420 likes | 704 Views
A Comparison of AirNow and AQS Particulate Matter Databases. Katina Gracien Brian Hare Graduate Assistants : Atina Brooks, John White & Andrew Moore Faculty Advisor : William F. Hunt Jr. & Dr. Kimberly Weems Clients: USEPA Chet Wayland David Mintz Tim Hanley Lewis Weinstock
E N D
A Comparison of AirNow and AQS Particulate Matter Databases Katina Gracien Brian Hare Graduate Assistants : Atina Brooks, John White & Andrew Moore Faculty Advisor : William F. Hunt Jr. & Dr. Kimberly Weems Clients: USEPA Chet Wayland David Mintz Tim Hanley Lewis Weinstock Scott Jackson
Background • Methods of Monitoring Fine Particulate Matter • FRM (federal reference method) • A single 24 hour measurement (midnight to midnight) • Continuous • TEOM ( Tapered Element Oscillating Microbalance) • Hourly measurements • RAMS • Speciated • Used Air Quality Index Reporting • Health Advisory • Reported in USA Today
Purpose • Compare AQS FRM with AirNow concentration levels on a national level • How well does the AirNow prediction agree with the FRM measurement? • How well do methods in different regions and/or states compare during the summer and winter seasons? • Which regions and/or states need improvement in the way they predict fine PM? • Is an Early Warning System needed to improve the accuracy of AiRNow reporting?
Analysis of the Federal Reference Method: Accuracy of AirNow Predictions
Map of Percentage Correct using AQI District of Columbia
Map of Percentage correct using AQI Percent Correct by Seasons (AirNow vs FRM)
Differences in Concentration Levels Average difference of AirNow – FRM by County (ug/m3)
Surrounding Lake Michigan Counties Average difference of AirNow – FRM
Chronic Under-reporting of AirNow • District of Columbia • Wrong Level =72/342 = 24.8 % • AirNow < FRM = 331/342 = 96.7% • Texas • Wrong Level 153/1241 = 15.16% • AirNow < FRM 1156/1241 = 93.2%
Chronic Over-reporting of AirNow • Michigan • Wrong Level 364/1913 = 19.02% • AirNow > FRM 1496/1913 = 78.2% • Indiana • Wrong Level 211/962 = 21.93% • AirNow > FRM 699/962 = 72.7%
AQS FRM vs. AirNow with AQI code Black = difference in AQI level
AirNow Levels vs. FRM by AQI Category 5 occurrences out of 8,666 obs. out of 23,262 obs. 7 occurrences 1 occurrence out of 286 obs. out of 19 obs.
Recommendations • Redo the Fine PM episode in the Midwest occurring in February 2005. • Explore the possibility of developing a QC procedure to pair “problem sites” with sites that consistently report data accurately. • 3. Provide an early warning system that data has not been properly adjusted at problem site.
Early Warning System Approach • Examine the correlation matrix for all PM fine sites using FRM(s): • Find sites in States with high reporting accuracy (using continuous PM fine instruments) that can be paired to sites in States with low reporting accuracy (Indiana, Michigan, etc.). • Develop regression equations predicting FRM measurements at a site in a county with poor reporting accuracy as a function of a PM fine measurements at an adjacent county with “good” reporting accuracy.
Average Differences by CountyWinter Ingham Washtenaw Wayne
Correlation Matrix for Selected FRM Sites 0.92821 0.92821 frm1=Washtenaw, frm2=Ingham, frm3=Wayne
Winter Wake Wayne
Average Differences by CountyWinter Ingham Washtenaw Wayne
STEP 1 Obtain the AirNow values for both of the sites, “good” and “bad”.
Step 2 Validation Predicted FRM Washtenaw, from the regression equation in previous slide Actual observed FRM Washtenaw
Step 3 {FRM Wayne= F(FRM Washtenaw)} Make sure this is predicted FRM Wayne value not the observed FRM Wayne value
Step 3 Validation Predicted FRM Wayne from the regression equation in previous slide Actual observed FRM Wayne
Next Steps & Recommendations • Continue to work on QC Procedure. • Further evaluate the effectiveness of this approach. • Identify other pairs of “good” & “bad” sites to apply to this QC procedure to.
Locations of the 5 Occurrences of when AirNow Reported Unhealthy for Sensitive and FRM forecasted Good Air Quality
Locations of the 7 Occurrences of when AirNow Reported Good and FRM forecasted Unhealthy for Sensitive Air Quality
Location of the 1 Occurrence of when AirNow Reported Good and FRM forecasted Unhealthy Air Quality
Map of Regression CoefficientsModel : AQS(frm-primary) = β*AirNow
Regression Coefficients by EPA Region with Season Model : AQS(frm-primary) = β*AirNow
Air Quality Monitoring Sites Accuracy AirNow Correctly Predicts AQI