This project analyzes weather quality data for Salt Lake City, Utah. Data was gathered every day between October and May each year from 1996-2001. Sixty-five different variables were collected each day. The critical variable for this analysis is PM2.5 which is the actual measure of air quality. The interesting data points are days considered "unhealthy", given by PM2.5 values of 64.5 or greater.

Many of the weather observations are correlated. For example, pressure and humidity relate to precipitation. Temperature relates to whether or not there is snow on the ground. Instead of showing all of these as predictors of PM2.5, only one interrelated value is generally shown. Six graphs below show five different correlations. Click each graph to see a larger version with an explaination.

PM2.5 as predicted by difference in surface temperature and 700mb temperature	PM2.5 predicted by date	PM2.5 predicted by yesterday's PM2.5 reading
PM2.5 predicted by daily maximum temperature	PM2.5 predicted by humidity	3D depiction of PM2.5 vs. maximum daily humidy and temperature

Spotfire was an excellent tool for analyzing this data. Because of the form in which the data was originally organized, I did not have any issues with formatting or compiling datasets. Once opened in SpotFire, I was able to do all of the analysis that I wanted, and some that I hadn't thought of.

I was particularly impressed with how many dimensions of data could be shown in one graph. While I did not use all of them in this analysis, supporting up to 5 dimensions (3D on graph, plus size and color coding) presents a great opportunity for analysis of even more complex patterns.

Jennifer Golbeck, March 2001