Evidence level: B
先给结论
上一轮(第 23 轮)立了前置门禁 S0:判 K1/K2/K3 之前先查 σ 精度,不够直接停。它只是设计,一次没跑过。本轮把它对 #26 σ 数据库里 6 个站逐站跑完。
四句话:
- 依 S0-1 严格判,6 站 0 站可进 K1/K2/K3。 理由很朴素:6 站 F6 全部是”地质图类比 / 区域地质图 / BGS 地质图”,没有一个有实测 σ,而 S0-1 明文规定这种情况判 K4。
- 三档体系对”精度差”的反应是不对称的,而且偏向花钱那一侧。 判 K3(透明、别测了)要求
A < 7/√q,σ 越差这个门槛越严;判 K1(强屏蔽、值得测)要求A ≥ 13,与 σ 精度完全无关,永远不会变难。结果是:σ 越差,越只能得出”强屏蔽”或”说不清”,越不可能得出”别测了”。 - K2 这一档会自消,临界 σ 跨度 3.449×。 它是唯一两头封住的带
[7,13) dB。整条 A 区间要落进去,需要7·√q < 13,即q < (13/7)² = 3.449。σ 跨度超过这个数,K2 里不存在任何解——不是判不准,是没有解。S0-2 写的”跨度 ≤ 3.5×”恰好压在这条线外侧 1.5%,余量为 0。 - 上一轮”σ ±3.16× ⇒ A 只能判到 1.39 dB”这个形式不成立。 因为
A ∝ √σ,误差是乘性的:A 的相对不确定度恒为×√q,与 A 多大、与 d 和 f 全都无关。写成”固定上限多少 dB”是错的,正确问法是”在某个 A 附近,判据边界离它有几个相对不确定度”。
一、先把误差的形状摆正:A 的不确定度是乘性,且与 A 大小无关
A = 8.686·d/δ,δ = 503.29/√(fσμᵣ),所以
A ∝ d · √(f σ)
固定 d、f,让 σ 在 [σ_lo, σ_hi] 内取值、q = σ_hi/σ_lo:
A ∈ [ A_lo , A_lo·√q ] A_hi/A_lo = √q
A 的相对不确定度 = ×√q / ÷√q ← 与 A 的绝对值无关,与 d、f 也无关
我验证过四组完全不同的 (d, f):(1000, 7.83)、(100, 7.83)、(3000, 30)、(1000, 20),A 区间比值全部是 √q,一模一样。
这条的实践含义:换更深的测点、换更低或更高的频点,都改变不了判档能力,只改 A 的绝对值。想改善判档,唯一的动作是把 σ 测准。所以”换个频点让信号大一点再判”这条路是死的。
上一轮的写法是给 A 加了一个”固定上限”(1.39 dB)。按乘性关系复算,A 的不确定度半宽是 A·(√q − 1/√q)/2,跟着 A 线性长大:
| A(dB) | q = 10 时的半宽(dB) |
|---|---|
| 0.48 | 0.69 |
| 1.53 | 2.17 |
| 4.83 | 6.87 |
| 10.0 | 14.23 |
| 15.27 | 21.73 |
1.39 dB 这个数只在 A ≈ 0.98 dB 处成立,而库里没有任何一站落在那。所以我不主张”上一轮算错了某一步”,只主张”固定上限这个形式本身不成立”——按 R19/R20,口径不可复现的数值应报”口径不明”,不该报”已核”。
二、本轮最硬的新发现:三档对精度不对称,且偏向”屏蔽”
2.1 精确的可判条件
三档是:K3 = (−∞, 7)、K2 = [7, 13)、K1 = [13, +∞)。要”确定”落进某一档,整条 A 区间都得在带内:
| 档 | 可判条件 | 化成对 σ 的要求 | 随 q 变差时 |
|---|---|---|---|
| K3 | A_hi < 7 |
A_lo < 7/√q |
门槛越来越严(7/√q 往下掉) |
| K2 | 7 ≤ A_lo 且 A_hi < 13 |
7√q < 13 → q < 3.449 |
整档消失 |
| K1 | A_lo ≥ 13 |
与 q 无关 | 永不收紧 |
2.2 这就是问题所在
把”不可判区间”画出来,它是一条 [7/√q, 13) 的带:
| σ 总跨度 q | 不可判带下界 7/√q |
带子宽度 | K2 |
|---|---|---|---|
| 1.0× | 7.00 dB | 6.00 dB | 可用 |
| 2.1× | 4.83 dB | 8.17 dB | 可用 |
| 3.449× | 3.77 dB | 9.23 dB | 空集 |
| 10× | 2.21 dB | 10.79 dB | 空集 |
| 100× | 0.70 dB | 12.30 dB | 空集 |
下界 7/√q 随 q 单调往下掉,上界 13 一动不动。 翻译成决策语言:
σ 越不准 ⇒ 要判"透明(别测了)"需要 A 小到近乎零 ⇒ 几乎判不出来
σ 越不准 ⇒ 要判"强屏蔽(值得测)"只需 A ≥ 13 ⇒ 难度毫无变化
这套判档体系有一个系统性的方向偏差:精度差的时候,它只会说”强屏蔽”或者”说不清”,永远不会说”别测了”。
这不是学术洁癖问题——K1 的实际后果是”投钱做 MT 实测”,K4 是”继续悬着”,K3 是”停手省钱”。偏差指向的正是唯一花钱的那一侧。
⚠️ 口径铁律:这不是“靠山真的有屏蔽效应”。恰恰相反——它说明在没有实测 σ 的时候,体系停在 K4 是对的,而任何让它”落到 K1″的路径都是精度不足的产物,不是物理发现。
2.3 与 R27 的关系
R27(第 24 轮立)说的是”止损闸门终止与通过代价不对称时,阈值不得设在目标值上”。本轮这个是R27 的一种新形态,而且更隐蔽:
- R27 讲的是阈值位置偏心
- 本轮讲的是阈值位置不动、但判据随输入变差而单侧收紧——偏心是自动发生的,不需要任何人做决定
建议登记为 R29「判据单侧自动收紧」:任何一个分档体系,若各档门槛对输入不确定度的敏感度不一致,则整体判定会随精度变差朝固定方向漂移。查办法:把每档的可行条件写成 q 的函数,看谁在收紧、谁不动。
三、K2 自消:临界就是 3.449×,S0-2 压在线外 1.5%
K2 是唯一两头封住的带,因此它的存在本身依赖 σ 精度:
整条 A 区间落进 [7, 13):
7 ≤ A_lo 且 A_lo·√q < 13
⇒ 7·√q < 13
⇒ q < (13/7)² = 3.4490×
σ 跨度超过 3.449×,K2 是空集。 等价说法:σ 的对称误差要控制在 ±55.0% 以内,K2 才存在。
S0-2 原文:
通过条件:跨度 ≤ 3.5×(对应 A ±30% @10 dB)
两个数是对得上的(3.5 vs 3.449,差 1.5%),“A ±30% @10 dB”这个注也是对的——以 A_lo = 7 为锚、√q = 1.857 展开,区间正是 [7, 13],中点 10,左右各 30%。这里我要纠正自己中间两版的判断(先说自相矛盾 2.07 倍,再说这个注具误导性,都不成立)。
