrated, including at stations where baseline normalization is difficult.","The HVG control implies the signal is tied to the depth, not just the ordering, of the decrease, so the method is aimed at resolving significant events rather than cataloging sub-percent Forbush effects.","Because the curvature is time-local and cheap to compute, it could be monitored in real time alongside established precursor indicators."],"fun_headline_variants":["Graph curvature marks both big Forbush onsets","Ricci curvature spots cosmic-ray storm onsets","Network geometry sharpens solar storm detection","Forman-Ricci curvature nails Forbush onset timing","Curvature spike signals two big cosmic-ray dips"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is that the curvature value at a timestamp is a local descriptor of the series at that time, even though the natural visibility graph links any mutually visible pair across the entire record; if long-range links from the deep-decrease hub contaminate the quiet-time baseline, the claimed onset-to-quiet contrast and timing are not purely local.","fun_headline_variants_meta":{"raw":{"variants":["Graph curvature marks both big Forbush onsets","Ricci curvature spots cosmic-ray storm onsets","Network geometry sharpens solar storm detection","Forman-Ricci curvature nails Forbush onset timing","Curvature spike signals two big cosmic-ray dips"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000179,"raw_usage":{"total_tokens":1249,"prompt_tokens":972,"completion_tokens":277,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":716,"completion_tokens_details":{"reasoning_tokens":217}},"tokens_in":716,"tokens_out":277,"duration_ms":3374,"temperature":1.0,"reasoning_tokens":217,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T05:46:45.219732+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute nodal curvature on a sliding window shorter than the main-phase gradient (e.g., 3 hours) for the May 2024 event; if the sharp onset minimum persists at the shock time, the signal is local, whereas if it shifts or disappears the whole-series result is contaminated by long-range visibility links. Alternatively, replace the FD interval in the record with quiet-time fluctuations, rebuild the visibility graph from the remaining samples, and check whether a curvature minimum still appears at the same time.","supporting_citations":[],"review_version":1}