Data Sources and APIs
(All made to be flexible and fetch user entered city)

NASA MERRA-2 Fetcher – retrieves aerosol, gas, and meteorological layers including SO₂, NO₂, CO, O₃, PM₂.₅, humidity, wind, temperature, and pressure.

OpenAQ API – integrated for real-time atmospheric pollutant sampling.

Google Air Quality API – provides localized AQI and pollutant concentrations.

Pollen API – adds seasonal and allergen data with forecasting to broaden health impact analysis.

OpenAI API – interprets and summarizes technical data into a readable context, with "unit conversion" into "spoons" of pollutants, as well as action steps for individuals + repost to local clean air initiatives.

Modeling and Prediction

(Using user entered city, distracting you with current data from OpenAQ, the prediction begins in the background)

1. Multivariate forecasting system with Python

2. Implements ridge regression for pollutant concentration forecasting and KNN classification for AQI category prediction.

3. Generates time series features including lag values, rolling statistics, and seasonal harmonics

4. Produces forecasts for pollutants and pollen and their AQI equivalents. (This adjusts depending on available data per city)

5. Combines satellite data (NASA) with ground-based measurements (Google, OpenAQ) to mitigate the effects of disruptions, like the current shutdown or just missing values.

Includes fallback persistence model for missing or incomplete datasets.

Application and Visualization

1. Built an interactive dashboard using Streamlit!

2. Displays pollutant-specific trends, AQI forecasts, and saturation-based analogies with my "spoon analogy" for easy understanding

3. Merges datasets dynamically for consistent updates/forecasts and visualization.

4. In advanced options, Outputs CSV summaries for external use and data sharing.

Technical Stack

Languages: Python (core), JavaScript (visualization hooks).

Libraries: Pandas, NumPy, Requests, Xarray, Matplotlib, Streamlit.

APIs and Sources: NASA Earthdata, Google Air Quality, EPA EDG, OpenAI.

Data Formats: CSV, JSON, NetCDF4.

Environment: Conda

Kaleidoscope: Turning air data into color, clarity, and care