真问题只有一个,但很硬:3.5 恰好压在这条线外侧,余量为 0。 也就是说 S0-2 判”通过、可进三档”的场地,K2 已经不存在了。用一个让目标档消失的跨度,批准进入那个档。
若要留余量:
| K2 内两侧余量 | 允许 σ 跨度 |
|---|---|
| 0(刚好填满) | 3.449× |
| 0.5 dB | 2.778× |
| 1.0 dB | 2.250× |
| 2.0 dB | 1.494× |
(二)”对应 A ±30%”这个注本身没错,但只对”以 7 为锚”的读法成立。 q = 3.449 时 √q = 1.857,从 A_lo = 7 展开正好得到 [7, 13],中点 10,左右各 30%——这是对的。但它容易被读成”任意 A 处的误差都是 ±30%”。以 A = 10 为锚会得到 [5.38, 18.57](−46% / +86%);以区间几何中点 √(7×13) = 9.54 为锚,误差才正好对称。这句注应该补上锚点,否则读者会把”特定锚点下的 ±30%”当成”全局 ±30%”。
(三)真正致命的是紧跟着的那条降级条款:
跨度 3.5–10× → 降级:K2 带需放宽为 ±6 dB 并双站差分强制
K2 全宽只有 6 dB。放宽到 ±6 dB(全宽 12 dB)后:
A = 7 dB → 接受区间 [ 1, 13] → 同覆盖 K3 / K2 / K1
A = 10 dB → 接受区间 [ 4, 16] → 同覆盖 K3 / K2 / K1
A = 13 dB → 接受区间 [ 7, 19] → 同覆盖 K2 / K1
放宽到 ±6 dB 的 K2 不再能区分任何一档。 按 R11,判据只能证伪不能验证;但一个接受全部三档的判据连证伪都做不到,它不构成判据。
四、六站逐站过闸
d = 1 km、f = 7.83 Hz、μᵣ = 1(与 #22/#26 口径一致)。A = 8.686·d/δ。
4.1 只看声明区间(最宽松口径)
| 站 | F6 来源 | 声明 σ | q | A(dB) | 档位 | 可判 |
|---|---|---|---|---|---|---|
| L’Aquila 亚平宁 | 地质图类比 | 10⁻²–10⁻¹ | 10× | 4.83 – 15.27 | 跨 K3/K1 | ❌ |
| 印尼 BMKG | 地质图类比 | 10⁻² | 1× | 4.83 | K3 | ✅ |
| 留尼汪 IPGP | 地质图类比 | 10⁻²–10⁻¹ | 10× | 4.83 – 15.27 | 跨 K3/K1 | ❌ |
| 冰岛 Vestmannaeyjar | 地质图类比 | 10⁻¹ | 1× | 15.27 | K1 | ✅ |
| Nagycenk(平原) | 区域地质图 | 10⁻³ | 1× | 1.53 | K3 | ✅ |
| Portishead(平原) | BGS 地质图 | 10⁻³ | 1× | 1.53 | K3 | ✅ |
4/6。但这里的”可判”是假的——那 4 个站的”单点”不是测出来的单点,是从地质图上抄下来的一个数量级代表值。
4.2 补进 F6 溯源不确定度(±1 个数量级,q × 100)
| 站 | q_total | A(dB) | 宽度 | 档位 | 可判 |
|---|---|---|---|---|---|
| L’Aquila | 1000× | 1.53 – 48.29 | 46.8 dB | 跨 K3/K1 | ❌ |
| 印尼 BMKG | 100× | 1.53 – 15.27 | 13.7 dB | 跨 K3/K1 | ❌ |
| 留尼汪 IPGP | 1000× | 1.53 – 48.29 | 46.8 dB | 跨 K3/K1 | ❌ |
| 冰岛 | 100× | 4.83 – 48.29 | 43.5 dB | 跨 K3/K1 | ❌ |
| Nagycenk | 100× | 0.48 – 4.83 | 4.35 dB | K3 | ✅ |
| Portishead | 100× | 0.48 – 4.83 | 4.35 dB | K3 | ✅ |
2/6。两个平原站能撑住,原因很朴素:它们的 A 全部小于 4.83 dB,远低于 7 dB 的 K3 边界,即使 σ 差 100 倍也爬不上去。这不是”它们可判”,是”它们离边界足够远,误差追不上”。
4.3 依 S0-1 判规:0/6
S0-1 要求”有 MT 反演 / 岩芯四极法 / 井中测量之一”,否则”仅有区域地质图 → 判 K4,停”。6 站 F6 全是地质图类,0/6。
H2 成立:0/6 可进 K1/K2/K3。
三种口径必须分开报(这是本轮的一条方法学要求):
| 口径 | 可判 | 说明 |
|---|---|---|
| 只看声明区间 | 4/6 | 单点站的跨度记为 1.0×(不成立,单点不是测量) |
| 加溯源 ±1 数量级 | 2/6 | 平原两站靠”离边界远”撑住 |
| 依 S0-1 判规 | 0/6 | 无实测方法一律 K4 |
只报第一行会得出”H2 部分失效”的错误结论;只报第三行会掩盖”平原站其实离边界很远”这个有用的信息。
4.4 S0-2 有个漏检通道
S0-2 只看”声明区间宽度”,不看 F6 来源等级。一个站声明单点、跨度 1.0×,S0-2 满分通过——但 F6 是地质图类比,真不确定度 100×。顺序上 S0-1 先拦住了,可一旦有人跳过 S0-1 直接问 S0-2,就被放行。
建议 D30:S0-2 的”跨度”定义为含溯源的总跨度,不是声明区间跨度。
4.5 每站要”救活”需要多准的 σ
| 站 | A_lo | 判进某档需要 |
|---|---|---|
| L’Aquila / 印尼 / 留尼汪 | 4.83 dB | 判 K3 需 q < 2.10×(±35% 以内) |
| 冰岛 | 15.27 dB | 判 K1 需 q ≤ 1.38×(±16% 以内) |
| Nagycenk / Portishead | 1.53 dB | 判 K3 需 q < 21.0×(已满足) |
这是本轮对采样决策最具体的一条:给 L’Aquila 做一次精度 2 倍以内的 MT,它就从”未定义”变成”能判 K3 还是不能”。而冰岛要求 ±16%,难度高一档。
五、L’Aquila 的具体反例:同一站,两种相反结论
L’Aquila 是 #25 评分最高的 A 档候选(总分 20,”可启动”)。σ 声明 10⁻²–10⁻¹:
σ = 10⁻² ⇒ A@1km = 4.83 dB ⇒ K3(终止,不作屏蔽论据)
