{"id":"ef2be8ec-f29f-4054-87a9-f062a57f37f0","arxiv_id":"2505.12881","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"WHALES, a two-pass subwaveform retracker with SWH-dependent weights, improves coastal significant wave height retrieval from satellite altimetry, increasing valid records by 30% at 5 km from the coast.","lead":"This paper describes WHALES, a new algorithm that estimates wave heights from satellite radar altimeter signals by fitting only part of the returned echo and weighting the fit by signal-dependent uncertainties. It reports that the algorithm, now used in ESA's Sea State Climate Change Initiative database, cuts measurement noise and increases the number of usable coastal measurements by 30%.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section 3.1's 30% data gain and 70% noise reduction compare Sea State CCI v3 against v1, but v3 also changes 20-Hz compression, editing, and cross-calibration; the WHALES-only contribution is never isolated, so the abstract's 'after applying the retracker' attribution is not established.","rationale":"Reader's formal weakest assumption is the Monte Carlo transferability of Brown+speckle simulations. That is a real robustness caveat, but it is partly mitigated by the round-robin comparison, the clean Section 3.2 retracker comparison, and the algorithm's documented ability to handle non-Brown cases. The more load-bearing threat to the central claim is internal to the validation design: the Section 3.1 numbers are database-version differences, not retracker effects. The same section explicitly names unrelated processing changes (negative-value treatment in 20-Hz-to-1-Hz compression) and a WHALES-specific Saral artifact, yet the abstract converts the net v3-v1 change into 'after applying the retracker discussed here.' A single ablation—same v3 pipeline with and without WHALES—would settle the attribution. This does not undermine the paper's genuine contributions: the algorithm is openly coded, the method description is detailed, and Section 3.2's same-data comparison of WHALES vs MLE4 vs Adaptive with common quality flags is a valid test that supports WHALES in the coastal zone. The conditional verdict should stand pending the ablation, so no verdict change is needed. My disagreement with the reader's stated weakest assumption is only about which weakness is most load-bearing; the overall conditionality is unchanged.","tokens_in":14675,"tokens_out":5309,"duration_ms":52244,"concrete_test":"On the same 30-day multi-mission sample used in Section 3.1, run the complete v3 processing chain (leading-edge detection, 20-Hz averaging, editing, 1-Hz compression, cross-calibration) twice, with WHALES and with the v1 baseline retracker (or MLE-4) substituted as the only change. Recompute valid-data counts versus distance-to-coast and SWH RMS versus SWH and versus coast distance. If the differences between the two runs closely reproduce the v3-v1 differences, the retracker is the causal driver; if the counts/RMS differences shrink substantially, the headline claims must be re-attributed to the v3 pipeline rather than WHALES alone.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.1 bases the two headline numbers on a version1-versus-version3 comparison of the full Sea State CCI database. The preceding paragraph (Section 3) states that v3 data were 'averaged at 1 Hz and cross-calibrated following the procedure described in Piolle and Dodet (2025)', and Section 3.1 itself attributes the different low-SWH behavior of SWH RMS to 'the treatment of negative values in the 20Hz to 1Hz compression scheme implemented in version3', and the Saral coastal degradation to an 'artifact of the WHALES retracker'. Hence v3 differs from v1 in retracker, editing/quality flags, 20-Hz-to-1-Hz compression, and cross-calibration simultaneously. The 30% valid-record increase at 5 km (Figure 5) and the up-to-70% SWH RMS reduction (Figure 4) are therefore not demonstrably caused by WHALES, and the abstract's 'after applying the retracker discussed here' overstates causal attribution. Section 3.2 is a cleaner comparison of retrackers on identical Jason-3 data with common flags and independently supports coastal skill, but it does not quantify the 30%/70% claims. The central claim requires an ablation that holds the rest of the processing chain fixed.