Evidence level: B
判定”运气周期相位漂移”前必须做一道归因工序:把 AMO/PDO 慢速海温振荡从序列里分离,再看 5-6 年频带与之交叉小波相干是否显著——相干显著 = 表观漂移(被大海牵着走),相干不显著 = 才是运气周期自己的事。
正文
五运六气把气候的五年、六年周期当成可推算的”准钟表”:火太过之年偏热、水太过之年偏寒,年年对表,用来预测疫病与农时。但现代气候学早就发现,地球气候不是一个孤立钟摆,而是被多个”慢速大钟”层层嵌套:太平洋十年涛动(PDO,约 20-30 年周期)、大西洋多年代际振荡(AMO,约 60-70 年周期),还有更长的趋势分量。当我们要检验五运六气的 5 年/6 年频带相位是否发生真实漂移时,最大的陷阱不是噪声,而是这些慢速振荡的”牵引效应”——大船拖着小船,慢波相位一变,高频小波的相位看起来就跟着动,而运气周期本身可能纹丝未动。这篇文章讲的就是怎么用交叉小波相干,把这道归因题做严谨。
第一步:为什么必须先去趋势、分离慢速分量
任何长序列里都同时躺着趋势(全球增暖)、慢速振荡(AMO/PDO)与目标频带(5-6 年运气周期)。不把它们分开,三个分量在谱域里互相污染,相位估计必然失真。气候学给出的标准工具是集合经验模态分解(EEMD):它自适应地把序列分解为不同尺度的本征模态函数(IMF)与一个趋势残差项,不预设基函数,能把几十年尺度的慢波从 5-6 年目标频带里干净剥出来。太湖流域 54 年(1965-2018)旱涝序列的研究就是标准示范:用 EEMD 提取年际-年代际振荡分量、EOF 降维、再用交叉小波(XWT)做遥相关分析,把趋势、慢速振荡与目标频带逐层切开(PMID 35226266)。这一道工序不做,后面所有的”相位漂移”结论都可能是趋势斜率变化伪造的。
第二步:交叉小波相干——把运气频带与 AMO/PDO 直接对表
分离出慢速分量后,关键问题变成:运气频带的相位变动,到底跟 AMO/PDO 的相位变动有多少共变性?答案要用交叉小波相干(XWT/WTC)来量化:它给出两个时间序列在时-频平面上各频带的相关强度与相位差。西南中国石笋重建的亚洲季风记录(PMID 40057570)提供了慢速振荡牵引的直接证据:小波分析识别出约 64 年主周期,与 AMO 的 60-70 年典型周期吻合;当 AMO 与 PDO 处于不同相位组合时,亚洲季风强度随之改变,雨带在”北湿南干”与”南湿北干”间切换——慢速海温振荡的相位变化,实实在在调制着区域气候频带。更早的长江中下游 300 年季风信号研究(PMID 26107375)进一步显示:夏季风信号强度与 PDO 密切相关,且 31 年频带季风同相、100 年频带反相(约 20 年相位差)——”相位”本身是时变、频带依赖的,慢波相位变化会系统性地叠加到目标频带上。
第三步:红噪声显著性检验——别把随机抖动当相干
小波相干谱上密密麻麻的”显著区”,很多是红噪声自相关造成的假阳性。标准做法是把两序列拟合成 AR(1) 红噪声过程,用蒙特卡洛生成上千对替代序列,得到每个(时间、尺度)点上相干的零分布,超过 95% 分位才算显著。这一检验在气候与生理信号里都是标配:例如多变量气候序列小波相关研究(PMID 33277562)用 WLMC 分析 AMO 与海温、热带气旋的千年相关,显著性全部建立在替代数据零分布之上;珠江流域植被-干旱研究(PMID 35303597)更进一步用偏小波相干(PWC),把 ENSO/PDO/太阳活动作为协变量偏掉之后,再检验植被与干旱的残余相干——这正是”分离表观与真实漂移”的统计骨架:把 AMO/PDO 偏掉,再看 5-6 年频带的残余相位是否仍显著。
把三步串成一条归因流水线
对任意中国气象站长序列,完整的归因检验是:①EEMD 分离趋势与 AMO/PDO 尺度慢速分量;②在 5-6 年频带提取去趋势前后的相位角,算相位差;③原始序列与 AMO/PDO 指数做交叉小波相干,红噪声替代数据检验显著性;④若相干显著,用偏小波相干把慢速指数偏掉,检验残余相位漂移是否仍通过显著性检验。结果会分叉成两种解读:若去慢速后残余漂移消失,说明观测到的”漂移”只是 AMO/PDO 相位变化的投影——表观漂移,运气周期本身依然稳定,五运六气作为稳定节律框架的假设反而被保全;若分离后仍有显著残余漂移,那才是真正需要严肃对待的”运气周期自身受扰”假说。先证伪表观漂移,再谈真实漂移——这正是运气周期检验走向可证伪科学的第一步。
这个区分对实践意义重大。若漂移主要是慢波牵引的表观效应,那么用五运六气做长期预测时,只需把 AMO/PDO 相位作为”背景场”校正,运气周期的钟表依然可靠;若分离后仍有真实残余,才需要警惕气候变暖是否真的扰动了运气周期本身。先分清是”表坏了”还是”你戴表的手在动”,再决定要不要修表。
证据与局限
已验证: EEMD+交叉小波是成熟的气候序列分离与相干检验工具(PMID 35226266);AMO(60-70 年)与 PDO(20-30 年)是真实可测的慢速海温振荡,其相位变化会调制亚洲季风强度与雨带位置(PMID 40057570);季风信号与 PDO 存在频带依赖的相位关联,相位关系时变(PMID 26107375);红噪声替代数据显著性检验与偏小波相干是标准方法(PMID 33277562, PMID 35303597)。
部分验证: 慢速振荡牵引导致高频频带”表观相位漂移”的机制在气候文献中有机制性描述,但”运气周期 5-6 年频带相位漂移是否可归因于 AMO/PDO”的直接检验尚无文献——这正是本课题要填补的空白。
待验证: ①对同一批中国多站长序列,EEMD 分离慢速分量前后 5-6 年频带相位漂移量是否显著缩小;②去慢速后残余相位漂移是否仍通过红噪声显著性检验;③AMO 与 PDO 哪个对运气频带相位的”牵引力”更大,是否存在东部 vs 西部站点差异。
相关阅读
证据等级: B(方法学与慢速振荡-季风机制有 PubMed 直接支撑,”运气频带 × AMO/PDO 分离”的直接检验尚属科学类比)
参考文献: PMID 26107375, PMID 40057570, PMID 33277562, PMID 35226266, PMID 35303597
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。
Attribution Testing of Phase Drift in Luck Band: How Strong Is the Cross-Wavelet Coherence Between Detrended 5-Year/6-Year Bands and AMO/PDO?
Evidence level: B
Before adjudicating “Yunqi cycle phase drift,” a rigorous attribution procedure must be performed: isolate the slow sea temperature oscillations of AMO/PDO from the series, then assess whether cross-wavelet coherence in the 5–6 year band is significant. Significant coherence implies apparent drift (dragged by the ocean); insignificant coherence indicates that the drift belongs to the Yunqi cycle itself.
Main Text
