About This Data

Data Source

Readings are collected hourly from the OpenWeatherMap Air Pollution API, which aggregates ground-station and satellite-derived measurements.

Pakistani government agencies (PMD, EPA Punjab) do not currently provide a public real-time API for air quality data, so this project uses the best available third-party aggregated source instead.

The AQI Scale

This dashboard uses OpenWeatherMap's 1-5 Air Quality Index:

• 1 - Good
• 2 - Fair
• 3 - Moderate
• 4 - Poor
• 5 - Very Poor

The Forecast Model

PM2.5 levels 6 hours ahead are predicted using a gradient-boosted model trained on historical readings, time-of-day patterns, and recent trends.

Known limitation: the model was initially trained on summer 2026 data, before Lahore's smog season (Oct-Feb) began. It has not yet seen the extreme pollution events typical of winter months. The live data pipeline keeps collecting through smog season, and the model is periodically retrained as more representative data accumulates.

Forecast accuracy is evaluated against a naive baseline (assuming no change from the current reading) - not just raw accuracy - since a forecasting model is only useful if it beats simply assuming things stay the same.