AAAR 31st Annual Conference
October 8-12, 2012
Hyatt Regency Minneapolis
Minneapolis, Minnesota, USA
Abstract View
A Case Study in Fusion of Surface PM2.5 Observations and 3D Air Quality Model Output
SINAN SOUSAN, Tiangfeng Chai, Jaemeen Baek, Scott Spak, Naresh Kumar, Jacob Oleson, Sarika Kulkarni, Gregory Carmichael, Charles Stanier, University of Iowa
Abstract Number: 425 Working Group: Aerosol Exposure
Abstract PM2.5 exposure estimates for United States at high spatial and temporal resolution are highly desirable for climate, visibility and health applications. In particular, accurate, high resolution, and spatially and temporally continuous datasets may enable advanced epidemiological studies focused on identifying more and less harmful types of particulate matter. Data assimilation via optimal interpolation (OI) was used to assimilate satellite AOD and surface PM2.5 measurements into Models-3 Community Multiscale Air Quality Model (CMAQ) model output. Data assimilation using MODIS satellite-based aerosol optical depth was conducted over the United States for 2002. Evaluation was conducted separately for six geographic regions of the U.S. The best combinations of error settings and averaging schemes led to a domain average improvement in fractional error from 1.2 to 0.97 at IMPROVE monitoring sites, and from 0.99 to 0.89 at STN monitoring sites. Somewhat larger improvements to fractional bias were observed. However, for 38% of the month-region combinations, MODIS OI degraded the forward model skill. Root causes of the limitations of the technique will be presented.
An alternate approach is to utilize surface measurements over the United States. The use of surface PM2.5 data will eliminates uncertainty associated from both satellite retrievals and calculations that enable the model and satellite data to be intercompared. But the point-nature of the surface observation versus the grid cell averaging of the 3D model (i.e. representation error) must be dealt with. Results from the Hollingworth-Lonnberg observational method (for determing model error covariance), and approaches for handling representation error will be presented together with results from OI of over 1500 monitoring locations. Data withholding schemes for evaluation of the OI output will also be presented.