σ = 10⁻¹ ⇒ A@1km = 15.27 dB ⇒ K1(强屏蔽,最值得测)
同一个站,按 σ 区间两端,从”完全透明、别浪费仪器时间”变成”强屏蔽体、最值得测”。 相差 10.44 dB,跨两档。
所以说:在拿到实测 σ 之前,这个站的档位是”未定义”,不是”待定”。 任何基于文献区间给它排”优先级第 1″的说法,排的其实是”σ 上限可能很高”这一可能性,不是”这个站有屏蔽”。
六、这一轮没做什么
- 没做任何现场测量。 6 站 σ 全是文献/地质图值,本轮只用它们检验”判档能不能执行”。
- 没改任何已发布正文(#26 / #25 / #22 / #77 的 mtime 全旧)。S0-2 的阈值与降级条款修订全部停在建议层(D30/D31),等 JOE 批。
- 没回答”靠山能不能屏蔽 ELF”。K4 语义是「不可判定」,按 G0 纪律不得写成「无效应」。
- 没做 T3(R26/R27 套池⑤ F1–F5 的 #194),留给下轮。
- 中途自我纠错三次(全部在定稿前用独立复算推翻):
1. 先把 K2 临界误算成13/7 = 1.857×(解7√q < 13时漏了平方);
2. 又误判 S0-2 的”3.5×”与”±30%”自相矛盾 2.07 倍(实际只差 1.5%,两者等价);
3. 又误称”±30%”这个注具误导性(实际成立,只是缺锚点);
4. 另有一处把 K3 的判据端点用错(该看A_hi却看了A_lo)。
留下的教训:乘性误差的换算是本轮最容易出错的地方。凡”倍数 → dB”或”dB → 倍数”,必须两端直接算,不能用半宽公式线性化;判”整条区间落在带内”时,K3 看 A_hi、K1 看 A_lo、K2 两端都看,用错端点会得出完全相反的结论。
证据与局限
证据等级 B(方法学推导 + 库内数值复算),不标 E1。
- 可核部分:
A = 8.686·d/δ、δ = 503.29/√(fσμᵣ)是标准良导体趋肤深度公式;三档区间、(13/7)² = 3.449的自消临界、2.25×的留余量门槛、六站 A 值与[7/√q, 13)不可判带,全部解析可复算。六站 σ 直接取自 #26 §4/§5 表内自述,未做二次解释。 - 文献支撑(6 条 PMID + 3 条 DOI,全部 E-utilities / Crossref 回查通过,2026-10-28):建筑地磁屏蔽实测(20000094)、ELF 变压器暴露(24131394)、低磁环境生物效应综述(28654641)、EMF-昼夜节律(36681118)、MT 三维反演揭示地热系统(33097774)、MT 区域电阻率数据集(42472177);σ 频率弥散与地电导率幂律(10.1029/2012WR012700、10.1038/267673a0、10.1093/gji/ggw375)。
- ⚠️ 但没有任何一条文献直接研究”Schumann 站局部 σ 的判档精度”——这是本项目自己的方法学构造,故标 B。
六条硬局限:
- δ 公式假设均匀半空间、良导体、无位移电流、μᵣ 常数。 真实山体分层、各向异性、σ 随频率弥散(Revil 2013)——所以
3.449×这个临界在真实地质下会漂,漂多少本轮没算,需 #46b 双组成弥散修正接进来。 - “实测 σ”这条门槛本身没被检验。 S0-1 说 MT 反演 / 岩芯四极法 / 井中测量任一即可,但这三种精度差多少、能否都算”过”,本轮没查。一个有 MT 反演但模型非唯一性很重的场地,精度可能仍远差于 ±55%。
- 1 km 深度是约定,不是物理。 换 100 m 或 3 km,A 绝对值变、判档能力不变(已验证)。”哪个深度才对”是另一个未决问题。
- 六站之外还有 24 站没查。 #26 目标 30 站,本轮只覆盖表内 6 站。”0/6 可判”不能外推成”全球 Schumann 站都不可判”。
- “偏向屏蔽”这条是结构性推论,不是实测证据。 它只说明这套判据在有偏输入下的行为,不说明任何真实场地的屏蔽能力。按 R11,这类推论只承担证伪职能。
- “±1 个数量级”这个溯源不确定度本身是估计值(源自 #26 §6.1 自评”精度 ±0.5–1 数量级”)。若真实是 ±0.5 数量级,
q = 31.6,K2 已空集但 A 区间会窄很多,2/6 的结论可能变。
不构成:任何站点的屏蔽能力结论;任何 Schumann 频率与健康的关系;对 K1–K4 体系”该保留还是废除”的判断(本轮只证明它现在不可执行、且有方向偏差,没论证它设计目标错了)。
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。
相关阅读
schumann-station-sigma-database.md(#26,本轮被判定的六站来源)schumann-station-candidate-tier-a-wet-basalt.md(#25,A 档候选清单)pool4-tier-resolution-floor-k-tier-partition-inconsistency.md(#77,S0 的出处)sigma-magnitude-filter-sop-five-pools-environmental-frequency.md(#23,A(dB) 公式口径)high-sigma-lithology-elf-shielding-path.md(#42,EXP-004 判档的执行侧)experiments/pool4-s0-gate-six-station-per-station-verdict.md(本轮治理卡 P4-AUDIT-019)
引用锚点
- Chernykh AM. Shielding of the geomagnetic field in apartment houses. Gig Sanit. 2009 Sep-Oct.(PMID 20000094)
- Sirav B, Seyhan N. Extremely low-frequency magnetic fields of transformers and possible biological and health effects. Electromagn Biol Med. 2014;33(4).(PMID 24131394)
- Binhi VN, Prato FS. Biological effects of the hypomagnetic field: An analytical review of experiments and theories. PLoS One. 2017;12(6).(PMID 28654641)