Yunqi treats the five- and six-year climate cycles as calculable “quasi-clocks”: years with excessive Fire are warmer, those with excessive Water are colder, and the system aligns annually to predict epidemics and agricultural timing. However, modern climatology has long established that Earth’s climate is not an isolated pendulum but is nested within multiple “slow large clocks”: the Pacific Decadal Oscillation (PDO, ~20–30 year cycle), the Atlantic Multidecadal Oscillation (AMO, ~60–70 year cycle), and longer trend components. When testing whether the 5/6-year Yunqi band phase undergoes genuine drift, the primary trap is not noise but the “drag effect” of these slow oscillations—a large ship dragging a small one; when the slow wave phase shifts, the high-frequency small wave phase appears to follow, even if the Yunqi cycle itself remains unchanged. This article details how cross-wavelet coherence can rigorously address this attribution problem.
Step 1: Why Detrending and Separating Slow Components Are Mandatory
Any long series simultaneously contains trends (global warming), slow oscillations (AMO/PDO), and the target band (5–6 year Yunqi cycle). Without separation, these components contaminate each other in the spectral domain, inevitably distorting phase estimates. The standard tool provided by climatology is Empirical Mode Decomposition with Ensemble Correction (EEMD): it adaptively decomposes the series into intrinsic mode functions (IMFs) of different scales and a trend residual without preset basis functions, cleanly stripping decades-scale slow waves from the 5–6 year target band. A 54-year (1965–2018) drought-flood sequence study in the Taihu Basin serves as a standard demonstration: using EEMD to extract interannual-decadal oscillation components, EOF dimensionality reduction, and then cross-wavelet (XWT) analysis for teleconnection studies, trends, slow oscillations, and the target band are sliced layer by layer (PMID 35226266). Without this procedure, all subsequent conclusions regarding “phase drift” may be fabricated by changes in trend slopes.
Step 2: Cross-Wavelet Coherence—Directly Aligning the Yunqi Band with AMO/PDO
After isolating slow components, the critical question becomes: to what extent is the phase variation of the Yunqi band covariant with the phase variation of AMO/PDO? The answer requires quantification via Cross-Wavelet Transform (XWT/WTC): it provides the correlation strength and phase difference between two time series across the time-frequency plane. A stalagmite reconstruction of Asian monsoon records from southwestern China (PMID 40057570) provides direct evidence of slow oscillation drag: wavelet analysis identifies a dominant ~64-year cycle, aligning with the typical 60–70 year AMO cycle; when AMO and PDO are in different phase combinations, Asian monsoon intensity changes accordingly, with rainbands switching between “north-wet south-dry” and “south-wet north-dry”—the phase shift of slow sea temperature oscillations genuinely modulates regional climate bands. Earlier research on the 300-year monsoon signal in the middle and lower reaches of the Yangtze River (PMID 26107375) further shows that summer monsoon signal strength is closely related to PDO, with the 31-year band in phase and the 100-year band anti-phase (approx. 20-year phase difference)—”phase” itself is time-varying and band-dependent; slow wave phase changes systematically superimpose onto the target band.
