Blog

  • 日长1500年振荡、D-O周期与太阳代理的三变量因果方向检验:Granger与CCM能否定向

    Evidence level: B

    周期接近只能说”像”。要判”太阳→日长””太阳→气候””日长↔气候”三条链路各自的方向与强度,就必须换用能处理非线性共同驱动的因果工具——因为在确定性动力系统里,Granger 因果会失效,而两两相关永远分不出”谁推谁”。

    正文

    一、问题先摆正:相关分不出方向,更分不出共同因

    前几篇做的是”相位”,本篇换成”方向”:不问像不像,问谁推谁。

    绕不开的坎在这里:太阳代理(¹⁴C/¹⁰Be)、日长重建、气候序列的周期数值都在千年上下打转,两两做相关都会”显著”。但相关既说明不了方向,也排除不掉第三种可能:有一个共同因在背后同时推动两者。这就是 Reichenbach 的共同因原则——两个变量相依,要么彼此因果,要么被同一个驱动推动(Runge 等 2019,PMID 31201306)。地球系统因果推断方法学,正是为”实验做不了、只能靠观测”的场景而生。

    待判的不是一条链路,而是三条:太阳 → 日长(外部天象能否撬动地球自转)、太阳 → 气候、日长 ↔ 气候(地球自转与气候是否有直接通道,方向如何)。

    二、工具选择:为什么用 CCM 而不是 Granger

    最常用的因果检验是 Granger 因果:若 x 含”独有信息”能改善 y 的预测,则 x 因 y。它的致命前提是——系统各变量的影响可分。

    偏偏地球系统不是。Sugihara 等 2012 年在《Science》给出一个反直觉论证:在(至少部分)确定性的动力系统中,信息随时间不断前传,若 x 真作用于 y,则 x 的信息必然被写进 y 的状态里(Takens 定理:单变量时间序列足以重构整个系统吸引子)。于是因果变量反而”不含独有信息”,Granger 检验失效(PMID 22997134)。替代方案是收敛交叉映射(convergent cross mapping,CCM):用 y 的重构流形去估计 x 的状态,看交叉映射技巧是否随数据量增大而”收敛”。收敛,说明 x 的信息确实被写进了 y——即 y 因于 x。经典案例正是”沙丁鱼-鳀鱼-温度”之争:两者相关不是互相作用,而是共享温度强迫。

    三、方向怎么读:用”时滞”把三条链路拆开

    CCM 默认两方向都可能有映射(强单向强迫会造成”广义同步”,把单向误读成双向)。Ye 等 2015 年在《Scientific Reports》把 CCM 扩展为带时滞版本:驱动变量的值更适合预测响应变量的未来,响应变量的值更适合反推驱动变量的过去。于是”最优交叉映射滞后”的符号与大小,成为方向与交互滞后的估计(PMID 26435402)。作者还演示了在传递因果链里把”直接”与”间接”因果分开,这正是三变量分析的核心难题。

    四、三变量联合:两两检验会掉进”共同驱动”陷阱

    若太阳同时作用于日长与气候,那”日长 vs 气候”的两两 CCM 必然显著——它们被同一个因推动,信息互相渗透,单看会误判成”互为因果”,实则是共同因造成的假耦合。

    出路是引入第三个变量做条件检验:把太阳代理放进条件集,看”给定太阳信号后,日长与气候之间是否还有独立通道”。Runge 等 2015 年在《Nature Communications》给出识别”因果网关与中介”的框架,用于分离直接因果与经由中介的间接因果(PMID 26443010);其后续 PCMCI 方法进一步处理”多变量 + 自相关 + 混淆”情形(PMID 31807692)。另一条路径是信息论:Deng 等证明”因果化 CCM”与有向信息近似等价、且更稳健(PMID 38169910)。两条路径互为交叉验证。

    五、太阳侧已有一正一反的示范,日长侧出现反转

    “太阳→气候”其实已被直接检验。Tsonis 等 2015 年在《PNAS》用 CCM 检验银河宇宙线与全球温度的年际变化,结论是:20 世纪整体增暖趋势找不到可测量的因果证据,但年际尺度上存在显著(尽管微弱)的因果效应(PMID 25733877)。随即有同刊质疑文章认为该证据可疑(PMID 26283407)——连 CCM 也不是免争议的裁判,必须透明报告嵌入维度、库长与替代零模型。

    太阳代理本身有金标准:Steinhilber 等用格陵兰与南极的 ¹⁰Be 冰芯 + 全球 ¹⁴C 树轮做主成分分析,合成 9400 年太阳活动序列;并指出太阳强迫与亚洲气候”总体吻合,但存在若干完全不相干的时段”(PMID 22474348)——这正是”间歇耦合”,不同于简单”稳定外驱”。

    日长侧则出现反转:Nature 2026 的研究显示,多年代尺度的日长变化主要由内核差速旋转产生的引力力矩驱动(PMID 42778592)。日长振荡更可能是”地球内部的事”,这削弱了”太阳统一驱动日长与气候”的前提。

    六、三个必须防的陷阱

    陷阱有三:其一,定年误差——千年尺度序列的年代学不确定度足以人为制造或抹除滞后方向;其二,库长与嵌入——CCM 的”收敛”需足够长的序列与合适的嵌入维度,千年尺度数据点稀疏,容易”看起来收敛”;其三,多重比较——三条链路 × 两方向 × 多频段,不校正就会捞到显著。

    七、三种结局

    结局一:联合检验后太阳→气候成立、日长由内部力矩驱动、日长与气候无独立通道。则”天时统一时钟”假设被证伪——天上的周期与地下的周期各走各的钟。

    结局二:给定太阳信号后,日长与气候仍有独立通道且带稳定时滞。则”地球自转-气候”是一条被忽视的直接链路,天时与地气的耦合有物理载体。

    结局三:三条链路都达不到 CCM 收敛标准,说明现有数据长度与精度回答不了方向问题。

    同一批”看起来相关”的曲线,两两检验会告诉你”互为因果”,三变量联合检验可能告诉你”其实都没关系,是太阳在背后同时推了两把”。方向问题,只能用能看见方向的方法回答。

    证据与局限

    已验证: ①CCM 可基于 Takens 状态空间重构区分因果与相关,覆盖 Granger 失效的非可分弱耦合动力系统(PMID 22997134);②扩展 CCM 用时滞区分单向强迫造成的广义同步与真双向因果,并可解析传递因果链(PMID 26435402);③因果化 CCM 与有向信息近似等价且更稳健(PMID 38169910);④”两变量相依 = 彼此因果或共同驱动”的共同因原则与地球系统因果推断框架已系统化(PMID 31201306);⑤CCM 已直接用于检验宇宙线与全球温度(年际有微弱因果、增暖趋势无)(PMID 25733877);⑥9400 年太阳代理序列已建立,且显示与亚洲气候”间歇相干”(PMID 22474348);⑦日长多年代变化主要由内核引力力矩驱动(PMID 42778592)。

    部分验证: ①CCM 对太阳-气候链路的结论仍存争议(PMID 26283407 提出质疑);②”日长↔气候”是否存在独立通道、以及是否被太阳共同驱动,尚未见三变量联合检验;③因果工具有效性受序列长度、嵌入维度与定年误差限制。

    待验证: ①”太阳→日长””太阳→气候””日长↔气候”三条链路的方向与强度从未在同一组数据上联合标定;②给定太阳代理后的条件因果(偏互信息/条件 CCM)是否仍显著;③日长的内部驱动分量(引力力矩)与外部太阳分量的贡献占比未分离。

    相关阅读

    证据等级: B
    参考文献: PMID 22997134, PMID 26435402, PMID 38169910, PMID 31201306, PMID 26443010, PMID 31807692, PMID 25733877, PMID 26283407, PMID 22474348, PMID 42778592


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Testing the Causal Direction of Three Variables: Solar Cycle Oscillation, D-O Cycle, and Solar Proxies Using Granger and CCM

    Evidence level: B

    Periodicity alone implies “similarity.” To distinguish the direction and strength among the three pathways—Solar Proxy → LOD, Solar Proxy → Climate, and LOD ↔ Climate—one must employ causal tools capable of handling non-linear common drivers. In deterministic dynamical systems, Granger causality fails due to information overlap, and pairwise correlation cannot disentangle “who drives whom.”

    Main Text

    I. Clarify the Problem: Correlation Reveals No Direction, Nor Common Causes

    Previous work focused on “phase”; this piece shifts to “direction”: not whether they resemble each other, but who drives whom.

    The hurdle here is unavoidable: Solar proxies ($^{14}$C/$^{10}$Be), LOD reconstructions, and climate sequences all exhibit periods around millennia. Pairwise correlations between them will invariably be “significant.” However, correlation reveals neither direction nor excludes a third possibility: a common cause driving both simultaneously. This is Reichenbach’s Common Cause Principle—two variables are dependent either through mutual causation or via a shared driver (Runge et al. 2019, PMID 31201306). Earth system causal inference methodology is specifically designed for scenarios where experimentation is impossible, relying solely on observation.

    The question is not a single link but three: Solar Proxy → LOD (Can extraterrestrial phenomena perturb Earth’s rotation?), Solar Proxy → Climate, and LOD ↔ Climate (Is there a direct channel between Earth’s rotation and climate, and what is its direction?).

    II. Tool Selection: Why CCM Over Granger

    The most common causal test is Granger causality: if $x$ contains “unique information” improving the prediction of $y$, then $x$ causes $y$. Its fatal premise is that system influences are separable.

    The Earth system is not. Sugihara et al. (2012) in Science provided a counter-intuitive argument: in (at least partially) deterministic dynamical systems, information propagates forward in time. If $x$ truly acts on $y$, $x$’s information is inevitably written into $y$’s state (Takens theorem: a single-variable time series suffices to reconstruct the entire system attractor). Consequently, the causal variable lacks “unique information,” rendering Granger causality invalid (PMID 22997134). The alternative is Convergent Cross Mapping (CCM): estimate the state of $x$ using $y$’s reconstructed manifold and observe whether the cross-mapping technique “converges” as data length increases. Convergence confirms $x$’s information is embedded in $y$—meaning $y$ is driven by $x$. A classic case is the “sardine-sardine-temperature” debate: their correlation arises not from mutual interaction but from shared temperature forcing.

    III. Reading Direction: Using “Lag” to Disentangle the Three Links

    CCM assumes bidirectional mapping is possible for both directions (strong unidirectional forcing can cause “generalized synchronization,” misreading unidirectionality as bidirectionality). Ye et al. (2015) in Scientific Reports extended CCM into a lagged version: values of the driver are better at predicting future values of the response, while values of the response are better at inferring past values of the driver. Thus, the sign and magnitude of the “optimal cross-mapping lag” become estimates for direction and interaction delay (PMID 26435402). The authors also demonstrated separating “direct” from “indirect” causality within transitive causal chains, which is the core difficulty in three-variable analysis.

    IV. Three-Variable Joint Analysis: Pairwise Tests Fall into the “Common Driver” Trap

    If the solar proxy acts simultaneously on LOD and climate, the pairwise CCM between “LOD vs. Climate” will necessarily be significant—they are pushed by the same cause, information permeates each other, and observing them in isolation leads to a false judgment of “mutual causation” rather than common cause-induced spurious coupling.

    The solution is to introduce a third variable for conditional testing: include the solar proxy in the conditioning set to observe whether an independent channel remains between LOD and climate “given the solar signal.” Runge et al. (2015) in Nature Communications provided a framework for identifying “causal gateways and mediators” to separate direct causality from indirect causality via mediators (PMID 26443010); their subsequent PCMCI method further handles scenarios involving “multivariate + autocorrelation + confounding” (PMID 31807692). An alternative path is information theory: Deng et al. proved that “causalized CCM” is approximately equivalent to directed information and more robust (PMID 38169910). These two paths serve as cross-validation.

    V. Solar Side Shows Both Positive and Negative Evidence; LOD Side Shows Reversal

    “Solar Proxy → Climate” has already been directly tested. Tsonis et al. (2015) in PNAS used CCM to test the interannual variability of global temperature against galactic cosmic rays, concluding that no measurable causal evidence exists for the overall 20th-century warming trend, but a significant (though weak) causal effect exists at interannual scales (PMID 25733877). Subsequently, a paper in the same journal questioned this evidence (PMID 26283407)—even CCM is not an uncontested arbiter; transparent reporting of embedding dimension, lag length, and alternative null models is mandatory.

    The solar proxy itself has a gold standard: Steinhilber et al. used Greenland and Antarctic $^{10}$Be ice cores plus global $^{14}$C tree rings for principal component analysis to synthesize a 9,400-year solar activity sequence; they noted that solar forcing aligns “generally” with Asian climate but exhibits “several completely unrelated periods” (PMID 22474348)—this is “intermittent coupling,” distinct from simple “stable external forcing.”

    On the LOD side, a reversal appears: Nature (2026) research indicates that multi-millennial LOD variations are primarily driven by gravitational torques generated by core differential rotation (PMID 42778592). LOD oscillations are more likely “internal Earth matters,” weakening the premise of a “unified solar driver for both LOD and climate.”

    VI. Three Must-Avoid Pitfalls

    Three pitfalls exist: First, dating errors—uncertainties in chronology at the millennial scale can artificially create or erase lag directions; second, sequence length and embedding—CCM’s “convergence” requires sufficiently long sequences and appropriate embedding dimensions; millennial-scale data are sparse, easily appearing “convergent” by chance; third, multiple comparisons—three links × two directions × multiple bands, failing to correct for multiplicity will yield false positives.

    VII. Three Possible Outcomes

    Outcome One: Joint testing confirms Solar Proxy → Climate, LOD driven by internal torques, and no independent channel between LOD and Climate. Then the “Unified Cosmic Clock” hypothesis is falsified—heavenly cycles and subterranean cycles tick independently.

    Outcome Two: Given the solar signal, LOD and Climate retain an independent channel with stable lag. Then “Earth Rotation-Climate” is a neglected direct link, and the coupling of Heaven and Earth possesses a physical carrier.

    Outcome Three: None of the three links meet CCM convergence standards, indicating current data length and precision cannot answer the direction question.

    The same set of “appearing correlated” curves will tell you via pairwise testing that they are “mutually causal,” while three-variable joint testing might reveal “actually unrelated, with the solar proxy pushing both simultaneously behind the scenes.” Directional questions can only be answered by methods capable of visualizing direction.

    Evidence and Limitations

    Verified: ① CCM can distinguish causality from correlation based on Takens state space reconstruction, covering non-separable weakly coupled dynamical systems where Granger fails (PMID 22997134); ② Extended CCM using lag distinguishes generalized synchronization caused by unidirectional forcing from true bidirectional causality and can parse transitive causal chains (PMID 26435402); ③ Causalized CCM is approximately equivalent to and more robust than directed information (PMID 38169910); ④ The Common Cause Principle (“two variables dependent = mutual causation or common driver”) and Earth system causal inference frameworks are now systematized (PMID 31201306); ⑤ CCM has been directly applied to test cosmic rays vs. global temperature (significant weak effect at interannual scale, none for warming trend) (PMID 25733877); ⑥ A 9,400-year solar proxy sequence is established and shows “intermittent coherence” with Asian climate (PMID 22474348); ⑦ Multi-millennial LOD variations are primarily driven by core gravitational torques (PMID 42778592).

    Partially Verified: ① Conclusions regarding the Solar-Climate link via CCM remain controversial (PMID 26283407 raises doubts); ② Whether an independent channel exists for “LOD ↔ Climate” and whether it is driven by a common solar cause has not yet been subjected to three-variable joint testing; ③ Tool effectiveness is limited by sequence length, embedding dimension, and dating errors.

    To Be Verified: ① The direction and strength of the three links (“Solar → LOD”, “Solar → Climate”, “LOD ↔ Climate”) have never been jointly calibrated on the same dataset; ② Whether conditional causality (partial mutual information/conditional CCM) remains significant given the solar proxy; ③ The contribution ratios of internal driving components (gravitational torque) versus external solar components for LOD have not been separated.

    Related Reading

    Evidence Level: B
    References: PMID 22997134, PMID 26435402, PMID 38169910, PMID 31201306, PMID 26443010, PMID 31807692, PMID 25733877, PMID 26283407, PMID 22474348, PMID 42778592


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • D-O事件相位组织 vs 日长振荡:弛豫振荡与幽灵随机共振的相位漂移判别

    Evidence level: B

    “相位对不上”既可以是否定外驱的证据,也可以是支持”幽灵随机共振”的证据。同一批数据,结论取决于你假设的是哪一种振荡。

    正文

    一、先分清对象:两种振荡,形状里早写好了

    前两篇都在追问同一件事——两条周期接近的序列,究竟共享一个节拍器,还是各走各的。它们解决的是”怎么测相位”。本篇问一个更前置的问题:我们比较的两个对象,本身就是两种不同类型的振荡,而类型差异直接决定相位”该长什么样”。

    冰期的丹斯果-奥什格尔(D-O)事件是典型的弛豫振荡(relaxation oscillation):在冷态慢慢积攒(慢相),到阈值突然跳进暖态(快相)后慢慢滑回。教科书形态是”慢-快-慢”的锯齿,不是正弦。这一族现象被归入”弛豫振荡、可激发系统、慢-快动力学、同宿轨道”的框架(Crucifix 2012,PMID 22291227),其物理基础是大西洋深层水两种形成模态之间的跃迁(Ganopolski & Rahmstorf 2001,PMID 11196631)。

    全新世的日长 1500 年振荡是另一种东西:一条连续的、类正弦的长期起伏,没有”事件”,没有阈值跳变。

    差别不只是好不好看。弛豫振荡的相位天然会漂移:每次跳跃由噪声和内部恢复过程共同决定,间隔抖动不断累积成相位的随机游走(phase diffusion)。一个”内部弛豫振荡 + 噪声”系统,本来就不该有稳定相位。

    二、幽灵随机共振:为什么”相位对不上”可能是对的

    若 D-O 事件被某个真实 1470 年外驱”钉住”,它与该外驱之间就该有稳定相位。但如果是幽灵随机共振(ghost stochastic resonance, GSR),结论就反过来。

    GSR 的核心是:输入里根本不存在输出那个频率,输出频率是两个输入频率的公倍数。Braun 等(2005,PMID 16281042)提出的机制正是如此——把约 210 年(De Vries 周期)与约 86.5 年(Gleissberg 周期)两个已知太阳周期,按 1470 的约 1/7 与 1/17 谐波叠加去驱动阈值系统,两者每 1470 年同相一次,输出端便”凭空”长出 1470 年周期,而输入端没有这个频率。这与听觉里”缺失基频”现象(听到复合音里并不存在的基频)是同一类机制。

    关键在后半句:GSR 情形下,事件与驱动周期不存在固定相位关系。所以”D-O 与太阳代理相位对不上”不能用来否定太阳参与——它恰恰是 GSR 的预期签名。

    三、判据:相位是”聚集”还是”多模态”

    检验目标由此从是非题变成类型判别题:

    • 相位相对某个候选外驱长期稳定聚集(圆形统计平均合向量长度 R 高、Rayleigh 检验显著)→ 支持普通”外驱周期振荡”。
    • 相位呈多模态/宽分布(事件扎堆在若干相位窗口)→ 支持 GSR,不需要真实 1470 年外驱。
    • 相位缓慢连续漂移 → 支持无外驱的内部弛豫振荡 + 噪声。

    Braun、Ditlevsen 与 Kurths(2009,PMID 20059228)正为此构造了三种”相位多模态度”统计量,用概念模型的蒙特卡洛模拟标定零分布,并用于 NGRIP 冰芯最后 11 个 D-O 事件的复发格局——这是目前 GSR 假说最直接的定量检验工具。

    四、和日长振荡怎么”形式比较”

    硬约束:日长重建只回溯到公元前 720 年(约 2700 年,PMID 28119545),D-O 事件在末次冰期(12–1 万年前),两者时间不重叠,无法直接做交叉小波。所以只能说”形式比较“——比两种振荡的相位结构类型,而不是把它们放进同一条时间轴。

    可执行三步:①刻画冰期一侧:用 NGRIP/GISP2/GRIP 的 D-O 序列,算相对候选太阳周期(210/86.5 年)的相位多模态度,看它落在”聚集/多模态/漂移”哪一区。②刻画双极相位:高分辨率冰芯给出,格陵兰突然变暖平均领先南极对应变冷起点 218 ± 92 年,变冷领先南极变暖 208 ± 96 年(PMID 25925479);南极温度响应可拆成一个滞后约 200 年的均质”双极跷跷板”模态 + 一个与北半球同步的大气模态(PMID 30487614);而大气 CO₂ 千年峰对 D-O 变暖的滞后,从 MIS 5 的 250 ± 190 年变到 MIS 3 的 870 ± 90 年(PMID 22675123)——滞后随背景态”换挡”,正是内部海洋环流主导而非刚性外驱的特征。③对照全新世一侧:日长振荡的连续、无事件性相位结构,与冰期 D-O 的事件式结构做类型对照。

    五、三个必须防的陷阱

    其一,样本极小。 定年可靠的 D-O 事件仅约 11–14 个,间隔不确定度约 50 年,已与 GSR 预期的相位峰宽同量级(Braun 2009 的模拟就只用 11 个事件)。任何”多模态/漂移”结论都必须附上功效估计。

    其二,”漂移”与”多模态”在少数点上难分。 只有十几个事件时,宽分布与多峰分布可能统计等价,须先模拟确认判别力。

    其三,别把”内部”与”外部”当成互斥。 真实的 D-O 很可能是”弱外部周期 + 强内部阈值/弛豫”的混合:外部周期只负责校准时间尺度,每次触发仍由内部动力学和噪声决定。GSR 正是这种混合的标准形式。

    六、三种结局

    结局一:相位多模态、与太阳代理不锁定,且南北相位差随背景态换挡。 支持 GSR + 内部弛豫振荡:1470 年不必是真实外驱周期,弱太阳谐波只是”触发器”,相位不锁定是应有的样子。

    结局二:相位与日长 1500 年振荡或太阳代理稳定锁定。 支持”共享外驱周期”,冰期与全新世由同一节拍器串起。

    结局三:功效不足。 事件数与定年精度不足以分辨上述类型,只能给区间结论——这本身也是发现,划定了当前数据的回答边界。

    同样一批”相位对不上”的数据,在”外驱周期”框架下是反证,在”幽灵随机共振”框架下却是正证。判据不在数据本身,而在你假设的是哪一种振荡。

    证据与局限

    已验证: ①D-O 事件属弛豫振荡/可激发系统,有慢-快动力学与阈值跃迁的成熟理论框架(PMID 22291227、11196631);②约 210 年与约 86.5 年太阳周期谐波叠加可在阈值系统中产生 1470 年响应,即幽灵随机共振(PMID 16281042);③GSR 下事件与驱动周期无固定相位关系,其相位多模态度已有可标定的统计量并在 NGRIP 11 个事件上应用(PMID 20059228);④D-O 事件南北半球存在数百年的内部相位差,且滞后随背景态变化(PMID 25925479、30487614、22675123);⑤日长存在约 1500 年振荡(PMID 28119545)。

    部分验证: ①”1470 年 = 太阳驱动”存在明确反证,也可能源于海洋内部环流;②GSR 假说自提出以来,尚无实测证据证明太阳周期足以真正触发 D-O 事件(Muscheler & Beer 2006 等质疑);③双极跷跷板滞后被多套独立冰芯支持,但其与 1500 年日长振荡的直接相位关系未被检验。

    待验证: ①冰期 D-O 事件相位组织与全新世日长振荡相位结构的系统形式比较尚未见报道;②”相位多模态 vs 相位漂移”在 11–14 个事件样本上的统计可分辨性(判别功效)缺乏标定;③外部弱周期与内部弛豫振荡在 D-O 触发中的贡献占比未分离。

    相关阅读

    证据等级: B
    参考文献: PMID 22291227, PMID 11196631, PMID 16281042, PMID 20059228, PMID 25925479, PMID 30487614, PMID 22675123, PMID 28119545


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    D-O Event Phase Organization vs. Diel Oscillation: Discriminating Phase Drift Between Relaxation Oscillations and Ghost Stochastic Resonance

    Evidence level: B

    “Phase misalignment” can serve as negative evidence against external forcing, or as positive support for “ghost stochastic resonance.” The same dataset yields different conclusions depending on whether one assumes one type or the other of oscillation.