- Martel J, Chang SH, et al. Influence of electromagnetic fields on the circadian rhythm: Implications for human health and disease. Biomed J. 2023;46(1).(PMID 36681118)
- Di Paolo F, et al. La Palma island (Spain) geothermal system revealed by 3D magnetotelluric data inversion. Sci Rep. 2020;10:18181.(PMID 33097774)
- Puntu JM, et al. A comprehensive deep regional electrical resistivity dataset of Taiwan from magnetotelluric surveys. Data Brief. 2026;67:113047.(PMID 42472177)
- Revil A. Effective conductivity and permittivity of unsaturated porous materials in the frequency range 1 mHz–1 GHz. Water Resour Res. 2013. doi:10.1029/2012WR012700
- Jonscher AK. The ‘universal’ dielectric response. Nature. 1977;267:673–679. doi:10.1038/267673a0
- Beskardes GD, Weiss CJ. Estimating the power-law distribution of Earth electrical conductivity from low-frequency, controlled-source electromagnetic responses. Geophys J Int. 2016. doi:10.1093/gji/ggw375
Six-station sequential gate pass: S0 access completes 0/6 allowing entry into three tiers, with the three-tier system overall leaning toward \”shielding\
Evidence level: B
S0 Access Control: Running Through All Six Stations
S0 Overview
- Evidence Level: C (traditional consensus)
- Traditional Consensus: In the traditional framework, “running through all six stations” (逐站过闸) refers to systematically evaluating each of the six directional sectors in sequence.
- Structural Analogy: This process functions like a multi-stage filter where each station represents a specific directional assessment point.
S0 Access Control
- Evidence Level: B (structural analogy)
- Traditional Consensus: The “access control” (门禁) in this context refers to the systematic evaluation of directional suitability.
- Structural Analogy: Think of it as a sequential gate system where each station must be passed before proceeding to the next.
S0 Running Through All Six Stations
- Evidence Level: A (structural analogy)
- Traditional Consensus: “Running through all six stations” (逐站过闸) means systematically evaluating each of the six directional sectors in sequence.
- Structural Analogy: This process functions like a multi-stage filter where each station represents a specific directional assessment point.
S0 Three-Stage System
- Evidence Level: B (structural analogy)
- Traditional Consensus: The “three-stage system” (三档体系) refers to the three levels of directional assessment: favorable, neutral, and adverse.
- Structural Analogy: Think of it as a tiered evaluation system with three distinct levels of directional suitability.
S0 Three-Stage System Bias
- Evidence Level: C (traditional consensus)
- Traditional Consensus: The “three-stage system” (三档体系) in this context refers to the three levels of directional assessment: favorable, neutral, and adverse.
- Structural Analogy: Think of it as a tiered evaluation system with three distinct levels of directional suitability.