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript describes WHALES, a two-pass adaptive subwaveform retracker for low-resolution-mode (LRM) altimetry that estimates significant wave height (SWH). The first pass is a leading-edge-only least-squares fit; the stopgate for the second pass is set from Eq. (3) as a linear function of the first-pass SWH, and the second-pass fit uses SWH-dependent weights derived from Monte Carlo simulations of Brown-model waveforms with Rayleigh speckle. The paper evaluates WHALES in two ways: a comparison of the ESA Sea State CCI version 3 database (produced with WHALES) against version 1, reporting a 30% increase in valid records at 5 km from the coast and up to 70% reduction in SWH RMS noise; and a comparison of Jason-3 SWH from WHALES, MLE-4, and the Adaptive retracker against 122 coastal buoys, reporting that WHALES achieves the best scatter index within 20 km of the coast. The paper also analyzes the effective along-track weighting kernel J_H of the cost function and discusses three waveform pathologies (sigma0 blooms, phenomenal seas, icebergs) with possible refinements.","tokens_in":15019,"tokens_out":5960,"duration_ms":58052,"significance":"If the headline results are attributable to WHALES, the paper documents a practical improvement in coastal SWH retrieval across multiple LRM missions, with potential value for the Sea State CCI climate record. The manuscript has real strengths: the retracker source code is openly available; the Sea State CCI v1 and v3 data are publicly accessible; the buoy comparison in Section 3.2 is an external, independent check on WHALES versus two operational retrackers; and the Section 4 kernel analysis connects design choices to measurement physics rather than relying solely on aggregate skill scores. These assets make the paper suitable for RSE if the attribution of the quantitative headline claims is strengthened.","major_comments":[{"comment":"The abstract states that valid data records increased by 30% at 5 km from the coast 'after applying the retracker discussed here', and Section 3.1 reports up to 70% SWH RMS reduction between version 3 and version 1. This comparison does not isolate the retracker: version 3 differs from version 1 simultaneously in the retracker, the 20-Hz-to-1-Hz compression scheme (including the flagging of negative SWH values), editing/quality flags, and cross-calibration, as the manuscript itself notes when it attributes the low-SWH behavior of the RMS curves to 'the treatment of negative values in the 20Hz to 1Hz compression scheme implemented in version3' and the Saral coastal degradation to an 'artifact of the WHALES retracker'. The 30% and 70% figures are therefore not demonstrably caused by WHALES, and the abstract overstates attribution. Please add an ablation that holds the rest of the processing chain fixed, for example by retracking the version 1 waveforms with WHALES while keeping the version 1 editing and compression, or by comparing version 3 processed with WHALES against version 3 processed with ALES or another retracker, and report the resulting data-count and RMS differences.","section":"Section 3.1, Figures 4-5, Abstract"},{"comment":"The stopgate coefficients in Eq. (3) and the SWH-dependent weight table are fitted to Monte Carlo simulations of Brown-model waveforms with multiplicative Rayleigh speckle, with the tolerance chosen as a 2 cm RMSE at 20 Hz. Consequently the improvement in SWH RMSE shown in Figures 2 and 3 is partly built into the design and does not by itself demonstrate that WHALES is near-optimal on real waveforms, especially coastal ones with non-Brown leading edges. The buoy validation is independent but tests SWH accuracy, not the along-track noise reduction claimed in Section 3.1. I suggest an out-of-sample or real-waveform check, for example comparing the RMSE of the fitted stopgate and weights against one or two alternative coefficient sets on the Jason-2 and Jason-3 waveforms used in Sections 3.2 and 5, or validating the noise reduction on 20-Hz real data with the editing and compression chain held fixed.","section":"Section 2.2-2.3, Figures 2-3"},{"comment":"The claim that WHALES gives the best scatter index within 20 km from the coast rests on a single median over an unspecified number of buoys in that distance bin, and the SI values are strongly correlated with the mean buoy SWH (used as the color scale), which is itself a function of distance to the coast. Please report the number of matchups per distance bin, the uncertainty of the median SI (e.g., a bootstrap or rank-sum test), and, if possible, a distance-binned or SWH-stratified breakdown so that the reader can assess whether the coastal advantage is statistically robust and not driven by one or two sheltered buoys.","section":"Section 3.2, Figure 6"}],"minor_comments":[{"comment":"Please state the units of SWH and 'Tracking point' explicitly; as written, a reader cannot tell whether 'Tracking point' is in gates or range units, and whether Eq. (3) is intended for all LRM missions or only for Jason-3.","section":"Section 