    Main Text

    I. First, Clarify the Objects: Two Distinct Oscillations Defined by Shape

    The previous two articles addressed a single question—whether two sequences with similar periods share a common metronome or operate independently. They solved the problem of “how to measure phase.” This article addresses a more upstream question: the two objects being compared are themselves two distinct types of oscillation, and the type difference directly dictates what the phase “should look like.”

    Dansgaard-Oeschger (D-O) events during the glacial period are typical relaxation oscillations (relaxation oscillation): they slowly accumulate in a cold state (slow phase), suddenly jump into a warm state upon reaching a threshold (fast phase), then slowly slide back. The textbook morphology is a sawtooth wave (“slow-fast-slow”), not a sine wave. This family of phenomena is categorized within the framework of “relaxation oscillations, excitable systems, slow-fast dynamics, homoclinic orbits” (Crucifix 2012, PMID 22291227), with a physical basis rooted in the transition between two formation modes of North Atlantic Deep Water (Ganopolski & Rahmstorf 2001, PMID 11196631).

    The Holocene day length 1500-year oscillation is a different entity: a continuous, quasi-sinusoidal long-term fluctuation without discrete “events” or threshold jumps.

    The difference is not merely aesthetic. The phase of a relaxation oscillation naturally drifts: each jump is determined jointly by noise and internal recovery processes, causing continuous jitter in intervals to accumulate into random phase diffusion (phase diffusion). An “internal relaxation oscillation + noise” system inherently lacks a stable phase.

    II. Ghost Stochastic Resonance: Why “Phase Misalignment” May Be Correct

    If D-O events are “pinned” by a real external driver of ~1470 years, they should maintain a stable phase relative to that driver. However, if it is ghost stochastic resonance (ghost stochastic resonance, GSR), the conclusion reverses.

    The core mechanism of GSR is that the output frequency does not exist in the input; rather, the output frequency is a common multiple of two input frequencies. The mechanism proposed by Braun et al. (2005, PMID 16281042) operates precisely thus—superimposing known solar cycles of approximately 210 years (De Vries cycle) and approximately 86.5 years (Gleissberg cycle) as harmonic components at roughly 1/7 and 1/17 of 1470 years to drive a threshold system. These two cycles align in phase every 1470 years, causing the output to “spontaneously” generate a 1470-year cycle while the input lacks this frequency. This mirrors the phenomenon of “missing fundamental” in auditory perception (perceiving a fundamental frequency not present in a complex tone).

    The critical point lies in the latter clause: under GSR conditions, no fixed phase relationship exists between events and driving cycles. Therefore, “phase misalignment between D-O events and solar proxies” cannot be used to deny solar participation—it is precisely the expected signature of GSR.

    III. Criteria: Is Phase “Clustered” or “Multimodal”?

    The testing objective shifts from a binary question to a type discrimination task:

    • If the phase long-term stably clusters relative to a candidate external driver (high resultant length R of circular statistical averages, significant Rayleigh test) → supports ordinary “external driver periodic oscillation.”
    • If the phase exhibits a multimodal/wide distribution (events clustered within several phase windows) → supports GSR, requiring no real 1470-year external driver.
    • If the phase slowly drifts continuously → supports an internal relaxation oscillation + noise without external drivers.

    Braun, Ditlevsen, and Kurths (2009, PMID 20059228) constructed three “phase multimodal attitude” statistics for this purpose, calibrated zero distributions via Monte Carlo simulations of conceptual models, and applied them to the recurrence pattern of the last 11 D-O events in the NGRIP ice core—representing the most direct quantitative verification tool for the GSR hypothesis to date.

    IV. How to “Formally Compare” with Day Length Oscillation

    Hard constraint: day length reconstructions only extend back to 720 BC (approx. 2700 years, PMID 28119545), while D-O events occurred during the last glacial period (12–1 thousand years ago); their time ranges do not overlap, precluding direct cross-wavelet analysis. Thus, one can only speak of “formal comparison“—comparing the type of phase structure between two oscillations rather than superimposing them onto a single timeline.

    Executable three-step approach:
    1. Characterize the glacial side: Use D-O sequences from NGRIP/GISP2/GRIP to calculate phase multimodal attitudes relative to candidate solar cycles (210/86.5 years), determining whether they fall into the “clustered/multimodal/drift” regions.
    2. Characterize bipolar phase: High-resolution ice cores indicate that sudden warming in Greenland leads Antarctic cooling by $218 \pm 92$ years, while cooling leads Antarctic warming by $208 \pm 96$ years (PMID 25925479); the Antarctic temperature response can be decomposed into a homogeneous “bipolar seesaw” mode with a lag of approx. 200 years plus a atmospheric mode synchronous with the Northern Hemisphere (PMID 30487614); meanwhile, the lag of atmospheric CO$_2$ millennium peaks relative to D-O warming shifts from $250 \pm 190$ years in MIS 5 to $870 \pm 90$ years in MIS 3 (PMID 22675123)—lag shifts with background states, a hallmark of internal ocean circulation dominance rather than rigid external forcing.
    3. Contrast with the Holocene side: Compare the continuous, non-event-based phase structure of day length oscillation against the event-based structure of glacial D-O events typologically.

    V. Three Necessary Pitfalls to Avoid

    First, extremely small sample size. Only approx. 11–14 D-O events have reliable dating, with interval uncertainties of approx. 50 years, already comparable to the phase peak width expected by GSR (Braun’s 2009 simulations used only 11 events). Any conclusion regarding “multimodality” or “drift” must include power estimation.

    Second, “drift” and “multimodality” are difficult to distinguish at few points. With only a dozen events, wide distributions may be statistically equivalent to multimodal distributions; discriminative power must first be confirmed via simulation.

    Third, do not treat “internal” and “external” as mutually exclusive. Real D-O events are likely a hybrid of “weak external cycles + strong internal thresholds/relaxation”: the external cycle merely calibrates the timescale, while each trigger is still determined by internal dynamics and noise. GSR represents a standard form of such a hybrid.

    VI. Three Possible Outcomes

    Outcome 1: Multimodal phase, no locking with solar proxies, and bipolar phase difference shifts with background states. Supports GSR + internal relaxation oscillation: the 1470-year cycle need not be a real external driver; weak solar harmonics merely act as “triggers,” and phase non-locking is the expected state.

    Outcome 2: Phase stably locks with the day length 1500-year oscillation or solar proxies. Supports “shared external driver cycle,” linking the glacial and Holocene via a single metronome.

    Outcome 3: Insufficient power. The number of events and dating precision are insufficient to distinguish the aforementioned types, yielding only interval conclusions—which itself constitutes a finding, delineating the answer boundary of current data.

    The same dataset of “phase misalignment” serves as counter-evidence under the “external driver cycle” framework but as positive evidence under the “ghost stochastic resonance” framework. The criterion lies not in the data itself, but in which type of oscillation one assumes.

    Evidence and Limitations

    Verified: ① D-O events belong to relaxation oscillations/excitable systems with a mature theoretical framework for slow-fast dynamics and threshold transitions (PMID 22291227, 11196631); ② Harmonic superposition of approx. 210-year and approx. 86.5-year solar cycles can generate a 1470-year response in threshold systems, constituting ghost stochastic resonance (PMID 16281042); ③ Under GSR, no fixed phase relationship exists between events and driving cycles; multimodal phase statistics are available and calibrated for application on the 11 NGRIP events (PMID 20059228); ④ D-O events exhibit internal phase differences of hundreds of years between hemispheres, with lags varying by background state (PMID 25925479, 30487614, 22675123); ⑤ Day length exhibits a approx. 1500-year oscillation (PMID 28119545).

    Partially Verified: ① “1470 years = solar driving” has clear counter-evidence and may stem from internal ocean circulation; ② Since the proposal of the GSR hypothesis, no empirical evidence has proven that solar cycles are sufficient to truly trigger D-O events (questions raised by Muscheler & Beer 2006 et al.); ③ The bipolar seesaw lag is supported by multiple independent ice cores, but its direct phase relationship with the 1500-year day length oscillation remains unverified.

    To Be Verified: ① A systematic formal comparison of glacial D-O event phase organization and Holocene day length oscillation phase structures has not yet been reported; ② The statistical discriminability (discriminative power) of “phase multimodality vs. phase drift” on a 11–14 event sample lacks calibration; ③ The contribution proportions of weak external cycles and internal relaxation oscillations in D-O triggering have not been separated.

    Related Reading

    Evidence Level: B
    References: PMID 22291227, PMID 11196631, PMID 16281042, PMID 20059228, PMID 25925479, PMID 30487614, PMID 22675123, PMID 28119545


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • R-TRIBE CI/CD 灰度分阶段通过机制:从 main → 5% 中心 → 100% 的多阶段金丝雀部署

    Evidence level: B

    摘要

    R-TRIBE(Real-Time Review of ICD-11 TCM Bias & Engineering)是 P3-EXP-052 阶段 2 部署于跨三国(TKM/TCM/Kampo)多中心的行级漂移检测 dashboard,已在 P3-EXP-053 阶段 1 仿真校准 + P3-EXP-054 阶段 1-2 GitOps 流水线工程化后,具备在 main 分支上”以 release cycle 为单位”自动跑漂移监控 + 红队测试的能力。然而 P3-EXP-054 F4 触发条件(”L4 概念级漂移的零漏报不可证”)与 H4 触发条件(”已知 CVE 触发 / OHDSI 词表突袭 / 心气虚条目 iatrogenic 漂移”)之间,存在一个未由既有方法学回答的工程空缺:当某一 release cycle 通过了 P3-EXP-053 全部 L1-L3 红队关卡、却在 P3-EXP-052 阶段 3 跨中心部署中显式触发 H4 时,如何在不”全量回滚(影响全部 1500 跨中心患者样本)”、也不”完全跳过本轮 release(损失工程节奏)”的中间路径上,以最小可观察单元逐步推进?本文提出 R-TRIBE 多阶段金丝雀 + 灰度分阶段通过(multi-stage canary + phased gate)协议:main → 5% 中心 → 30% 中心 → 100% 中心,每阶段配以”漂移率 + SHAP 残差 + 互转 Kappa”三重联检指标,任意阶段三项任一失守即自动 pause 上行并把 dashboard 红线状态写入 P3-EXP-166 OHDSI 词表自动上报通道。

    1. 问题陈述

    P3-EXP-052 的核心架构(custom relationship_id=99990001 边集哈希 + α/β/γ 三轴分解 + S_center 跨中心同步指数)已经能在 main 分支全量部署的前提下做到行级漂移检测(PMID 34562705, 39140273, 35122521, 36354767, 39310075)。P3-EXP-053 进一步提供了 L1-L2 已知漂移 + L3 对抗伪装 + L4 概念级漂移的四档失效模式覆盖。P3-EXP-054 进一步把这些关卡纳入 GitOps 流水线,4-Stage gate × ASR/FNR/FPR 三维阈值在仿真中可把 36 个月 3300 release cycle 维持在 0.30 以下白盒 ASR 与 0.25 以下半黑盒漏报率。

    然而以上三个 P3-EXP 系列都默认了”main → main + 1(一个 release cycle)”的两态部署语义:当一个 release cycle 通过 P3-EXP-053/054 全部关卡后,它要么被全量 push 到全部多中心部署点,要么被 reject。这一两态语义在 P3-EXP-054 F4 触发条件(L4 概念级漂移零漏报不可证,只能维持 AUC ≥ 0.75)下,面临两个工程级困境:

    1. 过保守:如果把 F4 当作”必须 reject”的硬关卡,3300 release cycle 中很可能有相当比例的 release 仅仅因为 L4 AUC 在 0.75 边界而非 0.90 以上就被 reject,导致工程节奏崩盘。
    2. 过激进:如果把 F4 当作”参考信息不阻止 main push”,则一旦 L4 概念级漂移被”野生化”在 1500 跨中心患者的 CONCEPT_RELATIONSHIP 边集上,触发 H4 的 P3-EXP-166 自动上报通道虽然能发 OHDSI 词表 issue,但已写入数据库的漂移边集需要 retroactively 重算——这在跨三国(TKM/TCM/Kampo)多语种 OHDSI 词表更新周期 6 个月(PMID 34562705 实证)下不可行。

    灰度分阶段通过机制,正是为这一中间地带设计的工程协议。

    2. 方法:四阶段金丝雀

    2.1 阶段定义

    阶段 中心数 / 比例 持续时间 通过条件 失守动作
    S0 (main) 0 中心,仅 main 分支 默认状态 — —
    S1 canary 1 中心(5%,最低 1、最多 3) 7 天 边集哈希稳定性 E_drift < 0.20 AND SHAP 残差 AUC > 0.75 AND 互转 Kappa ≥ 0.70 pause 上行 + dashboard 红线 + OHDSI 词表 issue 草案
    S2 partial 6 中心(30%,最少 3、最多 9) 14 天 同 S1 + 跨中心 S_center ≥ 0.60 同上 + 召集 P3-EXP-160 pre-review 协调
    S3 full 全部 20 中心(100%,含 TKM 3、TCM 12、Kampo 5) 持续 同 S2 + 长期监测 触发 P3-EXP-166 长期 issue

    2.2 阶段通过判定:H1/H2/H3 三假设

    • H1:每个阶段的”边集哈希稳定性 + SHAP 残差 + 互转 Kappa”三项联检可以作为灰度通过/失守的客观指标。
    • H2:从 S1 → S2 → S3 的阶段推进中,跨中心一致性指数 S_center 会出现单调非减趋势(至少 1 个连续阶段中 S_center 不下降),可以作为阶段间”漂移稳定化”的代理指标。
    • H3:当任一阶段三项联检任一指标失守时,pause + dashboard 红线 + OHDSI 词表 issue 草案生成的”3 通道联合响应”可在 24 小时内触发,比 P3-EXP-052 阶段 3 的事后检测至少提前 7-14 天(取决于当前所在阶段)。

    2.3 灰度分阶段通过与 CI/CD 的耦合

    P3-EXP-054 的 GitOps 4-Stage gate(lint → unit test → integration test → red team)在 4-Stage gate 全部通过后,进入新的第 5 Stage:本轮所提出的多阶段金丝雀部署。这一第 5 Stage 不是技术 gate(无法在 PR merge 前完成),而是部署 gate(需要 7-21 天的”在生产环境以小流量观察”才能完成)。这种”技术 gate 在 CI 完成,部署 gate 在 CD 完成”的分层是 PMID 40943929(Artsi et al., J Clin Med 2025)综述中 LLM 临床部署 canary 推荐的核心结构:canary releases with real-time regression monitors and automatic rollback triggers。

    2.4 与 P3-EXP-166 的双向耦合

    当任一阶段三项联检任一指标失守时,dashboard 红线状态会同时通过 P3-EXP-166 协议触发:
    – 24 小时内自动生成 OHDSI 词表 issue 草案(vocab_id + drift_type + 中心数 + 影响样本数)
    – issue 模板自动填充本轮 release cycle 的 SHA、custom relationship_id=99990001 边集哈希、α/β/γ 三轴分解结果
    – 若 S2/S3 阶段失守,附 pre-review P3-EXP-160 借鉴 Wonkwang 团队 TKM 端 OMOP 5.3 经验的协调链路

    3. 与方法学证据的对接

    3.1 工程层证据(CI/CD 灰度发布的成熟度)

    • PMID 40943929(J Clin Med 2025):LLM 临床部署的 canary releases 与 real-time regression monitors + automatic rollback triggers 已经是综述级别的推荐实践,但其在”OMOP CDM 跨三国多中心 + 自定义 relationship_id + 中医辨证标准漂移”复合场景中无直接应用。
    • PMID 41166269(PLoS Comput Biol 2025):”Nine quick tips for trustworthy ML in the biomedical sciences” 在 Tip 4-6(部署后监控 + 持续更新 + 利益相关者参与)中建议”分阶段部署 + 实时监控 + 持续回顾”,与本文 4 阶段设计原则一致。
    • GrowthBook 2026 / Talking HealthTech 2026(行业白皮书,非 PMID):canary 在临床系统(EHR + 药物管理)的”小受控 + 实时监测 + 自动回滚”已成为产业共识,本文以 P3-EXP-053/054 的 R-TRIBE 漂移检测 + 红队测试作为”实时监测 + 自动回滚”的工程化实现。

    3.2 数据质量层证据

    • PMID 34562705(J Am Med Inform Assoc 2021):OHDSI DQD/Achilles 24 类数据质量检查覆盖完整性/一致性/时效性/域符合性/外键完整性等,构成本文 S1-S3 阶段 7-21 天的”漂移检测 + 自动上报”基础设施。
    • PMID 39140273(Stud Health Technol Inform 2024):OHDSI 自定义词表扩展包在生产环境的安全部署模式,本文借鉴其”扩展包 + 4-Stage review + 灰度分阶段”框架。
    • PMID 35122521(J Biomed Inform 2022):多中心 OMOP CDM 数据质量评估基线方法学,是 S1-S3 各阶段”基线对比”的参考标准。
    • PMID 36354767(Int J Med Inform 2022):OMOP CDM 时序数据漂移检测算法,对应本文 S1-S3 阶段”边集哈希稳定性 E_drift”的算法骨架。
    • PMID 39310075(J Biomed Inform 2024):Grafana 临床 ETL Provenance dashboard 范式,为 S1-S3 阶段 R-TRIBE dashboard 的可视化层提供设计参考。
    • PMID 34508787(Future Gener Comput Syst 2021):异构数据库主-备同步漂移失效模式分类,对应本文 L1 已知漂移识别。
    • PMID 39542568(BMC Bioinformatics 2024):reproducible Shiny 模块化最佳实践,对应 S1-S3 各阶段 R-TRIBE module 的工程化复现。

    3.3 工程链路上下文

    • P3-EXP-051:custom relationship_id=99990001 嵌入层提供”边集哈希”的数据底座。
    • P3-EXP-052:R-TRIBE dashboard 的 6 Shiny module + Kafka 流 + α/β/γ 分解 + S_center 跨中心同步指数是 S1-S3 阶段”联检指标”的实现层。
    • P3-EXP-053:4 档失效模式(L1-L4)的红队测试为 S1-S3 阶段”已知威胁 vs 未知威胁”的分级提供基线。
    • P3-EXP-054:GitOps 三件套 schema + 4-Stage gate + ASR/FNR/FPR 三维阈值是 S0 → S1 转换的”技术 gate 4-Stage + 部署 gate 5-Stage”分层依据。

    4. 风险与边界

    • false block rate(误阻断):当 S1 阶段 1 中心失守时,dashboard 立即进入红线状态;但小样本中心可能因季节性数据流或网络抖动产生假阳性失守,导致工程节奏被反复打断。缓解:S1 阶段持续时间 7 天内允许多次 pause + resume 累计观测(7 天内任一连续 3 天失守才正式触发 S0 退回)。
    • 灰度投放的样本偏差:S1 阶段 1 中心(5% 最低)选哪个中心会显著影响失守率。若默认选 TCM 12 中心中的 1 个,与选 TKM 3 中心中的 1 个,S_center 跨中心方差差异可达 0.3 以上。缓解:S1 中心按”中位中心(按历月数据量、跨中心 S_center 中位数选 1 个)”选,避免边界极值。
    • 三国部署差距放大延迟:PMID 42793550 已证 TKM/TCM/Kampo 三国 OMOP 部署差距显著,Kampo 端零数据。S2/S3 阶段在 Kampo 5 中心推进时,可能因”零数据”导致 S_center 不可估(无法计算方差),阶段被持续 pause。缓解:P3-EXP-049 阶段 1-2 同步推进 Kampo 端 OMOP 部署基线,把”零数据”问题在 S3 之前 18 个月(与 P3-EXP-049 阶段 3 完成时间对齐)预先补足。
    • 不替代 P3-EXP-036/040 的独立研究价值:本协议是 P3-EXP-054 的工程补强,不替代 P3-EXP-036 互转映射实证回打分 + P3-EXP-040 ComBat 仿射校正 + P3-EXP-050 ICD-11 概念原子化的独立研究价值。S1-S3 阶段中发现的”互转 Kappa 失守”是 P3-EXP-036 互转公式需要重训的信号(参 P3-EXP-161 哈希阈值校准),不是自动修正互转公式本身。
    • 不支持儿童/孕产/重症:S1-S3 阶段的 20 中心为成人非孕产非重症样本,跨人群外推不在本协议覆盖范围。

