跳过导航至主内容
Zhenyu He · Jobs Stroustrup
English 中文
Skip to article

研究笔记 · Research Note

Aerosol Joint Retrieval Senior Thesis @ PKU

Zhenyu He · Jobs Stroustrup 阅读约 11 分钟

Aerosol Joint Retrieval Senior Thesis @ PKU

相关页面: 本页是 PKU-Undergraduate-Research umbrella 下 Aerosol senior thesis project 的独立 experience page (parallel 逻辑 to PKU-Walker-Circulation-Research 拆出); source page 见 .

基本信息

  • 时间: 2021.12 — 2022.06 (~7 months, 大四春 senior thesis period; active research 5.5 months + thesis writing 4 weeks + post-defense polish 2 weeks)
  • 角色: 本科毕业论文独立研究 (Sole author, PI-supervised)
  • 组织: 北京大学物理学院 大气与海洋科学系 (Jing-Li-李婧 group)
  • 指导教师: Jing-Li-李婧 (sole formal advisor, senior thesis PI)
  • 地点: 北京 / NSCC-GZ Tianhe-II (login as (undergraduate Tianhe-II account))
  • Senior thesis artifact: (senior thesis final docx).docx (+ .pdf, 2022-06-03 defended, 1,350 words core content / 192 段 / 7 tables / 4 refs / 21 images)

核心问题

首次 (first) 实现 MODIS + CALIPSO 迭代耦合联合反演气溶胶算法 — 通过把 CALIPSO 主动遥感反演的 aerosol extinction coefficient profile 用于替换 MODIS 被动遥感 Dark Target C5 算法中的 fixed scale-height assumption; 将 MODIS 反演得到的 column-total AOD 作为 CALIPSO 多条候选消光系数廓线的 total-mass constraint (boundary condition); 两者相互 iterate 直至 MODIS AOD 与 CALIPSO extinction profile 同时收敛, 互相 supplement 对方假设中最弱的环节.

🌟 Zhenyu 原创贡献

贡献 1 — First iteratively-coupled MODIS + CALIPSO joint retrieval algorithm (§4.2 自述):

“本研究首次实现了 MODIS 和 CALIPSO 的气溶胶联合反演算法的开发，同时该联合反演算法是将 MODIS 和 CALIOP 算法相互迭代、相互耦合的，不是孤立地进行改善，而是同时对于 MODIS 的反演 AOD 值和 CALIOP 的消光系数廓线进行改进，因而两卫星在联合反演的全程可以互相匹配、改善、验证。”

→ 不是单一 satellite 产品的 ad-hoc combination, 而是 convergent iterative coupling algorithm.

贡献 2 — CALIPSO L2 non-physical bug catch + 修正:

源532 nm 消光系数范围 (0-10 km)物理合理性
本研究联合反演 (image14)0 – 0.2 km⁻¹, 全程正值✅ 物理合理
CALIPSO L2 官方产品 (image15)-5000 至 -10000 km⁻¹ 突变负值❌ 严重非物理

→ 质的改进 (qualitative) — 不是把小误差降低, 而是修正 CALIPSO 官方 L2 产品的 non-physical 伪影 (从 -10⁴ km⁻¹ negative 修正到 0-0.2 km⁻¹ 合理正值范围).

贡献 3 — 4 Beijing case study validation (vs AERONET Beijing 站 550 nm AOD ground truth):

#日期AERONET 基准 AOD误差变化 (vs MODIS-only baseline)
12007-03-110.1430abs ↓ 0.0114, rel ↓ 7.97%
22008-04-140.7051⚠️ abs ↑ 0.4565, rel ↑ 64.74% (未收敛, 反向 preserved)
32008-04-301.0797abs ↓ 0.0322, rel ↓ 2.98%
42019-03-230.1004abs ↓ 0.0196, rel ↓ 19.52%

→ 3/4 成功改进 MODIS-only, absolute AOD error 减小 0.011-0.032, relative error 减小 3-20%, 高 AOD 区改善更显著. Case 2 反向结果完整保留 in thesis (not airbrushed) — 反向成为 algorithm applicability boundary 诊断 (see 贡献 4).