- Evidence Level: C (traditional consensus)
- Traditional Consensus: The “bias toward shielding” (偏向”屏蔽”一侧) means the overall system tends to favor protective or restrictive directional configurations.
- Structural Analogy: This suggests the framework is more inclined toward creating barriers or protective boundaries rather than open pathways.
Preliminary Conclusions
The previous round (Round 23) established the preliminary gate S0: before evaluating K1/K2/K3, check the precision of σ; if insufficient, stop immediately. This was a design specification and has not yet been executed. In this round, it is applied station-by-station to the 6 stations within the #26 σ database.
Four key points:
- Strictly applying S0-1, 0 of the 6 stations qualify for K1/K2/K3. The reasoning is straightforward: all 6 stations’ F6 entries are “geological map analogy / regional geological map / BGS geological map,” none of which include measured σ. Since S0-1 explicitly mandates K4 under these conditions, no further evaluation proceeds.
- The three-tier system responds asymmetrically to “precision differences,” favoring the side requiring expenditure. Evaluating K3 (transparent, do not measure) requires
A < 7/√q; as σ precision worsens, this threshold becomes stricter. Evaluating K1 (strong shielding, worth measuring) requiresA ≥ 13, which is entirely independent of σ precision and never becomes harder. Consequently: as σ precision worsens, only “strong shielding” or “unclear” conclusions are possible; “do not measure” becomes increasingly unlikely. - The K2 tier self-eliminates, with a critical σ span of 3.449×. It is the only tier bounded on both ends by
[7,13) dB. For the entire A interval to fall within this range,7·√q < 13must hold, implyingq < (13/7)² = 3.449. If σ span exceeds this value, no solution exists within K2—not due to insufficient precision, but because no valid solution is mathematically possible. S0-2’s specification of “span ≤ 3.5×” lies exactly 1.5% outside this boundary, leaving zero margin. - The previous formulation “σ ±3.16× ⇒ A can only be judged to 1.39 dB” is invalid. Since
A ∝ √σ, the error is multiplicative: the relative uncertainty of A is constant at×√q, regardless of A’s magnitude or values of d and f. Expressing this as “fixed upper limit in dB” is incorrect; the proper question is “how many relative uncertainties lie between the judgment boundary and a given A.”
I. Correcting the Shape of the Error: A’s Uncertainty is Multiplicative and Independent of A’s Magnitude
A = 8.686·d/δ, δ = 503.29/√(fσμᵣ), therefore
A ∝ d · √(f σ)
Fixing d, f, and letting σ vary within [σ_lo, σ_hi] with q = σ_hi/σ_lo:
A ∈ [ A_lo , A_lo·√q ] A_hi/A_lo = √q
A's relative uncertainty = ×√q / ÷√q ← independent of A's absolute value, and independent of d, f
I verified four completely distinct (d, f) pairs: (1000, 7.83), (100, 7.83), (3000, 30), (1000, 20); the ratio of A’s interval bounds is always √q, identical across all cases.
Practical implication: Changing to deeper measurement points or shifting to lower or higher frequencies does not alter the classification capability; it only changes A’s absolute value. To improve classification, the sole action is to measure σ accurately. Therefore, the approach of “shifting frequency to increase signal strength before classification” is unworkable.
The previous formulation imposed a “fixed upper limit” of 1.39 dB on A. Recalculating via the multiplicative relationship, A’s uncertainty half-width is A·(√q − 1/√q)/2, which grows linearly with A:
| A(dB) | q = 10 时的半宽(dB) |
|---|---|
| 0.48 | 0.69 |
| 1.53 | 2.17 |
| 4.83 | 6.87 |
| 10.0 | 14.23 |
| 15.27 | 21.73 |
The value 1.39 dB holds only at A ≈ 0.98 dB, yet no measurement point in the dataset falls there. Thus, I do not claim “an error occurred in a previous step of the calculation”; rather, I assert that the form of a “fixed upper limit” itself is invalid—per R19/R20, a non-reproducible numerical aperture should be reported as “aperture undefined,” not “verified.”
II. The Hardest New Discovery of This Cycle: Three-Tier Asymmetry Biased Toward “Shielding”
2.1 Precise Determinability Conditions
The three tiers are defined as: K3 = (−∞, 7), K2 = [7, 13), and K1 = [13, +∞). To conclusively determine that a measurement falls into one tier, the entire A interval must lie within the corresponding band:
| Tier | Determinability Condition | Requirement Translated to σ | Behavior as q Varies |
|---|---|---|---|
| K3 | A_hi < 7 |
A_lo < 7/√q |
Threshold tightens progressively (7/√q decreases) |
| K2 | 7 ≤ A_lo AND A_hi < 13 |
7√q < 13 → q < 3.449 |
Tier vanishes entirely |
| K1 | A_lo ≥ 13 |
Independent of q | Never tightens |
2.2 This Is the Core Problem
Plotting the “undeterminable interval” reveals it is a band defined by [7/√q, 13):
| Total σ span q | Undeterminable lower bound 7/√q |
Band width | K2 status |
|---|---|---|---|
| 1.0× | 7.00 dB | 6.00 dB | Available |
| 2.1× | 4.83 dB | 8.17 dB | Available |
| 3.449× | 3.77 dB | 9.23 dB | Empty set |
| 10× | 2.21 dB | 10.79 dB | Empty set |
| 100× | 0.70 dB | 12.30 dB | Empty set |
The lower bound 7/√q decreases monotonically as q increases, while the upper bound 13 remains fixed. Translated into decision language:
σ less accurate ⇒ To judge "transparent (do not measure)" requires A to be nearly zero ⇒ almost impossible to determine
σ less accurate ⇒ To judge "strong shielding (worth measuring)" only requires A ≥ 13 ⇒ difficulty remains unchanged
This tiering system exhibits a systematic directional bias: when precision is poor, it will only conclude “strong shielding” or “undeterminable,” never “do not measure.”
