v6.6.0

Earthquake Simulator Pro (open mirror) v6.6.0

Second release of the v6 line's multi-region era — the 13 historical regional presets are re-sourced from USGS ComCat authoritative values, inland waters can no longer raise tsunami alerts, and the regional alert panel says honestly what its thresholds are. Main content since v6.5.0:

Historical presets re-sourced from USGS ComCat

• The California / Chile / Italy preset events (13) now carry ComCat authoritative epicenters, magnitudes, depths, origin times and mechanisms instead of hand-filled values. Each event's query URL and ComCat id are frozen into snapshots (tools/data/comcat-presets-*.json, schema quake-sim-comcat-presets-v1) with full provenance and experiment-manifest registration — reproducible, never silently editable. All 13 event ids are frozen as test anchors, so a future re-resolution cannot silently swap events.

Mechanisms: dual GCMT product channels

• Mechanisms read BOTH ComCat moment-tensor and focal-mechanism products (the GCMT solutions for California events exist only as the latter), GCMT first. Of each double couple's two nodal planes, the one closer to the pack's current values is kept, with the selection rule and the alternate plane recorded in the snapshot.

Inland waters exempt from tsunami alerts

• Regional tsunami forecast controls now require an open-ocean eligibility mask (Physics.oceanConnectedMask): wet AND ≥10 m deep AND connected to the sea through water at least that deep (edge voting prevents grid-edge slicing artifacts). Enclosed below-sea-level basins (the Salton Sea) and tidal channels unresolvable at 0.05° are exempt — California goes 26→17 counties and 466→350 controls, so Imperial County no longer warns from a lake; San Francisco Bay survives via the deep Golden Gate channel. Chile (637→622) and Italy (142→136) are near-stable. The JMA builder stays mask-free — Japan paths byte-identical.

Honest boundaries

• Pre-instrumental events (1906 San Francisco, 1971 San Fernando, 1960 Valdivia, 1908 Messina) have no ComCat mechanism — the documented literature values stay, recorded as mechanism: null with a kept-mechanism note. No fabricated beachballs.
• Messina 1908 keeps its literature land-cell epicenter (the ComCat/ISC-GEM epicenter falls on strait water; the 1908 tsunami was submarine-slump driven and no point source applies) — an explicit KEEP_EPICENTER exception.
• The regional first-wave ETA panel now carries the note that alert levels approximate JMA thresholds and are not the region's official criteria (not SHOA/CAT/NOAA operational products).
• Labels track the authoritative values: Norcia M6.5→M6.6, Iquique M8.1→M8.2.

Quality

• Node test suite 1,216/1,216 green (new comcat-presets contract tests: snapshot schema/provenance, pack↔snapshot consistency, mechanism honesty, frozen event-id anchors); validate-release 24,591 checks pass; CI green. Japan code paths byte-identical. This release mints a new Zenodo version DOI under concept 10.5281/zenodo.22828696.

Provenance & citation

MIT license. Per-dataset provenance and DOIs: public/data-catalog.js (NIED K-NET/KiK-net 10.17598/NIED.0004, J-SHIS, JMA, GSI, GEBCO 2025); region packages: SCEDC/NCEDC/EarthScope FDSN station-text, INGV FDSN, USGS cb_2018 (public domain), openpolis/ISTAT provinces (CC-BY 4.0), Natural Earth (public domain), Yong et al. (2014) BSSA California Vs30 (USGS public domain), Heath et al. (2020) Earthquake Spectra 36(3) global Vs30 assembly (doi:10.1177/8755293020911137), USGS ComCat event/moment-tensor/focal-mechanism services (public domain). Validation including negative results: PHYSICS_BENCHMARKS.md, docs/SCIENTIFIC-REPORT.md. Changelog: [6.6.0] section of CHANGELOG.md.