贡献 4 — inconvenient evidence 独立 docx 保留:

独立 docx (supplementary inconvenient-evidence docx) preserve 完整 table: CALIPSO L2 柱总量 AOD 在 4/4 cases 都比 Zhenyu 联合反演的廓线积分 AOD 更接近 AERONET — 对 thesis main narrative disadvantageous.

实验L2 vs AERONET abs err联合 vs AERONET abs err谁更接近基准
1 (2007-03-11)0.06940.1942L2 更近 ⚠️
2 (2008-04-14)0.09171.6959L2 更近 ⚠️
3 (2008-04-30)0.32260.3044联合稍近 (持平)
4 (2019-03-23)0.05190.1541L2 更近 ⚠️

→ Zhenyu 没有把这个 comparison dimension 从 thesis 中 omit, 而是独立 docx full table preserve, 接受 defense 时可能被 point out 的 risk. 详见 Evidence Preservation Discipline (private companion) Instance 4.

贡献 5 — Failed-approach physical preservation (Instance 5 of Disadvantageous discipline):

CALIPSO_Retrieval_Changliang/ 25 versions 中 2 个 failed approaches 完整 preserved + explicit annotation:

  • try1_useH/ (2.5 GB, 早期 “useH” approach abandoned)
  • _之前错误的Cdistance版本/ (920 MB, C_distance.mlx bug 版本, 对应 2022-03-21 weekly PPT 中 “球面距离函数 bug + 前半周工作重做” 的 pre-fix 版本)
  • 顶层 txt 标注: 除了try_1_useH和_之前错误的Cdistance版本_其他文件夹中都是正确的函数.txt

→ 2 级冗余 annotation (folder name 自带”之前错误的” + 顶层 txt 显式列出错误 folders). 失败 approach 物理证据 preserved + explicit annotation, 不删除.

关键成就

  1. First iteratively-coupled MODIS + CALIPSO joint retrieval algorithm (本科毕业论文级别的 novel algorithm)

  2. CALIPSO L2 -10⁴ km⁻¹ non-physical bug catch + fix (0-10 km 高度 negative spike → 修正到 0-0.2 km⁻¹ 合理正值范围)

  3. 4 Beijing case AOD improvement (abs ↓ 0.011-0.032, rel ↓ 3-20%, vs MODIS-only baseline, AERONET ground truth)

  4. inconvenient evidence 独立 docx preservation ((supplementary inconvenient-evidence docx), 4/4 cases L2 柱总量更近 AERONET)

  5. 9-version thesis auto-diff (316 → 1,350 words in 27 days / 9 versions / 2 轮 李婧 edit pass / 9 条 substantive annotations); 316→1137 words in 3 天 (+821 words / +270 words/day = fastest content fill)

  6. 22 weekly group meeting PPTs sustained 6 月 (Dec 2021 - May 2022) at ~1 PPT/周 cadence; 含 Apr 11 dual-version (forme self-track + forteacher present) discipline

  7. 4-person lab scaffolding under Jing-Li-李婧 PI:

    • a lab senior collaborator — Fernald 1984 math pedagogy (Formula_derivation.docx 24 段) + lidar-ratio LUT (a 2020 AGU Fall Meeting paper, thesis ref [10]) + CALIPSO base code
    • Li-Chong-李充 师姐 — MODIS retrieval matlab code + 1 km 标高误差 → 30% AOD 误差 quantification + critical Caltech bridge (Li, C., Li, J., Dubovik, O., Zeng, Z. C., Yung, Y. L. 2020 Remote Sensing 12(9), 1524 paper 一作; 与 Zhenyu 同期 Caltech DSCOVR 项目 PI Yuk-Yung co-author)
    • Dong-Yueming-董悦明 — LAADS DAAC + wget MODIS L1B 下载 13-slide PPT (2020-09-04 onboarding tutorial)

