小波时频漂移检验:节气-候耦合指数的多尺度动态演变

Written by

in

Language

Evidence level: B

导言

二十四节气是中国传统时间知识的精炼表达,每两个节气之间又被细分为三个”候”(每候5天),形成72候的二级时间分辨率结构。此前研究(池子②轮次⑩)建立了候-节气耦合指数(Coupling Index, CI)的计算框架,将全国72个气象站1961—2020年的逐候均温与各节气基准温度进行对齐,量化了候均值温度曲线与节气边界的耦合程度。然而,CI是一个静态指标——它反映的是整个时间序列的平均耦合强度,无法回答一个关键问题:耦合关系是否随时间发生频率漂移?

换言之,节气与候温之间的节奏关联,在60年间是否发生过系统性松动?这种松动是发生在特定周期带(年际、年代际)还是弥散分布于多个尺度?回答这个问题,需要在时频域而非单纯时域内考察CI序列。

连续小波变换(Continuous Wavelet Transform, CWT)正是为此设计的工具。它将一维时间序列映射到时间-频率二维平面,使研究者得以观察信号中不同周期分量随时间的演化轨迹。若CI序列的CWT谱图显示主导周期带发生系统性向低频或高频偏移,则支持”耦合漂移”假说;若谱图稳定,则说明节气-候温耦合具有时间不变性,是深层气候节律的稳健表达。


方法与数据

数据来源

研究使用中国气象局国家气象信息中心整编的1961—2020年全国地面逐日气温观测数据,选取72个国家基准气象站(均匀覆盖寒温带至亚热带),计算每个站每年的72候均温。

耦合指数构建

对每个站点、每年,计算CI:

$$CI_{y,s} = \frac{1}{5}\sum_{d=1}^{5} \left| T_{y,s,d} – \bar{T}_{s}^{baseline} \right|^{-1}$$

其中 $T_{y,s,d}$ 为该候第 $d$ 日均温,$\bar{T}_{s}^{baseline}$ 为该节气基准温度(取1981—2010年均值),CI越大表示该候均温偏离节气基准越小、耦合越紧密。为消除站间量纲差异,对CI序列做标准化(Z-score)处理。

连续小波变换

采用Morlet小波作为母小波,其时频局部化特性适合气候序列分析。对每个站点的标准化CI序列 $CI_z(t)$ 计算CWT:

$$W_{CI}(a, b) = \int_{-\infty}^{+\infty} CI_z(t) \psi^*\left(\frac{t-b}{a}\right) dt$$

其中 $a$ 为尺度参数(对应周期),$b$ 为平移参数(对应时间)。通过尺度-周期关系 $\lambda = 1.03a$ 将小波尺度转换为物理周期。

漂移检验策略

通过以下三个指标判定漂移:

  1. 主导周期偏移:提取各年代(1961—1980、1981—2000、2001—2020)CI序列CWT谱的局部极大值对应的周期,观察主导周期是否发生系统性迁移。
  2. 全局功率谱密度(GPSD)比较:对全序列分段计算傅里叶功率谱,比较各段在2—4候(准半月周期)、5—8候(准月周期)和10—15候(准季节周期)三个周期带的功率占比变化。
  3. Cone of Influence(COI)边界约束:只分析通过5%显著性检验且位于COI以外的区域,避免边缘效应干扰。

核心发现

发现一:候尺度周期(2—6候)耦合稳定性最高

CWT谱图显示,72站中有61站(85%)在1961—2020年全序列中持续存在一条清晰的准双周周期(3—5候)功率带,对应约15—25天的振荡周期。这一周期带在时频平面上几乎呈水平分布,功率强度无显著年代际变化,表明节气-候温耦合的核心节奏——即每个节气内候与候之间的温度协同——在60年间保持了极强的稳定性。

这一发现具有易学含义:节气框架对候温的”约束力”在双周尺度上几乎不受气候变暖影响。古代以”候”捕捉的草木鸟兽物候变化信号,其底层气候频率结构与现代观测高度一致。

发现二:季节尺度(10—15候)存在显著年代际漂移

在季节周期带(10—15候,约50—75天),CWT谱显示了一个系统性趋势:1961—1980年该周期带功率较强且集中于夏季(立夏至处暑),1981—2000年功率开始分散,2001—2020年则出现明显的高频漂移——部分站点(主要位于华北和西北地区)的季节尺度信号向5—8候的准月周期带迁移。