---

中文：v6.6.0 = 区域历史预设改 USGS ComCat 权威值（13/13 命中，查询+id+provenance 冻结入快照与测试锚）+ 机制解双产品通道（moment-tensor 与 focal-mechanism，GCMT 优先）+ 内陆水体豁免（oceanConnectedMask 开海资格掩码：索尔顿湖类封闭盆地退出海啸预报区，加州 26→17 区 466→350 控制点，日本路径字节不变）+ 区域 ETA 面板「JMA 阈值近似」诚实注记 + 前仪器 4 事件文献机制诚实保留（不臆造沙滩球）+ 墨西拿 1908 文献陆格震中保留（KEEP_EPICENTER）+ 标签随权威值更新（Norcia M6.6/Iquique M8.2）；1216 测试全绿 + validate-release 24591，验证（含负结果）冻结于 PHYSICS_BENCHMARKS.md。

v6.5.0

Earthquake Simulator Pro (open mirror) v6.5.0

Regional tsunami alerts come online, and all three pilot regions are now at full parity: every one of California / Italy / Chile runs the nonlinear SWE tsunami solver on its own GEBCO 2025 bathymetry strip and warns through the JMA-grade alert chain. Main content since v6.4.0:

Regional tsunami alerts (Chile / Italy / California)

• Fixed the reason regional tsunamis never warned: every JMA 66-area offshore control sits in Japanese waters, so a Chile epicenter was always beyond the 1200 km forecast gate and tsunamiCircles stayed empty — the wave ran, zero alert output (user-reported, reproduced).
• While a region with an admin-areas package and a loaded bathy grid is active, the forecast-area registry swaps to that region's own polygons (16 Chilean regioni, Italian provinces, 58 California counties) snapped to wet cells of its regional grid (≤80 controls/area, area codes rNNN). The JMA chain, first-wave ETA countdown panel, alert chimes and TTS then run unchanged on regional codes — the shared physics is geography-neutral (basin transmission returns 1 for unknown basins; coastal heights sample the solver directly).
• The registry restores to the JMA 66 areas on region switch/deactivate; the forecast distance gate scales with the registry source (3500 km for country-scale areas).

Arrival-height fix (the "arrived" path)

• The tsunami worker builds its per-checkpoint peak cache only from the checkpoint list passed at solver creation, so a standalone regional solver created with none answered samplePeak with 0 forever and no physical arrival could ever register. Regional controls now ride with solver creation; verified live in the Chile run (Maule 13.4 m @393 s, Ñuble 17.9 m @142 s).

California bathymetry — tri-region parity

• New GEBCO 2025 California offshore strip (us-california, 0.05°, 220×210): NE Pacific deep water to −5026 m, Sierra crest to +3761 m. California's real Vs30 (Yong et al. 2014) predates this release; the strip completes the tsunami side. The 1906 San Francisco preset epicenter falls in a nearshore water cell (−21.6 m at 0.05° water-mean resolution), so its solver runs — a documented resolution artifact, not a tuned parameter; the other presets are on land and stay seismic-only.

Honest boundaries

• Regional alert thresholds reuse JMA semantics (advisory/warning/major-warning) — this is not a SHOA / CAT / NOAA operational product. Valdivia 1960 / Messina 1908 published epicenters sit on land cells at grid resolution and keep tsunami honestly off. The Chile/Italy grids run standalone (the global coarse grid covers only Japanese longitudes; no Japan checkpoints ride along). Sources and licenses are registered per resource — 14 SHA-256 entries in the research manifest (GEBCO 2025 via CEDA OPeNDAP, public domain with attribution requested; USGS global Vs30 assembled product, public domain).

Quality

• Node test suite 1,206/1,206 green (new region-tsunami-alert anchors lock the physics no-op assumptions and re-derive the regional registry over the real chile/italy/california data); validate-release 24,578 checks pass; CI (node 22 / 24) green. This release mints a new Zenodo version DOI under concept 10.5281/zenodo.22828696.

Provenance & citation

MIT license. New data this release: GEBCO Compilation Group (2025) GEBCO 2025 Grid, doi:10.5285/37c52e96-24ea-67ce-e063-7086abc05f29 (California / Chile megathrust / Messina strips via CEDA OPeNDAP); Heath, D.C., Wald, D.J., Worden, C.B., Thompson, E.M., Smoczyk, G.M. (2020), "A global hybrid VS30 map with a topographic slope-based default and regional map insets", Earthquake Spectra 36(3), doi:10.1177/8755293020911137 (Italy/Chile Vs30 from the assembled apps.usgs.gov product). Validation including negative results: PHYSICS_BENCHMARKS.md, docs/SCIENTIFIC-REPORT.md. Changelog: [6.5.0] section of CHANGELOG.md.

