This NASA-backed initiative aims to improve the accuracy and availability of air quality data by combining Earth observations, advanced modeling, and expanded ground-based sensor networks. A major component is the GEOS-CF-based localized forecasting tool, which empowers localized air quality prediction capabilities.
A lightweight forecasting tool leveraging NASA GMAO’s GEOS-CF model integrated with local observation data including Pandora.
- Enable high-resolution, site-specific air quality forecasts.
- Make GEOS-CF modeling outputs actionable and localized using real-time observations.
- Provide an accessible online platform to train, deploy, and evaluate forecast models per location.
- Pretrained models: Use existing models without retraining for immediate forecasting.
- Online training: Train and deploy models for specific locations in real time.
- Configuration exploration: View and modify location-specific forecast settings.
- Forecast visualization: Generate and export forecast plots using built-in tools.
- Community contributions: Support for saving models and sharing local data.
- Custom forecast generation: One-call function to generate outputs in various formats:
- 📊 DataFrames
- 📈 Forecast plots
- 🧩 SHAP (SHapley Additive exPlanations) values for uncertainty and feature impact analysis
- 🌐 Pandora data: Available via the Pandonia Global Network (PGN).
- 📱 PM2.5 Forecasts: Accessible through this website.
- 🔧 GEOS-CF Forecast Tool: Use the online tool for real-time model training and forecasting.
This project is supported by NASA and leverages contributions from the scientific community, developers, and atmospheric monitoring networks worldwide.
