Menu Close


Task 138

Comparison of Air Quality Models with Satellite Observations for Improved Model Predictive Capabilities

Principal Investigator(s):

C. Flynn


K. Pickering

Last Updated:

October 26, 2012 15:25:38

Description of Problem

Satellite observations provide several important benefits to air quality, including improved forecasting ability for air quality models, assessment of air quality for attribution to specific sources, and improved estimation of source emissions. However, many challenging problems remain for the use of satellite observations in diagnosing near-surface air pollution. The column-integrated quantities retrieved from satellite instruments for key trace gases and aerosols must be interpreted correctly to derive information about near-surface conditions. Despite these challenges, a major scientific goal remains the use of satellite observations to improve and validate current air quality models for more accurate predictive capability. The DISCOVER-AQ field project provides surface, in-situ aircraft, and remote sensing data that will aid in the interpretation of satellite data for air quality. This project was conducted in support of this overall goal, by comparing satellite observations, aircraft measurements, and surface air quality datasets with air quality model output. Such a comparison may lead to better understanding of the factors affecting the correlation of satellite observations with current models.

Scientific Objectives and Approach

The first deployment of the Earth Venture-1 DISCOVER-AQ project was conducted during July 2011 in the Baltimore-Washington region. The DISCOVER-AQ datasets, model output from the Community Multiscale Air Quality (CMAQ) model, and data from the Ozone Monitoring Instrument (OMI) onboard NASA’s Aura satellite were used in this analysis. The hourly time series of column O3 and NO2 from the aircraft data, ground-based Pandora UV/Vis spectrometer data, CMAQ output, and OMI dataset were analyzed for each of six Maryland Department of the Environment (MDE) air quality monitoring sites involved in the deployment. Two different column amounts for O3 and NO2 were computed from the aircraft data. Column_air was computed through integration of the trace gas profile after extension of the lowest aircraft mixing ratio value to the surface, while column_ground was computed in the same manner but after extension of the surface mixing ratio value to the profile, if available. Model column amounts for O3 and NO2 were computed through integration of the model profile from the model surface through the depth of the aircraft profiles. The column time series allow comparison of the column amounts obtained from the model, satellite, and in situ data. A correlation analysis was also performed between column amounts and surface mixing ratio data for O3 and NO2 for each surface site for the aircraft data, Pandora data, and CMAQ. The linear Pearson correlation coefficient (R) was computed for each site as a measure of the degree of fit of a linear relationship. Gao Chen (NASA/LaRC), Lok Lamsal (NASA/GSFC), and Jay Herman (NASA/GSFC) must be acknowledged for providing the aircraft, OMI, and Pandora column data, respectively.


The hourly column time series presented the column amounts of O3 and NO2 from multiple perspectives. The time series revealed the variation in the CMAQ model bias compared to the aircraft, Pandora, and OMI columns. During the July 4th weekend getaway days, CMAQ demonstrated a low bias relative to the observed columns for O3 and NO2, indicating that NOx emissions may be too low in the model for these days. Later in July, CMAQ demonstrated reasonable agreement with the observations during the late morning, but transitioned to a high bias by afternoon and evening. Mean ratios for July of the CMAQ column to the columns derived from the aircraft, Pandora, and OMI data demonstrated that for NO2, CMAQ has a general high bias for more urban sites, such as Beltsville, near Washington, D.C., and a low bias for more rural sites, such as Fair Hill, in northeastern Maryland. Example time series are displayed in the top panel of Figure 1.

Large correlation was obtained for both the aircraft O3 column_air and column_ground. Moderate correlation was obtained for the aircraft NO2 column_ground, while low correlation was obtained for NO2 column_air. These initial results indicate that with sufficient sensitivity to the lower troposphere, satellite observations may be meaningful for surface air quality analysis. Example scatter plots for the aircraft data are displayed in the middle panel of Figure 1. Large correlation was obtained for O3 in the CMAQ model, while moderate to large correlation was obtained for NO2 in CMAQ. Example plots for CMAQ are displayed in the bottom panel. For both O3 and NO2, CMAQ demonstrated larger correlation between the column and the surface than was apparent in the aircraft data, indicating that the CMAQ surface is more connected to the overlying column than observations support. This indicates that the model underestimates the variability observed in the lower troposphere, and that the model boundary layer is too well-mixed.

Other Publications and Conferences

Conference Publications:
Clare M. Flynn, Kenneth E. Pickering, Pius Lee, Youhua Tang, Andrew Weinheimer, Ronald Cohen, and James Szykman, “Correlation analysis of column-integrated P-3B data with surface mixing ratio for in situ observations and model for O3 and NO2”, 3rd International Workshop on Air Quality Forecasting Research, Potomac, MD, November 29, 2011.

Posted in Archive

Related Posts