. . . . "2026-05-21"^^ . . "Coastal-ROM trade-off reproduction outcome (Loveland et al. 2024)" . "Validated for hindcasting under the WSE-only-interest conditioning Loveland's §6 Conclusions states and the Quote carries verbatim. The conditional trade-off — reduced-order SWAN source terms (Gen1, Gen2) save computation versus the third-generation ST6 Gen3 package without compromising water-surface-elevation accuracy at NOAA gauges when WSE is of primary interest, with the wave-statistics caveat near hurricane tracks Loveland themselves quantify — reproduces from the public data with one per-storm divergence and one framing distinction worth surfacing.\n\nWall-clock cost: Loveland's §5.1 prose summary (\"around 1.5 times longer\") and §6 framing (\"about a 40 percent increase\") cover Hurricane Ike cleanly (Gen3/Gen1 = 1.44×, Gen3/Gen2 = 1.52×) but understate Hurricane Ida by approximately 20 percentage points relative to Loveland's own Table 4 (Gen3/Gen1 = 1.70×, Gen3/Gen2 = 1.76×). The Ida cost overshoot is in the paper's data but not in either prose summary; the replication therefore reproduces the cost direction unequivocally and the magnitude only at the per-storm granularity of Table 4.\n\nWSE-RMSE invariance: reproduces in full from Tables 5–6 (maximum with-wave Δ = 0.007 m for Ike, 0.001 m for Ida — within observational noise).\n\nWave-statistics sensitivity caveat: faithfully reflected by Loveland's own Tables 5–7 and Fig. 9 (Tp RMSE Ida Gen3 = 9.96 s versus Gen1 = 4.87 s, a factor-of-2 divergence; spatial Hs differences of several metres near hurricane tracks).\n\nScope of the Validated label, made explicit so that a CiTO-only citer cannot miss the conditioning: the verdict applies (a) to the hindcasting regime Loveland tested, not to operational forecasting where wind-field uncertainty dominates the error budget — a second conditioning Loveland states in §6 that the chosen Quote does not carry; (b) to the consistency of Loveland's published Δ values with the public observational baseline, not to model-level reproducibility of the ADCIRC+SWAN runs themselves, which were not re-executed." . . "Source files: results/headline_comparison.csv, results/tradeoffs.csv, results/obs_wse_peaks.csv, results/obs_wave_peaks.csv. Paper baselines transcribed verbatim from Loveland Tables 4-7.\n\nFigures supporting this verdict:\n- figures/main_result.png — storm-averaged trade-off summary (compute cost left panel; WSE-RMSE + Hs-RMSE right panel).\n- figures/fig10_11_obs_equivalent.png — observational side of Loveland Figs 10/11. Per-gauge observed peak WSE for both storms (14 Ike, 13 Ida) overlaid with Loveland's storm-averaged Gen1/Gen2/Gen3 RMSE bands. The three bands are visually indistinguishable on the same axis as the bars; the inter-config Δ (≤ 0.007 m Ike, ≤ 0.001 m Ida) is invisible at the scale of the inter-gauge peak variability (0.47-3.81 m). This is the WSE-invariance finding made visual.\n- figures/fig5_6_obs_equivalent.png — observational side of Loveland Figs 5/6. Per-buoy observed peak Hs and Tp for both storms with Loveland's storm-averaged Gen1/Gen2/Gen3 RMSE bands. The Ida Tp Gen3 = 9.96 s band is visibly separated from the Gen1/Gen2 bands at ~5 s — the factor-of-2 wave-statistics divergence that Loveland's text and the chain's Quote already carve out as a sensitivity caveat.\n- figures/fig7_8_buoy_winds_equivalent.png — observational side of Loveland Figs 7/8. NDBC buoy wind-speed time series for the 4 Ike-track buoys (42001, 42002, 42019, 42035 — all retrievable) and the 4 Ida-track buoys (42040, 42039 retrievable; 42001 retired 2021-07-25 before Ida; 42007 not in the 2021 NDBC stdmet archive). ADCIRC wind forcing not reproduced because the model output is not retrievable. Loveland's own Fig 8 panels visibly show ADCIRC's Ida winds diverging from the buoy series near landfall — direct visual evidence for the wind-field-accuracy conditioning Loveland states in §6 and that the Outcome's Conclusion carves out.\n\nRun-time (1064 Frontera cores), per storm, with Loveland's prose-summary framings as reference points:\n- Loveland §5.1: \"around 1.5 times longer\" (50 percent increase). Loveland §6: \"about a 40 percent increase\".\n- Ike: No SWAN 1.094 h; Gen1 2.837 h; Gen2 2.688 h; Gen3 4.086 h. Gen3/Gen1 = 1.44× (+44 percent); Gen3/Gen2 = 1.52× (+52 percent). Sits inside §5.1's \"1.5×\" framing; close to §6's \"40 percent\".\n- Ida: No SWAN 0.647 h; Gen1 1.881 h; Gen2 1.818 h; Gen3 3.206 h. Gen3/Gen1 = 1.70× (+70 percent); Gen3/Gen2 = 1.76× (+76 percent). Exceeds §5.1's \"1.5×\" framing by approximately 20 percentage points; exceeds §6's \"40 percent\" framing by approximately 30 percentage points. The Ida overshoot is in Loveland's own Table 4 but is not surfaced in either §5.1 or §6 prose.\n\nWSE RMSE at NOAA gauges (m):\n- Ike: No SWAN 0.210; Gen1 0.190; Gen2 0.196; Gen3 0.197. Max with-wave Δ = 0.007 m.\n- Ida: No SWAN 0.285; Gen1 0.282; Gen2 0.281; Gen3 0.282. Max with-wave Δ = 0.001 m.\n\n\nHs RMSE at NDBC buoys (m):\n- Ike: Gen1 0.900; Gen2 1.033; Gen3 0.802 — Gen3 best.