Earth and environment: ML domain track
You already work with spatial, seasonal, and messy observational data. This track orders the shared ML spine for earth science, climate-adjacent analysis, and environmental monitoring. It does not guarantee policy impact or a hire. Claims about the planet need careful evaluation language.
Core idea. Keep domain judgment. Add Python, regression or classification, time series, and honest metrics. Methods live in Modules 00–25.
Who this is for
- Earth science, geography, ecology, and environmental studies learners
- People analyzing sensors, stations, or remote-sensing derived tables
- Analysts who need forecasting habits without jumping straight to deep models
What you already bring
- Comfort with maps, seasons, and spatial dependence (even if informal)
- Respect for missing data and instrument outages
- Stakeholder questions that need clear uncertainty talk
ML problem types you will meet
- Time series forecasting and anomaly detection on station or sensor streams
- Regression on continuous environmental variables
- Classification of land cover or event labels from features
- Careful train/test splits that respect time and location
- Communication of error so decisions are not oversold
Gaps this track closes
- Module 01
- Modules 02–05
- Module 15
- Module 03 or Module 04 by task
- Module 21
- Module 19 when archives are tabular
Related resources
- Data Validation. Station and sensor series need checks before modeling.
- Ethics in ML. Environmental claims in public need caution.
- Common Errors. Time and space leakage are common failure modes.
- Model Explainability Cheatsheet. Quick sheet beside Module 21.
- Imbalanced Data Cheatsheet. Rare events and extremes need careful metrics.
Role emphasis (not destiny)
| Emphasis | Role | Why |
|---|---|---|
| Primary | Data Scientist | Models plus evaluation on observational data |
| Alternate | Data Analyst | Strong when reporting and SQL matter more than new models |
See Career Paths and Career Roadmap Guide. Treat times as emphasis maps only.
Intensity maps
Research support (tier B)
- Module 00 if needed
- Module 01
- Module 02
- Module 03 or Module 04
- Module 05
- Module 15
- Module 21
- Time-aware project practice via Module 15 exercises or a Module 16 project you adapt carefully
Job-oriented study (tier B)
- Complete Research support
- Add Module 19 and Module 07
- Add Module 20 for rare events
- Keep deep learning optional until baselines and Module 15 habits are solid
Ordered study checklist
- Pick intensity map
- Finish Module 01 with a public environmental CSV
- Write a time-based split before any fancy model
- Finish Module 05 and Module 15 for ordered data
- Practice one forecast plot with error bands described in plain words
- Document what the model must not be used for (policy overclaim)
Start here
Open Module 00 README if you need math or environment help. Otherwise open Module 01 README.
Honesty and traps
- Random shuffles destroy time series truth. Prefer time order.
- Spatial leakage is real when nearby points share train and test.
- Climate and environment claims in public need caution. This track teaches methods, not advocacy scripts.
Try next: Open Module 15: Time Series Analysis after Modules 01, 02, 05, and one supervised module (03 or 04).