{"id":"7aacc1dd-209a-4ff6-a218-0bd2535c3cd8","arxiv_id":"2501.14923","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":4,"one_line_summary":"Re-binned particle intensities can exceed spectral-binned intensities by up to a factor of five because they correspond to a spectral-index-dependent effective energy rather than the fixed log-centered energy.","lead":"This paper compares two ways to average solar energetic particle intensities over energy ranges: a standard 're-binned' average and a 'spectral binned' log-space average. It shows the two can differ by up to a factor of 5 and gives criteria for choosing which to use.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The observed factor-of-five ratio may be inflated because zero counts are replaced by upper-limit intensities before computing jlinlin; the paper does not quantify how much of the PSP ratio comes from this replacement.","rationale":"The paper is a clear methods contribution: the analytic derivation of jlinlin and jloglog for a power-law spectrum is correct aside from minor notation issues, and the conceptual distinction between an effective-energy intensity and a fixed log-centered-energy intensity is valuable. The most load-bearing part of the empirical claim is the PSP factor-of-five ratio, and that ratio depends on how zero-count measurements are treated. The reader identified the ad hoc zero-count replacement for jloglog, but the more specific concern is that the authors compute jlinlin on the same modified spectra, so zero-count bins are assigned positive upper-limit intensities. This biases jlinlin upward relative to a true count-space sum, especially at low count rates, and the paper's own text says the impact at low count times is strongly dependent on the replacement fraction. Because the analytic ratio Eq. (15) does not depend on this data-processing choice, the methodological message stands, but the observational magnitude is not yet quantitatively tied to the method difference. This is a condition that should be checked rather than a fatal flaw, so the reader's CONDITIONAL verdict remains appropriate and unchanged.","tokens_in":12984,"tokens_out":12037,"duration_ms":110342,"concrete_test":"Recompute the Figure 6 and Figure 7 intensities directly from the same PSP IS☉IS level-2 count data: set jlinlin = sum(N_i)/(G*dt*q*sum(dE_i)) with zero counts contributing zero, and set jloglog from Eq. (3) with zeros replaced by 0, 0.5x, and 1x the Gehrels upper limit. If the maximum jlinlin/jloglog ratio or its time dependence changes by more than the propagated uncertainties in low-count intervals, the factor-of-five observational claim is not robust. Separately, compare the observed ratio to Eq. (15) using per-time power-law fits to quantify finite-bin and spectral-curvature residuals.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Section 2.2, after Eq. (14), the authors state that measured zeros are replaced by their Gehrels upper-limit uncertainty before applying Eq. (3), and then 'For now, we compute jlinlin and jloglog on the same modified spectra.' This means the re-binned intensity used in the Section 4.2 PSP comparison is not the pure count-space quantity defined in Eq. (2), i.e., (sum of actual counts)/(G*dt*q*sum of dE); zero-count bins are assigned a positive intensity and therefore inflate jlinlin. The paper acknowledges that the replacement choice 'strongly dependent' affects very low count times, but it does not quantify how much of the factor-of-five ratio in Figure 6 arises from this ad hoc replacement rather than from the inherent jlinlin-versus-jloglog difference. The analytic Eq. (15) is unaffected by this issue, but the empirical 'up to a factor of 5' claim is not yet robust. A secondary assumption, that original bin intensities are point values at their log-centered energies and that the spectrum over the merged range is a single power law, is less severe for the narrow IS☉IS bins but still unquantified for real SEP spectra with curvature.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript contrasts two ways of combining a measured energy-differential intensity across several logarithmically spaced energy bins: the re-binned intensity jlinlin (Eq. 2), a linear average that returns to count space and corresponds to the value of the spectrum at an effective energy Eeff that depends on the spectral index gamma, and the spectral binned intensity jloglog (Eq. 3), a logarithmic average that corresponds to the value at the fixed log-centered energy Eg. For a single power law, the authors derive closed-form expressions for jlinlin, jloglog, Eeff, Eg, and the ratio jloglog/jlinlin (Eq. 15), which depends only on the spectral index, the merged energy range EN/E0, and the original bin width in log energy; no parameters are fitted. Modeled spectra illustrate the ratio (Fig. 3). The same definitions are then applied to PSP/ISIS proton data from the 26-30 August 2022 period (EPI-Lo, LET, HET), yielding time series in which jlinlin exceeds jloglog, in places by nearly a factor of