2.1, Eq. (3)"},{"comment":"The sentence '10,000 waveforms were simulated ... and then averaged to create a simulated high-rate waveform' is ambiguous: each simulated waveform already represents a high-rate echo, and the subsequent RMSE computation would be inconsistent with a single averaged waveform; please rephrase to describe the simulation and retracking procedure more precisely.","section":"Section 2.2"},{"comment":"The caption says 'unweighted (blue) and weighted (right)', but the panel labels and body text indicate 'weighted (red)'; please correct this.","section":"Figure 3 caption"},{"comment":"The text 'altimeters are in fact the only instruments that routinely report SHW values over 15 m' contains a typo: 'SHW' should be 'SWH'.","section":"Section 5.2"},{"comment":"The text refers to 'grey bars' for the number of valid data, but Figure 5 as printed appears to contain only RMS curves; please ensure the data-count overlay is visible or move it to a separate panel.","section":"Section 3.1, Figure 5"},{"comment":"Equation (1) uses H_s, while the paper later introduces SWH as a distinct quantity; please align the notation in the introduction to avoid confusing the theoretical wave height with the retracker-derived estimate.","section":"Section 1, Eq. (1)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a good fit for RSE, and the open code, public data, and external buoy check are notable strengths. The main risk is the attribution of the 30% and 70% headline numbers to WHALES, which needs an ablation before acceptance. No concerns about the citation pattern or novelty disclosure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Marcello and colleagues have written a careful description of WHALES, the retracker behind Sea State CCI v3. The algorithm itself is not new—it was benchmarked in Schlembach et al. (2020)—but this paper gives the first consolidated account of the two-pass design, the SWH-dependent weights, and the resulting effective footprint kernel. That Section 4 analysis, building on De Carlo and Ardhuin, is genuinely new and useful. The code is open, and the Jason-3 buoy comparison in Section 3.2 is a clean external check with common quality flags. On that evidence, WHALES does appear to be a strong coastal retracker, beating MLE-4 and Adaptive within 20 km of the coast.\n\nThe soft spot is the headline. The 30% valid-record increase and 70% noise reduction come from comparing Sea State CCI v1 and v3. As the paper itself notes, v3 also changes the 20-Hz-to-1-Hz compression, editing thresholds, and cross-calibration. So the abstract phrase 'after applying the retracker discussed here' attributes the gain to WHALES alone, but the numbers are product-level differences, not an ablation. The stress-test note is right. The Saral degradation is also confessed to be a WHALES artifact, which is honest but reinforces that the retracker has mission-specific issues. The Monte Carlo fit for the stopgate and weights is a reasonable design step, but it does include a tolerance choice that partly guarantees the synthetic improvement; that is not fatal because the buoy comparison is independent.\n\nAltogether, this is a useful methods paper for a community that needs to know exactly what WHALES does. It deserves peer review, but the authors should be asked to either rerun the comparison holding the rest of the chain fixed, or soften the causal wording. If that is fixed, I would be comfortable citing it.","headline":"Solid retracker description with a clean coastal validation, but the 30%/70% headline numbers mix the retracker's effect with unrelated database changes.","tokens_in":15540,"tokens_out":1511,"would_cite":true,"duration_ms":15166,"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":"WHALES, a two-pass weighted retracker for radar altimetry, retrieves significant wave height with up to 70% less along-track noise and 30% more valid coastal records than the previous Sea State CCI processing.","keywords":["satellite altimetry","significant wave height","retracking","coastal altimetry","weighted least squares","Brown model","subwaveform","Sea State CCI"],"falsifier":"Compute the actual standard deviation of WHALES fitting residuals at each range gate from a large set of real Jason-3 20 Hz waveforms, bin by SWH, and compare with the Monte Carlo table used for the weights; a large mismatch would show the weighting is miscalibrated for real echoes. A second decisive test is to rerun WHALES with unit weights on the same coastal and open-ocean datasets: if the weighted version does not consistently lower 20 Hz SWH RMS and improve scatter index against buoys, the simulated-noise premise fails on real