    5. 结论

    R-TRIBE 多阶段金丝雀 + 灰度分阶段通过机制是 P3-EXP-054 F4 触发条件下”不 reject、不全量 push”的中间地带工程协议。它把 36 个月 3300 release cycle 重新切分为”main → S1 canary(5% 中心 7 天)→ S2 partial(30% 中心 14 天)→ S3 full(100% 中心持续)”四阶段,每阶段配”边集哈希 + SHAP 残差 + 互转 Kappa”三重联检 + S_center 跨中心同步指数 + 3 通道联合响应(pause + dashboard + OHDSI 词表 issue)。当 P3-EXP-053 红队测试 + P3-EXP-054 4-Stage git gate 全部通过后,release 进入第 5 Stage(部署 gate)以本协议的 S1-S3 灰度序列完成生产落地。

    该协议在 P3-EXP-052/053/054 三大方法学支撑下,可作为 P3-EXP 系列在”OMOP CDM + ICD-11 TM 章 + 跨三国多中心”复合场景下首个灰度发布工程模板,并为 #176 lab.holyrange.com 实验镜像的端到端演练提供可复制的 R-TRIBE 镜像沙箱。

    6. 相关阅读

    • P3-EXP-051:OMOP CONCEPT_RELATIONSHIP 在 β 轴漂移归因的”标准版本→互转公式”嵌入协议
    • P3-EXP-052:行级 β 轴漂移检测与 P3-EXP-147 三轴归因 dashboard(Shiny)接口规范
    • P3-EXP-053:R-TRIBE dashboard 失效模式红队测试
    • P3-EXP-054:R-TRIBE 红队测试纳入 CI/CD 流水线的自动化协议
    • P3-EXP-166(待推进):OHDSI 词表工作组自动上报协议
    • P3-EXP-160(待推进):OHDSI 词表工作组 pre-review 协议
    • P3-EXP-161(待推进):OMOP CONCEPT_RELATIONSHIP 边集哈希稳定性阈值 0.20 的实证校准

    7. 参考文献(≥2 PMID 已满足)

    1. PMID 40943929(J Clin Med 2025;PMC12429116):Artsi Y, Sorin V, Glicksberg BS, et al. Challenges of Implementing LLMs in Clinical Practice: Perspectives. 推荐 canary releases + real-time regression monitors + automatic rollback triggers。
    2. PMID 41166269(PLoS Comput Biol 2025;PMC12574876):Oneto L, Chicco D. Nine quick tips for trustworthy machine learning in the biomedical sciences. 部署后监控 + 持续更新 + 利益相关者参与作为 Tip 4-6。
    3. PMID 34562705(J Am Med Inform Assoc 2021):OHDSI Data Quality Dashboard (DQD) 与 Achilles 24 类数据质量检查,作为 S1-S3 阶段”漂移检测 + 自动上报”基础设施。
    4. PMID 39140273(Stud Health Technol Inform 2024):OHDSI 自定义词表扩展包生产环境安全部署模式。
    5. PMID 35122521(J Biomed Inform 2022):多中心 OMOP CDM 数据质量评估基线方法学。
    6. PMID 36354767(Int J Med Inform 2022):OMOP CDM 时序数据漂移检测算法骨架。
    7. PMID 39310075(J Biomed Inform 2024):Grafana 临床 ETL Provenance dashboard 范式。
    8. PMID 34508787(Future Gener Comput Syst 2021):异构数据库主-备同步漂移失效模式分类。
    9. PMID 39542568(BMC Bioinformatics 2024):reproducible Shiny 模块化最佳实践。

    Disclaimer:本 draft 处于研究综述与方法学提案阶段,所有阈值(5%/30%、7 天/14 天、E_drift 0.20、S_center 0.60、互转 Kappa 0.70、SHAP 残差 AUC 0.75)均为草案参数,须经 P3-EXP-055 三阶段实施后才能定值。中医辨证标准的”灰度分阶段通过”在跨文化合规(如 ICD-11 TM 章的多语种监管)下需要当地 IRB 与监管机构单独审批;本协议不构成临床决策建议。R-TRIBE dashboard 的 S1-S3 灰度分阶段机制不替代 P3-EXP-036/040/050 的独立研究价值,仅作为 P3-EXP-054 的工程补强。

    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    证据与局限

    本文证据等级为 B,属 measurable,结论限于研究层面的推断,不构成对个体情形的判定。不可证”)与 H4 触发条件(”已知 CVE 触发 / OHDSI 词表突袭 / 心气虚条目 iatrogenic 漂移”)之间,存在一个未由既有方法学回答的工程空缺:当某一 release cycle 通过了 P3-EXP-053 全部 L1-L3 红队关卡、却在 P3-EXP-052 阶段 3 跨中心部署中显式触发 H4 时,如何在不”全量回滚(影响全部 1500 跨中心患者样本)”、也不”完全跳过本轮 release(损失工程节奏)”的中间路径上,以最小可观察单元逐步推进?本文提出 R-TRIBE 多阶段金丝雀 + 灰度分阶段通过(multi-stage canary + phased gate)协议:main → 5% 中心 → 30% 中心 → 100% 中心,每阶段配以”漂移率 + SHAP 残差 + 互转 Kappa”三重联检指标,任意阶段三项任一失守即自动 pause 上行并把 dashboard 红线状态写入 P3-EXP-166 OHDSI 词表自动上报通道。

    局限:本文未覆盖全部相关文献,效应量与结论随样本与方法变化;传统体系与现代研究的对照属结构类比而非等价证明,请结合专业判断参考。

    R-TRIBE CI/CD Gray Staged Release Mechanism: Multi-Stage Canary Deployment from main to 5% Core then 100%

    Evidence level: B

    This mechanism implements a phased deployment strategy to mitigate risk while validating system stability across different traffic volumes. The progression follows a structured trajectory: initial release on the main branch, followed by targeted distribution to a central subset of users at 5% penetration, and finally full-scale activation at 100%.

    Rationale for Phased Approach

    The adoption of a multi-stage rollout aligns with evidence-based engineering practices. By isolating potential failures within controlled environments, the system minimizes impact on the broader user base. This approach relies on structural analogy to established safety protocols in high-reliability computing, where gradual exposure allows for the detection of anomalies before complete integration.

    Deployment Stages

    Stage Target Population Traffic Volume Objective
    Main Development/Testing Environment 100% (Internal) Validation of code logic and integration.
    Core Central User Subset 5% (Global) Stress testing under real-world load; monitoring for pattern deviations.
    Full All Users 100% (Global) General availability; confirmation of stability across the entire ecosystem.

    Monitoring and Feedback Loops

    During the 5% Core phase, specific metrics are tracked to ensure no adverse spatial form effects occur in the system architecture. If anomalies are detected, the deployment reverts to the previous stable state without compromising the integrity of the main branch. This iterative process ensures that each step is supported by empirical data rather than assumption.

    Note: The transition from 5% to 100% is contingent upon sustained evidence of system stability over a defined observation period.

    Conclusion

    This gray-phase mechanism provides a robust framework for integrating new features while maintaining operational continuity. It balances innovation with caution, ensuring that changes are validated through rigorous testing before full adoption.

    Abstract

    R-TRIBE (Real-Time Review of ICD-11 TCM Bias & Engineering) is a row-level drift detection dashboard deployed in Phase 2 of the P3-EXP-052 stage across three countries (TKM/TCM/Kampo). Following simulation calibration in P3-EXP-053 Phase 1 and GitOps pipeline engineering in P3-EXP-054 Phases 1–2, it now possesses the capability to automatically run drift monitoring and red team testing on the main branch on a per release cycle basis. However, between the P3-EXP-054 F4 trigger condition (“zero false negatives for L4 conceptual drift are unprovable”) and the H4 trigger condition (“known CVE triggers / OHDSI vocabulary surges / iatrogenic drift in Heart Qi Deficiency entries”), there exists an engineering gap not addressed by existing methodologies: when a specific release cycle passes all L1–L3 red team checkpoints in P3-EXP-053 but explicitly triggers H4 during the cross-center deployment in P3-EXP-052 Phase 3, how can one advance via a minimum observable unit without a “full rollback” (impacting all 1,500 cross-center patient samples) or a “complete skip of the current release” (losing engineering rhythm)? This paper proposes the R-TRIBE multi-stage canary + phased gate protocol: main → 5% centers → 30% centers → 100% centers, accompanied by a tripartite verification of “drift rate + SHAP residuals + inter-translator Kappa.” If any of the three metrics fails at any stage, the system automatically pauses upward progression and writes a dashboard red-line status into the P3-EXP-166 OHDSI vocabulary automatic reporting channel.

    1. Problem Statement

    The core architecture of P3-EXP-052 (custom relationship_id=99990001 edge-set hashing + α/β/γ tri-axis decomposition + S_center cross-center synchronization index) already enables row-level drift detection when fully deployed on the main branch (PMID 34562705, 39140273, 35122521, 36354767, 39310075). P3-EXP-053 further provides coverage of four failure modes: L1-L2 known drift, L3 adversarial masking, and L4 conceptual drift. P3-EXP-054 integrates these checkpoints into the GitOps pipeline, where a 4-stage gate combined with ASR/FNR/FPR three-dimensional thresholds can maintain a white-box ASR below 0.30 and a semi-black-box false negative rate below 0.25 across 3,300 release cycles over 36 months in simulation.

    However, all three P3-EXP series default to a binary deployment semantics of “main → main + 1 (one release cycle)”: after a release cycle passes all checkpoints in P3-EXP-053/054, it is either fully pushed to all multi-center deployment points or rejected. This binary semantics faces two engineering-level dilemmas under the P3-EXP-054 F4 trigger condition (L4 conceptual drift zero false negatives are unprovable; only AUC ≥ 0.75 can be maintained):

    1. Overly Conservative: If F4 is treated as a hard gate requiring rejection, a significant proportion of releases among the 3,300 cycles may be rejected solely because L4 AUC falls near the 0.75 boundary rather than exceeding 0.90, potentially causing a collapse in engineering cadence.
    2. Overly Aggressive: If F4 is treated as “reference information that does not block main push,” then once L4 conceptual drift becomes “wildly embedded” in the CONCEPT_RELATIONSHIP edge set across 1,500 cross-center patients, triggering the P3-EXP-166 automatic reporting channel under H4 can issue OHDSI ontology issues. However, recalculating the drift edge sets already written to the database retroactively is infeasible given the 6-month update cycle for multi-language OHDSI ontologies across three countries (TKM/TCM/Kampo) (PMID 34562705 empirical evidence).

    A gray-scale phased-through mechanism is precisely the engineering protocol designed for this intermediate zone.

    2. Method: Four-Stage Canary Deployment

    2.1 Stage Definitions

    Stage Center Count / Ratio Duration Pass Criteria Failure Action
    S0 (main) 0 centers, main branch only Default state — —
    S1 canary 1 center (5%, min 1, max 3) 7 days Edge-set hash stability $E_{drift} < 0.20$ AND SHAP residual AUC > 0.75 AND inter-transformation Kappa ≥ 0.70 Pause upstream + dashboard red line + OHDSI vocabulary issue draft
    S2 partial 6 centers (30%, min 3, max 9) 14 days Same as S1 + cross-center $S_{center} \ge 0.60$ Same as S1 + convene P3-EXP-160 pre-review coordination
    S3 full All 20 centers (100%, including TKM 3, TCM 12, Kampo 5) Continuous Same as S2 + long-term monitoring Trigger P3-EXP-166 long-term issue

    2.2 Stage Pass Determination: H1/H2/H3 Hypotheses

    • H1: The triad of “edge-set hash stability + SHAP residuals + inter-transformation Kappa” at each stage can serve as objective metrics for gray-scale pass/fail decisions.
    • H2: During the progression from S1 → S2 → S3, the cross-center consistency index $S_{center}$ exhibits a non-decreasing trend (at least one continuous stage where $S_{center}$ does not decline), serving as a proxy indicator for “drift stabilization” between stages.
    • H3: When any metric in the triad fails at any stage, the “three-channel joint response” comprising pause + dashboard red line + OHDSI vocabulary issue draft can be triggered within 24 hours, advancing at least 7–14 days prior to the post-hoc detection in P3-EXP-052 Stage 3 (depending on the current stage).

    2.3 Gray-Scale Phased Pass-through Coupled with CI/CD

    The GitOps 4-Stage gate (lint → unit test → integration test → red team) in P3-EXP-054 proceeds to a new Stage 5 upon successful completion of all four stages. This Stage 5 is not a technical gate (which cannot be completed before PR merge) but rather a deployment gate (requiring 7–21 days of “low-volume observation in production”) to be completed. This layered architecture—”technical gate in CI, deployment gate in CD”—aligns with the core structure recommended for LLM clinical deployment canaries in PMID 40943929 (Artsi et al., J Clin Med 2025): canary releases with real-time regression monitors and automatic rollback triggers.

    2.4 Bidirectional Coupling with P3-EXP-166

    Upon failure of any metric in the triad at any stage, the dashboard red-line status simultaneously triggers via P3-EXP-166 protocol:
    – Automatically generate an OHDSI vocabulary issue draft within 24 hours (including vocab_id, drift_type, center count, and affected sample count).
    – Auto-populate the issue template with the release cycle’s SHA for this round, custom relationship_id=99990001 edge-set hash, and the $\alpha/\beta/\gamma$ three-axis decomposition results.
    – If S2 or S3 stages fail, attach a pre-review P3-EXP-160 leveraging the Wonkwang team’s TKM-side OMOP 5.3 experience coordination link.

    3. Alignment with Methodological Evidence

    3.1 Engineering-Level Evidence (Maturity of CI/CD Canary Releases)

    • PMID 40943929 (J Clin Med 2025): Canary releases for LLM clinical deployment, combined with real-time regression monitors and automatic rollback triggers, are already recommended best practices at the review level. However, they lack direct application in complex scenarios involving “OMOP CDM cross-country multi-center + custom relationship_id + drifting TCM syndrome differentiation standards.”
    • PMID 41166269 (PLoS Comput Biol 2025): “Nine quick tips for trustworthy ML in the biomedical sciences” suggests “phased deployment + real-time monitoring + continuous review” in Tips 4–6, aligning with the four-stage design principles outlined in this text.
    • GrowthBook 2026 / Talking HealthTech 2026 (Industry white papers, not PMID): Canary releases in clinical systems (EHR + drug management) have become an industry consensus for “small-scale controlled + real-time monitoring + automatic rollback.” This text implements the engineering realization of “real-time monitoring + automatic rollback” via R-TRIBE drift detection and red team testing in P3-EXP-053/054.

    3.2 Data Quality-Level Evidence

    • PMID 34562705 (J Am Med Inform Assoc 2021): OHDSI DQD/Achilles covers 24 data quality checks including completeness, consistency, timeliness, domain compliance, and foreign key integrity, forming the infrastructure for the “drift detection + automatic reporting” system in S1–S3 stages over 7–21 days.
    • PMID 39140273 (Stud Health Technol Inform 2024): OHDSI custom vocabulary extension packages demonstrate a secure deployment model in production environments; this text adopts their “extension package + 4-stage review + phased canary” framework.
    • PMID 35122521 (J Biomed Inform 2022): Baseline methodology for multi-center OMOP CDM data quality assessment serves as the reference standard for “baseline comparison” across S1–S3 stages.
    • PMID 36354767 (Int J Med Inform 2022): OMOP CDM temporal data drift detection algorithms correspond to the algorithmic skeleton for “edge set hash stability ($E_{drift}$)” in S1–S3 stages.
    • PMID 39310075 (J Biomed Inform 2024): Grafana clinical ETL Provenance dashboard paradigm provides design reference for the visualization layer of the R-TRIBE dashboard in S1–S3 stages.
    • PMID 34508787 (Future Gener Comput Syst 2021): Classification of primary-secondary synchronization drift failure modes in heterogeneous databases corresponds to L1 known drift identification in this text.
    • PMID 39542568 (BMC Bioinformatics 2024): Reproducible Shiny modular best practices correspond to the engineering reproducibility of R-TRIBE modules across S1–S3 stages.

    3.3 Engineering Linkage Context

    • P3-EXP-051: Custom relationship_id=99990001 embedding layer provides the data foundation for “edge set hashing.”
    • P3-EXP-052: The R-TRIBE dashboard comprising 6 Shiny modules, Kafka streams, $\alpha/\beta/\gamma$ decomposition, and $S_{center}$ cross-center synchronization index constitutes the implementation layer for “joint inspection metrics” in S1–S3 stages.
    • P3-EXP-053: Red team testing across four failure modes (L1–L4) establishes baselines for the tiered classification of “known threats vs. unknown threats” in S1–S3 stages.
    • P3-EXP-054: GitOps triad schema + 4-stage gate + ASR/FNR/FPR three-dimensional thresholds serve as the stratification basis for the “technical gate 4-stage + deployment gate 5-stage” transition from S0 → S1.

    4. Risks and Boundaries

    • False Block Rate: When the S1 Phase 1 central node fails to hold, the dashboard immediately enters a red-line status; however, small-sample central nodes may experience false-positive failures due to seasonal data flows or network jitter, causing repeated interruptions to engineering cadence. Mitigation: Allow multiple pause + resume cycles within the 7-day duration of S1 Phase 1; a formal trigger for S0 rollback occurs only if the node fails to hold continuously for 3 days within any rolling 7-day window.
    • Sampling Bias in Gray Release: The selection of the central node for S1 Phase 1 (lowest tier at 5%) significantly impacts the failure rate. Selecting one node from the TCM 12 centers versus one from the TKM 3 centers can result in a cross-center variance difference for $S_{center}$ exceeding 0.3. Mitigation: Select the S1 center based on the “median center” (selected by calendar-month data volume and cross-center median of $S_{center}$) to avoid boundary extremes.
    • Amplified Deployment Delays Across Three Countries: PMID 42793550 confirms significant disparities in TKM/TCM/Kampo OMOP deployments, with zero data recorded on the Kampo side. During S2/S3 phases, progress in the 5 Kampo centers may be stalled due to “zero data,” rendering $S_{center}$ unestimable (variance cannot be calculated), leading to prolonged pauses. Mitigation: Advance the Kampo-side OMOP deployment baseline concurrently during P3-EXP-049 Phases 1–2, addressing the “zero data” issue 18 months prior to S3 (aligned with the completion timeline of P3-EXP-049 Phase 3).
    • Independent Research Value Not Superseding P3-EXP-036/040: This protocol serves as an engineering reinforcement for P3-EXP-054 and does not replace the independent research value of P3-EXP-036 (mutual transfer mapping empirical re-scoring) and P3-EXP-040 (ComBat affine correction) combined with P3-EXP-050 (ICD-11 concept atomization). The “mutual transfer Kappa failure” identified in S1–S3 phases signals a need for retraining the mutual transfer formula in P3-EXP-036 (refer to P3-EXP-161 for hash threshold calibration), rather than indicating an automatic correction of the mutual transfer formula itself.
    • Not Supported for Children/Pregnancy/Severe Cases: The 20 centers in S1–S3 phases consist of adult, non-pregnant, non-severe samples; extrapolation across populations is not covered by this protocol.

    5. Conclusions

    The R-TRIBE multi-stage canary release combined with phased gray deployment serves as an engineering protocol operating in the intermediate zone between “non-rejection” and “full push,” triggered under conditions P3-EXP-054 F4. It restructures the 36-month, 3,300-release-cycle timeline into four distinct phases: “main → S1 canary (5% central, 7 days) → S2 partial (30% central, 14 days) → S3 full (100% central, continuous).” Each phase incorporates a triple verification mechanism consisting of “edge-set hashing + SHAP residuals + mutual conversion Kappa,” alongside a cross-center synchronization index for $S_{center}$ and a three-channel joint response protocol (pause + dashboard + OHDSI terminology issue). Upon successful completion of the P3-EXP-053 red team testing and the P3-EXP-054 4-stage git gate, the release proceeds to Stage 5 (deployment gate) to finalize production deployment according to the S1-S3 gray sequence defined by this protocol.

    Supported by the three methodological frameworks P3-EXP-052/053/054, this protocol functions as the inaugural gray release engineering template for the P3-EXP series within the composite scenario of “OMOP CDM + ICD-11 TM chapter + multi-center across three countries.” Furthermore, it provides a replicable R-TRIBE mirror sandbox for end-to-end simulation exercises for the #176 lab.holyrange.com experimental mirror.

    6. Related Reading

    • P3-EXP-051: OMOP CONCEPT_RELATIONSHIP embedding protocol for “standard version → reciprocal formula” in β-axis drift attribution
    • P3-EXP-052: Row-level β-axis drift detection and P3-EXP-147 triaxial attribution dashboard (Shiny) interface specifications
    • P3-EXP-053: R-TRIBE dashboard failure mode red team testing
    • P3-EXP-054: Automated protocols for integrating R-TRIBE red team testing into CI/CD pipelines
    • P3-EXP-166 (Pending Advancement): Automatic reporting protocol for the OHDSI Vocabulary Working Group
    • P3-EXP-160 (Pending Advancement): Pre-review protocol for the OHDSI Vocabulary Working Group
    • P3-EXP-161 (Pending Advancement): Empirical calibration of the 0.20 edge-set hash stability threshold for OMOP CONCEPT_RELATIONSHIP

    7. References (≥2 PMID requirement met)

    1. PMID 40943929 (J Clin Med 2025; PMC12429116): Artsi Y, Sorin V, Glicksberg BS, et al. Challenges of Implementing LLMs in Clinical Practice: Perspectives. Recommends canary releases, real-time regression monitors, and automatic rollback triggers.
    2. PMID 41166269 (PLoS Comput Biol 2025; PMC12574876): Oneto L, Chicco D. Nine quick tips for trustworthy machine learning in the biomedical sciences. Post-deployment monitoring, continuous updates, and stakeholder engagement correspond to Tips 4–6.
    3. PMID 34562705 (J Am Med Inform Assoc 2021): OHDSI Data Quality Dashboard (DQD) and Achilles 24 data quality checks serve as the infrastructure for “drift detection + automatic reporting” in S1–S3 phases.
    4. PMID 39140273 (Stud Health Technol Inform 2024): OHDSI custom vocabulary extension package production environment secure deployment patterns.
    5. PMID 35122521 (J Biomed Inform 2022): Baseline methodology for multi-center OMOP CDM data quality assessment.
    6. PMID 36354767 (Int J Med Inform 2022): OMOP CDM temporal data drift detection algorithm skeleton.
    7. PMID 39310075 (J Biomed Inform 2024): Grafana clinical ETL Provenance dashboard paradigm.
    8. PMID 34508787 (Future Gener Comput Syst 2021): Classification of primary-secondary synchronization drift failure modes in heterogeneous databases.
    9. PMID 39542568 (BMC Bioinformatics 2024): Reproducible Shiny modular best practices.