---

中文：v6.5.0 = 区域海啸警报上线（根因修复：JMA 66 区控制点全在日本导致区域事件永不告警；区域预报区注册表=活动区域行政区划吸附区域格网湿单元，JMA 三档/ETA 倒计时/警报音/语音原样生效）+ standalone 求解器到达波高修复（worker 峰值缓存按创建检查点建，空列表导致实测路径死透）+ 加州 GEBCO 2025 近海条带（三试点区域海啸+真实 Vs30 全部拉平）+ 意/智真实 Vs30（USGS 全球混合成品，Heath et al. 2020，DOI 10.1177/8755293020911137）+ 智利 megathrust/墨西拿海峡 GEBCO 2025 条带；诚实边界（JMA 语义近似非 SHOA/CAT/NOAA 产品、1906 旧金山近岸水格分辨率产物、陆格震中海啸如实关闭、14 资源 sha256 逐项登记）全部记录在案；node 测试 1206/1206、validate-release 24,578 项、CI 绿。

v6.4.0

Earthquake Simulator Pro (open mirror) v6.4.0

First tagged release of the v6.4 line — global real stations, a multi-region pilot, a 12-level intensity scale, ~775 KB off first paint, and real California site data. Main content since v6.3.3:

Global real stations become simulation stations

• Frozen FDSN packages (SCEDC 5,893 / NCEDC 337 / EarthScope backbone 2 = 6,232 real stations) now act directly as simulation stations (name = NET.CODE), with clickable metadata popups and per-package provenance; uncovered stations keep honestly-labelled estimates.

Multi-region registry: California / Italy / Chile

• Catalog-driven region activation with a live selector: California (USGS cb_2018 58 counties; presets 1906 San Francisco / 1971 San Fernando / 1989 Loma Prieta / 1994 Northridge), Italy (INGV 697 operational stations, ISTAT 110 provinces CC-BY; Norcia 2016 / Amatrice 2016 / L'Aquila 2009 / Irpinia 1980 / Messina 1908), Chile (CSN C1 84 stations via EarthScope, Natural Earth 16 regions; Valdivia 1960 M9.5 / Maule 2010 M8.8 / Illapel 2015 M8.3 / Iquique 2014 M8.1). Preset mechanisms are the published USGS/gCMT values.

Real California site parameters & county-level reports

• The USGS Yong et al. (2014, BSSA) California Vs30 grid (7.5″ resampled to 0.025°) feeds real Vs30 to 94.1% of the 6,232 stations; uncovered stations fall back to the labelled 500 m/s estimate with counts in the station note.
• County-level intensity reports: region-mode map coloring (region areas > JMA subdivisions > prefectures priority) and the「Region forecast / county」table with ±1σ and long-period columns.

12-level intensity scale (CSIS)

• Preference toggle between the JMA 10-level shindo scale and a 12-level CSIS scale (published MMI-aligned approximation; JMA 7 ≈ XI, XII deliberately unreachable). Intensity 1–12 voice cues lazy-load with a first-use download notice; the final bulletin speaks the 12-level sentence via browser TTS.

~775 KB off first paint

• Seven research/fault modules (observed-fault-models, finite-fault, waveform-analysis, moment-tensor, waveform-data, strong-motion-data/-waveforms) left the boot path for a lazy manifest with honest entry-point gating; the service worker caches them on first use, keeping the offline contract. bump-versions stage 2.5 auto-syncs the inline manifest hashes.

Rendering robustness

• Fixed the 0×0 waveform-canvas stall (self-healing), per-layer safeDraw (one throwing layer no longer kills the frame), view-keyed plate-boundary cache, JQuake-style small overview dots at low zoom, scale-aware station number icons.

Honest boundaries

• New regions ship without bathymetry (tsunami off); missing regional Vs30 = labelled estimates; CSIS is a documented approximation; region detect-mode colors observations only; realtime kmoni/EEW keeps official JMA semantics.

