研究笔记 · Research Note

Linear Superposition Framework (Zhenyu 原创)

Zhenyu He · Jobs Stroustrup 阅读约 6 分钟

Linear Superposition Framework (Zhenyu 原创, 2021)

页面用途: Zhenyu 在 2021 年 PKU 大四本科研究 (胡永云 / 聂绩指导) 期间原创提出的 conceptual framework, 用于解释 Walker Circulation 在不同 oceanic forcing 距离 $D$ 下的响应。

定义

Linear Superposition framework 是 Zhenyu 在 2021-07-13 于 PKU 大四 PKU-Walker-Circulation-Research 项目中原创提出的 conceptual framework, 用于解释 Walker Circulation 在不同 oceanic forcing 距离下的响应。

设 $F (W, C)$ 为大气对 Warm-pool forcing $W$ 与 Cold-pool forcing $C$ 联合响应的 operator。二阶 Taylor expansion:

$F (W, C) = F (0, 0) + \frac{\partial F}{\partial W}_{0} W + \frac{\partial F}{\partial C}_{0} C + \frac{1}{2} \frac{\partial ^{2} F}{\partial W ^{2}}_{0} W^{2} + \frac{1}{2} \frac{\partial ^{2} F}{\partial C ^{2}}_{0} C^{2} + \frac{\partial ^{2} F}{\partial W \partial C}_{0} W C + \dots$

关键简化: 当 $W$ 和 $C$ 在空间上相距较远 (forcing distance $D$ 大) 时, 非局域 forcing 对 cross derivative 的贡献衰减 → $\partial^{2} F / \partial W \partial C \to 0$ → drop interaction term:

$F (W, C) \approx F (W, 0) + F (0, C) = Response (W) + Response (C)$

即: 双源联合响应 ≈ Warm-only 响应 + Cold-only 响应。

核心论点

1. 4-field quantitative validation (D80-D180 bulk)

对 CAM3 aqua-planet 10 cases × 30-yr runs, 4 个 diagnostic field 的 bulk correlation (Warm+Cold 实际响应 vs Wonly+Conly 线性叠加预测):

Field	Bulk correlation	Spatial pattern
SST (30-yr climate)	0.9063	Double-pole warming, WP 侧增强
Precipitation	0.9215	Double-peak: WP + EP
ω @ 500mb	0.8282	Walker cell clearly visible
Shortwave radiation	0.9324	Subsidence region dark

所有 4 个 field bulk correlation > 0.82 → linear superposition framework quantitatively validated at $D \geq 80°$ .

2. Emergent non-monotonic Walker strength

Walker strength 定义为 $∣Δ ω @500 mb ∣_{WEP - EEP}$ (见 PKU-Walker-Circulation-Research §3.1 metric), 观察: 随 $D$ 从 60° → 180° 单调 increase 的 forcing 间距, Walker strength 非单调 — peaks at $D \approx 80$ – $100°$ 然后 decrease。

Linear superposition 解释: $D \approx 80$ – $100°$ 这个 peak 恰好对应 Wonly 的 natural descending area (200-220°E) 与 cold pool 几何重合的位置 → Wonly 下沉支 + Cold pool 冷却双重 reinforcement。不需要 nonlinear mechanism 解释这个 emergent non-monotonicity。

3. Framework 失效的边界 ( $D < 80°$ ): Low-cloud positive feedback

当 $D < 80°$ 时, Warm pool 与 Cold pool 几何上 “too close”, cross-derivative $\partial^{2} F / \partial W \partial C$ 不再 negligible → linear superposition 失效。

失效的 mechanism 被 Zhenyu 诊断为 low-cloud positive feedback loop:

CP descending ↑
  → suppress convection
  → trap humidity
  → low cloud ↑
  → albedo ↑
  → SW absorption ↓
  → SST ↓
  → ΔSST ↑
  → (loop back to intensify CP descending)

4-term energy budget (SW + LH dominate):

$Δ S W$ at $D = 60°$ = $- 66.5141$ W/m² → at $D = 180°$ = $- 110.6316$ W/m² (magnitude monotonically increasing with $D$ )
$Δ L H$ , $Δ L W$ , $Δ S H$ 相对小

这是 paper v6 abstract Key Point 3, verbatim:

“Low-cloud positive feedback dominates Walker suppression at small forcing widths (D<80°), diagnosed via 4-term energy budget with SW and LH as leading terms.”