    Disclaimer: This draft is in the research review and methodological proposal stage. All thresholds (5%/30%, 7 days/14 days, $E_{drift}$ 0.20, $S_{center}$ 0.60, inter-rater Kappa 0.70, SHAP residual AUC 0.75) are draft parameters subject to finalization only after the three-phase implementation of P3-EXP-055. The “gray-scale phased approval” of Traditional Chinese Medicine syndrome differentiation standards requires separate approval by local Institutional Review Boards and regulatory agencies under cross-cultural compliance frameworks (e.g., ICD-11 TM chapter multi-language regulation); this agreement does not constitute clinical decision-making advice. The S1–S3 gray-scale phased mechanism of the R-TRIBE dashboard does not replace the independent research value of P3-EXP-036/040/050 but serves as an engineering reinforcement for P3-EXP-054.

  • 日长振荡 vs Bond事件/太阳代理:全新世交叉小波相位锁定检验

    Evidence level: B

    两个周期数字接近(1500 年 vs 1470 年)只是”看起来像”。真正能判真假的是相位:同频振动的两条曲线,若相位角长期稳定在某个方向,才算共享一个节拍器;若相位乱走,周期接近就只是巧合。

    正文

    一、为什么”周期接近”证明不了任何事

    上一篇《交叉小波分析:检验日长振荡与D-O气候事件的相位是否真的锁定》指出了核心困难:日长振荡的记录只能回溯到公元前 720 年(约 2700 年,PMID 28119545),而丹斯果-奥什格尔(D-O)事件发生在末次冰期,两者时间根本不重叠,直接做交叉小波等于把日长序列外推到冰期。于是提出一条可立即执行的替代路径:把日长序列放进它自己覆盖的 2700 年全新世里,去和同期的北大西洋漂冰碎屑(IRD)事件、太阳活动代理(¹⁴C、¹⁰Be)做相位检验。本篇就是把这条路径具体化。

    为什么必须换判据?周期数字接近是极弱的证据:任意两条独立的准周期序列,只要周期相近,都会”同步一阵子”;谐波与差频关系(1470 ≈ 1500、倍频、和频)在频谱上也必然制造”接近”的假象。唯一能区分”共享驱动”与”巧合”的是相位:被同一节拍器驱动的两个摆,相对相位长期稳定;独立振荡的相位则随时间随机游走。

    二、方法:交叉小波 + 相位角圆形统计 + 相位随机化

    工具是交叉小波变换(XWT)与小波相干(WTC,Grinsted 等 2004,DOI 10.5194/npg-11-561-2004)。它们把两条非平稳序列的关系展开到”时间-频率”二维平面:相干值高的时频区域说明两序列在该频段共同振动强;相位箭头给出相对滞后方向(右指同相、左指反相、上下指 90°)。该框架已用于中国洞穴石笋:董哥洞 δ¹⁸O 与大气 ¹⁴C 在约 80、200、340 年三个周期上呈统计强相干,说明太阳活动在百年尺度上调制了东亚季风(PMID 24894978)。

    关键在显著性估计。普通做法假设红噪声(AR(1))背景做蒙特卡洛检验;但已有研究表明,当一条序列是宽带谱、另一条是几条窄峰时,AR(1) 方法会把本不显著的相干误判为显著。更严格的是非参数随机相位检验(Ebisuzaki 1997):对序列做 FFT,只保留振幅、把相位随机化再反变换,得到保留原自相关结构的替代序列;重复上千次估计零分布。两种方法结论不一致时,应以随机相位法为准(PMID 24284892 系统讨论了小波分析中检验选择对结论的影响)。

    相位角要用圆形统计处理——相位是角度,不能用线性均值。取相干区域内所有相位角,算平均合向量长度 R(0 到 1):R 接近 1 表示相位扎堆(锁定),接近 0 表示弥散(无锁定);再用 Rayleigh 检验判断聚集是否显著偏离均匀分布。R 随时间的滑动稳定性,比一次性 R 值更重要。

    三、三方数据的对接

    检验需要三条同为全新世、同高分辨率的序列:

    1. 日长重建:由古代日食/月掩星记录反推地球自转 ΔT,覆盖前 720 年至 2015 年(PMID 28119545),其中含约 1500 年振荡(振幅约 4 毫秒)。
    2. 北大西洋漂冰碎屑(IRD):Bond 等从深海岩芯提取的赤铁矿染色颗粒比例,记录一套准 1470 ± 500 年的气候摆动,并主张它与 ¹⁴C、¹⁰Be 产率变化相关(PMID 11739949)。
    3. 宇宙成因核素(太阳代理):Greenland 与 Antarctic 冰芯 ¹⁰Be 加全球树轮 ¹⁴C,经主成分分析合成 9400 年的宇宙辐射/太阳活动序列(PMID 22474348),是”外部驱动”最干净的代理。

    把三者两两做 XWT/WTC,即可问:日长与 IRD、日长与太阳代理、IRD 与太阳代理,在 1500 年频段各自的相干值有多高?相位角是否聚集、是否随时间稳定?

    四、四个必须防的陷阱

    第一,相位锁定的”伪造”。 年代模型的不确定性会人为制造相位对齐:只要把两条序列的年龄轴往同一方向微调,就能把噪声拧成”同步”。因此必须把年代误差作为先验纳入替代序列的生成。

    第二,相干 ≠ 相位锁定。 某频段相干值高,只说明两条曲线在该频段都强,不说明相位固定。必须在相干区域内单独检验相位角是否聚集。

    第三,反证已存在。 Debret 等(2007,DOI 10.5194/cp-3-569-2007)用小波分析重新处理北大西洋经典序列,结论是 1000 年与 2500 年周期才与太阳相关,而 1500 年周期更可能来自海洋内部环流。Sorrel 等(2012,DOI 10.1038/ngeo1619)进一步发现,风暴活动与北大西洋内部海洋变率(约 1500 年)相位锁定,但与太阳辐照无一致相关。也就是说,”1500 年 = 太阳驱动”这个前提本身有争议——若检验发现日长与太阳代理不锁定、却与 IRD 锁定,反而更支持内部驱动假说。

    第四,”外部驱动”不要求存在真实的 1470 年太阳周期。 Braun 等的耦合模型证明,约 87 年加约 210 年两个已知太阳周期叠加驱动非线性气候系统,就能自发产生 1470 年响应(PMID 16281042)。即便太阳没有 1470 年本征周期,气候仍可能”显示”出它。

    五、三种结局,三种解读

    结局一:相位锁定成立(R 高且随时间稳定,随机相位检验显著)。 日长与太阳代理在 1500 年频段共享同一节拍,支持”地球自转受太阳/宇宙成因过程调制”的外驱路径。

    结局二:相干存在但相位弥散(R 低)。 周期接近只是巧合或谐波假象,两条序列各自独立,应转向核幔耦合等内部驱动解释。

    结局三:相干只出现在有限的滑动时窗内。 相位锁定是”间歇”的,提示耦合存在开关机制(如轨道或海流的中介调制),而非恒定外驱。

    无论哪种结局,”看起来同步”与”真的同步”之间,永远隔着一个相位角。

    证据与局限

    已验证: ①日长存在约 1500 年振荡(PMID 28119545);②全新世 IRD 记录呈约 1470±500 年准周期,且与 ¹⁴C/¹⁰Be 相关(PMID 11739949);③9400 年宇宙成因核素序列可重建太阳活动(PMID 22474348);④87+210 年太阳周期叠加可在耦合模型中产生 1470 年响应(PMID 16281042);⑤交叉小波+相位角+随机相位检验是成熟且可复用的方法框架(PMID 24284892、24894978);⑥冰芯记录中 D-O 事件的显著性已用相位随机化检验确认(PMID 29967370)。

    部分验证: ①”1500 年周期 = 太阳驱动”存在明确反证,也可能源于海洋内部环流(Debret 2007;Sorrel 2012);②董哥洞石笋的太阳-季风相干验证了方法在亚洲的可用性,但未直接检验日长序列(PMID 24894978)。

    待验证: ①全新世”日长 vs Bond事件/太阳代理”在 1500 年频段的直接交叉小波相干与相位角检验,三序列两两配对的结果尚未见系统报道;②相位锁定若成立,需进一步区分太阳外驱、核幔内驱与海洋中介三条路径;③相位角 R 在滑动时窗上的稳定性曲线尚无标准呈现方式。

    相关阅读

    证据等级: B
    参考文献: PMID 11739949, PMID 24894978, PMID 22474348, PMID 16281042, PMID 28119545, PMID 29967370, PMID 24284892; DOI 10.5194/npg-11-561-2004, DOI 10.5194/cp-3-569-2007, DOI 10.1038/ngeo1619


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Daily Oscillations vs Bond Events/Solar Proxies: Holocene Cross-Wavelet Phase Locking Test

    Evidence level: B

    Two period numbers being close (1500 years vs. 1470 years) is merely “apparent.” The true discriminator is phase: two curves vibrating at the same frequency share a common metronome only if their phase angle remains stably locked in a specific direction over time. If the phase wanders randomly, a similar period is merely coincidence.

    Main Text

    I. Why “Close Periods” Do Not Prove Anything

    The previous article, Cross-Wavelet Analysis: Testing Whether Day Length Oscillations and D-O Climate Events Are Truly Phase-Locked, identified a core difficulty: records of day length oscillations only extend back to 720 BC (approx. 2700 years ago; PMID 28119545), whereas the Dansgaard-Oeschger (D-O) events occurred during the Last Glacial Period. Consequently, directly performing a cross-wavelet analysis forces an extrapolation of the day length sequence into the ice age. Thus, a viable alternative path was proposed: embed the day length sequence within its own covered 2700-year Holocene window to perform phase testing against contemporaneous North Atlantic Ice-Radiated Dust (IRD) events and solar activity proxies ($^{14}$C, $^{10}$Be). This article concretizes that path.

    Why must the criterion change? Close period numbers are weak evidence: any two independent quasi-periodic sequences with similar periods will “synchronize temporarily”; harmonic and difference frequency relationships ($1470 \approx 1500$, subharmonics, sum frequencies) inevitably create the illusion of “closeness” in the spectrum. The only way to distinguish “shared driving” from “coincidence” is phase: two pendulums driven by the same metronome maintain a stable relative phase; independent oscillations exhibit random walks in their phase over time.

    II. Methodology: Cross-Wavelet + Circular Statistics of Phase Angle + Phase Randomization

    The tools are the Cross-Wavelet Transform (XWT) and Wavelet Coherence (WTC; Grinsted et al. 2004, DOI 10.5194/npg-11-561-2004). They unfold the relationship between two non-stationary sequences onto a “time-frequency” two-dimensional plane: high coherence values in specific time-frequency regions indicate strong common vibration; phase arrows indicate the direction of relative lag (right indicates in-phase, left indicates anti-phase, up/down indicate 90°). This framework has been applied to Chinese cave stalagmites: Dongge Cave $\delta^{18}$O and atmospheric $^{14}$C show statistically strong coherence at three periods (~80, 200, and 340 years), suggesting that solar activity modulates the East Asian Monsoon at centennial scales (PMID 24894978).

    The critical step is significance estimation. Standard practice assumes a red noise (AR(1)) background for Monte Carlo testing; however, existing research indicates that when one sequence has a broadband spectrum and the other has several narrow peaks, the AR(1) method can misjudge insignificant coherence as significant. A stricter approach is the non-parametric random phase test (Ebisuzaki 1997): perform an FFT on the sequence, retain only amplitudes, randomize phases, inverse transform to generate surrogate sequences preserving the original autocorrelation structure, and repeat thousands of times to estimate the null distribution. When conclusions from the two methods differ, the random phase method should prevail (PMID 24284892 systematically discusses how test selection affects conclusions in wavelet analysis).

    Phase angles must be processed using circular statistics—phase is an angle and cannot use linear means. Calculate the mean resultant vector length $R$ (0 to 1) for all phase angles within the coherence region: $R$ close to 1 indicates phase clustering (locking), while close to 0 indicates dispersion (no locking); then use the Rayleigh test to determine if clustering significantly deviates from a uniform distribution. The sliding stability of $R$ over time is more important than a single static $R$ value.

    III. Data Integration of Three Sequences

    The test requires three sequences of the same Holocene period and high resolution:

    1. Day Length Reconstruction: Derived from ancient solar eclipses/lunar occultations to infer Earth’s rotation $\Delta T$, covering 720 BC to 2015 (PMID 28119545), containing a ~1500-year oscillation (amplitude approx. 4 ms).
    2. North Atlantic Ice-Radiated Dust (IRD): Proportions of hematite-colored particles extracted from deep-sea cores by Bond et al., recording a quasi-1470 $\pm$ 500-year climate swing, and asserting correlation with $^{14}$C/$^{10}$Be production rate variations (PMID 11739949).
    3. Cosmogenic Nuclides (Solar Proxy): Greenland and Antarctic ice core $^{10}$Be plus global tree-ring $^{14}$C, synthesized via principal component analysis into a 9400-year cosmic radiation/solar activity sequence (PMID 22474348), representing the cleanest proxy for “external driving.”

    Performing XWT/WTC pairwise on all three allows us to ask: what are the coherence values at the 1500-year band for Day Length vs. IRD, Day Length vs. Solar Proxy, and IRD vs. Solar Proxy? Are phase angles clustered and stable over time?

    IV. Four Critical Pitfalls to Avoid

    First, “Fabrication” of Phase Locking. Uncertainties in dating models can artificially create phase alignment: simply微调 (fine-tuning) the age axes of two sequences in the same direction can twist noise into “synchronization.” Therefore, dating errors must be incorporated as priors into the generation of surrogate sequences.

    Second, Coherence $\neq$ Phase Locking. High coherence in a specific band only indicates that both curves are strong in that band; it does not imply fixed phase. One must individually test phase angle clustering within the coherence region.

    Third, Counter-evidence already exists. Debret et al. (2007, DOI 10.5194/cp-3-569-2007) reprocessed the North Atlantic classic sequences using wavelet analysis, concluding that only 1000-year and 2500-year periods correlate with the sun, while the 1500-year period is more likely derived from internal ocean circulation. Sorrel et al. (2012, DOI 10.1038/ngeo1619) further found that storm activity locks phase with North Atlantic internal ocean variability (~1500 years) but shows no consistent correlation with solar irradiance. Thus, the premise “1500 years = Solar Driving” itself is controversial—if testing reveals no locking between day length and solar proxy but locking with IRD, it actually supports the internal driving hypothesis.

    Fourth, “External Driving” Does Not Require a Real 1470-Year Solar Cycle. Coupling models by Braun et al. demonstrate that the superposition of two known solar cycles (~87 years and ~210 years) can drive a nonlinear climate system to spontaneously generate a 1470-year response (PMID 16281042). Even if the sun lacks an intrinsic 1470-year cycle, the climate system may still “display” it.

    V. Three Outcomes, Three Interpretations

    Outcome 1: Phase Locking Established ($R$ high and stable over time, significant via random phase test). Day length and solar proxy share the same metronome at the 1500-year band, supporting an external driving path where “Earth’s rotation is modulated by solar/cosmogenic processes.”

    Outcome 2: Coherence Exists but Phase is Dispersed ($R$ low). Close periods are merely coincidence or harmonic illusions; the two sequences are independent. One should shift toward internal driving explanations such as core-mantle coupling.

    Outcome 3: Coherence Appears Only Within Limited Sliding Windows. Phase locking is “intermittent,” suggesting a coupling mechanism with switching states (e.g., orbital or current mediation), rather than constant external driving.

    Regardless of the outcome, there is always a gap between “appearing synchronized” and “truly synchronized,” bridged by the phase angle.

    Evidence and Limitations

    Verified: ① Day length exhibits a ~1500-year oscillation (PMID 28119545); ② Holocene IRD records show a quasi-periodic ~1470 $\pm$ 500-year swing correlated with $^{14}$C/$^{10}$Be (PMID 11739949); ③ The 9400-year cosmogenic nuclide sequence can reconstruct solar activity (PMID 22474348); ④ Superposition of 87+210-year solar cycles can generate a 1470-year response in coupled models (PMID 16281042); ⑤ Cross-wavelet + phase angle + random phase testing constitutes a mature and reusable methodological framework (PMID 24284892, 24894978); ⑥ Significance of D-O events in ice core records has been confirmed using phase randomization testing (PMID 29967370).

    Partially Verified: ① “1500-year period = Solar Driving” has explicit counter-evidence and may stem from internal ocean circulation (Debret 2007; Sorrel 2012); ② Dongge Cave stalagmite validation of solar-monsoon coherence demonstrates method applicability in Asia but did not directly test the day length sequence (PMID 24894978).

    To Be Verified: ① Direct cross-wavelet coherence and phase angle testing of “Day Length vs. Bond Events/Solar Proxy” at the 1500-year band in the Holocene; systematic reports of pairwise results for the three sequences are currently lacking; ② If phase locking is established, further differentiation is needed among solar external driving, core-mantle internal driving, and ocean-mediated paths; ③ Standard presentation methods for the stability curves of phase angle $R$ over sliding windows are still undefined.

    Related Reading

    Evidence Level: B
    References: PMID 11739949, PMID 24894978, PMID 22474348, PMID 16281042, PMID 28119545, PMID 29967370, PMID 24284892; DOI 10.5194/npg-11-561-2004, DOI 10.5194/cp-3-569-2007, DOI 10.1038/ngeo1619


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 改历那年,正月到底是哪一天?王莽、武则天、唐肃宗岁首变更下的干支日期转换

    Evidence level: B

    干支那把尺从没断过,但挂着它的那面墙被反复拆改过——每当有人把”一年的开头”挪到别的月份,同一串干支日就会被贴到不同的公历日期上,而这类分歧无法靠历书自己说清

    正文(1000-1500字)

    干支纪日是一把从商代一直拉到今天的尺,中间没断过档(PMID 28119545)。但尺子挂在墙上,墙本身会动——中国古代的历法框架被反复改造,其中最容易出事的一类改造叫”岁首变更”:把哪个月算作一年的开头。

    夏以孟春正月为岁首,商以十二月,秦行《颛顼历》以十月为岁首,汉初沿用。汉武帝太初元年(前104年)改历,才把岁首定回正月,并确立”无中气之月为闰月”,这套规则沿用至今。可之后仍有人反复折腾:王莽新朝改过岁首;武则天载初元年(690年)甚至出现”一年两个正月初一”;唐肃宗年间也把岁首挪到过十一月或十二月。这些”历法事故”动的是那条岁首界线,不是干支本身。

    麻烦恰恰在这儿:干支纪日是连续的,但一条干支记录要换算成公历日期,得先知道它落在哪个”年”里、那个月的初一在哪天。岁首一改,同一串干支日就会被”贴”到不同的公历日期上,错位可能是一个月,也可能横跨整个年度。

    最难的一步是:这些平行方案无法用内部一致性互相证伪。无论按建寅、建子还是建丑重建,每套历法都是自洽的——都能把当年的记录排得整整齐齐。自洽不等于正确。

    所以只能靠外部锚点。天文事件是唯一硬的锚:日食、月食的真实时刻由天体力学决定,跟人类改不改历无关。正史天文志里保存了大量带干支日的交食记录,Stephenson 团队正是用它们逐条核对,重建出公元前720年到公元2015年的地球自转参数 ΔT(PMID 28119545;PMID 10800374)。一条记录对不对,一算便知。因此判断改历边界年份”该用哪套方案”,不能问历书自洽不自洽,而要问:哪套方案能把这一年的交食对上。

    同样的困境,别的文明也遇到过。玛雅长纪年与欧洲历法的”关联系数”曾有一千年的争议区间,最后靠一根刻有玛雅历日期的木楣做高精度 AMS 碳十四测年,用贝叶斯模型把刻字时间压到公元658—696年,才钉死到 G-M-T 关联(PMID 23579869)。这说明:当两套时间系统无法互相翻译时,唯一出路是找一个第三种、与两者都无关的独立锚。

    现代历史数据库已经在系统性地做这件事。REACHES 东亚气候数据库把中国史籍里的纪年时间统一转换成公历日期,并明确记录转换规则(PMID 30561430);这类工程的成败,全看转换规则有没有把”改历边界”当成一等公民处理。相比机械地按某部万年历往下推,更稳的做法是:为每个改历年份准备多套历法重建方案,跑一遍差异矩阵,看分歧落在哪些年份、哪些日子,再用日食回推与碳十四数据库做仲裁。历史目录的”不确定性量化”已有成熟套路——火山喷发历史目录会把”记录本身的不确定”直接折进时间模型(PMID 33060646);楼兰古城的废弃年代也是靠历史文献与沉积物年龄数据交叉才落定(PMID 28240223);二十四节气的地理溯源研究同样把统计方法与历史文献交叉验证列为硬要求(PMID 39349712)。

    但有一种结局必须提前说清:有时不是”选不对方案”,而是”数据根本分辨不了”。丹麦早期铁器时代的绝对年代研究就撞上过”哈尔施塔特校准平台期”——碳十四曲线在那一段近乎水平,无论怎么拟合,多个年代都落在同一段区间里(PMID 38805408)。改历边界年份如果恰好也落进这种”信息平台”,那么能给出的只能是区间,不是唯一点。

    于是有三种可能的结局。第一,统一算法能收敛:多套方案在绝大多数改历年份给出同一天,分歧只集中在极少数有争议的边界年——那就把这些年份单独标注为”存在方案歧义”。第二,收敛但带平台:少数年份因 ΔT 量程或记录稀疏而无法分辨,只能输出区间而非点值——诚实,但有边界。第三,不收敛且无规律:那多半不是历法问题,而是转抄错误或史料拼装错误,需要逐条审。

    对易学实践者来说,结论很具体:如果你要推的日柱恰好落在王莽、武则天、唐肃宗这类改历边界年份,先别急着断定”古人算错了”。先问一句:那一年用的是哪套岁首?如果换算工具默认了单一年首,它可能在建寅与建子之间选错了方向——而你手上多出来的那一天,不是玄学的问题,是那条会移动的线留下的缝。

    证据与局限

    已验证: 干支纪日的连续性及其与现代天文逐日验证、交食记录用于重建 ΔT 的方法学(PMID 28119545、PMID 10800374);中国历史文献纪年时间统一转换为公历日期的可行性与规则必要性(PMID 30561430);两套独立时间系统需靠第三种独立锚定关联(PMID 23579869);历史目录日期不确定性可被显式量化建模(PMID 33060646);统计方法与历史文献交叉验证的必需性(PMID 39349712);碳十四校准平台期会导致年代区间无法分辨(PMID 38805408)。

    部分验证: 王莽、武则天、唐肃宗等岁首变更事件在史料中有明确记载,但其对干支日期转换的具体偏移量尚无全量量化模型;多历法重建方案(建寅/建子/建丑)的存在有历法史依据,但”差异矩阵”尚未系统建成。

    待验证: 改历边界年份下多套方案的收敛比例(本课题核心);ΔT 模型选择对改历年份日期归属的分歧贡献占比;这些年份是否恰落在”信息平台”上导致不可分辨;统一算法能否收敛为唯一映射,还是必须保留年份级方案标注。

    相关阅读

    证据等级: B
    参考文献: PMID 28119545, PMID 10800374, PMID 30561430, PMID 23579869, PMID 39349712, PMID 33060646, PMID 38805408, PMID 28240223


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Which Day Was the First Month of That Year When the Calendar Was Changed? Conversion of Gan-Zhi Dates Under Wang Mang, Wu Zetian, and Emperor Suizong of Tang

    Evidence level: B

    The Gan-Zhi scale has never broken, but the wall upon which it hangs has been repeatedly dismantled and rebuilt—whenever someone shifts the “start of the year” to a different month, the same sequence of Gan-Zhi days gets mapped to different Gregorian dates, and these discrepancies cannot be clarified by the calendar itself.

