Methodology

Current data sources

• Flood hazard: an adjustable, indicator-based model. Six environmental indicators are each reclassified into five hazard zones and combined into one flood hazard index: rainfall, distance from major rivers, slope, elevation, land use and land cover, and soil type. Sources are Copernicus DEM (elevation and slope), OpenStreetMap river networks, Open-Meteo reanalysis rainfall, ESA WorldCover land cover, and ISRIC SoilGrids soil data.

  The weights are not fixed. The map opens with published default weights (rainfall 25%, rivers 20%, slope 20%, elevation 15%, land cover 10%, soil 10%), and you can change any of them in the Flood panel. Every ward is recomputed live as you do, and “Reset to published weights” returns to the defaults. A ward's level is read from the model at the ward's geographic centre; “Model detail” shows the full underlying surface instead.

  Adjust the weights on the map

• Earthquake hazard (paused on the map while we refine it): modeled using a weighted overlay of fault line proximity (Global Earthquake Model active faults database), historical earthquake density (epicenter data from the Humanitarian Data Exchange), geology and lithology (ICIMOD), and a soil amplification layer (Vs30) derived from slope using the Wald and Allen (2007) proxy method. Combined into five hazard zones from very safe to very high hazard.

• Air quality: live station data from the World Air Quality Index (WAQI) network across Kathmandu Valley, refreshed every 30 minutes. A ward is coloured using its nearest active station when that station is within 4 km, wards farther away show no data rather than an estimated value. Classified into six AQI categories using US EPA breakpoints.

• Ward boundaries: official administrative boundaries covering Kathmandu, Lalitpur, and Bhaktapur districts.

A note on confidence

These are modeled estimates, not direct measurements at every point. Hazard zones are ward-level approximations built from multiple weighted environmental layers, not precise point-level readings, and a ward's flood level depends on the weights in use. Air quality values reflect the nearest monitoring station within 4 km, so a ward's reading is only as precise as that station, and wards beyond 4 km of any station show no data rather than an estimate.

This is v1 of the atlas. As better local data becomes available, through municipal partnerships, new surveys, or research collaborations, individual wards and layers will be updated and the confidence behind them will improve.