How karki2004/louisiana_oxygen_case_study works
This repository is a reproducible data-science case study, not a general-purpose software service. A researcher supplies public USGS observations to a single analysis script, which validates and prepares the time series, trains a model, simulates gaps, compares reconstruction methods, and writes results. A separate script assembles selected results into a PDF brief. Analysis stages are represented as distinct responsibilities within analyze.py; the README documents the experiment and outputs, but the sampled code is the main implementation evidence.
Main entry points: create_brief.py Brief generation · analyze.py Data validation · summary.csv Method summary · audit.json Run audit
Components 18
Analysis workflow
- Analysis run
analyze.py - Data validation
analyze.py - Feature preparation
analyze.py - Forest training
analyze.py - Complete-day cohort
analyze.py - Gap simulation
analyze.py - Reconstruction methods
analyze.py - Daily metrics
analyze.py
Analysis outputs
- Trial results
trials.csv - Method summary
summary.csv - Run audit
audit.json - Method comparison
- Data context
Research brief
- Brief generation
create_brief.py - Case-study PDF
Louisiana_Oxygen_Gap_Case_Study.pdf
Other components
- Researcher
- USGS observations
- USGS metadata
source_metadata.xml
Connections 21
- Researcher → Analysis run: runs analysis
- USGS observations → Analysis run: provides observations
- Analysis run → Data validation: loads and checks
- USGS metadata → Researcher: describes source
- Data validation → Feature preparation: prepares series
- Feature preparation → Forest training: supplies predictors
- Forest training → Complete-day cohort: predicts oxygen
- Data validation → Complete-day cohort: selects test days
- Complete-day cohort → Gap simulation: samples gaps
- Gap simulation → Reconstruction methods: masks observations
- Reconstruction methods → Daily metrics: supplies estimates
- Daily metrics → Trial results: writes trials
- Daily metrics → Method summary: aggregates metrics
- Analysis run → Run audit: writes run metadata
- Daily metrics → Method comparison: plots metric errors
- Data validation → Data context: plots observed context
- Researcher → Brief generation: runs brief
- Method summary → Brief generation: provides metrics
- Run audit → Brief generation: provides counts
- Method comparison → Brief generation: provides figure
- Brief generation → Case-study PDF: generates PDF