\n- Ida: Gen1 0.813; Gen2 0.883; Gen3 0.834 — Gen1 best.\n\nTp RMSE at NDBC buoys (s):\n- Ike: Gen1 2.822; Gen2 2.541; Gen3 2.413.\n- Ida: Gen1 4.865; Gen2 5.068; Gen3 9.960 — Gen3 a factor of 2 worse than Gen1.\n\nMean wave direction RMSE at NDBC buoys (degrees):\n- Ike: Gen1 43.238; Gen2 40.392; Gen3 37.076.\n- Ida: Gen1 62.380; Gen2 60.229; Gen3 67.730.\n\nObservational baseline (independently downloaded, NOT a re-derivation of the model-vs-obs RMSE): 14/14 NOAA CO-OPS gauges for Ike (5–14 September 2008); 13/13 for Ida (26 August – 4 September 2021); 9/10 NDBC buoys (42007 absent from NDBC historical archive for 2021). Storm peaks: Galveston Pier 21 = 3.20 m (Ike, 13 September 2008); Grand Isle = 1.65 m (Ida, 29 August 2021). These confirm that the observational dataset Loveland modelled against is publicly retrievable and consistent with NHC historical records. The model-vs-obs RMSE Δ values themselves were transcribed from Loveland's Tables 5–7, not independently re-computed (the model outputs are not retrievable — see Limitations item 1)." . "Three caveats limit how this Outcome should be cited:\n\n1. Stat-level reproduction, not model re-run. This replication does not independently re-run ADCIRC+SWAN at the paper's compute scale (1064 Frontera cores, 3-6 million-element meshes for Ike and Ida respectively). The model output side of the comparison is transcribed verbatim from Loveland et al.'s Tables 4-7; the observational side is independently downloaded from NOAA CO-OPS and NDBC. The Validated verdict therefore confirms consistency of Loveland's published numbers with the public observational baseline, not the reproducibility of the model run itself. Loveland's DesignSafe deposit (DOI 10.17603/DS2-7HBT-EF65) is access-restricted (HTTP 401 on the public listing API) and was not retrievable during this study.\n\n2. Wind-field-accuracy conditioning, dropped by the Quote. Loveland's §6 Conclusions adds a second conditioning that the chosen Quote does not carry: 'if the meteorological forcing is not sufficiently accurate, which is common in forecasting scenarios, then the additional computational cost associated with the detailed Gen3 source terms may not improve accuracy of the model.' The Validated verdict therefore covers the hindcasting regime (Loveland used Oceanweather Inc. validated hindcasts for Ike and HURDAT2-derived GAHM winds for Ida) and should not be cited as supporting reduced-order source terms in operational forecasting where wind-field uncertainty dominates the error budget. This condition lives in the Replication Study's Methodology / Scope field as well.\n\n3. Wave-statistics sensitivity is context, not independent corroboration. The second sentence of the Quote (large Hs sensitivities near hurricane tracks) is Loveland's own quantified finding — Tables 5-7 and Fig. 9, including the Tp RMSE Ida factor-of-2 divergence noted above. The replication confirms these published numbers but does not surface new wave-statistics evidence. Outcome-level citers should not use this chain as independent corroboration of the wave-statistics caveat; that caveat traces to Loveland's own data.\n\n4. The independently downloaded NOAA observations are a sanity check on Loveland's source data, not an independent re-derivation of the model-vs-obs RMSE. What this replication independently verified is that (a) the NOAA gauge/buoy time series Loveland used are publicly retrievable end-to-end (14/14 Ike + 13/13 Ida WSE gauges, 9/10 NDBC wave buoys) and (b) the storm peaks at those gauges are consistent with NHC historical records. The model-fidelity Δ values (Tables 5–7) were not independently re-computed — they were transcribed from Loveland's published tables. A downstream citer should not infer model-level verification of the model output from this Outcome's evidence trail.\n\n5. Wall-clock-cost framing granularity. Loveland's §5.1 prose summary (\"around 1.5 times longer\") and §6 framing (\"about a 40 percent increase\") cover Ike but understate Ida by approximately 20 percentage points relative to Loveland's own Table 4. The verdict on the run-time direction is unequivocally supported; the magnitude is supported at per-storm granularity but not at the prose-summary level. This is visible in Loveland's published data; the replication surfaces but does not author it.\n\nThe replication's main figure (figures/main_result.png) summarises the trade-off as 'save 30-40 percent wall-clock at less than 5 percent added WSE-RMSE cost', which is faithful to Loveland's framing for Ike but is the lower bound of the actual per-storm range. A revised summary that includes Ida — 'save 30-70 percent wall-clock at less than 5 percent added WSE-RMSE cost' — is more accurate but less catchy; both are honest given the figure shows the per-storm bars." . . . . . "Anne Fouilloux" . "2026-05-22T21:20:05.896Z"^^ . . . . . "Coastal-ROM trade-off reproduction outcome (Loveland et al. 2024)" . . "RSA" . "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" . "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" . . .