Pollution Data with Eureka Table Lens

CMSC 838F: Information Visualization

David T. Wang

Application Project: Ozone Level Weather Data Exploration Using Eureka Table Lens

March 6, 2002

Motivation and Background

In this study, we are using the Eureka Table Lens application in applying the "Exploratory Data Analysis" technique to weather data provided by a Meteorolgist. We hope to find some correlation between days with poor Air Qulity Index against various weather attributes. We used this data set in hopes of comparing and contrasting two different table based visualization programs in using the same dataset.

My Dataset

I have chosen to examine two sets of weather data provided by Dianne S. Miller, Meteorologist at Sonoma Technology, Inc. (SLC site Excel file) (MSP site Excel file). For a description of the what the fields represent see SLC variable description and MSP variable description.

Using Eureka Table Lens from Inxight, I have attempted to look for correlation between the Ozone levels and various weather attributes.

For SLC site, these Air Quality Index categories are used. The AQI is determined by the PM2.5 or PM25Max number.
Good -- 0-15.5 ug/m3
Moderate -- 15.6-40.5 ug/m3
USG -- 40.6-64.5 ug/m3
Unhealthy -- >64.6 ug/m3

For MSP site, Air quality is reported using the Air Quality Index, which is based on the concentration of ozone levels, reported in parts per billion, separated into five categories, Good, Moderate, Unhealthy for Sensitive Groups, Unhealthy, and Very Unhealthy.
Good -- 0-64 ppb
Moderate -- 65-84 ppb
USG -- 85-105 ppb
Unhealthy -- >105ppb

Interesting information about the data

SLC Site Data
- We first spotlight the top 20 days with poorest AQI Data is reported as PM25Max, or maximum observed 24-hr PM2.5 in ug per M^3. (Figure 3)
- Strong relationship between PM25Max and Maximum temperature. Most high PM25Max data correlate to days with very low Max temperatures. (Figure 4)
- Strong relationship between PM25Max on consecutive days. A high PM25Max day is likely correlated with a high yestpm25, which is the high PM25Mx for yesterday. (Figure 5)
MSP site Data
- We first spotlight the top 20 days with highest Ozone concentration. Data is reported in 03Max8Hr, or maximum observed in any 8 Hr period. (Figure 6)
- No relationship between 03Max8Hr and Wind Speed. Result is all over the place (Figure 7)
- Strong relationship between 03Max8Hr and Wind Direction. Many high Ozone period focused in narrow band of wind directions. (where's the ozone coming from?) (Figure 8)
Very strange and counter intuitive result. AQI has inverse relationship between SLC and MSP sites. At SLC, colder temps correlate to lower AQI, while lower MSP site, higher temps correlate to lower AQI.

Critique and suggestions for improvement

Eureka Table Lens is a good tool to play with if your data set is limited to 10 to 20 variables, and hundreds of entries. When the variable gets to be more than a few tens, or if the number of data points gets past a thousand, the visualization of the complete dataset becomes a problem. In these cases, we would still have to revert back to some sort of filtering. We resorted to simply creating focal areas for the first few columns, and moved the rest completely out of focus. Without a better understanding of the data, we do not know if we ignored something important.

On the other hand, we were pleasantly surprised to discover the correlation between ozone level and wind direction, indicating that there may be a source for the ozone from a certain direction.

The entire application has been setup to keep all of the data on a single display without scrolling. However, this paradigm may not be necessary, and perhaps scrolling should be allowed. Also, it may be nice to incorporate interpolation algorithms, and allow the program to generate and superimpose interpolated curves to aid the user in pattern detection.