Quality

• Node test suite 1,187/1,187 green; validate-release 24,555 checks pass; CI (node 22 / 24) green. This release mints a new Zenodo version DOI under concept 10.5281/zenodo.22828696.

Provenance & citation

MIT license. Per-dataset provenance and DOIs: public/data-catalog.js (NIED K-NET/KiK-net 10.17598/NIED.0004, J-SHIS, JMA, GSI, GEBCO 2025); region packages: SCEDC/NCEDC/EarthScope FDSN station-text, INGV FDSN, USGS cb_2018 (public domain), openpolis/ISTAT provinces (CC-BY 4.0), Natural Earth (public domain), Yong et al. (2014) BSSA Vs30 (USGS public domain). Validation including negative results: PHYSICS_BENCHMARKS.md, docs/SCIENTIFIC-REPORT.md. Changelog: [6.4.0] section of CHANGELOG.md.

---

中文：v6.4.0 = 全球真实台站测站化（6,232 FDSN 站）+ 多区域试点注册表（加州 58 县 / 意大利 110 省 / 智利 16 大区，历史预设采用公开震源机制）+ 加州实测定标（Yong 2014 Vs30 覆盖 94.1% 站）+ 县级震度速报与涂色 + 12 级烈度 CSIS（MMI 对齐公开近似）+ 首屏懒加载 ~775KB + 渲染鲁棒性修复；诚实边界（新区域无水深=海啸关闭、缺 Vs30 包=估计值如实标注、区域检测模式仅观测涂色）全部记录在案，验证（含负结果）冻结于 PHYSICS_BENCHMARKS.md。

v6.3.3

Earthquake Simulator Pro (open mirror) v6.3.3

Sync of upstream quake_sim research batches through 2026-09-15. Main content since v6.3.2 — the conditional-spectrum (CS) broadband research line, closed end-to-end:

CS v4 executed — P-SV hypothesis refuted

• Pre-registered 30-shard quad-double P-SV arm (~2.2 days wall clock) completed; all three shape-band gates FAIL (P-SV 0.794 / 0.398 / 0.727 vs SH-only 0.776 / 0.432 / 0.574 log10 bias at 0.1–0.5 / 0.5–2 / 2–5 s). Production broadband stays SH-only. Frozen report: tools/data/cs-pipeline-v4-report.json.

Attribution arc v5→v8 closed

• Target audit PASS (max muC delta 0.0047); stress lever DEAD (S* = 0); station-column swap SOURCE-OWNED (0.198); aligned-config retest on 13 real events × ~700 stations: alignedDs(0.2 s) = +0.02 vs legacy bare arm −0.238 (CONFIG-OWNED); per-class kappa recalibration KILLED by its pre-registered scenario non-regression gate (real-event and scenario optima have opposite signs; anchor-at-seam coupling).
• Final role of the CS shape gate = synth-vs-GMPE consistency monitor. Five single-parameter cures (kappa / stress / P-SV / LF gain / per-class kappa) all measured-dead, each locked by frozen tripwires.

Also carried in these syncs

• P-SV terminal chapter in docs/SCIENTIFIC-REPORT.md; PHYSICS_BENCHMARKS.md expanded (105 test files / 1,142 tests green); experiment manifest at 79 frozen artifacts.
• Retained from earlier v6.3.x syncs: PSHA deaggregation UI + BPT time-dependent engine, RTWave live multi-station waveform panel (server-side miniSEED STEIM1/2 decoder), waveform-analysis tooling, segmented/listric fault geometry, rupture-view card, dynamic-rupture scenario CLI.

Provenance & citation

MIT license. Per-dataset provenance and DOIs: public/data-catalog.js (NIED K-NET/KiK-net 10.17598/NIED.0004, J-SHIS, JMA, GSI, GEBCO 2025). Validation including negative results: PHYSICS_BENCHMARKS.md, docs/SCIENTIFIC-REPORT.md.