waveforms.","tokens_in":14507,"feed_emoji":"🌊","tokens_out":9306,"duration_ms":85242,"temperature":0.7,"pith_summary":"The paper describes and validates WHALES, a two-pass retracker that estimates significant wave height from conventional radar altimetry waveforms. The central claim is that by fitting only a subwaveform built around the leading edge and by weighting the fit residuals inversely to their Monte-Carlo-derived uncertainty, WHALES produces lower-noise SWH estimates than previous processing while retrieving more valid measurements in coastal waters. Applied to the ESA Sea State CCI v3 database, it increases valid records by 30% at 5 km from the coast and reduces along-track SWH noise by up to 70% relative to version 1. Against coastal buoys, WHALES gives the lowest scatter index within 20 km of the coast among the retrackers compared. A sympathetic reader would take this as evidence that the leading-edge subwaveform plus SWH-dependent weighting is a better default for coastal sea-state products.","feed_headline":"A two-pass weighted retracker gains 30% more coastal wave-height data","feed_subtitle":"Applied to the Sea State CCI v3 archive, WHALES lowers along-track significant wave height noise by up to 70 percent.","key_machinery":"The load-bearing object is a weighted least-squares cost function $F(\\theta) = \\sum_i w_i (y_i - \\hat{y}_i)^2$ used with the Brown waveform, where the weights $w_i$ are the inverse of the Monte-Carlo-estimated standard deviation of the fitting residuals at each range gate for a given SWH. The SWH-dependent stopgate relation controls how much trailing edge is included, and the two-pass procedure makes the subwaveform and weights adaptive to each echo. This combination carries the argument because it reduces speckle-noise influence without importing trailing-edge contamination from land or bright targets; it also changes the effective averaging kernel of the SWH estimate, shortening the zero-crossing scale compared with ordinary least squares.","core_discovery":"The discovery is that the statistical weighting of residuals, derived from the standard deviation of residuals over 10,000 simulated Brown waveforms per SWH level, makes a subwaveform retracker simultaneously less noisy and more usable near land. WHALES first fits the leading edge with unit weights, uses the resulting SWH to extend the subwaveform according to the stopgate relation, then refits with weights equal to the inverse of the Monte Carlo residual standard deviation. In simulations the weighted fit improves SWH RMSE by a factor of 1.5 to 2 across SWH values above 1 m. In the Sea State CCI v3 database the along-track SWH RMS falls by up to 70% compared to v1 and the number of valid records rises 30% at 5 km from the coast; in coastal-buoy comparisons within 20 km, WHALES has the lowest scatter index of the tested retrackers.","pith_inferences":["An implication the paper leaves implicit is that the 30% valid-record gain may partly reflect v3's edited compression scheme rather than WHALES alone; a controlled rerun with identical editing would separate the retracker's contribution.","Because the weights are tuned on Brown-model waveforms, the noise advantage should transfer most faithfully to open-ocean echoes; for strongly non-Brown coastal echoes the gain may shrink, consistent with the paper's own report that Saral 40 Hz data did not improve in the 0-5 km band.","One testable extension is to use the same inverse-variance weighting for the epoch parameter, since leading-edge residuals carry the same speckle statistics; the paper does not report that extension.","Another consequence not drawn by the authors is that the stopgate tolerance acts as a tunable resolution knob: lowering the allowed RMSE difference would trade more noise for a shorter effective footprint, and raising it would do the opposite."],"forward_implications":["The Sea State CCI v3 LRM record contains about 30% more valid SWH records at 5 km from the coast and 15% more at 10 km than version 1.","Along-track SWH noise, measured by the 20 Hz SWH RMS, drops by up to 70% over the 0-10 m SWH range and becomes more similar across the six missions analyzed.","In coastal matchups with 122 buoys, WHALES achieves the lowest scatter index among the three retrackers compared within 20 km of the coast.","Because the weighting changes the effective footprint kernel, WHALES resolves shorter along-track scales of wave-height variability than a plain three-parameter least-squares fit, which matters for wave-group and coastal studies.","The same two-pass weighted scheme can be ported to other waveform models and to missions not yet in the archive, extending three decades of consistent SWH