  8. Mid-term commitment → final delivery 完整闭环: 2021-12-14 mid-term answer 预 commit 4 features (相关系数提升 + RMSE 降低 + 拟合斜率 → 1 + 廓线 smoother no negative), May 2022 final 全部 confirmed + 加 1 新 (CALIPSO L2 -10⁴ 非物理修正)

5.5-month trajectory 关键 milestones

日期事件状态
2021-12-10第 1 个正式 progress PPTMODIS code 跑 NaN + 主动 ask 学长学姐 (honest assessment)
2021-12-14中期答辩 PPT (12 slides) + 讲稿”Replace MODIS LUT entirely” ambitious framework + 4 features pre-commit
2021-12-17李充讨论 PPT (7 slides)line-level + path-level MODIS NaN demonstration to 师姐
2022-01-17weekly reportMODIS NaN 1-month debug closure (root cause: 暗像元算法 + 云雪亮地表 automatic NaN = expected behavior; diagnostic: cross-validate via MODIS L2 Mean_Reflectance_Land)
2022-02-21weekly reportCALIPSO single-stack closure — “反演代码中已读通, 测试中的 bug 已解决. Sigma_532, Sigma_1064, 雷达比廓线均可以得到”
2022-03-14weekly report (374 KB)联合算法 first 实例 (2019-08-01 北京 case) + 2 关键 questions: (Q1) CALIPSO vs MODIS 量级差异, (Q2) AOD 对 scale height 是否敏感 = “假迭代平衡” seed
2022-03-21weekly report🌟 C_distance.mlx bug fix — “球面距离函数 bug, 前半周工作重做” (对应 P3 _之前错误的Cdistance版本/ preserved folder)
2022-03-28weekly report (17 slides)AERONET 9 组合 methodology + self-reject 2_2 (早-晚 13 小时差)
2022-04-04weekly report (15 slides)log-linear 全廓线拟合 vs 1/e 法 multi-method comparison 加入 (见 Multi-Method-Comparison-Spine); 2019-08-01 case 两种方法 0.66 km vs 4.1 km = 6× 差异
2022-04-11dual-version forme (35 slides) + forteacher (34 slides)8 cases (2007 Jan-Apr) systematic sweep: 5 舍弃 + 1 不收敛 + 2 success, 每 rejection 有 documented reason (CALIPSO NaN / 单廓线 / 与 MODIS 差距过大 / AERONET 缺测)
2022-05-02weekly report (28 slides)AERONET 验证完全加入 + 4 final cases lock (3_6 / 2_1 / 4_8 / 4_9) = thesis Section 3.4 source-of-truth
2022-05-07thesis writing 起步0507 模板版 (316 words)
2022-06-03defense + 终稿1,350 words / 192 段 / 7 tables / 4 refs / 21 images

技术深度 (resume-grade specifics)