---

中文：本发布为上游 quake_sim CS 宽频带研究线闭卷同步（v4 全门阵列执行 + v5–v8 归因弧线收官：P-SV 补全假设按预登记否证，逐类 κ 重校准被情景非回退门判死，生产线保持 SH-only；详见 CHANGELOG [6.3.3] 节）。

---

📚 引用 / Cite

Zenodo 概念 DOI / concept DOI（引用所有版本 · cites all versions，始终解析到最新已归档版 · always resolves to the latest archived release）：10.5281/zenodo.22828696 · 机器可读引用 / machine-readable: CITATION.cff
本版 / this release: version DOI 10.5281/zenodo.22828697（v6.3.3 是首个 Zenodo 归档版本，更早的 release 标签未入档 / first archived version; earlier release tags are not archived）

v6.3.2

v6.3.2 — P-SV 分层柔度研究五批同步

上游五批研究内容（ crest 噪声批 → Schur 导纳批 → 裁决批 → 三项 continuation 批 → det 主值扣除批），纯研究线，无应用行为变化。

新增

• Schur 导纳柔度表示（psv.psvSchurCompliance）：逐层导纳步进替代长程 A/W 传输���生产接线 params.schurCompliance（complianceAt 分发器，legacy 缺省位级不动）
• FD-BVP 独立裁判（crest-fd-referee.js）：box 格式边值求解器——分级网格 + 层边界强制节点 + 剖面输出；与所有链代数完全独立
• 闭合式半空间 BVP 参考（crest-halfspace-closed.js）：自由面 + 辐射 + 源跳条件 6×6 直接解
• P-SV 守卫除数旋钮（params.psvGuardDiv）：div10 已过 SH 家族门（max 0.011 ≤ 0.10），缺省保持字节兼容
• CS 管线 v4 预登记（PRE_REG_V4）：双臂定义 + 三个具名运行前置，运行撤回至前置落地
• 自适应 k 中点细化（params.adaptiveK，opt-in）

修复

• crest-ode-referee 未排序 stops 回归：多层配置末段反向积分（HALF 单层验证盲区）；R11 点 0.79 → 3.9e-3

研究结论（psv-scale-diagnosis v11，全部实测冻结）

• 深配置裁决：三个互证参考（RK4 外推 / 特征分解输运 / 闭合解）两两一致 4e-3；闭合解 ≡ −Schur（跳取向符号）；expm 无罪（7e-13）；legacy A/W 链 O(1) 偏离且 cap 敏感
• Schur 臂 dk 级数仍不收敛（194/194/523/680/541）——阻塞精确化为带内极点求积（解析 det 主值扣除），自适应细化实测不足已记录
• 分层柔度裁决危机（OPEN）：R11 三层上链家族与网格收敛的 FD 相差 30%，无存活独立分层裁判可裁决——登记关键路径 = FD 的 SH 退化对照（shake91 锚）

测试

• psv-schur.test.js（4 锚，含裁决锁：legacy 若对上闭合解即触发重冻）+ psv-alias-ladder.test.js（2 锚）入测试链
• psv-scale-diagnosis v11 tripwire 更新（v10 判决保留）

Full Changelog: v6.3.1...v6.3.2

---

📚 引用 / Cite

Zenodo 概念 DOI / concept DOI（引用所有版本 · cites all versions，始终解析到最新已归档版 · always resolves to the latest archived release）：10.5281/zenodo.22828696 · 机器可读引用 / machine-readable: CITATION.cff

v6.3.1 — 应用身份修复（标题/帮助显示 v6.2 → v6.3）

v6.3.1 — 应用身份修复（标题/帮助显示 v6.2 → v6.3）

补丁版本，无任何物理/功能代码变更。 v6.3.0（npm）发布时包内应用身份仍为 v6.2，本版正式携带修复：

🏷️ 修复内容

• app.title 三语 →「地震模拟器 v6.3 / Earthquake Simulator v6.3」；场景快照 appVersion → v6.3
• 使用帮助新增 v6.3 更新章节（时间依赖 PSHA / 危险分解 / J-SHIS 对照·时间依赖臂 / P-SV ��尼修复 / 诚实边界，五对 ×3 语言）
• guide.html / report.html 页脚版本 → v6.3