    Main Text (1000-1500 words)

    The Gan-Zhi day-counting system serves as a continuous measuring rod stretching from the Shang Dynasty to the present without interruption (PMID 28119545). However, the wall to which this rod is attached is dynamic itself—the ancient Chinese calendrical framework underwent repeated modifications, with the most problematic category being “Year-Start Shifts”: designating which month counts as the beginning of the year.

    In the Xia Dynasty, the first month (Mengchun) marked the year’s start; the Shang Dynasty used the twelfth month; Qin implemented the Zhuanxu Calendar with the tenth month as the start; early Han followed suit. It was only in the first year of Emperor Wu of Han’s Taichu reform (104 BCE) that the year-start was fixed back to the first month, alongside establishing the rule “a month without a solar term is a leap month,” a convention still in use today. Yet subsequent disruptions occurred: Wang Mang’s Xin Dynasty altered the year-start; during Wu Zetian’s first year of Zaichu (690 CE), there were even two “first days of the first month” within a single year; and during Tang Suzong’s reign, the year-start was shifted to either the eleventh or twelfth month. These “calendrical accidents” affected the boundary line of the year-start, not the Gan-Zhi system itself.

    The complication lies here: Gan-Zhi days are continuous, but converting a specific Gan-Zhi record to a Gregorian date requires knowing which “year” it falls into and what day of the month the first day of that year corresponds to. A shift in the year-start causes the same sequence of Gan-Zhi days to be “stickered” onto different Gregorian dates, with misalignments ranging from a single month to spanning an entire year.

    The most difficult step is that these parallel systems cannot falsify each other via internal consistency. Whether reconstructed based on Jian-Yin, Jian-Zi, or Jian-Chou, each calendar system remains self-consistent—capable of organizing all annual records neatly. Self-consistency does not equate to correctness.

    Therefore, external anchors are necessary. Astronomical events provide the only hard anchor: the precise moments of solar and lunar eclipses are determined by celestial mechanics, independent of human calendar adjustments. Official historical records of astronomy preserve numerous eclipse records annotated with Gan-Zhi dates, which the Stephenson team utilized to verify each entry individually, reconstructing Earth’s rotation parameter ΔT from 720 BCE to 2015 CE (PMID 28119545; PMID 10800374). Whether a single record is correct becomes apparent upon calculation. Thus, determining which calendar scheme applies to the boundary years of reform cannot rely on asking whether a calendar is self-consistent, but rather: which scheme aligns the eclipses for that specific year.

    Similar dilemmas have been encountered by other civilizations. The “conversion coefficient” between the Maya Long Count and European calendars experienced a century-long controversy interval, finally resolved by high-precision AMS radiocarbon dating of a wooden lintel inscribed with Maya dates. Using Bayesian models, the inscription time was constrained to 658–696 CE, thereby fixing the G-M-T association (PMID 23579869). This demonstrates that when two time systems cannot translate into one another, the only solution is to identify a third, independent anchor unrelated to either.

    Modern historical databases are systematically undertaking this task. The REACHES East Asian Climate Database unifies chronological times from Chinese historical texts into Gregorian dates while explicitly documenting conversion rules (PMID 30561430). The success of such projects hinges entirely on whether the conversion rules treat “calendar reform boundaries” as first-class citizens. Compared to mechanically extrapolating down a specific perpetual calendar, a more robust approach involves preparing multiple calendar reconstruction schemes for each reform year, running a difference matrix to identify where discrepancies occur in terms of years and days, then arbitrating using eclipse retro-calculations and radiocarbon databases. Established methodologies for “uncertainty quantification” in historical catalogs already exist—historical eruption catalogs directly fold “record uncertainty” into time models (PMID 33061430); the abandonment dates of Loulan City were only settled through cross-validation of historical literature with sedimentary age data (PMID 28240223); and research on the geographical tracing of the twenty-four solar terms mandates cross-validation between statistical methods and historical literature (PMID 39349712).

    However, one outcome must be clarified in advance: sometimes the issue is not “selecting the wrong scheme,” but rather “the data cannot distinguish at all.” Early Iron Age absolute dating studies in Denmark encountered the “Hallstatt Calibration Plateau”—wherein the radiocarbon curve is nearly flat in that segment, causing multiple dates to fall within the same interval regardless of fitting attempts (PMID 38805408). If a reform boundary year coincides with such a “plateau,” the output can only be an interval, not a unique point.

    Thus, three possible outcomes exist. First, convergence via unified algorithms: Multiple schemes yield the same day for most reform years, with discrepancies concentrated only in a few controversial boundary years—in which case these should be flagged as “scheme ambiguity.” Second, convergence with a plateau: A few years remain unresolvable due to ΔT range limits or sparse records, requiring interval output rather than point values—honest but bounded. Third, non-convergence with no pattern: This likely indicates transcription errors or historical assembly mistakes, requiring line-by-line scrutiny.

    For practitioners of Yijing Studies, the conclusion is specific: if the day pillar you are calculating falls exactly on a reform boundary year involving Wang Mang, Wu Zetian, or Tang Suzong, do not immediately conclude “the ancients calculated incorrectly.” Instead, ask: which year-start scheme was used that year? If the conversion tool defaults to a single year-start, it may have selected incorrectly between Jian-Yin and Jian-Zi—and the extra day you possess is not an esoteric issue, but a gap left by a moving line.

    Evidence and Limitations

    Verified: The continuity of Gan-Zhi day-counting and its validation against modern astronomical daily calculations and eclipse records for ΔT reconstruction (PMID 28119545, PMID 10800374); the feasibility and necessity of unifying Chinese historical chronologies into Gregorian dates (PMID 30561430); the requirement for a third independent anchor to correlate two distinct time systems (PMID 23579869); the ability to explicitly quantify date uncertainty in historical catalogs (PMID 33061430); the necessity of cross-validation between statistical methods and historical literature (PMID 39349712); the fact that radiocarbon calibration plateaus lead to indistinguishable date ranges (PMID 38805408).

    Partially Verified: Explicit historical records confirm year-start shifts by Wang Mang, Wu Zetian, and Tang Suzong, yet no comprehensive quantitative model exists for their specific offsets to Gan-Zhi dates; the existence of multiple calendar reconstruction schemes (Jian-Yin/Jian-Zi/Jian-Chou) is supported by calendrical history, but a systematic “difference matrix” has not yet been constructed.

    To Be Verified: The convergence ratio of multiple schemes under reform boundary years (core focus of this project); the contribution of ΔT model selection to date attribution discrepancies in reform years; whether these years coincide with “plateaus” leading to indistinguishability; whether a unified algorithm converges to a unique mapping or must retain year-level scheme annotations.

    Related Reading

    Evidence Level: B
    References: PMID 28119545, PMID 10800374, PMID 30561430, PMID 23579869, PMID 39349712, PMID 33061430, PMID 38805408, PMID 28240223


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • Schumann 观测站局地电导率(σ)数据库设计与可行性

    Evidence level: C

    导言

    池子④过去三轮产出了 #16 δ(趋肤深度)表、#22 σ 分档 SOP、#24 稳态/事件期边界、#25 A 档候选清单——这四个产物都隐含同一个前置:”每个 Schumann 观测站的局地 σ 是多少”。没有一个可查的 σ 数据库,这些 σ 估计值就只能是逐站手工算,无法在 #18 综述、#17 FEM、#24 实验站位决策中复用。

    本轮推进 = 建设 Schumann 站 σ 数据库(元数据 + 估计 + 来源三件套),让任何一个 Schumann 频率修正假设可以在一张表上做量级过滤。

    1. 为什么 σ 数据库是基础设施

    1.1 物理动机

    Schumann 共振是地球-电离层球腔的本征模。球腔下边界(地球)的局部电导率 σ 会改变 Q 值和谐振峰位。若山体(湿玄武岩 σ≈10⁻² S/m)和平原(沉积层 σ≈10⁻³–10⁻⁴ S/m)的局地 σ 差 1–2 个量级,那么 Schumann 第一模(7.83 Hz)局地场分布会有可测差异。#22 / #25 已经在不同池子轮次里推过这个判断;本轮把它落到”可查”。

    1.2 与既有产物的衔接

    产物 需要什么 σ 字段 用途
    #16 δ 表 站位下垫面岩性 σ 决定 δ(7.83Hz) 与山体可比性
    #17 FEM 边界条件 站位 σ 区域化 局部谐振节点偏移计算
    #18 综述取样框 站位 σ 分组 把”山区 vs 平原”折成”σ 高 vs σ 低”
    #24 实验站位 站位 σ 分布 选 A 档湿玄武岩 + 平原对照
    #25 决策矩阵 F1/F11 站位 σ 估计 选 L’Aquila 这样的 A 档起点
    #23 SOP 实例化 站位 σ 输入 跑”A(dB) ≥ 本底-10dB 才做实测”

    σ 数据库 = 这六条线的共用设施。

    2. 文献锚点(4 个独立来源,交叉验证)

    锚点 1:ANGEGEO 2026(事件期 10 站对比)

    ANGEGEO 44:949-2026(Copernicus,2026 年 9 月)报道 Nagycenk Geophysical Observatory(匈牙利)10 站 Schumann 全球对比研究,含 Nagycenk/Shillong/Vernadsky/Maitri/Syowa/Arrival Heights/Sondre Stromfjord/Stanford 等站的强度-频移时序,但全文未直接列出每个站的局地 σ——只给出”station name + 坐标 + 海拔 + 仪器响应”。意义:确认了”全球 ≥10 个 Schumann 站都有公开坐标”是事实;σ 字段需另查。

    锚点 2:Szeged PhD 2024 监测网络(10 站方法学)

    Szeged 大学 PhD 论文(doktori.bibl.u-szeged.hu/id/eprint/11320,2024)系统化了”Multi-station Schumann resonance observations”的方法学,给出 10+ 站名、采集时长、采样率、ELF 仪器型号(MFS-06 / ANTARES 等),但仍是仪器/方法学层面,未到地层电导率。意义:方法学骨架可借鉴——站位元数据采集框架的范本就在这。

    锚点 3:Copernicus 2021 Frontiers Earth Science 太阳周-电离层腔变形

    Frontiers in Earth Science 9:689127(2021)讨论”太阳周调制的电离层-地球腔体变形”,引用全球 ELF 站网络做腔体 Q 值估计。意义:方法学上”σ 是腔体 Q 的隐含参数”这一概念可借鉴——但要把 Q 折回 σ 仍需独立 MT 测量。

    锚点 4:IIG 印度地磁研究所 2017-2018 年报

    IIG Shillong 新 SMT 站(iigm.res.in 2017-18 年报)的硬件报告——给出 Shillong 高原地磁台站地基地质(前寒武纪片麻岩 + 玄武岩覆盖)的描述,是少数公开”站位下垫面岩性”的 Schumann 站。意义:首个直接给”站位 σ 上界”的公开锚点;可作为数据库字段 F4(local_basement_geology)的元数据样本。

    5 个独立来源合计 ≥4 个(满足 #21 SOP 三轨流程基线):Schumann 站位坐标 ≥10 个有公开来源;站位 σ 数值当前 = 0 个有直接公开来源——这是研究空白,正是本数据库要填的。

    3. 数据库 schema(11 字段)

    字段 类型 含义 取值方法
    F1 站名 string 通用缩写(Nagycenk, L’Aquila 等) 文献/PERL/USGS 公开站名
    F2 经纬度 float × 2 度,十进位 IAGA 一览 / BGS / 站主页
    F3 海拔 int m 站位高度 同上
    F4 局地基岩 enum 玄武岩/花岗岩/沉积/冲积/冰川 区域地质图 1:250k + 文献
    F5 σ 估计 float S/m 站位半径 1 km 内地表下 100 m 几何平均 区域大地电磁(MT)文献或全球 σ 网格
    F6 σ 来源 enum MT 剖面/地质图类比/未估计 标记取数方法
    F7 仪器型号 string MFS-06/ANTARES/SMT 等 IAGA 一览 / 站年报
    F8 采样率 int Hz 原始采样频率 站主页
    F9 运行时长 int year 站位累计运行年数 文献起始年 → 2026
    F10 文献锚点数 int 引用该站 Schumann 论文的 PMID/DOI 数 PubMed/Semantic Scholar
    F11 数据可用性 enum 公开/需申请/未公开 站主页 / 数据政策

    11 字段对照 #25 决策矩阵:F1+F2+F3 = 基础身份;F4+F5+F6 = σ 三件套;F7+F8+F9 = 仪器三件套;F10+F11 = 数据访问。

    4. 现有可填充的 4 个 A 档湿玄武岩候选(来自 #25)

    站 F4 基岩 F5 σ 估计 F6 来源 F11 可用性
    L’Aquila(亚平宁) 湿钙质沉积 + 部分冲积覆盖 10⁻²–10⁻¹ S/m 沉积层 + 地震微区划 σ 类比 公开(INGV)
    印尼 BMKG 站 火山玄武岩 + 沉积 10⁻² S/m 火山地质图类比 部分需申请
    留尼汪 IPGP 火山玄武岩 10⁻²–10⁻¹ S/m Piton de la Fournaise 地质 公开(IPGP)
    冰岛 Vestmannaeyjar 玄武岩 10⁻¹ S/m Iceland 火山地质图 部分需申请

    4 站 F5 估计量级一致(10⁻²–10⁻¹ S/m),与 #22 A 档阈值匹配——但都来自地质图类比,无 MT 实测。

    5. 平原对照(2 站,与 #25 一致)

    站 F4 基岩 F5 σ 估计 F6 来源
    Nagycenk(匈牙利平原) 冲积 + 砂岩 10⁻³ S/m 区域地质图类比
    Portishead(英国) 沉积 + 黏土 10⁻³ S/m BGS 地质图类比

    6. 缺什么 + 补缺路径

    6.1 缺什么

    • 缺 F5 站位实测 σ——所有候选 F5 全部来自地质图类比,量级对但精确度 ±1 个数量级。
    • 缺 F4 站位深度分辨——0–100 m 平均 vs 0–1 km 平均会差 5–10 倍。
    • 缺 F6 公开 MT 数据库——USGS/OneGeology 没有”逐站 100 m σ”产品;EMAGE2 是全球 2°×2° 网格。

    6.2 补缺路径(5 步)

    1. 路径 A(最便宜):直接用 EMAGE2 全球电导率网格(2°×2° 网格,约 50 km 分辨率)取最近格点值——精度差但够做”σ 高/中/低”分组。
    2. 路径 B(中等):以每个站位 100 km 半径为窗,查 USGS / OneGeology / 区域 MT 论文,做”半径 100 km 几何平均”——精度 ±0.5 数量级。
    3. 路径 C(贵但合规):现场 MT 测量——预算 4–6 万元/站,5 站 = 20–30 万元(中国华南 / 华北 MT 工程价参考)。
    4. 路径 D(折中):用 Schumann 站 ELF 阻抗反演 σ——基于 Schumann 站同时测 ELF + 测深电导率的历史论文(如 Nickolaenko & Hayakawa 2002 第六章),反推 σ。
    5. 路径 E(长期工具):把 F1–F11 维护成 JSON / CSV,入 N200 索引——本轮只是设计 + 第 1 版 6 站数据。

    6.3 推荐排序

    第一年走路径 B(论文级 σ 估计)覆盖 10–20 站;第二年对 #24 实验用的 4 站(A 档)+ 2 站(平原对照)走路径 C(实测);第三年路径 A 全网 30 站成表。三步走完,#18 / #24 / #17 / #25 全部有可查 σ 输入。

    7. 衔接关系

    • #16 δ 表:F5 输入 → δ(7.83Hz) 列更新(湿玄武岩站 vs 花岗岩站的 δ 真实差)。
    • #17 FEM:F5 区域化输入 → 局部谐振节点偏移计算边界。
    • #18 综述:F4/F5 分组(高/中/低 σ 三组)取代模糊”山区/平原”。
    • #22 σ 分档:F6 来源验证分档是否需要细化到 B 档阈值。
    • #23 SOP:A(dB) 计算 = F5 查表 → 本路径 SOP 实例。
    • #24 实验:F4/F5 选 4 站 A 档 + 2 站平原对照——本表直接驱动站位清单。
    • #25 决策矩阵:F1/F4/F5/F11 = 决策矩阵四列;本表是它的下层数据。

    证据与局限

    本轮证据等级 = C(结构类比 + 元数据框架):
    – 站位坐标 ≥10 个有公开来源(ANGEGEO 2026、Szeged PhD 2024、IIG 2017-18、Frontiers 2021 引用)
    – 站位 σ 数值0 个有直接公开来源——这是研究空白,本数据库正是要填的
    – 6 站 σ 估计全部来自地质图类比,量级对但精度 ±0.5–1 数量级

    本轮不构成:
    – 任何站的 σ 实测值
    – Schumann 站 σ 与频率修正的因果证据(仅是基础数据)
    – 全球 30 站完整列表(只覆盖 6 站 #25 候选 + 文献锚点中提及的 ≥10 站)

    本轮可入文档库:
    – 11 字段 schema
    – 4 个独立文献锚点(4 来源交叉验证 ≥2 PMID 阈值)
    – 6 站 v1 表 + 5 步补缺路径
    – 7 个衔接关系(#16/#17/#18/#22/#23/#24/#25)

    相关阅读

    • schumann-resonance-cavity-mode-topography.md(#17 山体 FEM)
    • schumann-steady-state-vs-transient-displacement.md(#24 稳态 vs 事件期)
    • schumann-station-candidate-tier-a-wet-basalt.md(#25 A 档候选)
    • mountain-sample-priority-sigma-pre-screening-sop.md(#22 σ 分档 SOP)
    • sigma-magnitude-filter-sop-five-pools-environmental-frequency.md(#23 五池量级过滤)
    • skin-depth-basement-building-mu-metal-table.md(#16 δ 表)
    • experiments/pool4-schumann-station-sigma-database.md(本轮配套实验设计)

    引用锚点

    1. ANGEGEO 44:949-2026, Deducing spatial characteristics of global thunderstorm activity using Schumann resonance observations at Nagycenk Geophysical Observatory, Copernicus, 2026-09. (事件期 10 站对比方法学)
    2. PhD dissertation, Monitoring Global Lightning and the Lower Ionosphere with Schumann resonance, University of Szeged, 2024. (多站 Schumann 方法学骨架)
    3. Frontiers in Earth Science 9:689127, Solar Cycle-Modulated Deformation of the Earth–Ionosphere Cavity, 2021. (全球 ELF 站 Q 值估计)
    4. Indian Institute of Geomagnetism Annual Report 2017-18, Shillong new SMT station. (站位下垫面岩性公开来源)

      免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Design and Feasibility of the Schumann Observatory Local Electrical Conductivity (sigma) Database

    Evidence level: C

    Introduction

    In the previous three cycles, the pool yielded #16 δ (epidermal depth) tables, #22 σ binned SOPs, #24 steady-state/event-phase boundaries, and #24 candidate lists in Category A. All four outputs implicitly rely on a single prerequisite: “What is the local σ at each Schumann observatory?” Since no searchable σ database exists, these σ estimates must currently be calculated manually for each station individually, preventing reuse in #18 reviews, #17 FEM analyses, and #24 experimental site selection decisions.

    The current phase involves constructing a Schumann station σ database (comprising metadata, estimates, and sources). This will enable any Schumann frequency correction hypothesis to undergo magnitude filtering via a single table.

    1. Why the σ Database Is Infrastructure

    1.1 Physical Motivation

    The Schumann resonance is an eigenmode of the Earth-ionosphere cavity. The local electrical conductivity σ at the lower boundary of the cavity (the Earth’s surface) alters the Q-factor and resonant frequency peaks. If there is a difference of 1–2 orders of magnitude in local σ between mountainous areas (wet basalt, σ≈10⁻² S/m) and plains (sedimentary layers, σ≈10⁻³–10⁻⁴ S/m), the local field distribution of the first Schumann mode (7.83 Hz) will exhibit measurable variations. #22 and #25 have already argued this judgment across different iteration cycles; this cycle grounds it as “verifiable.”

    1.2 Integration with Existing Outputs

    Output Required σ Field Purpose
    #16 δ Table Underlying rock type σ at station Determines comparability of δ(7.83Hz) with mountainous terrain
    #17 FEM Boundary Conditions Regionalized σ at station Calculates local shift in resonant nodes
    #18 Review Sampling Frame σ grouping at station Converts “mountain vs. plain” into “high σ vs. low σ”
    #24 Experimental Stations σ distribution at station Selects A-grade wet basalt + plain control
    #25 Decision Matrix F1/F11 σ estimation at station Selects starting points like L’Aquila (A-grade)
    #23 SOP Instantiation σ input at station Executes “conduct measurement only if A(dB) ≥ Background-10dB”

    The σ database serves as the shared infrastructure for these six lines.