five, with strongly correlated time evolution (Fig. 6). The paper concludes with practical guidance on when each measure is appropriate, and it explicitly acknowledges the zero-count complication in computing jloglog.","tokens_in":13243,"tokens_out":17197,"duration_ms":135792,"significance":"The distinction drawn here is practically important: re-binning to improve counting statistics is ubiquitous in SEP work, and plotting jlinlin at the log-centered energy silently conflates two different energies whenever gamma is not 0 or 2. The analytic development is a genuine strength: Eqs. (4)-(15) are derived in closed form with no free parameters, the ratio formula is falsifiable and could be used to correct published re-binned intensities, and the effective energy and log-centered energy are clearly distinguished. The manuscript is also candid about the zero-count issue and about the interpolation procedure for EPI-Lo thick-foil apertures. The observational section is illustrative rather than definitive: the factor-of-five comparison is computed on spectra in which zero-count bins have been replaced by upper-limit intensities, and the paper does not quantify the resulting bias (see major comments). If the empirical comparison withstands a count-space cross-check, this will be a useful methods reference for the SEP community; the parameter-free analytic ratio and the decision criteria do not depend on the contested zero-count handling.","major_comments":[{"comment":"The observational comparison in Fig. 6 does not implement the count-space re-binning defined in Eq. (2). Section 2.2 states that measured zeros are replaced by their Gehrels 84.13% upper-limit intensities before applying Eq. (3), and that \"for now, we compute jlinlin and jloglog on the same modified spectra.\" A zero-count bin contributes nothing to Eq. (2) before replacement; after replacement it contributes a positive intensity, inflating jlinlin precisely in the low-count epochs where the same paragraph admits the result is \"strongly dependent\" on the replacement fraction. The abstract's \"up to a factor of 5\" therefore characterizes the modified-spectra quantities, not the pure jlinlin-versus-jloglog difference, and the magnitude of the inflation is unquantified. To make this headline empirical claim robust, please report the fraction of 1-minute samples in the 26-30 August 2022 interval that contain zero-count bins within each merged energy range, recompute jlinlin directly from summed counts with zeros retained as a cross-check, and state how much of the ratio in the right column of Fig. 6 survives that recomputation.","section":"Section 2.2 (zero-count paragraph); Section 4.2, Fig. 6"},{"comment":"The observed ratio is matched to the analytic prediction only qualitatively (\"as expected from the dependence of the effective energy on the spectral index\"). Because Eq. (15) is a parameter-free prediction once the time-varying spectral index is known, the right column of Fig. 6 can be directly overlaid with Eq. (15) evaluated at the fitted gamma per sample. Such an overlay would separate the intrinsic binning effect from the zero-count contamination identified above and would convert the factor-of-five statement from an illustration into a quantitative validation; it would also reveal whether the epochs of largest deviation from Eq. (15) coincide with the zero-count epochs, which is the decisive test of the paper's core empirical claim.","section":"Section 4.2, Fig. 6 (right column)"}],"minor_comments":[{"comment":"The abstract reports results \"for two SEP events observed by PSP,\" but Section 4.2 describes \"a period consisting of a series of SEP events\" on 26-30 August 2022 and never isolates or names exactly two events; please reconcile the event count and the event identification between the abstract and the body.","section":"Abstract; Section 4.2"},{"comment":"The factor A * E_{g,i}^{-gamma} in Eq. (7) carries the summation index i outside the sum; to be consistent with Eqs. (8)-(10), it should read A * E_{g,0}^{-gamma}.","section":"Eq. (7)"},{"comment":"Eqs. (9)-(10) and (15) have removable singularities at gamma = 1, and Eq. (10) is also singular at gamma = 0, yet the model grid in Fig. 3 includes gamma = 1; the limiting forms should be stated explicitly, since the expressions as printed evaluate to 0/0 at that spectral index.","section":"Eqs. (9)-(10) and (15); Fig. 3"},{"comment":"The sentence about \"the intensity at the log-centered energy that is independent of the spectral index and remains constant over time\" can be misread as claiming that the binned intensity value is time-invariant; the intended statement is that the energy Eg is independent of gamma and constant in time, and the wording should be adjusted accordingly.","section":"Abstract"},{"comment":"The statement that jloglog represents \"the intensity of a single power law fit at the log-centered energy\" even \"in the presence of nonlinear behavior of the spectrum\" is only exactly true when the fitted line is evaluated at the arithmetic