🔖 tag 出处说明

• v6.3.0 已移回 npm 实际发布内容（commit 1e675f9）——tag 与已发布工件重新一致；
• v6.3.1 = v6.3.0 内容 + 应用身份批（c41bc58）+ 包版本号（本提交）。

✅ 质量

• node 测试 1071 / 0 失败；validate-release 24470 检查通过；CI（node 22 / 24）全绿。

Full Changelog: v6.3.0...v6.3.1

---

📚 引用 / Cite

Zenodo 概念 DOI / concept DOI（引用所有版本 · cites all versions，始终解析到最新已归档版 · always resolves to the latest archived release）：10.5281/zenodo.22828696 · 机器可读引用 / machine-readable: CITATION.cff

v6.3.0 — PSHA 危险分解 UI · BPT 时间依赖 PSHA · P-SV 阻尼修复

v6.3.0 — PSHA 危险分解 UI · BPT 时间依赖 PSHA · P-SV 阻尼修复

✨ 新功能

• PSHA 危险分解（Deaggregation）：Physics.deaggregate 接入信息页「地震危险性」卡第三画布。0.5 级震级档堆叠柱（按构造类着色：板内 / 板块边界 / 板内深源）+ 模态 bin 与 C/I/S 占比题注 + top-5 震级×Rrup 距离贡献表（ε = 相对 GMPE 中位数的标准差偏移）。泊松路径复用已算危险曲线反演 IM 目标，无额外全量积分；BPT 开启时分解保持泊松语义（卡片注记言明）。
• BPT 时间依赖 PSHA：南海海槽 3 个情景源（M8.9 全域 / M8.2 东段 / M8.3 西段）按 Brownian passage time 更新过程评估——ERC 平均复发 117 年，上次全域事件 1707 宝永，单段 1944/1946。PSHA 卡开关切换，危险曲线带 50 年条件概率徽标；UHS 透传 timeDependent。

🔬 科研线

• J-SHIS 对照 v2（时间依赖臂）：泊松中段亏损在 BPT 下全场址翻正——预登记诊断的实证；RP475 尾部倍率（中位 4.5×）按非同口径如实记录，不做等价声明。
• P-SV 复模量修复：psvEigenvectors 牵引行 实μ → 复 μ*（此前每层 Q 被静默丢光，阻尼只剩半空间导纳一条路）。R10 阻尼半空间 RK4 对拍 1.4%；无阻尼路径逐位不变。
• R11 分层独立验证：首条跨界面 RK4 对拍；参考器两处缺陷（层深单位 / W 对向下传）如实记录并修复，全锚 11/11。
• SH 混叠守卫收官：除数阶梯按预登记门选型 div 80（远场 max 0.070 ≤ 0.10）；暴露报告 v3 + scorecard 重冻结（带级影响 ≤ 0.05 log10��brune 臂逐位不变）。
• P-SV 尺度诊断 v9 重定基：修正内核上 fullTensor 系列呈干净单调发散——crest 带 OPEN 重确认；登记残差提取 / 逐层 Schur 导纳为下一步。负结果照冻。

🏷️ 版本说明

• ⚠️ 本版发布时包内应用身份仍为 v6.2（网站标题 / 使用帮助章节 / whatsnew 弹窗未更新）——身份修复随 v6.3.1 补丁发布，v6.3.1 = 本版内容 + 身份批 + 包版本号，无物理/功能变更。
• manifest v6.3；package 6.3.0。

⚠️ 诚实边界

• opts.psv（P-SV 三分量混合宽带）保持研究态，生产 BLOCKED；PSHA 绝对水平与 J-SHIS 官方曲线对照进行中。

✅ 质量

• node 测试 1071 / 0 失败；validate-release 24459 检查通过；CI（node 22 / 24）全绿。

Full Changelog: v6.2.2...v6.3.0

---

📚 引用 / Cite

Zenodo 概念 DOI / concept DOI（引用所有版本 · cites all versions，始终解析到最新已归档版 · always resolves to the latest archived release）：10.5281/zenodo.22828696 · 机器可读引用 / machine-readable: CITATION.cff