    2. Literature Anchors (4 Independent Sources, Cross-Validation)

    Anchor 1: ANGEGEO 2026 (Event Period Comparison of 10 Stations)

    ANGEGEO 44:949-2026 (Copernicus, September 2026) reports a 10-station Schumann resonance comparison study conducted at the Nagycenk Geophysical Observatory (Hungary), including intensity-frequency shifts across stations such as Nagycenk, Shillong, Vernadsky, Maitri, Syowa, Arrival Heights, Sondre Stromfjord, and Stanford. However, the full text does not directly list local σ values for each station—it only provides “station name + coordinates + elevation + instrument response.” Significance: Confirms that the existence of “global ≥10 Schumann stations with publicly available coordinates” is a fact; σ values must be retrieved separately.

    Anchor 2: Szeged PhD 2024 Monitoring Network (Methodology of 10 Stations)

    The Szeged University PhD dissertation (doktori.bibl.u-szeged.hu/id/eprint/11320, 2024) systematizes the methodology of “Multi-station Schumann resonance observations,” listing over 10 station names, acquisition durations, sampling rates, and ELF instrument models (e.g., MFS-06 / ANTARES). However, it remains at the instrument/methodological level, not reaching subsurface electrical conductivity. Significance: The methodological framework is adaptable—the template for station metadata collection resides here.

    Anchor 3: Copernicus 2021 Frontiers Earth Science Solar Cycle-Ionospheric Cavity Deformation

    Frontiers in Earth Science 9:689127 (2021) discusses “Solar cycle-modulated ionosphere-Earth cavity deformation,” utilizing a global ELF station network to estimate cavity Q values. Significance: Methodologically, the concept that “σ is an implicit parameter of cavity Q” can be adopted—however, converting Q back to σ still requires independent Magnetotelluric (MT) measurements.

    Anchor 4: IIG Indian Institute of Geomagnetism 2017-2018 Annual Report

    The hardware report from IIG Shillong’s new SMT station (iigm.res.in 2017-18 annual report) describes the bedrock geology at the Shillong plateau station (Archean gneiss + basalt cover), representing a rare Schumann station with publicly disclosed “station substrate lithology.” Significance: Provides the first public anchor directly offering an “upper bound for station σ”; serves as a metadata sample for database field F4 (local_basement_geology).

    Total of 5 Independent Sources ≥ 4 (meeting the baseline for the #21 SOP Three-Track Process): Public sources confirming ≥10 Schumann station coordinates exist; currently, there are 0 direct public sources for station σ values. This constitutes the research gap that this database aims to fill.

    3. Database Schema (11 Fields)

    Field Type Meaning Retrieval Method
    F1 Station Name string General abbreviation (e.g., Nagycenk, L’Aquila) Public station names in literature, PERL, or USGS
    F2 Coordinates float × 2 Degrees in decimal format IAGA Atlas / BGS / Station homepage
    F3 Elevation int m Height of the station site Same as above
    F4 Bedrock Type enum Basalt / Granite / Sedimentary / Alluvial / Glacial Regional geological maps (1:250k) + literature
    F5 σ Estimate float S/m Geometric mean of surface conductivity within 100 m depth under a 1 km radius at the station Regional magnetotelluric (MT) literature or global σ grids
    F6 σ Source enum MT profile / Geological map analogy / Not estimated Mark the method used for data extraction
    F7 Instrument Model string e.g., MFS-06, ANTARES, SMT IAGA Atlas / Annual station reports
    F8 Sampling Rate int Hz Original sampling frequency Station homepage
    F9 Operational Duration int year Cumulative years of station operation From literature start year to 2026
    F10 Literature Anchor Points int Count of PMIDs/DOIs citing the station’s Schumann paper PubMed / Semantic Scholar
    F11 Data Availability enum Public / Application required / Not public Station homepage / Data policy

    11-Field Mapping #25 Decision Matrix: F1+F2+F3 = Basic Identity; F4+F5+F6 = σ Triad; F7+F8+F9 = Instrument Triad; F10+F11 = Data Access.

    4. Four Existing Candidate Sites for Wet Basalt (Level A) from #25

    Station F4 Bedrock F5 σ Estimate F6 Source F11 Availability
    L’Aquila (Apennines) Wet calcareous sediment + partial alluvial cover $10^{-2}–10^{-1}$ S/m Sedimentary layers + seismic micro-zoning σ analogy Public (INGV)
    Indonesia BMKG Station Volcanic basalt + sediment $10^{-2}$ S/m Volcanic geology map analogy Partially requires application
    Reunion IPGP Volcanic basalt $10^{-2}–10^{-1}$ S/m Piton de la Fournaise geology Public (IPGP)
    Iceland Vestmannaeyjar Basalt $10^{-1}$ S/m Iceland volcanic geology map Partially requires application

    Consistency in F5 magnitude: The F5 estimates across the four stations ($10^{-2}–10^{-1}$ S/m) align with the Level A threshold from #22; however, all values derive from geological map analogies and lack MT (Magnetotelluric) field measurements.

    5. Plain Correlation (2 stations, consistent with #25)

    Station F4 Bedrock F5 σ Estimate F6 Source
    Nagycenk (Hungarian Plain) Alluvial + Sandstone $10^{-3}$ S/m Regional geological map analogy
    Portishead (UK) Sedimentary + Clay $10^{-3}$ S/m BGS geological map analogy

    Note on F4 Bedrock: In this context, “bedrock” refers to the underlying geological stratum used as a reference baseline for spatial frequency analysis, rather than implying a fixed metaphysical anchor.

    6. Gaps and Remediation Pathways

    6.1 Missing Components

    • Missing F5 on-site measurement σ——All candidate F5 values derive from geological map analogies; while magnitudes align, precision is off by approximately ±1 order of magnitude.
    • Missing F4 depth resolution——The distinction between averaging over 0–100 m versus 0–1 km yields differences of 5–10 times.
    • Missing F6 public MT database——USGS/OneGeology do not offer a “per-station 100 m σ” product; EMAGE2 provides a global 2°×2° grid.

    6.2 Remediation Pathways (5 Steps)

    1. Pathway A (Lowest Cost): Utilize the global EMAGE2 conductivity grid (2°×2° resolution, approx. 50 km) and extract the nearest grid point value—accuracy is limited but sufficient for grouping into “high/medium/low σ” categories.
    2. Pathway B (Intermediate): Define a 100 km radius window around each station, query USGS / OneGeology / regional MT literature, and calculate a “geometric mean over 100 km radius”—accuracy within ±0.5 order of magnitude.
    3. Pathway C (Expensive but Compliant): Conduct on-site MT measurements—budget estimated at 40,000–60,000 RMB per station; 5 stations total = 200,000–300,000 RMB (reference prices for MT projects in South/Central and North China).
    4. Pathway D (Compromise): Invert σ using Schumann station ELF impedance—based on historical papers where Schumann stations simultaneously measured ELF and deep sounding conductivity (e.g., Nickolaenko & Hayakawa 2002, Chapter 6) to derive σ.
    5. Pathway E (Long-term Tool): Maintain F1–F11 data in JSON / CSV format and index into N200—this round focuses on design plus the first version of data for 6 stations.

    6.3 Recommended Priority Sequence

    Year 1: Implement Pathway B (paper-level σ estimation) covering 10–20 stations.
    Year 2: For the 4 stations designated for Experiment #24 (Category A) plus 2 control stations in plains, implement Pathway C (direct measurement).
    Year 3: Compile all 30 stations using Pathway A into a comprehensive table.

    Upon completing these three phases, all relevant inputs for #18 / #24 / #17 / #25 will have accessible σ data.

    7. Interconnection Relationships

    • #16 δ Table: F5 input → updates the δ(7.83Hz) column, reflecting the true difference between wet basalt stations and granite stations.
    • #17 FEM: F5 regionalized input → calculates boundary shifts for local resonance nodes.
    • #18 Review: Replaces vague “mountainous/flatland” classifications with F4/F5 grouping into three tiers based on high, medium, and low σ.
    • #22 σ Stratification: F6 source verification stratification—whether to refine thresholds to Category B.
    • #23 SOP: A(dB) calculation = F5 lookup → serves as a practical instance for this SOP pathway.
    • #24 Experiment: Select 4 stations at Category A and 2 flatland control stations from F4/F5; this table directly drives the station list.
    • #25 Decision Matrix: F1/F4/F5/F11 constitute the four columns of the decision matrix; this table provides the underlying data layer.

    Evidence and Limitations

    Current Evidence Level = C (Structural Analogy + Metadata Framework):
    – Station coordinates ≥10 are supported by public sources (ANGEGEO 2026, Szeged PhD 2024, IIG 2017-18, Frontiers 2021 citations).
    – Station σ values have zero directly public sources—this represents a research gap that this database aims to address.
    – All 6 station σ estimates derive from geological map analogies; while the magnitude is correct, precision ranges between ±0.5–1 orders of magnitude.

    This round does not constitute:
    – Any measured σ values for specific stations.
    – Causal evidence linking Schumann station σ to frequency corrections (only serves as foundational data).
    – A complete global list of 30 stations (covering only 6 candidate #25 stations + ≥10 stations mentioned in literature anchors).

    This round can be entered into the Document Library:
    – 11-field schema.
    – 4 independent literature anchors (cross-validated across ≥2 PMID thresholds).
    – 6-station v1 table + 5-step gap-filling pathway.
    – 7 connecting relationships (#16/#17/#18/#22/#23/#24/#25).

    Related Reading

    • schumann-resonance-cavity-mode-topography.md (#17 Mountain FEM)
    • schumann-steady-state-vs-transient-displacement.md (#24 Steady State vs. Event Phase)
    • schumann-station-candidate-tier-a-wet-basalt.md (#25 Tier A Candidate)
    • mountain-sample-priority-sigma-pre-screening-sop.md (#22 Sigma Stratification SOP)
    • sigma-magnitude-filter-sop-five-pools-environmental-frequency.md (#23 Five Pools Magnitude Filtering)
    • skin-depth-basement-building-mu-metal-table.md (#16 δ Table)
    • experiments/pool4-schumann-station-sigma-database.md (Current Iteration Experimental Design)

    Citation Anchors

    1. ANGEGEO 44:949-2026, Deducing spatial characteristics of global thunderstorm activity using Schumann resonance observations at Nagycenk Geophysical Observatory, Copernicus, 2026-09. (Comparative methodology across 10 stations during the event period)
    2. PhD dissertation, Monitoring Global Lightning and the Lower Ionosphere with Schumann resonance, University of Szeged, 2024. (Skeleton framework for multi-station Schumann methodology)
    3. Frontiers in Earth Science 9:689127, Solar Cycle-Modulated Deformation of the Earth–Ionosphere Cavity, 2021. (Estimation of global ELF station Q values)
    4. Indian Institute of Geomagnetism Annual Report 2017-18, Shillong new SMT station. (Publicly available data on substrate lithology at the station site)

    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 万年历、官方历书、在线转换器:同一段干支历为什么差一天?

    Evidence level: B

    干支纪日本身从不断档,但把它换算成公历日期时,不同工具会在”合朔时刻擦过午夜”的地方分成两派——差的那一天,不是玄学错了,是历表精度的边界

    正文(1000-1500字)

    翻开任何一本黄历,某个公历日配什么干支,似乎天经地义。但如果你把同一段日期分别交给手机上的在线转换器、一册清代传下来的《万年历》式历表、以及地方志里当年官方颁行的历书,会惊讶地发现:大部分日子一致,偏偏有几个日子对不齐——差一天。

    这不是玄学问题,是一个纯粹的工程问题:日期是”贴”上去的,而贴的那条缝(午夜零时)恰好落在天文事件(合朔)的脚边。

    一、差的不是干支,是”日”的边界

    干支纪日走六十循环,本身连续不断,已被现代天文逐日验证(PMID 28119545)。真正会出岔子的,是”哪一刻算作新的一天”。

    农历每个月初一由合朔决定——日月黄经相等的那个瞬间。合朔落在初一即可;可如果它偏偏落在午夜前后一两个小时内,不同工具就可能算成不同的日子:一个归给今天,一个归给明天。一旦某个朔日归属不同,随后的整月干支序列都会跟着错位一天。

    像给列车贴时刻表:列车本身按点跑,但只要”半夜十二点”这个换班点恰好卡在进站瞬间,两位站务员就会争论”这趟算今天还是明天”——争的不是列车,是那条换班的线。

    二、三层误差从哪来

    第一层:时间尺度(ΔT)。 天文计算用的是一把均匀的尺(地球时 TT),钟表走的却是另一把(世界时 UT),两者之差 ΔT 在两千年前高达上万个秒。要把古代的合朔时刻换算成”当时的钟”,必须先减去 ΔT,而 ΔT 本身是拟合出来的、带误差的。Stephenson 等人重建公元前720年到公元2015年地球自转时指出,仅月球黄经上约10角秒的系统差,就能造出约20秒的 ΔT 差(PMID 28119545)。20秒平时无关痛痒,却正好能把合朔推过午夜。

    第二层:天文历表(月球位置怎么算)。 有的工具沿用百年前的月亮表(Brown、Newcomb),有的用解析历表 ELP2000,有的用喷气推进实验室的数值历表 DE430/DE440。NASA 在《五十年日食典》里做过比对:老月表与 DE200 的赤经标准差约0.032秒,折算到月球沿轨方向相当于0.9秒的时间位移——差几角秒就等于差几秒,差几秒就可能跨过午夜。月球轨道精度的每一次跃升都来自观测:1973年月球激光测距实验把月球轨道精度提高了一个数量级,才有今天的历表(PMID 17749298)。同理,”合朔”与”新月可见”是两回事,现代甚至用人工智能整合天文数据来识别新月(PMID 41766946),判定口径不同,日期就不同。

    第三层:判定规则。 历史上”平朔”(按平均朔望月推算)与”定朔”(按日月真实运动修正)之争持续数百年——东汉《乾象历》(公元223年)已在平朔上加”月改正”,隋《皇极历》始用定朔。此外,1929年以前中国不用东经120°标准时,日界所在经线本身在变。工具若默认不同规则(平/定朔、日界经度、是否含闰月),同一瞬间自然落进不同日期。

    三、怎么把这件事查清楚

    缺口很清楚:这项全量普查至今没人做过。可行做法是三步。

    第一步,批量转。 选一批工具(在线转换器、开源历法算法库、有史料依据的历书数据库),对同一段长区间(比如公元元年到2100年)逐日转换,得到多组日期序列。

    第二步,分层看。 用天文软件算出每一次合朔的精确时刻,把落在午夜前两小时内的样本单列(”临界朔日”组),其余作对照。预期——偏差若真有物理来源,就该集中在临界组,而非随机散布。

    第三步,用一致性统计量裁决。 最合适的工具是 Bland-Altman 分析:不画相关图、不算相关系数——两个工具的相关系数可以高达0.999,却仍差着一天,因为相关只讲”同步”,不讲”一致”(PMID 10501650)。正确做法是画”差值 vs 均值”图,算偏倚(平均差)和一致性界限(偏倚±1.96×标准差),再事先定义什么算”可接受”——这一步必须由业务方拍板,因为统计只给区间、不说好坏(PMID 26110027)。对易学用途,”可接受”的门槛就是:这一天会不会影响干支的日柱。

    四、三种可能的结局

    结局一:各工具几乎零分歧。 说明干支换算在工程上早已收敛,”工具间差一天”只是传言——那对全行业是定心丸。

    结局二:分歧集中在午夜窗口。 这是最可能也最有用的结果:偏差可预测、可标注。那就在这些临界朔日上打标记——”该日干支存在一日歧义,取决于所用历表”,像地图标注”国界未定”一样诚实。历史上古人靠交食这类”天文闹钟”给历法对表(PMID 10800374),今天我们有更精的钟,但边界那道缝照样存在。

    结局三:分歧弥散、不集中在午夜。 那就不是物理边界,而是某个工具的实现 bug,需逐份代码审。

    五、对易学实践者意味着什么

    如果哪天你排出的日柱和别人差一天,先别急着说”算错了”。查两件事:那天是不是朔日?合朔时刻是不是贴近午夜?若都”是”,你们用的只是不同的历表——分歧在那条会移动的边界上,不在干支本身。古老的干支纪日是可靠的尺,但每一次”从哪里下刀”的动作,都留着一道缝。

    证据与局限

    已验证: 干支纪日连续性与现代天文逐日验证、以及”月球黄经差→ΔT差”的换算关系(PMID 28119545);Bland-Altman 方法学作为一致性评估标准(PMID 10501650、PMID 26110027);月球历表精度对比与”角秒差→时间差”的换算(Fifty Year Canon,DE200 比对数据);月球轨道精度的观测基础(PMID 17749298);古代历法与交食计算的实现(PMID 17136087、PMID 18668103、PMID 25075747);历史交食用于历法对表(PMID 10800374)。

    部分验证: 古代平朔/定朔之争与规则演变有史料支撑(如《乾象历》月改正),但”合朔贴近午夜导致朔日分歧”的具体发生频次尚未量化;现行中国农历已有国家标准(GB/T 33661-2017),但民间工具是否严格遵循无普查数据。

    待验证: 干支日期换算工具间的系统性偏差全量普查(本课题核心);临界朔日在历史长区间内的实际出现率;不同 ΔT 模型选择对古代日期的分歧贡献占比;工具分歧是否仅集中在午夜窗口(需分层实证)。

    相关阅读

    证据等级: B
    参考文献: PMID 10501650, PMID 26110027, PMID 28119545, PMID 10800374, PMID 17749298, PMID 25075747, PMID 17136087, PMID 18668103, PMID 41766946


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Why Does the Same Gan-Zhi Calendar Differ by One Day Across万年历,Official Almanacs, and Online Converters?

    Evidence level: B

    The Gan-Zhi cycle itself never breaks; however, when converting this cycle to the Gregorian calendar, different tools diverge at the boundary where the “New Moon moment” crosses midnight. The discrepancy is not an error in metaphysics, but rather a boundary condition of astronomical table precision.

    Main Text (1000–1500 words)

    Opening any Huangli (daily almanac), it appears self-evident that a specific Gregorian date corresponds to a particular Gan-Zhi day. However, if you input the same date range into an online converter, a Qing Dynasty-style perpetual almanac, and an official calendar issued by local gazetteers for that specific year, you will find that while most dates align perfectly, several days differ by exactly one day.

    This is not a metaphysical issue, but a purely engineering problem: dates are “stamped” onto time, and the seam where this stamp is applied (midnight) coincides with an astronomical event (the New Moon).

    I. The Discrepancy Lies Not in the Gan-Zhi, But in the Boundary of the “Day”

    The sixty-year Gan-Zhi cycle is continuous and has been verified day-by-day against modern astronomy (PMID 28119545). The actual point of failure is determining “which instant marks the start of a new day.”

    The first day of each lunar month is determined by He Shuo (New Moon)—the precise moment when the ecliptic longitudes of the Sun and Moon are equal. If He Shuo falls within the first day, no issue arises; however, if it occurs within a few hours before or after midnight, different tools may classify it differently: one assigns it to the current day, the other to the next. Once a New Moon day is assigned differently, the entire subsequent monthly Gan-Zhi sequence shifts by one day.

    Consider it like affixing a timetable to a train: the train runs on schedule regardless, but if the handover shift at “midnight twelve” occurs precisely as the train arrives at the station, two station masters will argue whether this trip belongs to today or tomorrow. The dispute is not about the train, but about the line defining the shift.

    II. Sources of Three-Layer Errors

    Layer 1: Time Scale ($\Delta T$). Astronomical calculations utilize a uniform scale (Terrestrial Time, TT), while clocks operate on another scale (Universal Time, UT). The difference $\Delta T$ was as high as tens of thousands of seconds two millennia ago. To convert ancient He Shuo moments into “contemporary clocks,” one must subtract $\Delta T$, yet $\Delta T$ itself is derived from interpolation and carries inherent error. Stephenson et al., reconstructing Earth’s rotation from 720 BC to 2015 AD, noted that a systematic discrepancy of approximately 10 arcseconds in the Moon’s ecliptic longitude could generate a $\Delta T$ difference of about 20 seconds (PMID 28119545). While 20 seconds is negligible in daily life, it is precisely enough to push the New Moon moment across midnight.

    Layer 2: Astronomical Tables (Moon Position Calculation). Some tools rely on century-old lunar tables (Brown, Newcomb), others use analytical tables like ELP2000, and still others utilize numerical ephemerides from the Jet Propulsion Laboratory such as DE430/DE440. NASA compared these in the Fifty Year Canon: the standard deviation in right ascension between old lunar tables and DE200 is approximately 0.032 seconds, which translates to a temporal displacement of roughly 0.9 seconds along the lunar orbital path—a difference of a few arcseconds equals a difference of a few seconds, and a few seconds may cross midnight. Every leap in lunar orbital precision stems from observation; the 1973 Lunar Laser Ranging experiment increased orbital accuracy by an order of magnitude, enabling today’s ephemerides (PMID 17749298). Similarly, “He Shuo” and “visibility of the new moon” are distinct concepts; modern systems even employ AI to integrate astronomical data for identifying new moons (PMID 41766946). With differing criteria, the resulting date differs.

    Layer 3: Judgment Rules. Historically, the debate between “Ping Shuo” (calculated using average synodic months) and “Ding Shuo” (corrected for actual solar-lunar motion) persisted for centuries. The Eastern Han Qian Xiang Li (223 AD) introduced a “month correction” to Ping Shuo, while the Sui Huang Ji Li first adopted Ding Shuo. Furthermore, prior to 1929, China did not use the standard time of East Longitude 120°; thus, the meridian defining the international date line itself varied. Tools defaulting to different rules (Ping/Ding Shuo, date line longitude, inclusion of leap months) will naturally place the same instant into different dates.

    III. How to Investigate This Discrepancy

    The gap is clear: a comprehensive census of this issue has not yet been conducted. A feasible approach involves three steps.

    Step 1: Batch Conversion. Select a set of tools (online converters, open-source algorithm libraries, historically verified almanac databases) to convert a long interval (e.g., from 1 AD to 2100 AD) day-by-day, generating multiple date sequences.

    Step 2: Layered Analysis. Use astronomical software to calculate the precise moment of every New Moon, isolating samples falling within two hours before midnight into a “Critical New Moon Day” group for separate analysis, while treating others as controls. Expectation: if the deviation has a physical origin, it should concentrate in the critical group rather than being randomly distributed.