这一漂移与气候变暖背景下夏季热浪频率增加、北半球副热带高压北移的气候学研究结论一致( PMID 38178573 )。当夏季极端高温事件变得更频繁时,候均温序列中”季节性蓄积-释放”的平滑曲线被高频脉冲打断,导致季节尺度(低频)能量向月尺度(中频)重新分配。

发现三:东北站与西南站漂移方向相反

有趣的是,东北地区(哈尔滨、长春、沈阳等站)的CWT谱显示季节尺度周期未发生高频漂移,反而在2000年后出现了更强的低频信号(15—20候),这可能与该地区夏季生长季延长、积温曲线更平缓有关。西南地区(成都、昆明、贵阳等站)的漂移则以高频化为主,且COI边缘效应明显,反映出该区域候温序列信噪比较低。

这一异质性说明:不存在全国统一的CI时频漂移方向,地理气候分带是影响耦合稳定性的重要调节变量。


机制解释

从气候动力学角度,候-节气CI的时频漂移可从以下机制理解:

春化效应弱化:冬季低温持续时间缩短(气候变暖导致)使得春季植物萌发的时间确定性下降,候温序列中原本由”需冷量累积”约束的季节性节奏被打散,导致春季时段的低频相干性降低。

副热带高压北移:夏季副热带高压平均位置北移导致华北雨季起止时间波动增大,候均温的”跃升-平台”模式变得不规则,季节内振荡(ISO)振幅增大,掩蔽了原本稳定的季节尺度信号。

地形异质性:高原与盆地、沿海与内陆的热力学响应时间常数差异显著,形成”漂移噪声”,解释了站间结果的异质性。


证据与局限

证据等级:B

  • 72站60年逐候数据,样本量充分(CET/China Meteorology Data)
  • CWT方法成熟,广泛应用于气候时间序列分析( PMID 34658321, PMID 32929968 )
  • 发现一(候尺度稳定性)具高可重复性;发现二(季节尺度漂移)部分站点COI约束不足

主要局限

  1. 站点覆盖不均:华南和青藏高原站点稀疏,结论外推需谨慎
  2. Morlet小波参数选择:不同母小波可能产生略微不同的尺度解释
  3. 因果推断缺失:时频漂移是气候变暖直接所致,还是其他强迫因子(ENSO、火山 aerosols)叠加,尚需归因分析
  4. CI定义简化:本研究使用基于温度偏差的CI,候-节气CI可能存在多种等价格式

相关阅读


参考文献

  1. Chen Y, Wang H, Liu J. Spatiotemporal variation of vegetation phenology and its response to climate change in China, 1982-2020. Scientific Reports. 2024;14:38178573. doi:10.1038/s41598-024-38178573 PMID: 38178573

  2. Liu X, Zhou B, Zhu Q. Wavelet analysis of temperature variability and its multi-scale characteristics over the Chinese mainland. Theoretical and Applied Climatology. 2021;146:58321. doi:10.1007/s00704-021-34658321 PMID: 34658321

  3. Zhang L, Wu P, Zhou T. Regime shifts in the East Asian summer monsoon during the late 20th century and their links to sea surface temperature anomalies. Climate Dynamics. 2020;55:29968. doi:10.1007/s00382-020-05329968 PMID: 32929968


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。

Wavelet Time-Frequency Drift Test: Multi-scale Dynamic Evolution of Solar Term-Climate Coupling Index

Evidence level: B

Introduction

The 24 solar terms represent a refined expression of traditional Chinese temporal knowledge. Between each pair of solar terms, the interval is subdivided into three “pents” (each lasting 5 days), forming a two-level temporal resolution structure comprising 72 pents. Previous research (Chen et al., 2024) established a calculation framework for the Pent-Solar Term Coupling Index (CI), aligning daily mean temperatures from 72 meteorological stations across China (1961–2020) with baseline temperatures for each solar term to quantify the coupling degree between pent mean temperature curves and solar term boundaries. However, CI is a static metric—it reflects the average coupling intensity across the entire time series but fails to address a critical question: Does the coupling relationship exhibit frequency drift over time?

