How elhachimi-ch/climatefiller works
Project type and purpose
• ClimateFiller is a Python framework (library + CLI tool) for manipulating in-situ climate time-series: gap-filling, bias-correction (ML + ERA5-Land), outlier detection (Isolation Forest, LOF, quantiles).
• It is not a multi-service web app but a modular data-processing toolkit with external dependencies (CDS API, Google Earth Engine).
File structure
• Top-level scripts:
• climatefiller.py (core class/module)
• kitchen.py, kitchen_v2.py (main entrypoints/orchestrators)
• quantilesdetector.py, gap_filling.py, et0_models_calibrations.py, temp_test.py (specialized workflows)
• Library modules: lib.py, utils.py
• Data folder: raw CSVs (climate_ts.csv, los_angeles_…, ws*.csv)
• catboost_info/: ML model training logs, checkpoints
• Config & docs: environment.yml, requirements.txt, .readthedocs.yaml, docs/ & source/ (Sphinx)
• Test artifacts: test-utils.ipynb
README for additional insights
• External services:
• Climate Data Store API (cdsapi)
• Google Earth Engine (earthengine API)
• Installation flows: conda env + CLI run (kitchen.py)
• Demo usage: DataFrame → ClimateFiller class → missing_data → outlier elimination
Key recommendations
• Flow from raw input → processing → output → model storage.