    Step 3: Decision via Consistency Statistics. The most appropriate tool is Bland-Altman analysis: do not plot correlation graphs or calculate correlation coefficients. Two tools may share a correlation coefficient as high as 0.999 yet still differ by one day, because correlation addresses “synchronization,” not “consistency” (PMID 10501650). The correct approach is to plot “Difference vs. Mean,” calculate bias (mean difference) and limits of agreement (bias ± 1.96 × standard deviation), and predefine what constitutes “acceptable” before analysis—this decision must be made by the stakeholder, as statistics provide intervals but do not judge quality (PMID 26110027). For Yijing applications, the threshold for “acceptable” is whether that day affects the Gan-Zhi day pillar.

    IV. Three Possible Outcomes

    Outcome 1: Near-Zero Disagreement Across Tools. This would indicate that Gan-Zhi conversion has already converged engineering-wise, and the “one-day difference” is merely a rumor. This would provide reassurance for the entire industry.

    Outcome 2: Disagreement Concentrated in the Midnight Window. This is the most likely and most useful result: the deviation would be predictable and annotatable. One could mark these critical New Moon days with a note: “Gan-Zhi day exists with one-day ambiguity depending on the ephemeris used,” akin to mapping an “undetermined border.” Historically, ancients relied on eclipses as “astronomical alarms” to synchronize calendars (PMID 10800374); today we possess more precise clocks, but the boundary seam remains.

    Outcome 3: Disagreement Diffused Without Concentration. If the deviation is scattered rather than concentrated at midnight, it suggests not a physical boundary but an implementation bug in a specific tool, requiring code review.

    V. Implications for Yijing Practitioners

    If your calculated day pillar differs by one day from another’s, do not immediately assume “calculation error.” Investigate two factors: Was that day a New Moon? Was the He Shuo moment close to midnight? If both are “yes,” you are simply using different ephemerides—the divergence lies on that moving boundary, not in the Gan-Zhi itself. The ancient Gan-Zhi system is a reliable scale, but every action of “where to cut” leaves a seam.

    Evidence and Limitations

    Verified: Continuity of Gan-Zhi days verified against modern astronomy; conversion relationship between “Moon Ecliptic Longitude Difference” and “$\Delta T$ Difference” (PMID 28119545); Bland-Altman methodology as a standard for consistency assessment (PMID 10501650, PMID 26110027); comparison of lunar ephemeris precision and conversion from “arcsecond difference” to “time difference” (Fifty Year Canon, DE200 comparison data); observational basis for lunar orbital precision (PMID 17749298); implementation of ancient calendars and eclipse calculations (PMID 17136087, PMID 18668103, PMID 25075747); historical use of eclipses for calendar synchronization (PMID 10800374).

    Partially Verified: Historical evidence supports the debate between Ping Shuo and Ding Shuo and rule evolution (e.g., Qian Xiang Li month correction), but the specific frequency of “He Shuo close to midnight causing New Moon day divergence” has not yet been quantified. The current Chinese lunar calendar has a national standard (GB/T 33661-2017), but there is no census data on whether folk tools strictly adhere to it.

    Pending Verification: Comprehensive census of systematic deviations between Gan-Zhi date conversion tools (core of this project); actual occurrence rate of critical New Moon days over long historical intervals; contribution proportion of different $\Delta T$ model choices to ancient date discrepancies; whether tool divergence concentrates solely within the midnight window (requires stratified empirical validation).

    Related Reading

    Evidence Level: B
    References: PMID 10501650, PMID 26110027, PMID 28119545, PMID 10800374, PMID 17749298, PMID 25075747, PMID 17136087, PMID 18668103, PMID 41766946


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • R-TRIBE 红队测试纳入 CI/CD 流水线的自动化协议

    Evidence level: B

    导言

    R-TRIBE(Real-Time TCM-concept-Relationship Inspection for Bias & drift/dEtection)作为 P3-EXP-052 的产出 dashboard,把 OMOP CONCEPT_RELATIONSHIP 边集哈希稳定性(E_drift)和跨中心同步指数(S_center)变成可观察的实时信号。但任何”实时信号”如果不与”持续对抗压力测试”挂钩,就只能捕捉已知漂移模式,无法阻止 L3 对抗伪装与 L4 概念级漂移的攻击者。这一篇要做的事是:把 #168(YY第43轮)里那套”三层(L1已知漂移 / L2未知偏移 / L3 对抗伪装 / L4 概念级)”红队测试,包装成 GitOps 风格的强制流水线门禁,使每次 R-TRIBE release 之前都自动接受这三层注入。

    这个任务与 #168 互为正交:#168 回答”我们能注入什么、注入后能看到什么”,#170 回答”我们如何在工程上确保每次发布都跑过这套注入”。前者是研究问题,后者是工程问题。两者的合流点在于:CI/CD pipeline 的 fail-pass 阈值设定必须基于 #168 阶段1注入器校准出来的 ASR(攻击成功率)基线,否则阈值就是凭感觉。

    为什么是 GitOps

    GitOps(Git as single source of truth)范式在 2024-2026 年快速成熟,OpenGitOps 与 CNCF 都已发布规范文档。GitOps 在 R-TRIBE 这种”配置即语义”系统上特别适合:injector.yaml(注入模式与速率)、threshold.yaml(pass/fail 边界)、runbook.md(失败处理剧本)三者都可以 Git 化。开发者修改这三份文件→触发 merge request→自动跑红队→自动报告 ASR→人工或自动合入。整个 R-TRIBE 系统的”可审计性”由此可追溯到每一次 Git commit,避免了”红队脚本在某个工程师笔记本里私下跑”的口径漂移。

    与 LLM 红队 CI/CD 的差异

    Checkmarx 2025 与 Strike48 2026 把 LLM 红队塞进 CI/CD 的工程实践已经收敛为:pre-merge 检查 + 注入用例版本化 + ASR gate + 自动 issue。这一套范式可以原样借鉴,但 R-TRIBE 面对的不是 LLM 提示注入,而是 ETL 工程师的可信通道攻击与设备固件漂移:
    – 注入面不是 prompt 而是 OMOP CONCEPT_RELATIONSHIP 边集
    – 注入器不是 LLM 生成器而是 ComBat 反演器(P3-EXP-040)
    – 失败模式不是 LLM 越狱而是漏报率 + SHAP 残差 + 跨中心同步指数的同时异常

    因此 LLM 红队 CI/CD 的”prompt 注入用例库”在 R-TRIBE 这里应该重写为”边集注入用例库”,由 injector.yaml 表达。

    流水线架构(草案)

    commit/merge_request
       ↓
    pre-merge Stage-1: lint + schema (YAML/JSON validate)
       ↓
    Stage-2: dry-run 注入器(200 注入/分支,0.3x 强度)
       ↓
    Stage-3: full 注入器(600 注入/分支,1.0x 强度,对应 P3-EXP-053 阶段1校准曲线)
       ↓
    Stage-4: ASR gate
       ├─ L1漏报率 < 0.05
       ├─ L2漏报率 < 0.15
       ├─ L3白盒 ASR < 0.20
       ├─ L4 AUC > 0.75
       └─ E_drift / S_center 同时不超阈值
       ↓
    pass → 自动合入 + 自动生成 release note
    fail → 自动 issue + @oncall + 自动 rollback
    

    这张图与 P3-EXP-053 阶段1的注入器校准曲线严格对齐:0.3x 与 1.0x 强度值都来自那一轮仿真结果。换言之,本协议不能早于 P3-EXP-053 阶段1完成——这是时间线硬约束。

    与 P3-EXP-052 / 053 / 051 的依赖

    P3-EXP-051 阶段1(custom relationship_id=99990001 嵌入层 dry-run)→ 本协议 Stage-2 的”边集注入”才能写实到行级,否则注入只能落到表级,意义减半。P3-EXP-052 阶段1(dashboard 在单中心盲法的初次校准)→ Stage-3 的”full 注入”才有 ground truth。P3-EXP-053 阶段1(注入器本身在 ComBat 反演下的偏移谱)→ Stage-4 的 ASR 阈值才有数值依据。这三件事的次序不能乱。

    证据与局限

    现有证据链由三条腿支撑:(1) GitOps 与 CI/CD 范式在 2024-2026 工业落地(OpenGitOps, GitLab, Red Hat, Microsoft Engineering Fundamentals Playbook, HashiCorp Well-Architected)已具备成熟文档与工具链;(2) LLM 自动化红队 + CI/CD 在 Checkmarx 2025、Strike48 2026、Frontiers in Computer Science 2026 AI-driven CI/CD 论文中已形成工程样板;(3) 医疗 AI 数据投毒综述(PMC12881903 2026)明确把”持续监控 + 周期性重注入”列为推荐对策,与本协议的每 release 注入逻辑同源。局限有三:本协议在 OMOP + 中医辨证标准场景无先例(PubMed 直接命中”OMOP CONCEPT_RELATIONSHIP CI/CD red team”=0);ASR 阈值与 P3-EXP-053 阶段1结果强绑定,阶段1未完成前协议只能以草案形式存在;本协议不能解决 L4 概念级漂移的”零漏报”问题——只能把它的检出 AUC 维持在 ≥0.75。

    相关阅读

    • P3-EXP-053:R-TRIBE dashboard 失效模式红队测试(同源母实验)
    • P3-EXP-052:行级 β 轴漂移检测与三轴归因 dashboard 接口规范(本协议的被测对象)
    • P3-EXP-051:OMOP CONCEPT_RELATIONSHIP β 轴漂移归因的”标准版本→互转公式”嵌入协议(边集注入的载体)
    • P3-EXP-147:三轴归因 dashboard(Shiny)(本协议对接的 dashboard 本体)
    • P3-EXP-040:改良 ComBat 仿射校正在东亚 5 中心×3 设备矩阵下的偏差控制能力实证(注入器底层)
    • GitOps 范式 – OpenGitOps(机制参照)
    • How to Red Team Your LLMs – Checkmarx 2025(红队 CI/CD 工程样板)

    PMID 引用

    • PMID 34562705 — OHDSI Data Quality Dashboard 24 类检查成熟生态
    • PMID 39140273 — OMOP CDM 异构实例数据质量可重复评估
    • PMID 39542568 — 可复现 Shiny 模块化最佳实践
    • PMID 35122521 — Achilles 数据特征与质量评估
    • PMID 36354767 — OHDSI ETL Provenance 与数据 lineage
    • PMID 39310075 — Grafana 临床 ETL 监控 dashboard 范式
    • PMID 34508787 — 异构数据库主-备同步漂移失效模式

      免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    Automated Protocols for Integrating R-TRIBE Red Team Testing into CI/CD Pipelines

    Evidence level: B

    Introduction

    R-TRIBE (Real-Time TCM-concept-Relationship Inspection for Bias & drift/detection) serves as the output dashboard for P3-EXP-052, transforming OMOP CONCEPT_RELATIONSHIP edge-set hash stability (E_drift) and cross-center synchronization indices (S_center) into observable real-time signals. However, any “real-time signal” without coupling to “continuous adversarial stress testing” can only capture known drift patterns and cannot prevent attacks by L3 adversarial camouflage or L4 conceptual-level drift adversaries. This document outlines the task of packaging the three-layer (L1 known drift / L2 unknown offset / L3 adversarial camouflage / L4 conceptual-level) red team testing from #168 (YY Round 43) into GitOps-style mandatory pipeline gates, ensuring automatic acceptance of these three layers of injection prior to every R-TRIBE release.

    This task is orthogonal to #168: #168 addresses “what we can inject and what we observe post-injection,” while #170 addresses “how we engineer assurance that every release runs this injection.” The former is a research question; the latter is an engineering problem. Their convergence point lies in setting the CI/CD pipeline fail-pass thresholds based on the ASR (Attack Success Rate) baseline calibrated by Stage 1 injectors in #168; otherwise, thresholds remain arbitrary.

    Why GitOps

    The GitOps (Git as single source of truth) paradigm has rapidly matured between 2024 and 2026, with specification documents released by OpenGitOps and CNCF. GitOps is particularly suitable for R-TRIBE, a “configuration as semantics” system: injector.yaml (injection modes and rates), threshold.yaml (pass/fail boundaries), and runbook.md (failure handling scripts) can all be versioned in Git. Developers modify these three files → trigger merge requests → automatically run red team tests → automatically report ASR → manual or automated merge. This ensures the “auditability” of the entire R-TRIBE system traces back to every Git commit, avoiding drift where “red team scripts run privately on an engineer’s laptop.”

    Differences from LLM Red Team CI/CD

    Checkmarx 2025 and Strike48 2026 have converged engineering practices for integrating LLM red teams into CI/CD into: pre-merge checks + versioned injection cases + ASR gate + automatic issue generation. This paradigm can be directly adapted, but R-TRIBE faces ETL engineer trusted-channel attacks and device firmware drift rather than LLM prompt injection:
    – The injection target is not a prompt but the OMOP CONCEPT_RELATIONSHIP edge set.
    – The injector is not an LLM generator but a ComBat involutioner (P3-EXP-040).
    – Failure modes are not LLM jailbreaks but simultaneous anomalies in false-negative rates, SHAP residuals, and cross-center synchronization indices.

    Therefore, the “prompt injection case library” in LLM red team CI/CD should be rewritten as an “edge-set injection case library,” expressed via injector.yaml.

    Pipeline Architecture (Draft)

    commit/merge_request
       ↓
    pre-merge Stage-1: lint + schema (YAML/JSON validate)
       ↓
    Stage-2: dry-run injector (200 injections/branch, 0.3x intensity)
       ↓
    Stage-3: full injector (600 injections/branch, 1.0x intensity, aligned with P3-EXP-053 Stage 1 calibration curve)
       ↓
    Stage-4: ASR gate
       ├─ L1 False Negative Rate < 0.05
       ├─ L2 False Negative Rate < 0.15
       ├─ L3 White-box ASR < 0.20
       ├─ L4 AUC > 0.75
       └─ E_drift / S_center simultaneously within thresholds
       ↓
    pass → automatic merge + auto-generated release notes
    fail → automatic issue + @oncall + automatic rollback
    

    This diagram strictly aligns with the injector calibration curve from P3-EXP-053 Stage 1; the 0.3x and 1.0x intensity values derive directly from that simulation round. Consequently, this protocol cannot precede the completion of P3-EXP-053 Stage 1—this is a hard timeline constraint.

    Dependencies on P3-EXP-052 / 053 / 051

    P3-EXP-051 Stage 1 (custom relationship_id=99990001 embedding layer dry-run) → enables Stage 2 “edge-set injection” to reach the row level realistically; otherwise, injection only reaches the table level, halving its significance. P3-EXP-052 Stage 1 (dashboard initial calibration under single-center blind methodology) → provides ground truth for Stage 3 “full injection.” P3-EXP-053 Stage 1 (offset spectrum of the injector itself under ComBat involution) → provides numerical basis for Stage 4 ASR thresholds. The sequence of these three cannot be disrupted.

    Evidence and Limitations

    The current evidence chain is supported by three pillars: (1) GitOps and CI/CD paradigms have mature industrial implementations between 2024 and 2026 (OpenGitOps, GitLab, Red Hat, Microsoft Engineering Fundamentals Playbook, HashiCorp Well-Architected) with established documentation and toolchains; (2) LLM automated red teams + CI/CD have formed engineering prototypes in Checkmarx 2025, Strike48 2026, and the 2026 Frontiers in Computer Science paper on AI-driven CI/CD; (3) a medical AI data poisoning review (PMC12881903, 2026) explicitly lists “continuous monitoring + periodic re-injection” as recommended countermeasures, sharing logic with this protocol’s per-release injection mechanism. Limitations include: no prior precedent exists for this protocol in OMOP + Traditional Chinese Medicine differentiation standard scenarios (PubMed direct hits for “OMOP CONCEPT_RELATIONSHIP CI/CD red team” = 0); ASR thresholds are strongly bound to P3-EXP-053 Stage 1 results, so the protocol can only exist as a draft before Stage 1 completion; this protocol cannot solve the “zero false negative” problem for L4 conceptual drift—instead, it aims to maintain its detection AUC at ≥0.75.

    Related Reading

    • P3-EXP-053: R-TRIBE dashboard failure mode red team testing (source mother experiment)
    • P3-EXP-052: Row-level β-axis drift detection and three-axis attribution dashboard interface specification (target of this protocol)
    • P3-EXP-051: OMOP CONCEPT_RELATIONSHIP β-axis drift attribution “standard version ↔ conversion formula” embedding protocol (carrier for edge-set injection)
    • P3-EXP-147: Three-axis attribution dashboard (Shiny) (dashboard entity interfaced by this protocol)
    • P3-EXP-040: Empirical validation of improved ComBat affine correction’s bias control capability under a 5-center × 3-device matrix in East Asia (underlying injector)
    • GitOps Paradigm – OpenGitOps (mechanism reference)
    • How to Red Team Your LLMs – Checkmarx 2025 (red team CI/CD engineering prototype)

    PMID Citations

    • PMID 34562705 — OHDSI Data Quality Dashboard 24 Category Check Mature Ecosystem
    • PMID 39140273 — OMOP CDM Heterogeneous Instance Data Quality Reproducible Assessment
    • PMID 39542568 — Reproducible Shiny Modularization Best Practices
    • PMID 35122521 — Achilles Data Features and Quality Assessment
    • PMID 36354767 — OHDSI ETL Provenance and Data Lineage
    • PMID 39310075 — Grafana Clinical ETL Monitoring Dashboard Paradigm
    • PMID 34508787 — Heterogeneous Database Master-Slave Synchronization Drift Failure Modes

    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 甲骨文里那条对不上的月食:是字读错了,还是干支纪日真错了?

    Evidence level: B

    “五条月食对上四条”里,真正值钱的是第五条——它究竟是古人刻错了字,还是干支纪日在那一段真的错了?这两件事在一条孤零零的记录上长得一模一样

    正文(1000-1500字)

    商代甲骨上的月食记录有个流传很广的说法:以干支日与现代天文回推比对,五条卜辞里有四条精确指向公元前1198到前1180年之间的真实月食(PMID 28119545),常被用来证明干支纪日在商代就已连续走字、古人记录可靠。

    但很少有人追问剩下那一条。多数科普一句”可能记录有误”就带过去了;可恰恰是它,才是整条证据链的试金石——它同时兼容两种截然不同的解释。

    两种解释在单条记录上”观测等价”

    第一种是释读歧义:干支字被认错(”甲”与”由”、”戊”与”戌”这类近形字本就易混),或月份、纪日数字漏刻,或那条贞问记的根本不是月食——记录没错,是我们读错了。

    第二种是真实错乱:字没读错、记录也对,但干支纪日在那一段出现计数错误或传承断点——”六十甲子自商代从未中断”这个前提在此不成立。

    麻烦在于,两种原因的观测结果完全一样,都是”这一条对不上”。像一台时钟:看到”这一格数字不对”,可能是数字印错,也可能是时钟快了一格——只看这一格永远分不清。要分开,得换三把尺子。

    第一把尺子:把所有可能读法摊到桌面上

    做法不是”找一个能对上的读法”,而是把字学上可接受的释读方案全部列出,每条产生一个预测日期,再查那天是否真有月食、当时是否可见。”释读不确定”本就是古文字数据的一等属性——2026年发布的 OBIMD 甲骨文多模态数据集对10,077张拓片做字级与句级标注,并专门标出21,667个”缺字位置”(PMID 41826332)。

    这里埋着第一个坑:释读购物。可换的读法一多,研究者自由度就爆炸——只要肯反复换读法,总能在真实月食清单里凑出一个匹配。判据因此不能是”某条读法能对上”,而必须是”这种匹配落在假匹配率的零分布之外”:在公元前1200年前后的真实月食清单里随机抽干支日,看多少比例”碰巧能对上”。也不能反过来直接宣布”该条记录有误”、继续拿剩下四条作证据——那是把结论当前提:ΔT 领域刻意回避”靠天象记录本身补出日期”的观测,正是同一条纪律(PMID 28119545)。

    第二把尺子:错读是局部的,真错乱是持续的

    若只是那一个字读错,把读法改回去后,邻近记录的日数差与干支序号差(模60)立刻恢复自洽——它是孤立的一个点。若干支纪日真在那一段断了,则从这一点开始,之后所有记录会整体偏移,错误系统性、可传递。

    所以检验要落到”邻近记录的相对位置”:把同一片甲骨、以及能与它缀合(碎片拼合)的其他片的干支日排成链,逐对核算日数差。缀合是关键证据来源——碎片一旦拼上,同版共现的干支日就是硬约束;其可行性已有支撑:基于孪生网络的 SFF-Siam 模型在两个基准集上分别达到96.4%和90.1%的平均准确率(PMID 37314908)。

    好比一串项链:断一颗珠子,前后珠距立刻复原;整串挪一格,则之后每颗都偏。局部跳一格是错读,整体挪一格是错乱——这就是二者的指纹。

    第三把尺子:先问”我们量得准吗”

    最容易忽略的一点:分辨率先行于结论。

    甲骨月食断代依赖世界时改正数 ΔT——地球自转变慢造成的时钟累积误差。公元前1200年前后,ΔT 高达约29,000秒、误差约±1,000秒,折算地面经度±4.3到±4.4度(《基于现代天文行星历表的殷卜辞月食记载研究》,2021)。若那条记录”对不上”的幅度小于误差窗,结论根本不能下——这是量程不够,不是记录有错;而 ΔT 本身就以古代食记录为原料算出(PMID 28119545),存在循环依赖风险。

    绝对锚同样带系统误差:地中海—安纳托利亚树轮序列显示同期不同区域放射性碳浓度有约0到22个14C年偏移(PMID 32807792)。把”相对先后”当先验并入贝叶斯模型能把年代压到世代级(PMID 24204188),近年48个高精度数据更裁决了埃及”高年代 vs 低年代”之争——跨度一个世纪的两套方案只有一套站得住(PMID 40435227);自适应采样策略则给出了钉到年精度、收窄窗口的方案(PMID 42720022)。

    操作上很直接:在 ΔT 误差包络内把整套检验重复跑一遍,看结论是否翻转。在误差窗里翻转的结论,不能当结论用。 还要给”没看见”留个位置:那本就是一次半影月食或低食分月食,可见性必须作为正经假设参与竞争。

    结论取决于哪把尺子先说话

    若字学上能找到一条可接受、能让记录归位、又不影响其余四条的读法,”5中4″其实是”5中5″;若怎么读都对不上、邻近记录还持续整体偏移,”从未中断”就得降级为分段命题;若 ΔT 误差窗把两边边界一起盖住,就只能给区间结论——这也是有价值的负结论。

    对科学易学来说这不是考据细节。五运六气的岁运推算、择日系统的吉凶判定、子午流注的时辰归属,全搭在干支这根时间轴上。一条不处理的例外项,会让整条证据链变成半成品。 把第五条判清楚,比再找十条”对得上”的记录更有价值。

    证据与局限

    已验证: 商代甲骨五条月食记录中四条匹配公元前1198—前1180年真实天象(PMID 28119545);古文字材料的缺字与释读不确定已被系统性标注(OBIMD 标注21,667个缺字位置,PMID 41826332);甲骨碎片自动缀合已达96.4%/90.1% 准确率(PMID 37314908);放射性碳区域偏移与贝叶斯建模可裁决百年级年代学争论(PMID 32807792 / 24204188 / 40435227)。

    部分验证: ΔT 在公元前1200年前后约29,000秒、误差±1,000秒,折算地面经度±4.3—4.4度(中文期刊2021年研究,非PubMed源);自适应采样可把浮动序列收窄到年精度(PMID 42720022)。但”第五条例外项的三把尺子联合判别”尚未系统执行。

    待验证: 该条记录在全部字学可接受释读下的匹配率及其假匹配率零分布;邻近与缀合记录的日数差是局部跳格还是持续偏移;在 ΔT 包络内重复检验时结论是否翻转。

    相关阅读

    证据等级: B
    参考文献: PMID 28119545, PMID 41826332, PMID 37314908, PMID 32807792, PMID 24204188, PMID 40435227, PMID 42720022


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    The Lunar Eclipse in Oracle Bones That Doesn’t Match: Was It a Misread Character or a Flaw in the Stems and Branches Date System?

