Backgound
Air Pollution in Los Angeles is a fact of life. Enormous improvements have been made since the 70s. At the same time, the long term health hazards effects of air pollution are now known to be much higher. By measuring air quality we are able to better understand the underlying processes that cause air pollution and take effective measures to reduce it. By forecasting air pollution on a daily basis, individuals can reduce the ill-effects by planning their lives. They can remove themselves from areas on days when particularly high pollution occurs, they can stay indoors or plan their exercise schedules that avoid the hours when high pollution occurs.
Pollutants
The major pollutants in Los Angeles are Ozone, Nitrous-Oxides, Carbon Monoxide and Particulates.
Forecasting
At this time we only forecast Ozone pollution. Ozone pollution is also known as photochemical smog. The factors that increase ozone are sunlight and a complex reaction between nitrogen-dioxide, nitrogen oxide and VOCs. Ozone is dispersed by wind. Our forecasting model uses a heuristic algorithm that "learns" how past local Ozone, NO, NO2 and current wind direction, speed, humidity,temperature, time of day, day of week ( commuter traffic patterns) affects current Ozone levels. Having learnt this "pattern", we apply the pattern to the future. Given today's ozone, NO, NO2 levels and the forecast windspeeds, directions, temperatures and humidity we forecast tomorrow's Ozone conditions.
Teaching Data
The California Air Resources Board http://www.arb.ca.gov provides historic (1980 and onwards) hourly data on pollutant levels are various measuring stations around the Los Angeles area. This provided us with past pollutant levels to learn from.
The California Irrigation Management Information System (CIMIS) http://wwwcimis.water.ca.gov/cimis provides past hourly data on wind speeds, direction, temperatures and dewpoints at various measuring stations dotted around the Los Angeles Area. This provided us with the meteorological data to learn from
Forecasting Data
The Califonia ARB again provides current pollutant level data. The NOAA National Weather Service http://www.nws.noaa.gov/ provides wind, humidty and temperature forecasts for the next few days. We use their data of predicted future weather conditions, combine it with the current pollutant level data and come up with a pollution forecast.
Accuracy
We backtested our forecasting model. Ozone pollution is classified in 6 categories from good to extremely hazardous. We achieved a 89% accuracy rates for the first 4 categories, a 80% accuracy with the 5 the category and only a 33% accuracy with the extremely hazardous category. This last low rate can be attributed to the fact that extremely hazardous levels are very very rare and there were insufficient past cases to learn from.
Forecasting Schedule
TBD