  • 5 种语言 integrated: MATLAB (CALIPSO + MODIS retrieval 主 stack) + Python (MODIS data + plotting, MPL_Extinction ref) + Bash (batch processing) + R + NCL
  • ~6,400 MATLAB lines across hzy_experi folders (aggregate)**: 33 个 MODIS hzy_experi folders (each ~1,057-1,606 lines, 4-6 .m files) + 24 个 CALIPSO hzy_experi* folders (each ~14 mlx files) = 大部分 copy-paste per-experiment, 真正 unique ~1,500-2,000 lines core code + hardcoded path/data adjust per folder
  • 48 GB Modis_scripts + 5.7 GB CALIPSO_Retrieval_Changliang (aggregate storage) — 含 raw HDF data + per-experiment cached .mat; 贺震昱_code_整理/ final clean reproducible version 仅 56 MB (source + 18 LUT + 2 instruction txt)
  • 22 weekly PPTs / Dec 2021 - May 2022 / ~1 PPT/周 sustained cadence (PPT size 演化反映 work intensity: Jan-Feb 30-70 KB code debug → Mar-Apr 400-900 KB experiment 图 → Apr 11 dual-version peak)
  • 9-stage experimental-design selection cascade: CALIPSO 数据质量 / 50 km 距离 / 35 min 时间差 / MODIS-CALIPSO AOD range overlap / MODIS 预反演 30 km 平均 / CALIPSO 沿轨 28 km (~163 profiles, ~8 s 飞行) 平均 / AERONET 同时段 availability / 吸收型主导 regime (Jan-Apr) / log-linear 全廓线拟合 (而非 2-point 1/e 法) — 多年 satellite dataset distilled down to 4 rigorously-validated successful experiments
  • 18 LUT .mat files (3 bands × 6 aerosol types = MODIS Dark Target C5 forward RT pre-computed tables, earlier lab production) + 12 chinaha_X.mat (1-12 月 China monthly scale height LUT) + Table.mat (Angstrom ↔ lidar ratio LUT (earlier lab production))
  • NSCC-GZ Tianhe-II HPC (login as (undergraduate Tianhe-II account), 李婧 lab account) — 与 Walker 项目 (undergraduate Tianhe-II account) (Ji Nie lab account) 同 物理 infrastructure + 不同 lab account (cross-project HPC continuity)
  • 2 次 PaperPass 旗舰版 查重 trial + 1 次知网正式查重 (post-knowledge-graph plagiarism check discipline) + 2 份 PKU 官方 forms (导师评阅表 + 审查表, 5-3 created 1 个月 buffer ahead of 6-3 defense) = PKU 本科毕业论文 administrative discipline
  • 2 GB 北京 + 中国 OpenStreetMap shapefile (over-prepared for BOUNT_line.shp Beijing boundary subset 仅用; preservation of future-work data) + 李万彪 大气物理 教材 自批注扫描版 24 MB (本科课堂教材 carry-over 到 senior thesis 阶段 reference)

学术 / 职业意义

  • 本科 portfolio 3-track parallel Sep-Dec 2021 组成部分: Aerosol thesis drafting + PKU-Walker-Circulation-Research paper v3→v6 + Caltech-DSCOVR-Research MCMC 反演 + nwp_hw 4 道作业 ≈ 3-4 deliverable/week sustained (与 Caltech DSCOVR 项目 Feb 2022 paper draft 终止后 Aerosol 单项目 full focus 不同 cadence pattern, 但同 sustained discipline)
  • Caltech institutional bridge via Li-Chong-李充: 李充 2020 Remote Sensing paper co-author 包含 Yuk-Yung — 这是 Zhenyu 在 2021.03 进入 Caltech DSCOVR 项目的 institutional pathway 之一. PKU 李婧 group → 李充 paper → Caltech Yuk Yung group 是 2021.03 Zhenyu 与 Caltech onboarding 的 co-author chain.
  • Senior thesis 2022-06-03 defense → 2022-08 UC Berkeley 入学 timeline: 毕业论文 Jun 3 defended, 9 月 start Berkeley Romps lab PhD (详见 )
  • Aerosol research → Inez Fung methane mapping (2023.12 — 2024.06) preparation training: satellite remote sensing discipline 一脉相承 (CALIPSO L1/L2 处理 → Sentinel-5P TROPOMI CH₄ 处理; iterative coupled algorithm 方法论 → emission inventory matching)
  • 致谢段 cross-project 确认 mentorship sequence: thesis 终稿致谢 (paragraph 178-186) 列出全部 mentorship 网络 — 李婧老师 (4th, Aerosol PI) [detailed mentorship] → 聂绩老师 + 胡永云老师 (5th, Walker) “本科生科研期间指导我的聂绩老师、胡永云老师” — explicit 确认: Zhenyu 本科生科研 = Walker 项目 (Nov 2020 - Oct 2021), 与 thesis = Aerosol (Dec 2021 - Jun 2022) 是 sequential 而非 parallel