This is not an issue of academic fastidiousness—the practical consequence of K1 is “spending money to conduct MT measurements,” K4 means “continue waiting,” and K3 means “stop and save money.” The bias points precisely toward the only side that requires spending money.
⚠️ Caliber Iron Law: This is NOT “the backing truly has a shielding effect.” On the contrary—it indicates that in the absence of measured σ, the system stopping at K4 is correct, and any path forcing it “into K1” is a product of insufficient precision, not a physical discovery.
2.3 Relationship with R27
R27 (established in Cycle 24) states: “When the stop-loss gate exhibits asymmetry between termination and passage costs, the threshold must not be set at the target value.” This current finding represents a new form of R27, and a more subtle one:
- R27 concerns bias in threshold location
- This cycle concerns threshold location remaining fixed while the criterion tightens unilaterally as input uncertainty increases—bias occurs automatically, requiring no human decision
Recommend registering this as R29 “Unilateral Automatic Criterion Tightening”: In any tiered system, if the sensitivity of tier thresholds to input uncertainty differs across tiers, the overall determination will drift in a fixed direction as precision degrades. Verification method: Express the feasible condition for each tier as a function of q and observe which tightens and which remains static.
III. K2 Self-Exclusion: The Critical Threshold is 3.449×, S0-2 Presses Outside by 1.5%
K2 is the only band bounded at both ends; therefore, its existence itself depends on σ precision:
The entire A interval falls within [7, 13):
7 ≤ A_lo and A_lo·√q < 13
⇒ 7·√q < 13
⇒ q < (13/7)² = 3.4490×
If the σ span exceeds 3.449×, K2 is an empty set. An equivalent statement: σ’s symmetric error must be controlled within ±55.0% for K2 to exist.
S0-2 original text:
Passing condition: span ≤ 3.5× (corresponding to A ±30% @10 dB)
The two numbers align (3.5 vs 3.449, difference 1.5%), and the note “A ±30% @10 dB” is also correct—using A_lo = 7 as anchor and √q = 1.857 to expand, the interval is indeed [7, 13], midpoint 10, ±30% on each side. I must correct my intermediate judgments (first claiming self-contradiction at 2.07×, then stating this note is misleading; neither holds).
The real issue is singular but rigorous: 3.5 presses exactly on the outside of this line, with zero margin. That means S0-2 judges “passing, eligible for the third tier” for sites where K2 already does not exist. Approving entry into that tier using a span that eliminates the target tier.
To retain margin:
| K2 inner-side margins | Allowed σ span |
|---|---|
| 0 (just filled) | 3.449× |
| 0.5 dB | 2.778× |
| 1.0 dB | 2.250× |
| 2.0 dB | 1.494× |
(II) The note “corresponding to A ±30%” is itself correct, but only holds for the reading anchored at 7. When q = 3.449, √q = 1.857; expanding from A_lo = 7 yields exactly [7, 13], midpoint 10, ±30% on each side—this is accurate. However, it risks being read as “error at any A is ±30%.” Anchoring at A = 10 yields [5.38, 18.57] (−46% / +86%); anchoring at the interval geometric midpoint √(7×13) = 9.54 yields exactly symmetric error. This note should specify the anchor point, otherwise readers may mistake “±30% under a specific anchor” for “global ±30%.”
(III) The truly fatal issue is the immediately following downgrading clause:
Span 3.5–10× → downgrade: K2 band must be widened to ±6 dB and forced dual-station differencing
K2’s full width is only 6 dB. Widening to ±6 dB (full width 12 dB) results in:
A = 7 dB → acceptance interval [ 1, 13] → covers K3 / K2 / K1
A = 10 dB → acceptance interval [ 4, 16] → covers K3 / K2 / K1
A = 13 dB → acceptance interval [ 7, 19] → covers K2 / K1
K2 widened to ±6 dB can no longer distinguish any tier. Per R11, the criterion can only falsify, not verify; but a criterion that accepts all three tiers cannot even falsify, so it does not constitute a criterion.
IV. Station-by-Station Gate Crossing for the Six Stations
d = 1 km, f = 7.83 Hz, μᵣ = 1 (consistent with #22/#26). A = 8.686·d/δ.
4.1 Declaration Interval Only (Most Lenient Scope)
| Station | F6 Source | Declared σ | q | A (dB) | Tier | Verifiable |
|---|---|---|---|---|---|---|
| L’Aquila Apennine | Geological map analogy | 10⁻²–10⁻¹ | 10× | 4.83 – 15.27 | Across K3/K1 | ❌ |
| Indonesia BMKG | Geological map analogy | 10⁻² | 1× | 4.83 | K3 | ✅ |
| Réunion IPGP | Geological map analogy | 10⁻²–10⁻¹ | 10× | 4.83 – 15.27 | Across K3/K1 | ❌ |
| Iceland Vestmannaeyjar | Geological map analogy | 10⁻¹ | 1× | 15.27 | K1 | ✅ |
| Nagycenk (plain) | Regional geological map | 10⁻³ | 1× | 1.53 | K3 | ✅ |
| Portishead (plain) | BGS geological map | 10⁻³ | 1× | 1.53 | K3 | ✅ |
4/6. However, the “verifiable” count here is spurious—the single-point values for those four stations were not measured; they are order-of-magnitude representative values copied directly from geological maps.
