MP Climate Intelligence — Village-Level Climate Hazard Dashboard
Village and district-level heatwave, drought, and extreme precipitation
analytics for five Madhya Pradesh districts (Bhopal, Indore, Jabalpur, Rewa,
Sidhi), computed from IMD 0.05° gridded daily data (2000–2024), with UNDP
DiCRA NDVI and a CMIP6 (NEX-GDDP, SSP2-4.5) workflow for 2040 projections.
Data policy: every displayed value is traceable to a verified source.
Synthetic district, village, and cadastral records were removed in the
2026-08 cleanup. See CONTRIBUTING.md and docs/DATA_SOURCES.md.
Repository layout
cNARMADA_IITIndore/
├── app.py Streamlit host (iframe + GitHub raw data)
├── requirements.txt full pinned dependency set
├── .env.example data paths / keys (copy to .env)
├── scripts/ reproducible pipeline
│ ├── config.py env-driven paths, thresholds, base periods
│ ├── common.py NetCDF loader + samplers
│ ├── 01_extract_village_timeseries.py IMD daily series per village
│ ├── 02_compute_indices.py heatwave / SPI / ETCCDI per village-year
│ ├── 02_finish.py district rollup helper
│ ├── 03_build_chart_data.py chart JSON
│ ├── 04_build_dashboard_json.py dashboard payload
│ ├── 05_gee_cmip6_2040.js Earth Engine CMIP6 export (paste into GEE)
│ ├── 06_convert_gee_export.py GEE CSV → dashboard JSON
│ ├── 07_build_dicra_forecast.py DiCRA NDVI + OLS trend forecast
│ └── run_all.bat Windows one-click runner
├── dashboard/
│ ├── index.html single-page dashboard (Leaflet + Chart.js)
│ ├── mp_climate_loader.js binds climate JSON
│ ├── dicra_ndvi_loader.js NDVI charts
│ ├── cadastral_loader.js stub — awaits MP Bhulekh official records
│ ├── india_boundaries_loader.js all-India state/district overlays
│ └── data/ verified JSON/GeoJSON payloads
├── tools/ one-off inspection/debug scripts
├── tests/ pytest unit tests for index math
├── docs/
│ ├── METHODOLOGY.md index definitions and caveats
│ ├── DATA_SOURCES.md provenance register for every layer
│ ├── REQUIREMENTS_ROADMAP.md 45-point requirements → implementation plan
│ └── DEPLOYMENT.md hosting instructions
└── .github/workflows/ci.yml syntax check + tests + synthetic-data guard
Quickstart
pip install -r requirements.txt
cp .env.example .env # point IMD_* paths at your NetCDF folders
cd scripts && python 01_extract_village_timeseries.py
python 02_compute_indices.py && python 03_build_chart_data.py
python 04_build_dashboard_json.py
cd ../dashboard && python -m http.server 8000 # open localhost:8000
Windows: scripts\run_all.bat. Tests: python -m pytest tests/.
What is computed
| Hazard | Indices | Definition |
|---|
| Heatwave | heatwave_days, severe_heatwave_days | IMD plains criteria: Tmax ≥ 40°C with departure ≥ 4.5°C (severe ≥ 6.5°C) or Tmax ≥ 45°C, runs ≥ 2 days, Mar–Jun |
| Drought | SPI-3/6/12, drought_months, drought_probability | McKee et al. (1993) gamma-fitted SPI |
| Extreme precipitation | R95p, R99p, Rx1day, Rx5day, CDD, CWD | ETCCDI; percentile thresholds from fixed 2000–2014 base period |
| Future | 2040 delta vs 2000–2014 baseline | 8-model CMIP6 NEX-GDDP ensemble, SSP2-4.5 (scripts 05–06) |
Known limitations (stated deliberately)
- Village values are nearest-pixel samples of a 5.5 km grid; most villages
are smaller than one pixel.
VILLAGE_SAMPLE_METHOD = "polygon" is the
planned upgrade for polygon zonal means.
- The 25-year record is below the WMO 30-year climate normal; SPI gamma
fits use ≥ 15 samples per calendar month and are noisier than standard.
forecast_2040.json served to the dashboard is an OLS trend
extrapolation labeled "indicative"; CMIP6 projections require running
the GEE workflow.- Cadastral (khasra) parcels are disabled until official MP Bhulekh /
Revenue Department records are integrated.
License
MIT (code). IMD, DiCRA, Census boundary data retain provider licenses.