使用的技能

  • 卫星遥感与数据处理 (MODIS L1B Dark Target C5 + CALIPSO L1 Fernald 方法 + AERONET L2 插值; NASA LAADS/LARC/AERONET 三 data center; 5 种插值 methodology comparison)
  • 气候物理与大气科学 (Fernald 1984 雷达方程 + Angstrom 指数与 lidar ratio 物理约束 + Kaufman & Tanré 1998 Dark Target + Twomey/Albrecht 间接辐射强迫; 吸收型 aerosol scale-height 敏感性物理 motivation)
  • Python-科学计算 (MODIS data 下载 + AERONET 插值 plotting + MPL ref code)
  • MATLAB — 未 separate skill page, 但 ~6,400 lines across hzy_experi folders (aggregate)* + CALIPSO + MODIS 双 stack 主要 implementation 是 capability evidence; 可在 Python-科学计算 skill page 补充 MATLAB 技能 sub-section

相关页面

  • PKU-Undergraduate-Research — 本科研究 umbrella (Aerosol section 缩成 cross-link to 本页)
  • Jing-Li-李婧 — sole formal advisor, reference methodology signature source
  • a senior lab collaborator — 师兄, Fernald 1984 math pedagogy + CALIPSO base code
  • Li-Chong-李充 — 师姐, MODIS code + 1km 标高 quantification + critical Caltech bridge via Yuk Yung co-author
  • Dong-Yueming-董悦明 — 师姐/师兄, MODIS download tutorial (2020-09-04)
  • Yuk-Yung — Caltech PI, 与李婧通过李充 paper paper attribution connected (bridge context for Caltech-DSCOVR-Research)
  • PKU advisor methodology companion (private) — 李婧 reference methodology signature (mirror with Ji-Nie-聂绩 Walker project (private context))
  • Evidence Preservation Discipline (private companion) — Instance 4 (难以解释.docx) + Instance 5 (try1_useH/ + _之前错误的Cdistance版本/) + Instance 6 (8-case rejection cascade)
  • Multi-Method-Comparison-Spine — log-linear 全廓线拟合 vs 1/e 法 2-point multi-method; 5 个 numerical method (linear 外插 + pchip 外插 + log 拟合 + 1/e 法 + multi-AOD_c 平均) 并列 active
  • Independent Judgment narrative (held until 2026-06-10) — Moment 5 artifact (Aerosol 难以解释 docx)
  • — thesis source page (终稿 docx + .pdf)
  • Academic-Honors-PKU — 本科毕业论文 archive
  • PKU-Walker-Circulation-Research — parallel Walker experience page (sequential predecessor; 致谢段 explicit 确认 sequence)
  • Caltech-DSCOVR-Research — 同期 Caltech 项目 (2021.03 - 2022.03 MCMC 反演; bridge via 李充 Yuk Yung co-author)
  • — Aerosol satellite remote sensing → methane mapping successor PhD 项目

来源

  • research_wiki/projects/aerosol_retrieval/overview.md (~46k words 完整, P1-P5 deep-dive 全)
  • research_wiki/resume_angles/problem_solving.md §Project 4 Aerosol (若 mention)
  • research_wiki/resume_angles/cross_disciplinary.md §Section 9 李婧 methodology (private)
  • raw artifact: raw/pku项目文件/pku_research/Lijing/毕业论文写作/(senior thesis final docx).docx (+ .pdf)
  • raw artifact: raw/pku项目文件/pku_research/Lijing/毕业论文写作/supplementary inconvenient-evidence section.docx (inconvenient evidence independent docx)
  • raw artifact: raw/pku项目文件/pku_research/Lijing/code/ (48 GB Modis_scripts + 5.7 GB CALIPSO_Retrieval_Changliang + 56 MB 贺震昱_code_整理 final clean)
  • raw artifact: raw/pku项目文件/pku_research/Lijing/毕业论文写作/ppt_pre合集/ (22 weekly PPTs, Dec 2021 - May 2022)