Presented at the 7 th Annual CMAS Conference

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Identifying Optimal Temporal Scale for the Correlation of AOD and Ground Measurements of PM 2. 5 to Improve the Model Performance in a Real-time Air Quality Estimation System Hui Lia, Fazlay Faruquea , Worth Williamsa, Mohammand Al. Hamdanb, Jeffrey Luvallb a. University of Mississippi Medical Center, Jackson, Mississippi 39216 b. NASA Marshall Space Flight Center, Huntsville, Alabama 35812

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Introduction n n NASA founded project on developing an DSS for asthma surveillance, intervention, and prevention Real-Time PM 2. 5 Estimation System: 3 components Originally developed NASA Marshall Space Flight Center (MSFC) – – AOD-PM 2. 5 linear regression models A Surface Model to Interpolate AOD-derived and ground PM 2. 5 to continue surfaces – Approach to integrate the above two interpolated surfaces into a final output surface based on the weight (90% for ground surface via 10% for AOD -derived surface)

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Introduction: continue n MODIS AOD shows great promise in improving estimate of PM 2. 5 – Gupta et al. , 2006; Kumar et al. , 2007 n Challenging on using satellite data in a real-time pollution system – Affected by many factors – Vary widely in different regions and different seasons

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 AOD-PM 2. 5 Relationship in 2004 AOD-PM 2. 5 Relationship in 2005

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Introduction: continue n Two major aspects worth consideration in a real-time air quality system – Approach to integrate satellite data with ground data for the pollution estimation – Identification of an optimal temporal scale for calculating the correlations of AOD and ground data

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Goal n Goal: identify the optimal temporal scale on determining AOD-PM 2. 5 correlation coefficients to improve PM 2. 5 estimation using satellite AOD data Within the last 3 days 08/10/08 08/12/08 Calculated date

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Model domain and monitoring stations

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Methodology n Five different temporal scales on utilizing satellite data and evaluating their impact on the model performance – – – n Within the last 3 days Within the last 10 days Within the last 30 days Within the last 90 days Time period with the highest correlation in a year Statistics for performance evaluation – – – Mean Bias (MB) Normalized Mean Bias (NMB) Root Mean Square Error (RMSE) Normalized Mean Error (MNE) Index of Agreement (IOA)

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 AOD sample data n Within a radius of 0. 9 degree inside a station Pixel Point Station Range of a station AOD=(AOD 1+AOD 2+AOD 3)/3

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008

Distribution of R-Squared values across different temporal scales in 2004 and 2005

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Discussion n Impact of Data Integration on the Model Performance – Model performance show only slight difference among the five selected temporal scales for the correlation of AOD and ground data – The weight of satellite data should be dependent on their relationship with ground data n Optimal Temporal Scale for the Correlation of AOD and Ground data – The optimal temporal scale: within the latest 30 days suggests that it might be a good strategy to build AOD-PM 2. 5 regression models on a monthly basis – The conclusion might not be able to apply to other areas considering different atmosphere conditions n Areas to Improve – Incorporate others factors to determine the optimal temporal scale using satellite data

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Conclusion The best optimal temporal scale is not the last 3 or 10 days in the solution n The temporal scale of the latest 30 days displays the best model performance n This conclusion does not consider the confounding impact of weather conditions n

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Acknowledge n Funding Agency – NASA Stennis Space Flight Center

Presented at the 7 th Annual CMAS Conference, Chapel Hill, NC, October 6 -8, 2008 Questions or Comments?