ghe-bomad/ecotrace-pilot-data-kiambu

# ecoTrace biogas-sensor pilot, Kiambu County, Kenya Sensor telemetry, reference measurements and de-identified household survey from a household-biogas monitoring pilot on the ABPL digester fleet in Kiambu County, Kenya, together with the code that derives corrected flow and gas composition from the raw device output. **22 devices · 2,975 device-days · 5,499,810 rows · 2026-01-27 to 2026-07-19** This repository is the **dataset and its derivation code, released together under one licence**. Analysis and figure code for the papers lives in separate repositories (see Related work). ## Quick start ```bash pip install -r requirements.txt python3 src/derive.py # raw Parquet -> data/derived_data/, ~15 s python3 -m pytest src/test_sensor_correction.py ``` ## Layout ``` data/ raw_data/SN_*/YYYY-MM-DD.parquet archival device telemetry (207 MB) derived_data/ state-0 rows with corrected flow and composition; regenerated by src/derive.py, not committed metadata/ device inventory, state codes, calibration card biogas_composition.json reference gas-composition spot measurements survey.csv de-identified household survey (n = 20) docs/field_correction.md the correction method and its limits src/ derivation code and tests ``` Every directory has its own README describing its contents, units and limits. Start with `data/README.md`. ## The field correction: read this before using the data The deployed firmware crossed the two thermistor ADC channels. In the raw files **`powerC` holds the FLOW thermistor and `powerF` holds the COMP thermistor**, the opposite of what the names say. As a result the firmware's `flow` column is magnitude-wrong and its `comp` column is inverted. No over-the-air update was possible, so the fix is applied in post-processing. `src/derive.py` keeps only rows in device state 0 (see below), undoes the swap and recom …