v6.2.1 — CS 短周期仲裁者 · P-SV 内核 · 海啸观测 direct 策展 · PSHA 活取复核

核心变更（自 v6.2.0）

CS 短周期仲裁者（预登记规则先于测量）

• 13 事件 Kyoshin 三分量配对谱形残差测量：目标侧 INCONCLUSIVE（zhao 短周期形状被观测支持，配对中位 −0.03）
• κ 杠杆否证（obs−syn = −0.008）；几何平均计分排除（gm/trans ≈ 0）
• 合成−观测诊断（−0.24 @0.2 s）与 CS 门禁方向（+0.4~0.78）矛盾：巨震档目标侧欠估与合成侧超额并存，无 M8.5+ 近场记录不可分账——如实登记

P-SV 离散波数内核（psv.js，7/7 锚定）

• 4×4 特征向量传播器 + 每层 Gram-Schmidt 再正交化（解决低速层导波模态 1e97 动态范围）+ 厚层细分 + 深度有限差分偶极 + 方位 Bessel 矩恒等式装配
• 锚定：Rayleigh 色散根（3212 vs 三次方程 3214 m/s）、独立 RK4 ODE 交叉验证 <0.4%、辐射节点、α 数值积分交叉、FD 收敛、频带稳定、k→0 SI 一维柱绝对尺度 0.7%
• hybrid.js opts.psv 三分量 opt-in（默认关 = 字节兼容）；生产接线待极点积分（主值积分+残差提取）

海啸观测 direct 策展（R5-6，数据集翻绿）

• 5 事件 / 80 条全 direct / 到时字段；认证门翻转（researchReady=true + directCount 守恒）
• 记分卡重冻：��中率 48%、0 误报、首批到时残差 bias −37 min
• 数据管道入库（JMA 月报 PDF 字符坐标解析 + TTJT 调查 CSV + 1993 官方页面）

PSHA 活取复核

• J-SHIS Y2024 官方曲线活取与内嵌逐位一致（RP475 中位 1.827 不变）

验证：全量测试链绿 · validate-release 24,452 checks · bump/manifest 同步

MIT License · 数据来源与引用见 README

---

📚 引用 / Cite

Zenodo 概念 DOI / concept DOI（引用所有版本 · cites all versions，始终解析到最新已归档版 · always resolves to the latest archived release）：10.5281/zenodo.22828696 · 机器可读引用 / machine-readable: CITATION.cff

v6.2.0 — 实时波形 · 波形分析 · 断层系统升级 · 动力学破裂管线

核心更新（自 v6.1.0）

实时监测

• 实时多台站活波形面板（RTWave）：4 个环绕日本的开放 GSN 台站（IU MAJO 松代 / YSS 南萨哈林斯克 / INCN 仁川 / TATO 台湾坪林）实时波形，5/10/20 分钟窗口，EEW 事件叠加 P/S 预计到时刻度；仪器原始计数（未标定）诚实标注
• 服务端 /api/waveform/live 代理 + 纯 JS miniSEED v2.4 STEIM1/2 解码器（libmseed 权威语义移植��三方验证 = 官方夹具 + ObsPy 逐样本零失配 + 四站活流）

波形分析

• 体验报告页三站波形分析节（最强/最近/中等距离自动选站）：滑移/能量/周期指标 + 矩速率曲线 + 波形 PGA 反演震级 vs 设定震级
• 信息页波形分析工具：PGA/PGV/PGD、Arias 强度、CAV、D5-95、卓越周期、视拐角频率 + 震级反演对比

断层系统

• von Kármán 滑移谱（Mai & Beroza 2002 型各向异性，H=0.75，网格无关）+ 浅部滑移亏损旋钮（矩守恒）
• 断层破裂详情卡：断层面剖面（沿走向×深度，滑移/破裂时间着色，成核星，运行中动画）+ 矩速率/累积矩图 + 9 行统计
• 分段弯折断层：地图路径编辑 → 共节点段平面链，连续滑移场跨弯折，3D 走时（诚实限制：平面段近似深部弯折间隙）
• 铲式几何（faultListricDip，默认 0 完全兼容）
• 动力学破裂管线：run-scenario.js 场景 CLI（导出前过应用自带验证器）+ 捆绑模型（SH 走滑 Mw7.45 / PSV 倾滑 Mw6.71）+ 导入卡一键加载