Step 3: Red Noise Significance Testing—Do Not Mistake Random Fluctuations for Coherence
The dense “significant regions” on wavelet coherence spectra are often false positives caused by red noise autocorrelation. The standard approach involves fitting the two series to an AR(1) red noise process, generating thousands of surrogate pairs via Monte Carlo simulation to obtain the zero distribution of coherence at each (time, scale) point; only values exceeding the 95th percentile are considered significant. This test is standard in both climatology and physiological signals: for instance, multi-variable climate series wavelet correlation studies (PMID 33277562) use WLMC to analyze millennial correlations between AMO, sea temperatures, and tropical cyclones, with significance entirely built upon the surrogate data zero distribution; a vegetation-drought study in the Pearl River Basin (PMID 35303597) goes further by using Partial Wavelet Coherence (PWC), removing ENSO/PDO/solar activity as covariates before testing the residual coherence between vegetation and drought—this precisely constitutes the statistical skeleton of “separating apparent from real drift”: removing AMO/PDO, then testing whether the residual phase shift in the 5–6 year band still passes significance testing.
Stringing the Three Steps into an Attribution Pipeline
For any long Chinese meteorological station series, the complete attribution test is: ① EEMD separation of trend and AMO/PDO scale slow components; ② extracting phase angles before and after detrending in the 5–6 year band and calculating the phase difference; ③ performing cross-wavelet coherence between the original series and AMO/PDO indices with red noise surrogate data significance testing; ④ if coherence is significant, using PWC to remove the slow indices and testing whether residual phase drift still passes significance testing. Results bifurcate into two interpretations: if residual drift disappears after removing slow components, the observed “drift” is merely a projection of AMO/PDO phase changes—apparent drift, and the Yunqi cycle itself remains stable, thereby preserving the assumption of the Five Movements and Six Qi as a stable rhythmic framework; if significant residual drift persists after separation, that is the genuine hypothesis requiring serious attention regarding “Yunqi cycle self-disturbance.” Disprove apparent drift first, then discuss real drift—this is the first step toward making Yunqi cycle testing falsifiable science.
This distinction carries significant practical weight. If drift is primarily an apparent effect driven by slow waves, then when using Five Movements and Six Qi for long-term forecasting, AMO/PDO phases should be treated as a “background field” for correction, and the Yunqi cycle clock remains reliable; if significant residual drift persists after separation, only then should one be vigilant about whether climate warming truly disturbs the Yunqi cycle itself. First distinguish whether “the watch is broken” or “your hand wearing the watch is moving,” then decide whether to repair the watch.
Evidence and Limitations
Verified: EEMD + cross-wavelet are mature tools for climate series separation and coherence testing (PMID 35226266); AMO (60–70 years) and PDO (20–30 years) are real measurable slow sea temperature oscillations whose phase changes modulate Asian monsoon intensity and rainband positions (PMID 40057570); monsoon signals exhibit phase relationships dependent on frequency bands with time variability (PMID 26107375); red noise surrogate data significance testing and partial wavelet coherence are standard methods (PMID 33277562, PMID 35303597).
Partially Verified: The mechanism by which slow oscillation drag causes “apparent phase drift” in high-frequency bands is described mechanistically in climatology literature, but there is no direct literature testing whether “Yunqi cycle 5–6 year band phase drift can be attributed to AMO/PDO”—this is precisely the gap this topic aims to fill.
To Be Verified: ① Whether phase drift in the 5–6 year band significantly shrinks before and after EEMD separation of slow components for the same set of Chinese multi-station sequences; ② whether residual phase drift after removing slow components still passes red noise significance testing; ③ which of AMO or PDO exerts a greater “drag force” on the Yunqi band phase, and whether there are differences between eastern vs. western stations.
Related Reading
- Truth Behind Yunqi Cycle Phase Drift: Scientific Separation of Apparent and Real Drift
- Statistical Framework for Yunqi Cycle Phase Drift Detection
- Five Movements and Six Qi = Climate Cycle Medicine
Evidence Level: B (Methodology and slow oscillation-monsoon mechanisms have direct PubMed support; “Yunqi band × AMO/PDO separation” direct testing remains a scientific analogy)
References: PMID 26107375, PMID 40057570, PMID 33277562, PMID 35226266, PMID 35303597
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.