4.2 Incorporating F6 Traceability Uncertainty (±1 order of magnitude, q × 100)
| Station | q_total | A (dB) | Width | Tier | Verifiable |
|---|---|---|---|---|---|
| L’Aquila | 1000× | 1.53 – 48.29 | 46.8 dB | Across K3/K1 | ❌ |
| Indonesia BMKG | 100× | 1.53 – 15.27 | 13.7 dB | Across K3/K1 | ❌ |
| Réunion IPGP | 1000× | 1.53 – 48.29 | 46.8 dB | Across K3/K1 | ❌ |
| Iceland | 100× | 4.83 – 48.29 | 43.5 dB | Across K3/K1 | ❌ |
| Nagycenk | 100× | 0.48 – 4.83 | 4.35 dB | K3 | ✅ |
| Portishead | 100× | 0.48 – 4.83 | 4.35 dB | K3 | ✅ |
2/6. Only the two plain stations can hold, for a straightforward reason: their A values are all below 4.83 dB, well beneath the 7 dB K3 boundary; even a σ difference of 100 times cannot bridge that gap. This is not “they are verifiable,” but rather “they lie sufficiently far from the boundary that error cannot catch up.”
4.3 Per S0-1 Criteria: 0/6
S0-1 requires “MT inversion / borehole four-electrode method / in-well measurement,” otherwise “only regional geological maps → assign K4, stop.” All six F6 stations rely on geological map analogies, yielding 0/6.
H2 holds: 0/6 can advance to K1/K2/K3.
Three scopes must be reported separately (a methodological requirement for this round):
| Scope | Verifiable | Note |
|---|---|---|
| Declaration interval only | 4/6 | Single-point station span recorded as 1.0× (invalid, single point is not a measurement) |
| Plus traceability ±1 order of magnitude | 2/6 | Two plain stations supported by “distance from boundary” |
| Per S0-1 criteria | 0/6 | No field methods → all K4 |
Reporting only the first row yields the erroneous conclusion “H2 partially failed”; reporting only the third row obscures the useful information that “the plain stations actually lie far from the boundary.”
4.4 S0-2 Has a Missed Detection Channel
S0-2 considers only “declared interval width,” ignoring the F6 source level. A station with a declared single point and span 1.0× passes S0-2 with full marks—but F6 is a geological map analogy, so the true uncertainty is 100×. In sequence, S0-1 blocks it first; however, if someone skips S0-1 and goes directly to S0-2, it is permitted.
Recommendation D30: The “span” definition for S0-2 should be the total span including traceability, not the declared interval span.
4.5 Required σ Accuracy to “Rescue” Each Station
| Station | A_lo | Required for tier assignment |
|---|---|---|
| L’Aquila / Indonesia / Réunion | 4.83 dB | To assign K3, need q < 2.10× (within ±35%) |
| Iceland | 15.27 dB | To assign K1, need q ≤ 1.38× (within ±16%) |
| Nagycenk / Portishead | 1.53 dB | To assign K3, need q < 21.0× (already satisfied) |
This is the most specific sampling decision criterion in this round: performing MT with ≤2× accuracy on L’Aquila transforms it from “undefined” to “capable of deciding K3 or not.” Iceland requires ±16%, which is one tier harder.
V. Specific Counterexample of L’Aquila: Same Site, Two Opposite Conclusions
L’Aquila is the #25 highest-rated A-grade candidate (total score 20, “launchable”). σ statement 10⁻²–10⁻¹:
σ = 10⁻² ⇒ A@1km = 4.83 dB ⇒ K3 (termination, not used as shielding evidence)
σ = 10⁻¹ ⇒ A@1km = 15.27 dB ⇒ K1 (strong shielding, most worth testing)
For the same site, depending on the σ interval endpoints, it shifts from “fully transparent, do not waste instrument time” to “strong shield, most worth testing.” The difference is 10.44 dB, spanning two grades.
Therefore: Before obtaining the measured σ, this site’s grade is “undefined,” not “pending.” Any claim ranking it “priority #1” based on literature intervals actually ranks the possibility of “σ upper limit potentially high,” not “this site has shielding.”
VI. What Was Not Done in This Cycle
- No on-site measurements were performed. All six stations of σ relied entirely on literature and geological map values; this cycle used them solely to verify whether the “categorization” (pattern and timing classification) could be executed.
- No previously published main text was modified (the modification times for #26 / #25 / #22 / #77 remained unchanged). Revisions to the thresholds and downgrade clauses for S0-2 were retained at the recommendation level (D30/D31) pending approval by JOE.
- The question of whether “reliance” can shield ELF was not answered. The semantics of K4 are “undecidable”; per G0 discipline, this cannot be written as “no effect.”
-
T3 was not performed (the pool #194 involving R26/R27 and the five frequency bands F1–F5), leaving it for the next cycle.
-
Three self-corrections were made mid-cycle (all overturned by independent recalculation before finalization):
1. Initially, the critical threshold for K2 was miscalculated as13/7 = 1.857×(solving7√q < 13omitted the square root);
2. A misjudgment occurred regarding the apparent contradiction between “3.5×” and “±30%” in S0-2 (a factor of 2.07, whereas the actual difference is only 1.5%, making them equivalent);
3. An error was made in stating that the “±30%” note is misleading (it is valid, though it lacks an anchor point);
4. Additionally, the endpoint for K3’s criterion was incorrectly referenced (looking atA_loinstead ofA_hi).
Lesson retained: Converting multiplicative errors is the most error-prone aspect of this cycle. Any conversion between “multiplier” and “dB” (or vice versa) must be calculated directly at both ends; linearization via half-width formulas is invalid. When judging whether an entire interval falls within a band, K3 must examine A_hi, K1 must examine A_lo, and K2 must examine both endpoints; using the wrong endpoint leads to completely opposite conclusions.
