hks-epod / ddc-demo

Delhi pollution. Predictive model + demo visualization.

ddc-demo: Predicting, visualizing Delhi air pollution

Summary

4 November 2015. This project will have two goals: to use air pollution data to (1) predict future pollution, and (2) visualize those predictions.

Resources

• AirNow (US Gov)
• SAFAR - India
• An example of linked vizi
• Mike Bostock: Using tooltips to find x-values on a line chart
• Mike Bostock: Multi-series line chart
• Air quality prediction hackathon winning model

TODO

1. Get git set up.
2. Lit review of existing time-series predictive models for environmental factors.
3. Line chart of past pollution data.
   • aqi_data_old.dta to two .csvs: full data, only PM2.5 (aqi_pm25.csv).
   • Template: IndiaSpend.
   • Tooltips?
   • Style tooltips.
   • Why the weird line restart?
   • chart g?
   • Generate some example predictions.
     • Harder than I thought...
   • Restrict time range.
   • Check margins and how they're working for the axes and the line charts?
   • Move chart down a bit/some styling.
   • Axes.
   • Merge: Average of all stations, or just most reliable? Or most conservative estimate?
   • Color the stations.
   • Add a legend.
   • Fix NaN showing up (e.g. in Shadipur station data) - they pop up in the top-left corner of viz.
   • Semi-transparent white rectangle for title/subtitle/line-labels?
4. Heat map of pollution around each of the 11 monitors. Which monitors are broken? Which have constant data?

• TopoJSON?
• Get a map of Delhi up, dots for stations.
• How to link the two maps?

5. Get station geodata.