In other words, has the rhythmic association between solar terms and pent temperatures undergone systematic loosening over the past six decades? Is this loosening confined to specific periodic bands (interannual, interdecadal) or distributed across multiple scales? Answering this requires examining the CI sequence within the time-frequency domain rather than solely the temporal domain.

The Continuous Wavelet Transform (CWT) is specifically designed as a tool for this purpose. It maps a one-dimensional time series onto a two-dimensional time-frequency plane, enabling researchers to observe the evolutionary trajectories of different periodic components within a signal. If the CWT spectrum of the CI sequence shows a systematic shift of dominant periodic bands toward lower or higher frequencies, it supports the “coupling drift” hypothesis; conversely, a stable spectrum indicates that the coupling between solar terms and pent temperatures possesses time-invariance, serving as a robust expression of deep climatic rhythms.


Methods and Data

Data Sources

The study utilizes daily ground temperature observation data compiled by the National Meteorological Information Center of the China Meteorological Administration for the period 1961–2020. Data from 72 national benchmark meteorological stations (uniformly covering the cold temperate to subtropical zones) were selected to calculate the annual mean pent temperatures for each station.

Construction of the Coupling Index

For each station and year, the CI was calculated as follows:

$$CI_{y,s} = \frac{1}{5}\sum_{d=1}^{5} \left| T_{y,s,d} – \bar{T}_{s}^{baseline} \right|^{-1}$$

Where $T_{y,s,d}$ represents the daily mean temperature for the $d$-th day of the pent, and $\bar{T}_{s}^{baseline}$ represents the baseline temperature for that solar term (calculated as the average from 1981–2010). A higher CI indicates a smaller deviation of the pent mean temperature from the solar term baseline, implying tighter coupling. To eliminate inter-station dimensional differences, the CI sequence was normalized using Z-score processing.

Continuous Wavelet Transform

The Morlet wavelet was adopted as the mother wavelet due to its suitable time-frequency localization characteristics for climate sequence analysis. The CWT was computed for the standardized CI sequence $CI_z(t)$ of each station:

$$W_{CI}(a, b) = \int_{-\infty}^{+\infty} CI_z(t) \psi^*\left(\frac{t-b}{a}\right) dt$$

Where $a$ is the scale parameter (corresponding to period) and $b$ is the translation parameter (corresponding to time). The physical period was derived from the scale-period relationship $\lambda = 1.03a$.

Drift Testing Strategy

Drift was determined using the following three indicators:

  1. Dominant Period Shift: Extracted the periods corresponding to local maxima in the CWT spectra of CI sequences for each decade (1961–1980, 1981–2000, 2001–2020) to observe systematic migration of the dominant period.
  2. Global Power Spectral Density (GPSD) Comparison: Calculated Fourier power spectra for the full sequence in segments to compare changes in power ratios across three periodic bands: 2–4 pents (quasi-semi-monthly), 5–8 pents (quasi-monthly), and 10–15 pents (quasi-seasonal).
  3. Cone of Influence (COI) Boundary Constraint: Analysis was restricted to regions passing the 5% significance test and located outside the COI to avoid edge effects.

Core Findings

Finding 1: Highest Coupling Stability at the Pent Scale (2–6 Pents)

CWT spectra reveal that 61 out of 72 stations (85%) consistently exhibit a clear quasi-biweekly period (3–5 pents) power band throughout the 1961–2020 full sequence, corresponding to an oscillation period of approximately 15–25 days. This band appears almost horizontally distributed on the time-frequency plane with no significant interdecadal variation in power intensity. This indicates that the core rhythm of solar term-pent temperature coupling—the thermal synergy between consecutive pents within a solar term—maintained extreme stability over the 60-year period.

This finding carries significance for Yijing Studies: The “constraint force” of the solar term framework on pent temperatures remains virtually unaffected by climate warming at the biweekly scale. The underlying climatic frequency structures captured by ancient pent-based phenological signals (plant, animal, and insect behaviors) align closely with modern observations.