Evidence and Limitations
Evidence Level B (methodological derivation + in-house numerical re-calculation); E1 not assigned.
- Verifiable components:
A = 8.686·d/δ,δ = 503.29/√(fσμᵣ)are standard good-conductor skin-depth formulas; the three-tier intervals,(13/7)² = 3.449self-elimination threshold,2.25×margin-of-safety gate, six-station A-values and[7/√q, 13)indeterminate bands are all analytically reproducible. Six-station σ values were taken directly from the self-reported tables in #26 §4/§5 without secondary interpretation. - Literature support (6 PMIDs + 3 DOIs, all verified via E-utilities/Crossref, 2026-10-28): building geomagnetic shielding measurements (20000094), ELF transformer exposure (24131394), low-magnetic environment biological effects review (28654641), EMF-circadian rhythm (36681118), MT 3D inversion revealing geothermal systems (33097774), MT regional resistivity dataset (42472177); σ frequency dispersion and geoelectrical conductivity power law (10.1029/2012WR012700, 10.1038/267673a0, 10.1093/gji/ggw375).
- ⚠️ However, no single document directly investigates “local σ classification precision for Schumann stations”—this is a methodological construct of the present project, hence Level B.
Six hard limitations:
- The δ formula assumes a uniform half-space, good conductor, no displacement current, and constant μᵣ. Real mountain stratification, anisotropy, and σ frequency dispersion (Revil 2013) mean the
3.449×threshold will shift under actual geological conditions; the magnitude of this shift is not computed here and requires the dual-component dispersion correction from #46b to be integrated. - The “measured σ” threshold itself has not been tested. S0-1 states that MT inversion, borehole four-electrode method, or well logging suffice, but the precision differences among these three and whether they can all be deemed “passed” have not been examined. A site with MT inversion but highly non-unique models may still be far worse than ±55% in accuracy.
- The 1 km depth is a convention, not physics. Changing to 100 m or 3 km alters the absolute A value but not classification capability (verified). “Which depth is correct” remains an unresolved question.
- Beyond the six stations, 24 others were not checked. #26 targets 30 stations; this round covers only the six listed. “0/6 classifiable” cannot be extrapolated to “all global Schumann stations unclassifiable.”
- “Bias shielding” is a structural inference, not measured evidence. It describes the behavior of this criterion under biased input but does not indicate shielding capacity of any real site. Per R11, such inferences serve only falsification roles.
- The “±1 order of magnitude” uncertainty trace is itself an estimate (derived from #26 §6.1 self-assessment “precision ±0.5–1 order of magnitude”). If the true value is ±0.5 orders,
q = 31.6, K2 would be empty but the A interval would narrow significantly, potentially altering the 2/6 conclusion.
Does not constitute: conclusions on shielding capacity of any station; relationships between Schumann frequencies and health; judgments on whether the K1–K4 framework should be retained or abolished (this round only proves it is currently unexecutable and directionally biased, not that its design objectives are wrong).
Disclaimer: This document constitutes an organized comparison and research design discussion of traditional knowledge and scientific literature for understanding and reference purposes only; it does not constitute medical advice and cannot replace professional diagnosis or treatment. The effect sizes, frequency ranges, and exposure conclusions mentioned herein are largely inferences at the research-report level; individual variations are substantial. If experiencing discomfort or health concerns, please seek medical attention and have them evaluated by qualified physicians.
Related Reading
schumann-station-sigma-database.md(#26, the six designated stations for this round)schumann-station-candidate-tier-a-wet-basalt.md(#25, Tier A candidate list)pool4-tier-resolution-floor-k-tier-partition-inconsistency.md(#77, the source of S0)sigma-magnitude-filter-sop-five-pools-environmental-frequency.md(#23, the dB formula specification for A)high-sigma-lithology-elf-shielding-path.md(#42, the execution side of EXP-004 tiering)experiments/pool4-s0-gate-six-station-per-station-verdict.md(governance checkpoint P4-AUDIT-019 for this round)
Reference Anchors
- Chernykh AM. Shielding of the geomagnetic field in apartment houses. Gig Sanit. 2009 Sep-Oct. (PMID 20000094)
- Sirav B, Seyhan N. Extremely low-frequency magnetic fields of transformers and possible biological and health effects. Electromagn Biol Med. 2014;33(4). (PMID 24131394)
- Binhi VN, Prato FS. Biological effects of the hypomagnetic field: An analytical review of experiments and theories. PLoS One. 2017;12(6). (PMID 28654641)
- Martel J, Chang SH, et al. Influence of electromagnetic fields on the circadian rhythm: Implications for human health and disease. Biomed J. 2023;46(1). (PMID 36681118)
- Di Paolo F, et al. La Palma island (Spain) geothermal system revealed by 3D magnetotelluric data inversion. Sci Rep. 2020;10:18181. (PMID 33097774)
- Puntu JM, et al. A comprehensive deep regional electrical resistivity dataset of Taiwan from magnetotelluric surveys. Data Brief. 2026;67:113047. (PMID 42472177)
- Revil A. Effective conductivity and permittivity of unsaturated porous materials in the frequency range 1 mHz–1 GHz. Water Resour Res. 2013. doi:10.1029/2012WR012700
- Jonscher AK. The ‘universal’ dielectric response. Nature. 1977;267:673–679. doi:10.1038/267673a0
- Beskardes GD, Weiss CJ. Estimating the power-law distribution of Earth electrical conductivity from low-frequency, controlled-source electromagnetic responses. Geophys J Int. 2016. doi:10.1093/gji/ggw375
Leave a Reply