Motivation

Air pollution is a mix of particles and gases that can reach harmful concentrations both outside and indoors. Its effects can range from higher disease risks to rising temperatures. Soot, smoke, mold, pollen, methane, and carbon dioxide are a just few examples of common pollutants.

Challenge

Accurate early warnings of poor air quality are useful because they give people the option to reduce their risk of exposure to poor air by limiting outdoor activity at these times. Our challenge is to integrate NASA data, ground-based air quality data, and citizen science data to create an air quality surface that displays the most accurate data for a location and time.

Method

The main idea is to use multiple satellite data sources from NASA and Copernicus missions as well as data from local observations to create machine learning algorithm, which will predict Air Quality Index based on the input data.

Result

As a result of our method we got raster map for whole world with Air Quality index. The data is presented as Web-page and REST API.

Data used

• OMI/Aura Near UV Aerosol Optical Depth and Single Scattering Albedo 1-orbit L2
• MODIS/Terra Aerosol 5Min L2 Swath 10km
• OMI/Aura Near UV Aerosol Optical Depth and Single Scattering Albedo 1-orbit L2
  
• MLS/Aura Near-Real-Time L2 Nitric Acid (HNO3) Mixing Ratio V004
  
• MLS/Aura Near-Real-Time L2 Ozone (O3) Mixing Ratio V004
  
• AIRS/Aqua L1B Near Real Time (NRT) Infrared (IR) geolocated and calibrated radiances V005
• Data from Sentinel 5 mission

Technology used

• Python
• Numpy
• Scipy
• Requests
• Tensorflow
• Node.js
• Mongodb
• Mapbox

Github repository