科学批

• 分段南海震源 v2（J-SHIS 官方对照：RP475 高估中位 5.97×→1.83×）+ CS 形状门诊断三因分解 + LOEO 更正

---

测试 1026/1026 · validate-release 24,449 全过 · 详细变更见 CHANGELOG.md

---

📚 引用 / Cite

Zenodo 概念 DOI / concept DOI（引用所有版本 · cites all versions，始终解析到最新已归档版 · always resolves to the latest archived release）：10.5281/zenodo.22828696 · 机器可读引用 / machine-readable: CITATION.cff

v6.1.0 — PSHA/UHS + 宽带时程合成

v6.1.0 — 概率地震危险性（PSHA/UHS）+ 宽带时程合成研究管线

新特性

• PSHA 震源模型与危险性引擎：冻结 USGS ComCat 目录（1923 年起、M5+、8,883 事件，公有领域）→ 时空去簇、完备性扫描（Mc=5.0@1980）、逐构造类 Aki 极大似然 b 值、自适应平滑 → 0.25° 网格 GR 震源 + 南海海槽 M9 / 首都直下 M7.3 情景源（公开发表复发间隔换算年率）。Physics.hazardCurve Poisson 年超越率积分，含逻辑树三分支认知不确定性带。
• 信息页"地震危险性 (PSHA)"卡：PGA 年超越率曲线 + 475/1000/2500/5000 年重现期一致危险谱（UHS，CSV 可导出）。Zhao2006 系数表逐字扩至 hazardlib 同源的全 21 周期（sha256 复验），谱周期逻辑树塌缩为单模型（公开记录的能力边界）。
• SH 离散波数 Green 函数内核（tools/broadband）：Thomson–Haskell 传播矩阵 + 双力偶源的 Bouchon 求和；五解析锚（半空间闭式 / 静态极限 / Love 频散 / 垂直入射 / 波数收敛）全部通过。
• 混合宽带时程合成 + 预登记记分卡：1 Hz 以下确定性 DW 低频 + 绝对定标 Boore 随机高频（集合标定冻结 50 MPa / κ0.02），对 13 个冻结 Kyoshin 观测包（USGS 矩张量逐事件解析）评估预登记验收门。结果如实冻结：长周期（2–10 s）较 Brune 基线改善 +0.143，通过；绝对门全部未过——包括参照 GMPE 自身在强度优选台站集上带约 −0.5 偏差地板的发现。
• 科学修复：绝对标定链抓出并修复 ×ω² 过微分、黎曼填充因子、IASP91 延续层密度 1000× 三个单位级错误。
• 月度冻结数据刷新（连续第二个月漂移序列）与月份键本地历法修复。

诚实的边界

• PSHA 绝对水平在与 J-SHIS 官方曲线对照（待完成的外部门）之前不做等价性声明。
• 宽带混合器无 P-SV（瑞利/SV），无法给出三分量 JMA 计测震度；全部负结果冻结于报告并测试锁死。

验证

npm test 939/939 · validate-release 24368 项检查全过 · 冻结报告/实验工件内容哈希身份（qsx1-*）入回归。

---

EN summary: Probabilistic seismic hazard (self-built USGS ComCat source model + Poisson engine, logic-tree epistemic band) with Info-tab hazard curves and uniform hazard spectra (verbatim 21-period Zhao2006 table); SH discrete-wavenumber Green's-function kernel with five analytical anchors; hybrid broadband synthesis (deterministic DW LF + absolutely calibrated stochastic HF) scored against 13 frozen Kyoshin packages under pre-registered gates — the long-period improvement gate passes (+0.143), absolute gates fail honestly (including the −0.5 selection-bias floor finding for the reference GMPE). Monthly frozen-data refresh included. Full test suite green.

---

📚 引用 / Cite

Zenodo 概念 DOI / concept DOI（引用所有版本 · cites all versions，始终解析到最新已归档版 · always resolves to the latest archived release）：10.5281/zenodo.22828696 · 机器可读引用 / machine-readable: CITATION.cff