Research Note

Satellite Remote Sensing & Data Processing

Zhenyu He · Jobs Stroustrup 1 min read

Expert

Sentinel-5P/TROPOMI L2 product processing (CH4)
Joint MODIS-CALIPSO retrieval
DSCOVR single-point light curve processing
Data formats: netCDF, HDF5
Spatial data processing: QA filtering, spatial remapping, KD-tree indexing, gridded aggregation
Benchmarking against ground-truth emission inventories (e.g., EPA)

MODIS+CALIPSO joint retrieval signatures (PKU senior thesis, 2021-2022)

First iteratively-coupled MODIS+CALIPSO algorithm (not single-satellite or ad-hoc combination)
CALIPSO L2 non-physical bug catch + correction: -10⁴ km⁻¹ negatives → corrected to physically reasonable 0-0.2 km⁻¹ range
4 Beijing case study: absolute AOD error ↓ 0.011-0.032, relative ↓ 3-20% (vs single-satellite baseline)
Lab scaffolding under Jing Li with collaborator support on Fernald math, MODIS, and tutorials, plus Caltech bridge via Yuk L. Yung

DSCOVR single-point Earth retrieval signatures (undergraduate, 2021-2022)

L1 → L2 light-curve time-series processing ((remote workstation) remote-server bootstrap)
Inverse-problem regularization (4 methods compared, see MCMC & Bayesian Inversion + Caltech 4 contributions in DSCOVR Inverse Problem + Regularization Methods @ Caltech / UCR)

TROPOMI methane signatures (PhD, 2023-2024)

Vernier caliper aggregated precision discovery: np.unique(lats/lons) 9.76e-5° ≈ 10.87m effective precision (raw ~5.5×3.5 km pixel)
Negative-slope regression inversion diagnosis + density+quantile reformulation (Nebraska R²=0.633)
Texas puzzle resolved via free-roaming hypothesis (advisor input from Inez Fung)