不同观点 / 与其他 framework 的对比

vs Nonlinear Walker mechanism literature

传统 literature (e.g., Cessi & Young 1992, Dijkstra 2005) 常把 Walker strength 的 spatial dependence 归因于 nonlinear feedback (e.g., Bjerknes feedback, ocean-atmosphere coupling nonlinearity)
Linear Superposition framework 说: 大部分 parameter regime ( $D \geq 80°$ ) 的 emergent behavior 可以只用 linear superposition 解释, 不需要 nonlinear; nonlinear mechanism 只在 $D \in [60°, 80°]$ boundary regime 才 kick in (low-cloud feedback)
这是一个 conceptual simplification 而非 extension — 减少 explanatory burden

vs Perturbation theory in general

任何 small-signal regime 都可以用 Taylor expansion + drop high-order term; 但通常 interaction term 是主导 (e.g., climate sensitivity feedback)
Walker/width 问题的特殊性: forcing 是空间分隔的 (W 在 WP, C 在 EP), cross-derivative 是 spatial-integrated $\partial^{2} F / \partial W (x_{W}) \partial C (x_{C})$ , decay 随 $∣ x_{W} - x_{C} ∣$ (non-local operator 性质)

vs atmospheric general circulation model sensitivity literature

Held & Soden 2006, Neelin 2011 等经典综述强调 feedback coupling 的 global integration
Linear Superposition 是 spatial-decomposition perspective — forcing 可以按空间 split, response 也可以 split, under the “forcings far apart” assumption

开放问题 (待 Zhenyu 深入或未来 research extension 时补充)

Forcing shape generalization: 目前只 test Gaussian-shape Warm pool + Gaussian-shape Cold pool; 非 Gaussian forcing (如 real-world 地形 + land-sea contrast) 下 linear superposition 是否仍然成立?
Coupled vs atmosphere-only: 目前 aqua-planet CAM3 是 “slab ocean + fixed SST” framework; coupled ocean model 下 ocean dynamics 会 introduce additional nonlinearity — linear superposition 的适用边界缩小多少?
Application to ENSO: Walker circulation 的 ENSO variability 是否也可以 decompose 为 “mean + linear superposition + residual nonlinear” 3-term?
Non-equilibrium state: 所有 validation 是 30-yr climate average; transient response (如 abrupt CO2 increase) 下 linear superposition 还 hold 吗?

应用场景

Zhenyu 已用的场景:

CAM3 aqua-planet 10 cases × 30-yr runs 验证 (Walker paper v6)

可迁移的场景:

Hadley circulation 对 meridional forcing 的 response
Madden-Julian Oscillation 的 superposition decomposition
北极 amplification 的 spatial source attribution (Arctic warming 的 tropics vs high-lat forcing decomposition)
Tropical cyclone track forcing decomposition
更广: 任何 “multiple forcings spatially separated” 的 GCM diagnostic experiment

[待补充] 未来 Hadley / ENSO decomposition 具体 case studies 可在此扩充 (Berkeley Romps Lab 或 future research extension 中若应用此 framework, 应回填实际 case + 数值验证)

开发历史 (breakthrough timeline — 7-month intuition accumulation)

2020-12-29 slide 5 (W1.001 seed): “Walker 左右非对称 framing” — 早期 framework seed
2021-04-26 master PPT slides 32-65 (W1.034 拼贴萌芽): first systematic superposition attempt
2021-05-16 胡组 15 slides slide 15: D180 self-consistent linear superposition explicit
2021-05-26 / 5-28 / 6-2 进展报告: 84 figure groups systematic (5-28 拼贴系统化到 6 fields)
🌟 2021-07-13 进展报告 29 slides slide 24 (breakthrough):

After advisor discussion, the linear superposition framework proposed by Zhenyu provided a satisfactory explanation, and was endorsed.
Prof. Ji-Nie-聂绩当场 endorsement “之后就可以开始写论文”
2021-07-16 Hu group (胡组) 28 slides: post-breakthrough 给 Hu group 的 echo, slide 20 “线性叠加可以解释 Q1”

7-month intuition accumulation: 从 2020-12 framing 到 2021-07-13 breakthrough, breakthrough 是 weekly-writeup articulate 的 accumulate 出来, 不是 lucky inspiration。

来源

research_wiki/projects/walker_hadley/overview.md §Linear Superposition (math derivation + 4-field corr table + emergent prediction + low-cloud feedback failure boundary)
research_wiki/projects/walker_hadley/overview.md §Authorship + §Paper v6 Key Points + §9-version auto-diff table
research_wiki/resume_angles/problem_solving.md §Project 5
research_wiki/resume_angles/technical_depth.md §Section 9 Walker 5-stage HPC pipeline
research_wiki/resume_angles/independent_judgment.md — 此 framework 是 “本科生提框架 + advisor 当场 endorse” 的独立判断时刻
Paper draft: internal-only manuscript draft (not submitted) (Oct 22, 2021, 3663 words, 11 refs, 未投稿)
Breakthrough artifact: 2021-07-13 breakthrough PPT (private artifact)
[待补充] 进一步的 Walker 项目 .docx 学习笔记 + 数值实验 artifact (Wonly / Conly case individual run log + 4-term energy budget breakdown PPT)