NCEP’s operational Gridpoint Statistical Interpolation (GSI) three-dimensional variational (3DVAR) system has been extended to analyze aerosol species from WRF/Chem GOCART module. The current implementation permits a synergic assimilation of surface PM2.5 (particulate matter with diameters less than 2.5 microns) observations from the AIRNow network and MODIS aerosol optical depth (AOD) retrieval products in a unified 3DVAR data assimilation framework. In this new aerosol data assimilation framework, the analysis variables are 15 aerosol species from the GOCART module (p25, hydrophobic and hydrophilic, organic and black carbon; sulfate; sea salt in four particle-size bins; and dust in five particle-size bins). The system analyzes 3D mass concentrations of these 15 variables in a one-step 3DVAR minimization procedure without assumptions regarding how total aerosol mass is partitioned among the species, unlike most previous studies that required these assumptions. Observation operators and their Jacobian for MODIS AOD and surface PM2.5 were integrated into the GSI system.
This newly developed aerosol data assimilation system was firstly applied to a dust storm case occurred in March 2010 over East Asia region. This dust storm is under-predicted by the WRF-Chem model initialized from NCEP global meteorological field. The statistics of the amplitude and length-scale of the aerosol background error is obtained using the “NMC” method. Assimilating MODIS AOD observations substantially improves the aerosol analyses and forecasts when comparing to independent AOD observations from AERONET and CALIPSO as well as surface PM10 data.
Recently the system was also applied over the continental United States with 20-km horizontal grid spacing. Three data assimilation experiments were carried out for the CalNex 2010 period (02 June ~ 14 July 2010). Two of three experiments assimilate only PM2.5 or MODIS AOD and the third one assimilates both PM2.5 and AOD. 3DVAR analysis of aerosols was produced every 6-hrs, and 48-hr WRF/Chem model forecasts were initialized from the analyses. Generally, assimilation of both AOD and PM2.5 led to improved aerosol analyses and forecasts when comparing to independent ground-based and airborne aerosol observations. These results and the application of this approach to air quality and aerosol forecasting will be discussed.