mean of log E in an unweighted fit; for a curved spectrum jloglog is simply the geometric mean of the binned intensities, and the sentence should be rephrased to avoid implying equality with the fitted value.","section":"Section 4.2 (Fig. 5 discussion)"},{"comment":"The description \"Upper and lower uncertainties are propagated individually using the inverse variance method\" is not sufficiently detailed to reproduce the shaded regions in Fig. 6; one sentence on how the asymmetric Gehrels limits enter the inverse-variance weighting, and how the 11-minute smoothing interacts with the propagation, should be added.","section":"Section 4.2 (error propagation)"},{"comment":"Minor language issues include \"as been demonstrated\" (should be \"has been demonstrated\"), \"using jlinlin at his energy\" (should be \"at this energy\"), and \"does not need to be applied\" (should agree in number with \"steps\"); the citation for the dust-affected aperture exclusion is given as Shen et al. 2024 \"in prep.\" and should be updated or replaced with an in-line description.","section":"Sections 2.1, 2.2, and Appendix A"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern does land on reading: Section 2.2 explicitly says both measures are computed on the same modified spectra, so the Fig. 6 comparison is not the pure Eq. (2)-versus-Eq. (3) comparison advertised in the abstract, and the factor-of-five headline is therefore not yet robust. The fix is within the manuscript's scope: a zero-count fraction statistic plus a count-space cross-check for jlinlin would settle it. The analytic core (Eqs. (4)-(15)) is sound, the paper is a good fit for ApJS as a methods reference, and the authors' own limitation statements are honest; I would not reject on the current evidence, but the empirical claim needs the requested quantification before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the paper does the algebra right, and the distinction it draws between count-space (re-binned) and spectral (log-log) averaging is genuinely useful. The empirical factor-of-five claim is real but not clean: the zero-count replacement inflates both quantities, and the paper doesn't say how much of the ratio comes from that.\n\nWhat is new: the combined-bin analytic expressions for jlinlin, jloglog, their ratio, and the effective energy. The effective-energy concept exists in Kronberg & Daly for single bins, and the log-log average is old, but nobody had spelled out the merged-bin case with the ratio and the selection criteria. That is worth having. The derivations in Eqs. (4)-(15) are algebraically consistent, and the presentation is clear. The flowchart in Figure 8 is a practical summary that people will use.\n\nSoft spots, in order of importance. First, the zero-count treatment. Section 2.2 replaces zeros with Gehrels upper limits before applying Eq. (3), and then computes jlinlin and jloglog on the same modified spectra. That means the jlinlin used in Figure 6 is not the pure count-space sum defined in Eq. (2); it includes a positive contribution from bins that had zero counts. The paper admits the replacement's effect on low-count times is 'strongly dependent' on the chosen fraction, but it never quantifies how much of the factor-of-five in Figure 6 comes from this. The analytic ratio, Eq. (15), is unaffected, and for gamma~4-5 and wide energy ranges it already predicts factors above five, so the central claim doesn't depend on the observational comparison. But the observational illustration is muddied, and a referee should ask for a sensitivity run (e.g., excluding zero-count times, or using a different replacement) to show the factor-of-five is robust.\n\nSecond, a consistency issue: the abstract says 'two SEP events,' while Section 4.2 describes a period with a series of SEP events and shows six instrument panels. The count should be reconciled. Third, minor typos ('his energy' in Section 3, 'as been demonstrated' in Section 2.1). Fourth, for a methods paper, releasing the processing code would help; the data are public and the steps are described in enough detail that a motivated reader could reproduce, but code would lower the barrier.\n\nThe paper is a methods contribution, not a discovery paper. It will be useful to anyone who merges SEP energy bins and wants to know what their averaged intensity actually means. The central distinction is correct, and the limitations are stated honestly—the zero-count issue is acknowledged, not hidden. I'd send it to peer review with a request to quantify the zero-count sensitivity and clean up the event count. After that, it should be published.","headline":"Solid methods paper: correct analytic distinction between count-space and spectral averaging, but the empirical factor-of-five is muddied by an unquantified zero-count replacement.","tokens_in":13831,"tokens_out":4483,"would_cite":true,"duration_ms":47630,"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":"The two standard procedures for merging energy bins measure different things: one follows a