Research Note

Python Scientific Computing

Zhenyu He · Jobs Stroustrup 2 min read

Expert

Python as primary tool for scientific computing and data analysis:

Processing and visualization of numerical simulation data (netCDF, HDF5)
Data pipeline construction (QA filtering, spatial remapping, temporal aggregation)
Spatial data acceleration (KD-tree indexing, gridded aggregation)
Statistical analysis and sensitivity experiment design
Implementation of inversion methods (MCMC, etc.)

— plume dynamics diagnostic pipeline, sensitivity sweeps
— Sentinel-5P CH4 pipeline, KD-tree acceleration
Undergraduate Research @ Peking University & Caltech/UCR — MCMC retrieval, climate simulation analysis
Walker Circulation Dynamics @ PKU — figure polishing + diagnostic
Aerosol Joint Retrieval Senior Thesis @ PKU — MATLAB auxiliary
DSCOVR Inverse Problem + Regularization Methods @ Caltech / UCR — 5-language context switch
— 62 ipynb aggregate

Aggregate scale signature

~175k+ lines total aggregated across 7 years of research (PKU Walker 27,882 MATLAB+NCL lines + PKU Aerosol ~6,400 MATLAB lines + Caltech DSCOVR ~3,000+ lines + PhD 115,363 lines / 62 ipynb / 3,793 code cells — Python primary + cross-language auxiliary)
62 ipynb notebooks in PhD research notebook (UC Berkeley Romps Lab aggregate)

Multi-language fluency (context-switch discipline)

5-language fluency within the 11-month Caltech DSCOVR project: IDL (Yuk Yung lab OOP MCMC class) + MATLAB (Hansen csvd) + Python (Kawahara CUDA + emcee + healpy) + Mathematica (symbolic derivations) + R (Metropolis-Hastings)
Fortran 77/90 in PhD stage (NCAR CAM3 + DAM LES source mods, see Numerical Model Source Code Modification)

Filename-as-version-control discipline

Jupyter notebooks use filename to encode methodology version (e.g., enforceEpsilon_z_iteration_v3.ipynb — filename encoding methodology iteration)
Aerosol code directory: hzy_experi1_package_collaborator_revised_now_no_error_above_height_4 (filename gloss: “hzy_experi1_package_handed-to-collaborator_revised_now-no-error-when-elevation-over-4”; filename encodes collaboration history)
See Inconvenient Evidence Retention Discipline (private)

Research-grade Python libraries