records."],"supporting_citations":[{"why":"This reference provides the theoretical Brown waveform that WHALES fits to each measured echo.","marker":"Brown (1977)"},{"why":"This reference supplies the Brown-model formulation and initial conditions used in the WHALES fitting procedure.","marker":"Passaro et al. (2014a)"},{"why":"This reference introduces the ALES subwaveform strategy from which WHALES inherits its leading-edge focus.","marker":"Passaro et al. (2014b)"},{"why":"This reference provides the Nelder-Mead simplex estimator that WHALES uses for iterative convergence.","marker":"Nelder and Mead (1965)"},{"why":"This reference supports the choice of inverse-variance statistical weighting when fitting uncertainties vary across data points.","marker":"Wolberg (2006)"},{"why":"This reference defines the footprint kernel and the effective-resolution analysis used in Section 4.","marker":"De Carlo and Ardhuin (2024)"},{"why":"This reference reports the round-robin assessment in which WHALES performed best for coastal significant wave height.","marker":"Schlembach et al. (2020)"},{"why":"This reference describes the Sea State CCI version 1 database against which version 3 is compared.","marker":"Dodet et al. (2020)"},{"why":"This reference defines the Adaptive retracker used in the coastal-buoy comparison.","marker":"Tourain et al. (2021)"},{"why":"This reference defines the MLE-4 retracker used in the coastal-buoy comparison.","marker":"Thibaut et al. (2010)"}],"fun_headline_variants":["Coastal wave data jumps 30% with WHALES retracker","WHALES cuts coastal wave-height noise, boosts data 30%","Smart weighting in retracker yields 30% more coastal data","Retracker's two-pass trick recovers 30% more coast waves","WHALES: 30% more valid coastal wave records"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes that the error statistics of real altimeter waveforms, including contaminated coastal ones, match the Brown-model-plus-Rayleigh-speckle simulations used to set the stopgate coefficients and the SWH-dependent weights, so that the two-pass adaptive fit stays near-optimal.","fun_headline_variants_meta":{"raw":{"variants":["Coastal wave data jumps 30% with WHALES retracker","WHALES cuts coastal wave-height noise, boosts data 30%","Smart weighting in retracker yields 30% more coastal data","Retracker's two-pass trick recovers 30% more coast waves","WHALES: 30% more valid coastal wave records"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000166,"raw_usage":{"total_tokens":1217,"prompt_tokens":873,"completion_tokens":344,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":489,"completion_tokens_details":{"reasoning_tokens":255}},"tokens_in":489,"tokens_out":344,"duration_ms":3844,"temperature":1.0,"reasoning_tokens":255,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:24:20.209909+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the actual standard deviation of WHALES fitting residuals at each range gate from a large set of real Jason-3 20 Hz waveforms, bin by SWH, and compare with the Monte Carlo table used for the weights; a large mismatch would show the weighting is miscalibrated for real echoes. A second decisive test is to rerun WHALES with unit weights on the same coastal and open-ocean datasets: if the weighted version does not consistently lower 20 Hz SWH RMS and improve scatter index against buoys, the simulated-noise premise fails on real waveforms.","supporting_citations":[{"cited_title":", year 1977","cited_arxiv_id":null,"evidence_quote":"This reference provides the theoretical Brown waveform that WHALES fits to each measured echo."},{"cited_title":", author Mead, R","cited_arxiv_id":null,"evidence_quote":"This reference provides the Nelder-Mead simplex estimator that WHALES uses for iterative convergence."},{"cited_title":", year 2006","cited_arxiv_id":null,"evidence_quote":"This reference supports the choice of inverse-variance statistical weighting when fitting uncertainties vary across data points."},{"cited_title":", author Ardhuin, F","cited_arxiv_id":null,"evidence_quote":"This reference defines the footprint kernel and the effective-resolution analysis used in Section 4."},{"cited_title":", author Passaro, M","cited_arxiv_id":null,"evidence_quote":"This reference reports the round-robin assessment in which WHALES performed best for coastal significant wave height."},{"cited_title":", author Piolle, J.F","cited_arxiv_id":null,"evidence_quote":"This reference describes the Sea State CCI version 1 database against which version 3 is compared."},{"cited_title":", author Poisson, J","cited_arxiv_id":null,"evidence_quote":"This reference defines the MLE-4 retracker used in the coastal-buoy comparison."}],"review_version":1}