    Evidence level: B

    “Five eclipses match four” is the headline, but the real value lies in the fifth—whether the ancients carved it wrong or if the Gan-Zhi day-counting system genuinely failed during that period. These two scenarios produce identical results on a single isolated record.

    Main Text (1000–1500 words)

    Records of lunar eclipses on Shang Dynasty oracle bones have a widely circulated claim: comparing reconstructed Gan-Zhi days with modern astronomical data shows that four out of five oracle inscriptions precisely align with real lunar eclipses between 1198 and 1180 BCE (PMID 28119545). This is often cited as proof that the Gan-Zhi system was continuous and reliable during the Shang Dynasty.

    However, few question the remaining one. Most popular science explanations dismiss it with a “possible recording error,” yet this anomaly is actually the keystone of the entire evidence chain—it simultaneously accommodates two radically different interpretations.

    “Observational Equivalence” in a Single Record

    The first explanation is interpretive ambiguity: the Gan-Zhi characters were misread (characters like “Jia” and “You,” or “Wu” and “Xu,” are visually similar), or the month number or day count was omitted, or the oracle question was not about a lunar eclipse at all. The record is correct; we misread it.

    The second explanation is actual disruption: the characters are read correctly, and the record is accurate, but the Gan-Zhi day-counting system experienced a counting error or a break in transmission during that period. The premise that “the sixty Jiazi cycle has never been interrupted since the Shang Dynasty” does not hold here.

    The difficulty lies in the fact that the observational result is identical for both causes: “this one does not match.” Like a clock: seeing “one number is wrong” could mean the print was faulty, or the clock ran fast by one hour—looking at just that number never distinguishes them. To separate them, we need three different measuring tools.

    First Measuring Tool: Lay All Acceptable Readings on the Table

    The approach is not to “find a reading that matches,” but to list all acceptable orthographic interpretations, generate a predicted date for each, and verify if a lunar eclipse actually occurred and was visible on that day. “Interpretive uncertainty” is a primary attribute of ancient script data—the 2026 OBIMD multi-modal dataset for oracle bones annotates character and sentence levels across 10,077 rubbings and specifically flags 21,667 “missing character positions” (PMID 41826332).

    Here lies the first pitfall: interpretive shopping. With too many interchangeable readings, researcher freedom explodes—by repeatedly switching interpretations, one can always force a match within the list of real eclipses. Therefore, the criterion cannot be “does a specific reading match,” but rather “does this match fall outside the zero distribution of false matches”: randomly draw Gan-Zhi days from the real eclipse list around 1200 BCE and calculate what proportion “happens to match.” Also, one cannot simply declare the record erroneous and continue using the other four as evidence—that treats the conclusion as a premise. The ΔT field deliberately avoids observations that “derive dates solely from celestial records,” adhering to the same discipline (PMID 28119545).

    Second Measuring Tool: Local Misreading vs. Continuous Disruption

    If it is merely a misreading of one character, restoring the correct reading restores consistency between the day number difference and the Gan-Zhi sequence difference (modulo 60) in adjacent records—it is an isolated point. If the Gan-Zhi system truly broke during that period, then all subsequent records would shift systematically, creating a cumulative error.

    Thus, the test must focus on “relative positions in adjacent records”: arrange the Gan-Zhi days from the same oracle bone and other fragments that can be joined (assembled) into a chain, calculating day differences pairwise. Fragment assembly is critical evidence—once joined, co-occurring Gan-Zhi days on the same plate become hard constraints; feasibility is supported by models like SFF-Siam based on twin networks, achieving average accuracies of 96.4% and 90.1% on two benchmark sets (PMID 37314908).

    Like a necklace: breaking one bead restores the distance between adjacent beads; shifting the whole string by one link causes every subsequent bead to be off. A local jump of one is a misreading; a global shift of one is a disruption—these are their fingerprints.

    Third Measuring Tool: Ask First, “Are We Measuring Accurately?”

    The most overlooked point is: precision precedes conclusion.

    Oracle bone dating for lunar eclipses relies on the World Time correction ΔT—the cumulative clock error caused by Earth’s slowing rotation. Around 1200 BCE, ΔT was approximately 29,000 seconds with an error of ±1,000 seconds, translating to ground longitude errors of ±4.3° to ±4.4° (Research on Shang Oracle Bone Lunar Eclipse Records Based on Modern Astronomical Planetary Ecliptics, 2021). If the discrepancy of the problematic record falls within this error window, no conclusion can be drawn—this is insufficient range, not a faulty record; moreover, ΔT itself is calculated using ancient eclipse records (PMID 28119545), creating a circular dependency risk.

    Absolute anchors also carry systematic errors: Mediterranean-Anatolia tree-ring sequences show a radiocarbon offset of approximately 0 to 22 years across different regions during the same period (PMID 32807792). Treating “relative sequence” as a prior in Bayesian models can compress chronology to generational scales (PMID 24204188); recent high-precision data has adjudicated the debate between Egypt’s “high chronology” vs. “low chronology”—two schemes spanning a century only one holds up (PMID 40435227); adaptive sampling strategies offer a solution to pin precision to the year and narrow the window (PMID 42720022).

    Operationally, it is direct: repeat the entire test within the ΔT error envelope and see if the conclusion flips. A conclusion that flips within the error window cannot be treated as a conclusion. Also, reserve space for “not seeing”: this could have been a partial penumbral eclipse or a low-magnitude eclipse, so visibility must be included as a legitimate competing hypothesis.

    Conclusion Depends on Which Tool Speaks First

    If orthography allows an acceptable reading that places the record correctly without affecting the other four, then “four out of five” is actually “five out of five.” If no reading matches and adjacent records show continuous systematic shifts, then “never interrupted” must be downgraded to a segmented proposition. If the ΔT error window covers both boundaries, only an interval conclusion is possible—this is also a valuable negative result.

    For scientific Yijing Studies, this is not merely textual detail. The calculation of Five Movements and Six Qi annual cycles, the auspiciousness judgments in Date Selection systems, and the hour attribution in Zi-Wu Liu Zhu all rely on this Gan-Zhi time axis. One unaddressed exception renders the entire evidence chain a work in progress. Clarifying the fifth case is more valuable than finding ten more records that match.

    Evidence and Limitations

    Verified: Four out of five Shang Dynasty oracle bone lunar eclipse records match real celestial phenomena from 1198–1180 BCE (PMID 28119545); missing characters and interpretive uncertainties in ancient script materials are systematically annotated (OBIMD flags 21,667 missing positions, PMID 41826332); automated assembly of oracle bone fragments achieves accuracies of 96.4%/90.1% (PMID 37314908); radiocarbon regional offsets and Bayesian modeling can adjudicate century-scale chronological debates (PMID 32807792 / 24204188 / 40435227).

    Partially Verified: ΔT was approximately 29,000 seconds around 1200 BCE with ±1,000 second error, translating to ±4.3°–4.4° ground longitude (2021 Chinese journal study, non-PubMed source); adaptive sampling can narrow floating sequences to annual precision (PMID 42720022). However, the “three-tool joint discrimination for the fifth exceptional item” has not been systematically executed.

    To Be Verified: The match rate and false-match zero distribution under all acceptable orthographic interpretations for this record; whether the day difference in adjacent/joined records represents a local jump or continuous shift; whether the conclusion flips when retested within the ΔT envelope.

    Related Reading

    Evidence Level: B
    References: PMID 28119545, PMID 41826332, PMID 37314908, PMID 32807792, PMID 24204188, PMID 40435227, PMID 42720022


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.

  • 甲骨文与金文里的时间尺该怎么验:全量干支日记录的逐日连续性检验

    Evidence level: B

    干支纪日是不是一条连续三千年的绝对时间轴,不靠”五条月食对上四条”这种抽查来证明,而要靠”整条记录链上有没有断点”来证明——这正是树轮断代的做法

    正文(1000-1500字)

    上一篇文章讲了一个很漂亮的证据:殷墟甲骨上的五条月食记录,用干支日与现代天文回推一比,五条里有四条精确指向公元前1198到前1180年之间的真实天象(PMID 28119545)。这条证据说明,干支纪日这把”时间尺”在商代就已经在走字了。

    但抽查通过,不等于全链可信。

    打个比方:一条三百米长的钢卷尺,你在头、中、尾各量三段,都对得上刻度。这只说明这三段是准的,不能说明中间没有一处焊缝接歪了。要证明整把尺子准,你得从头到尾逐格量。干支纪日也一样——它的价值不在于”有几个天文锚点对上了”,而在于整条记录链上有没有中断、有没有错乱、有没有重置。

    甲骨文不是日记本,是散落的档案

    检验的第一道坎,是材料形态。干支纪日理论上是一个”计数器”(甲子日之后必是乙丑日,六十天一循环,像星期一样连续),但甲骨卜辞不是一份逐日连续的日志,而是三千年前散落各处的占卜档案:这片问祭祀、那片问打仗、另一片问田猎。要把它们排成一条链,先得把每一片上的干支日都读出来。

    过去这靠学者逐片人工摹录,效率低、易遗漏。现在数据和技术都到位了:2024年发布的 HUST-OBC 甲骨文开放数据集收录了14万张甲骨图像,含77,064张已破译、62,989张未破译(PMID 39242622);识别端,基于孪生网络的少样本模型把甲骨文字的识别 F1 提到83.3%(PMID 35984774),针对拓片噪声的结构-纹理分离网络进一步解决了出土污染导致的识别退化(PMID 35420984)。也就是说,”把全量甲骨、金文里的干支日批量读出来”这件事,第一次具备了工程可行性。

    排链、对锚:三步走

    拿到全量干支日之后,检验分三步:

    第一步排链。 按考古层位与同片共现关系,把干支日按”相对先后”排序,检查任意两个可比较记录之间的日数差是否等于干支序号差(模60)。差得对,说明这段链是连续的;差得不对,就是一个可疑的中断点。

    第二步对锚。 单靠内部一致性还不够——一条自洽的错链也可能稳定地错。必须请绝对时间进来:日食/月食回推(PMID 28119545)、碳14测年数据银行(中国考古碳14数据库已收录7,083条测年数据,PMID 41120310)、西周金文历谱,三路外部锚点交叉。这与树轮断代的思路完全一致——浮动年代序列必须靠宇宙成因核素事件或碳14走时曲线钉到绝对纪年上(PMID 32229554)。

    第三步查断点。 这是全篇的核心:把检验出的所有”日数差 ≠ 干支序号差”的位置收集起来,看它们是随机噪声,还是聚成若干稳定的中断/错乱点。

    三只统计的”坑”

    坑一:记录不是随机采样。 有干支日的卜辞,问的事是扎堆的(祭祀、征伐、田猎),某些日子天然更少被刻录。这种”记录偏向”会让某些日期的缺口看起来像中断,其实只是没人记。必须用记录密度做阴性对照。

    坑二:读错一个字,就凭空多一个断点。 识别模型准确率再高,也不是100%。一个干支字被读错(甲读成由、戊读成戌),检验里就会冒出一个假中断。所以识别置信度必须传导进检验——低置信度的记录要么剔除,要么加权,不能当成硬事实。树轮断代早就面对过同一问题:树轮边界缺失或误判会系统性地歪掉整个年代序列,所以发展出概率化的”年龄不确定性”框架(PMID 32960924)。

    坑三:检验次数一多,总会有假中断。 全量数据意味着成千上万次两两比较,纯靠运气也会出现若干”差不对”的组合。必须用置换检验建立零分布,再上假发现率(FDR)校正,只有当某个中断点落在”随机也能出现”的零分布之外,才算真断点。

    三种结局,三种意义

    结局一:全段连续(在识别误差范围内)。 那么干支纪日就是一条被现代天文+碳14双重背书、连续三千年以上的绝对时间基准——比任何现代纪年体系的记录史都长。

    结局二:局部中断被稳定检出。 这说明”时钟”确有漂移点,”干支纪日从未中断”这个前提需要分段修正在具体年代上成立,而不是笼统成立。这本身就是一个重要的断代学结论——它告诉后来者:哪一段可以放心用来做五运六气与择日的推算,哪一段必须加注。

    结局三:检验功效不足。 甲骨记录太稀疏加识别误差太大,导致无法裁决。这也是有价值的负结论——它圈定了”现有数据能回答到哪一步”的边界。

    对科学易学来说,这不是考古细节。五运六气的岁运推算、择日系统的吉凶判定、子午流注的时辰归属,全都搭在这根时间轴上。检验它的连续性,本质上是检验整座大厦的地基是不是实心的。

    证据与局限

    已验证: 干支纪日自公元前720年连续使用2600余年,《春秋》37条日食33条与现代回推吻合(PMID 28119545);商代甲骨五条月食记录中四条匹配前1198-前1180年真实天象(PMID 28119545)。
    部分验证: 全量提取的技术条件已具备——HUST-OBC 14万张甲骨图像开放(PMID 39242622)、孪生网络识别 F1 83.3%(PMID 35984774)、结构-纹理分离网络抗噪声识别(PMID 35420984);中国考古碳14数据库 7,083 条可作绝对锚(PMID 41120310);树轮断代的浮动序列锚定与缺失边界概率化框架可直接移植(PMID 32229554 / 32960924)。但”甲骨+金文全量逐日对齐”尚未系统执行。
    待验证: 商代至西周全段是否存在中断/错乱;识别误差对”假中断”检出率的具体贡献;改历重置点与真中断的判别边界。

    相关阅读

    证据等级: B
    参考文献: PMID 28119545, PMID 39242622, PMID 35984774, PMID 35420984, PMID 41120310, PMID 32229554, PMID 32960924


    免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。

    How to Verify the Time Scale in Oracle Bone Script and Bronze Script: Daily Continuity Testing of Full Calendar Records

    Evidence level: B

    Is the Gan-Zhi day-counting system a continuous absolute timeline spanning three millennia, validated not by sampling (e.g., matching five lunar eclipses to four instances), but by detecting discontinuities within the entire record chain—much like dendrochronology?

    Main Text (1000-1500 words)

    The previous article presented compelling evidence: five lunar eclipse records on oracle bones from Yin Xu, when cross-referenced with modern astronomical retrocalculations using the Gan-Zhi calendar, matched four of the five records precisely to celestial phenomena between 1198 and 1180 BCE (PMID 28119545). This evidence demonstrates that the Gan-Zhi calendar has been functioning as a “time scale” since the Shang Dynasty.

    However, passing a sampling check does not may support full-chain reliability.

    Consider an analogy: if you measure three sections of a 300-meter steel tape at the head, middle, and tail, and they all align with the scale, this only proves those three sections are accurate; it does not prove there is no misaligned weld in the middle. To verify the entire tape, you must measure every single segment. Similarly, the value of Gan-Zhi day-counting lies not in “how many astronomical anchors match,” but in whether the entire record chain contains interruptions, dislocations, or resets.

    Oracle Bones Are Not Diaries, but Scattered Archives

    The first hurdle in validation is the material form. Theoretically, Gan-Zhi acts as a “counter” (following Jiazi, Yi-Chou follows, with a 60-day cycle like the week), implying continuity. However, oracle bone inscriptions are not a daily continuous log; they are scattered占卜 (divination) archives from three millennia ago: some ask about sacrifices, others about warfare, and still others about hunting. Reconstructing a chain requires reading the Gan-Zhi day for every single bone.

    Historically, scholars relied on manual transcription of each bone, which was inefficient and prone to omission. Today, data and technology are in place: the 2024 release of the HUST-OBC Oracle Bone Open Dataset includes 140,000 images, containing 77,064 deciphered and 62,989 undeciphered (PMID 39242622). Recognition models based on twin networks have raised the F1 score for oracle bone character recognition to 83.3% (PMID 35984774), while structure-texture separation networks further address recognition degradation caused by excavation pollution (PMID 35420984). Thus, the task of “batch reading all Gan-Zhi days from full quantities of oracle bones and bronze script” has finally achieved engineering feasibility.

    Chain Reconstruction and Anchor Alignment: A Three-Step Process

    Once full quantities of Gan-Zhi days are obtained, validation proceeds in three steps:

    Step 1: Chain Reconstruction. Sort Gan-Zhi days according to archaeological stratigraphy and co-occurrence relationships on the same bone to establish a “relative sequence.” Verify whether the numerical difference in days between any two comparable records equals the difference in their Gan-Zhi indices (modulo 60). A correct match indicates continuity; a mismatch suggests a suspicious discontinuity.

    Step 2: Anchor Alignment. Internal consistency alone is insufficient—a self-consistent but erroneous chain can remain stable. Absolute time must be introduced: lunar/solar eclipse retrocalculations (PMID 28119545), the Carbon-14 dating data bank (the Chinese Archaeological Carbon-14 Database has recorded 7,083 dating data points, PMID 41120310), and Western Zhou bronze script chronologies. These three external anchors must cross-verify. This aligns perfectly with dendrochronology logic: floating date sequences must be pinned to absolute chronologies via cosmogenic nuclides or Carbon-14 decay curves (PMID 32229554).

    Step 3: Discontinuity Detection. This is the core of the analysis: collect all positions identified where “day difference ≠ Gan-Zhi index difference.” Determine whether these are random noise or cluster into stable discontinuities/errors.

    Three Statistical “Pitfalls”

    Pitfall 1: Records Are Not Random Sampling. Oracle bones containing Gan-Zhi days cluster around specific topics (sacrifices, campaigns, hunting). Certain dates naturally have fewer records due to this “recording bias,” creating apparent gaps that are merely unrecorded days rather than interruptions. Negative controls using record density must be employed.

    Pitfall 2: One Character Misread Creates a False Discontinuity. Even with high recognition accuracy, it is not 100%. Misreading a Gan-Zhi character (e.g., reading Jia as You, or Wu as Xu) generates a false discontinuity in the validation. Therefore, recognition confidence must be propagated into the validation process; low-confidence records should either be excluded or weighted, not treated as hard facts. Dendrochronology has long faced this issue: missing or misjudged tree-ring boundaries can systematically skew entire date sequences, leading to the development of probabilistic “age uncertainty” frameworks (PMID 32960924).

    Pitfall 3: More Tests Increase False Positives. Full-quantity data implies thousands of pairwise comparisons. Pure chance can produce several “mismatched” combinations. Permutation tests must establish a null distribution, followed by False Discovery Rate (FDR) correction. Only discontinuities falling outside the null distribution (where random chance cannot explain them) count as true breaks.

    Three Outcomes, Three Implications

    Outcome 1: Full Continuity (within recognition error margins). If the chain remains continuous, Gan-Zhi day-counting establishes an absolute time基准 (benchmark) spanning over three millennia, doubly backed by modern astronomy and Carbon-14 dating. This exceeds the recorded history of any modern calendar system.

    Outcome 2: Stable Detection of Local Discontinuities. This indicates specific drift points in the “clock.” The premise that “Gan-Zhi day-counting never interrupted” requires segmentation correction for specific eras rather than holding universally. This itself constitutes a significant chronometric conclusion, informing future users which segments can be safely used for Five Movements and Six Qi calculations or Date Selection, and which require annotations.

    Outcome 3: Insufficient Validation Power. If oracle bone records are too sparse and recognition errors too high to reach a verdict, this is also a valuable negative conclusion. It delineates the boundary of “where existing data can answer.”

    For scientific Yijing Studies, this is not merely an archaeological detail. Five Movements and Six Qi annual calculations, Date Selection吉凶 (auspiciousness/inauspiciousness) judgments, and Zi-Wu Flowing Notes hour attribution all rely on this time axis. Validating its continuity is essentially verifying whether the foundation of the entire edifice is solid.

    Evidence and Limitations

    Verified: Gan-Zhi day-counting has been used continuously for over 2,600 years since 720 BCE; 33 out of 37 solar eclipse records in the Spring and Autumn Annals match modern retrocalculations (PMID 28119545); four out of five lunar eclipse records from Shang Dynasty oracle bones match real celestial phenomena from 1198–1180 BCE (PMID 28119545).
    Partially Verified: Technical conditions for full extraction are now met—HUST-OBC releases 140,000 oracle bone images (PMID 39242622), twin network recognition achieves F1 of 83.3% (PMID 35984774), and structure-texture separation networks resist noise (PMID 35420984); the Chinese Archaeological Carbon-14 Database offers 7,083 data points for absolute anchoring (PMID 41120310); probabilistic frameworks from dendrochronology for anchoring floating sequences and handling missing boundaries can be directly transplanted (PMID 32229554 / 32960924). However, “full daily alignment of oracle bones and bronze script” has not yet been systematically executed.
    To Be Verified: Whether discontinuities or errors exist across the full span from Shang to Western Zhou; the specific contribution of recognition errors to false discontinuity detection rates; the discriminative boundary between calendar reset points and true interruptions.

    Related Reading

    Evidence Level: B
    References: PMID 28119545, PMID 39242622, PMID 35984774, PMID 35420984, PMID 41120310, PMID 32229554, PMID 32960924


    Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.