Breathing easier: Local Air Quality Forecasting for Columbus, Ohio

The 2023 Canadian wildfires served as a powerful reminder of how deeply air quality can affect our lives, particularly those with existing health conditions. Witnessing the far-reaching consequences firsthand inspired me to develop a solution for my community.

I've created a dedicated LSTM model, specifically tailored to Columbus, Ohio. This advanced deep learning approach forecasts air quality up to three days in advance, giving residents the information they need to plan ahead. By focusing exclusively on Columbus, my model takes into account the unique factors that influence our local air quality, such as traffic patterns, industrial emissions, and meteorological conditions.

I chose an LSTM model for its ability to excel in time-series forecasting. Air quality data is inherently sequential, with current conditions influenced by past patterns. LSTMs, with their memory cells, can capture these long-term dependencies and learn the complex relationships within the data. This makes them particularly well-suited for predicting how air quality might evolve over time, providing more accurate and reliable forecasts for the Columbus area. My aim is to provide a reliable and easily accessible source of forecasting, empowering you to make informed decisions about your day and prioritize your health. Together, let's stay ahead of air quality concerns and ensure a healthier future for all.

  • The model is updated every hour. Not all data is parsed by hour, but the model runs on up-to-date data as it becomes available.

  • Data sources include:

    • Air Quality - EPA/AirNow.gov

    • Traffic - Tom Tom

      • Data harvesting ongoing

    • Wildfire - NASA/FIRMS

    • Energy production - EIA

    • Weather - Weatherstack

  • Currently the training and prediction is centered in downtown Columbus (zip code - 43215). Peripheral zip codes will be modeled at a future date.