spectral-index-dependent effective energy, the other a fixed log-centered energy, and they can differ by a factor of five.","keywords":["solar energetic particles","particle intensity","re-binned intensity","spectral binned intensity","effective energy","log-centered energy","power-law spectra","Parker Solar Probe"],"falsifier":"Take a high-count simulated SEP spectrum with a known power-law index $\\gamma$ and known bin geometry, generate Poisson count samples, compute both intensities, and compare each to the true spectrum at the predicted $E_{\\rm eff}$ and $E_g$; if the sample values do not reproduce $j(E_{\\rm eff})$ and $j(E_g)$ within counting error, the point-value and single-power-law assumptions fail. On real PSP/IS☉IS data, one could check whether the observed ratio of the two intensities tracks Eq. (15) when $\\gamma$ is independently fitted at each time step.","tokens_in":12819,"feed_emoji":"☀️","tokens_out":11634,"duration_ms":87896,"temperature":0.7,"pith_summary":"This paper asks a deceptively simple question: when narrow detector energy bins are merged into one wider bin, what is the correct single intensity to report? The authors show that the two standard answers are not interchangeable. The re-binned intensity, a linear average of intensity in energy, is actually the intensity at an effective energy that moves with the time-varying spectral index; the spectral-binned intensity, an average of log intensity in log energy, is the intensity at the fixed log-centered energy of the merged bin. For power-law SEP spectra the two values can differ by up to a factor of five, matching what the authors find in Parker Solar Probe IS☉IS proton data from two solar energetic particle events. The practical criterion that follows is simple: use the re-binned intensity when the question is about counts in a wide energy range, and the spectral-binned intensity when the question is about the intensity at a well-defined, time-invariant energy.","feed_headline":"Two ways to average solar particle intensities can differ by up to 5x","feed_subtitle":"Re-binning tracks counts at a moving effective energy; spectral binning reads the fixed log-centered energy.","key_machinery":"The machinery is the generalized averaging identity of Eq. (1), which separates the transformation of the energy axis ($\\hat{X}$) from the transformation of the intensity ($\\hat{Y}$) before averaging; setting both to identity gives $\\overline{j}_{\\rm linlin}$, and setting both to logarithms gives $\\overline{j}_{\\rm loglog}$. Under the two assumptions named above, the paper derives closed forms for both averages, for the effective energy $E_{\\rm eff}$ (Eq. 10), for the log-centered energy $E_g$ (Eq. 14), and for the ratio $\\overline{j}_{\\rm loglog}/\\overline{j}_{\\rm linlin}$ (Eq. 15). The ratio formula is what carries the argument: it turns the difference between the two measures into a function of three controllable parameters, and it is the expression used to interpret the observed factor-of-five gap in PSP/IS☉IS data.","core_discovery":"The paper's central claim is that $\\overline{j}_{\\rm linlin}$ and $\\overline{j}_{\\rm loglog}$ are different physical measures, not two estimators of the same averaged intensity. Treating each original bin's intensity as a point value at its log-centered energy and assuming a single power law $j(E)=A E^{-\\gamma}$ across the merged range, the re-binned intensity equals $j(E_{\\rm eff})$ with $E_{\\rm eff}$ given in Eq. (10), while the spectral-binned intensity equals $j(E_g)$ with $E_g=\\sqrt{E_0E_N}$, a purely geometric quantity. Their ratio, Eq. (15), depends only on the spectral index $\\gamma$, the range ratio $E_N/E_0$, and the original logarithmic bin width $\\Delta\\log E$; it equals one only for $\\gamma=0$ and $\\gamma=2$. In PSP/IS☉IS proton measurements of two August 2022 SEP events, the re-binned intensity is consistently larger than the spectral-binned intensity, up to a factor of about five, even though the two time series are strongly correlated. The paper also shows that the zero-count treatment for $\\overline{j}_{\\rm loglog}$ introduces a bias that grows as counting statistics drop.","pith_inferences":["A natural step the paper leaves implicit is to derive the analogous ratio for differential energy flux $E\\,j$, since the paper only notes that the loglin average is the count-space operation for that quantity; the same power-law machinery would give a corresponding correction factor.","The zero-count replacement by the Gehrels upper limit is one reasonable convention; a Bayesian estimator that treats zero counts as censored Poisson samples could reduce the low-count bias of $\\overline{j}_{\\rm loglog}$ and would be a testable improvement.","The same derivation could be repeated for non-uniform or overlapping energy bins, which the paper explicitly excludes; the factor-of-five bound may differ, and quantifying it would extend the method to instruments with irregular bin spacing.","The paper's energy ranges from different IS☉IS instruments (EPI-Lo, LET, HET) are separate; placing them on a common $\\overline{j}_{\\rm loglog}$ grid would make cross-instrument spectra directly comparable at fixed energies, an application the paper does not pursue."],"forward_implications":["Any study that labels a merged-bin intensity by the bin's log-centered energy will overstate the intensity at that energy by up to a factor of several for typical SEP spectral indices, unless it uses the spectral-binned intensity.","Because the two time series are strongly correlated, studies of relative intensity evolution are robust to either choice, while studies comparing magnitudes across energy ranges or across times are not.","The spectral-binned intensity keeps a time-invariant energy label, making it the appropriate measure for spectral evolution and for comparisons of spectra at a fixed energy; the re-binned intensity's energy label shifts as the spectral index changes, shortening the time scale over which it can be compared meaningfully.","Equation (15) gives a ready conversion factor: with a fitted spectral index and known bin geometry, one can translate between the two measures or decide when the difference is negligible.","Using one measure instead of the other can change the inferred time at which a spectrum transitions from falling to rising, which matters for identifying spectral roll-ups and velocity dispersion effects in SEP events."],"supporting_citations":[{"why":"supplies the earlier estimate of the error in assigning a single-bin intensity to its log-centered energy, which the paper extends to merged bins.","marker":"Kronberg & Daly 2013"},{"why":"provides the upper-limit uncertainties used to replace zero-count measurements before logarithmically averaging.","marker":"Gehrels 1986"},{"why":"describes the IS☉IS instrument suite whose proton data are used for the observational comparison.","marker":"McComas et al. 2016"},{"why":"describes the Parker Solar Probe mission that supplied the SEP observations.","marker":"Fox et al. 2016"},{"why":"documents the observed PSP spectral-index range of about 0 to 7 that the modeled ratios are meant to cover.","marker":"Mitchell et al. 2023"},{"why":"identifies the 26–30 August 2022 event period with three-stage particle acceleration that the paper analyzes.","marker":"Chen & Li 2024"}],"fun_headline_variants":["Re-binning vs spectral binning: SEP intensity differs up to 5x","SEP intensity binning: two methods, up to 5x difference","Re-binned vs spectral binned SEP intensities: up to 5x apart","SEP intensity: re-binning vs spectral binning differ by 5x","Parker Solar Probe data: binning choice changes SEP intensity 5x"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire analytic structure treats each original narrow bin as if its measured intensity sits exactly at the bin's central energy on a log scale, and it assumes the spectrum across the merged bin is a single power law; if real SEP spectra are strongly curved or the bins are wide, the ratio in Eq. (15) is only approximate.","fun_headline_variants_meta":{"raw":{"variants":["Re-binning vs spectral binning: SEP intensity differs up to 5x","SEP intensity binning: two methods, up to 5x difference","Re-binned vs spectral binned SEP intensities: up to 5x apart","SEP intensity: re-binning vs spectral binning differ by 5x","Parker Solar Probe data: binning choice changes SEP intensity 5x"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0011,"raw_usage":{"total_tokens":4670,"prompt_tokens":1108,"completion_tokens":3562,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":724,"completion_tokens_details":{"reasoning_tokens":3458}},"tokens_in":724,"tokens_out":3562,"duration_ms":20073,"temperature":1.0,"reasoning_tokens":3458,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T14:46:38.381282+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a high-count simulated SEP spectrum with a known power-law index $\\gamma$ and known bin geometry, generate Poisson count samples, compute both intensities, and compare each to the true spectrum at the predicted $E_{\\rm eff}$ and $E_g$; if the sample values do not reproduce $j(E_{\\rm eff})$ and $j(E_g)$ within counting error, the point-value and single-power-law assumptions fail. On real PSP/IS☉IS data, one could check whether the observed ratio of the two intensities tracks Eq. (15) when $\\gamma$ is independently fitted at each time step.","supporting_citations":[{"cited_title":"A., & Daly, P","cited_arxiv_id":null,"evidence_quote":"supplies the earlier estimate of the error in assigning a single-bin intensity to its log-centered energy, which the paper extends to merged bins."},{"cited_title":"G., Cohen, C","cited_arxiv_id":null,"evidence_quote":"documents the observed PSP spectral-index range of about 0 to 7 that the modeled ratios are meant to cover."},{"cited_title":"2024, ApJL, 967, L33, doi: 10.3847/2041-8213/ad4a79","cited_arxiv_id":null,"evidence_quote":"identifies the 26–30 August 2022 event period with three-stage particle acceleration that the paper analyzes."}],"review_version":1}