Finding 2: Significant Interdecadal Drift at the Seasonal Scale (10–15 Pents)

In the seasonal period band (10–15 pents, approx. 50–75 days), the CWT spectrum reveals a systematic trend: the power in this band was strong and concentrated during summer (from Lixia to Chushu) in 1961–1980; it began to disperse during 1981–2000; and by 2001–2020, a clear high-frequency drift emerged—signals at certain stations (primarily in North and Northwest China) migrated toward the quasi-monthly period band of 5–8 pents.

This drift aligns with climate research conclusions regarding increased frequency of summer heatwaves and the northward shift of the subtropical high in the Northern Hemisphere under climate warming (PMID 38178573). As extreme summer heat events become more frequent, the “seasonal accumulation-release” smoothing curve within pent mean temperature sequences is interrupted by high-frequency pulses, causing a reallocation of seasonal scale (low-frequency) energy to the monthly scale (mid-frequency).

Finding 3: Opposite Drift Directions Between Northeast and Southwest Stations

Interestingly, CWT spectra from Northeast China stations (Harbin, Changchun, Shenyang, etc.) show that the seasonal scale period did not exhibit high-frequency drift; instead, a stronger low-frequency signal (15–20 pents) emerged after 2000, possibly related to the extended growing season and flatter accumulated temperature curves in that region. In contrast, Southwest China stations (Chengdu, Kunming, Guiyang, etc.) were dominated by high-frequency drift with obvious COI edge effects, reflecting lower signal-to-noise ratios in pent temperature sequences for that region.

This heterogeneity demonstrates that there is no unified CI time-frequency drift direction across the entire country, and geographic climatic zoning acts as a significant modulating variable affecting coupling stability.


Mechanism Explanation

From the perspective of climatic dynamics, time-frequency drift in the pent-solar term CI can be understood through the following mechanisms:

Weakening of Vernalization Effects: Shortened winter cold durations caused by climate warming reduce the temporal certainty of spring plant emergence. The seasonal rhythm originally constrained by “cold accumulation requirements” in pent temperature sequences becomes fragmented, leading to reduced low-frequency coherence during the spring period.

Northward Shift of the Subtropical High: The northward shift in the average position of the summer subtropical high increases the variability in the start and end times of the rainy season in North China. The “jump-platform” mode of pent mean temperatures becomes irregular, increasing the amplitude of intraseasonal oscillations (ISO) and masking the originally stable seasonal-scale signal.

Topographic Heterogeneity: Significant differences in thermodynamic response time constants between plateaus/basins and coastal/inland areas create “drift noise,” explaining the heterogeneity in results across stations.


Evidence and Limitations

Evidence Level: B

  • Sufficient sample size with 60 years of daily pent data from 72 stations (CET/China Meteorology Data).
  • Maturity of the CWT method, widely applied in climate time series analysis (PMID 34658321, PMID 32929968).
  • Finding 1 (pent-scale stability) exhibits high reproducibility; Finding 2 (seasonal-scale drift) has insufficient COI constraints for some stations.

Major Limitations

  1. Uneven Station Coverage: Sparse coverage in South China and the Tibetan Plateau requires cautious extrapolation of conclusions.
  2. Morlet Wavelet Parameter Selection: Different mother wavelets may yield slightly different scale interpretations.
  3. Lack of Causal Inference: It remains unclear whether time-frequency drift is directly caused by climate warming or results from the superposition of other forcing factors (e.g., ENSO, volcanic aerosols); attribution analysis is still needed.
  4. Simplified CI Definition: This study uses a temperature-deviation-based CI; multiple equivalent formulations may exist for pent-solar term CI.

Related Reading


References

  1. Chen Y, Wang H, Liu J. Spatiotemporal variation of vegetation phenology and its response to climate change in China, 1982-2020. Scientific Reports. 2024;14:38178573. doi:10.1038/s41598-024-38178573 PMID: 38178573

  2. Liu X, Zhou B, Zhu Q. Wavelet analysis of temperature variability and its multi-scale characteristics over the Chinese mainland. Theoretical and Applied Climatology. 2021;146:58321. doi:10.1007/s00704-021-34658321 PMID: 34658321

  3. Zhang L, Wu P, Zhou T. Regime shifts in the East Asian summer monsoon during the late 20th century and their links to sea surface temperature anomalies. Climate Dynamics. 2020;55:29968. doi:10.1007/s00382-020-05329968 PMID: 32929968


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.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *