Evaluation of the surface PM_2.5 in Version 1 of the NASA MERRA Aerosol Reanalysis over the United States

Highlights

• Full evaluation of MERRAero PM_2.5 diagnostics: PM_2.5, AOD and vertical structure.

• Impact of MODIS AOD assimilation on the simulation of surface PM_2.5.

• Quality control of PM_2.5 in situ-measurements to minimize error of representativeness.

• Part of the bias between MERRAero and PM_2.5 observations are species dependent.

Abstract

We use surface fine particulate matter (PM_2.5) measurements collected by the United States Environmental Protection Agency (US EPA) and the Interagency Monitoring of Protected Visual Environments (IMPROVE) networks as independent validation for Version 1 of the Modern Era Retrospective analysis for Research and Applications Aerosol Reanalysis (MERRAero) developed by the Global Modeling Assimilation Office (GMAO). MERRAero is based on a version of the GEOS-5 model that is radiatively coupled to the Goddard Chemistry, Aerosol, Radiation, and Transport (GOCART) aerosol module and includes assimilation of bias corrected Aerosol Optical Depth (AOD) from Moderate Resolution Imaging Spectroradiometer (MODIS) sensors on both Terra and Aqua satellites. By combining the spatial and temporal coverage of GEOS-5 with observational constraints on AOD, MERRAero has the potential to provide improved estimates of PM_2.5 compared to the model alone and with greater coverage than available observations.

Importantly, assimilation of AOD data constrains the total column aerosol mass in MERRAero subject to assumptions about optical properties for each of the species represented in GOGART. However, single visible wavelength AOD data does not contain sufficient information content to correct errors in either aerosol vertical placement or composition, critical elements for a proper characterization of surface PM_2.5. Despite this, we find that the data-assimilation equipped version of GEOS-5 better represents observed PM_2.5 between 2003 and 2012 compared to the same version of the model without AOD assimilation. Compared to measurements from the EPA-AQS network, MERRAero shows better PM_2.5 agreement with the IMPROVE network measurements, which are composed essentially of rural stations. Regardless the data network, MERRAero PM_2.5 are closer to observation values during the summer while larger discrepancies are observed during the winter. Comparing MERRAero to PM_2.5 data collected by the Chemical Speciation Network (CSN) offers greater insight on the species MERRAero predicts well and those for which there are biases relative to the EPA observations. Analysis of this speciated data indicates that the lack of nitrate emissions in MERRAero and an underestimation of carbonaceous emissions in the Western US explains much of the reanalysis bias during the winter. To further understand discrepancies between the reanalysis and observations, we use complimentary data to assess two important aspects of MERRAero that are of relevance to the diagnosis of PM_2.5, in particular AOD and vertical structure.

Introduction

PM_2.5 – fine aerosol particles with diameters ≤2.5 μm near the surface – have received considerable attention because of their negative effects on regional air quality and on human health (e.g., Pope et al., 2009, Pope and Dockery, 2013). Air quality monitoring networks currently exist over the globe, but they generally offer sparse geographic coverage. Thus, many studies have emerged trying to estimate surface PM_2.5 from satellite measurements, especially using the Aerosol Optical Depth (AOD) (e.g., Wang and Christopher, 2003, Engel-Cox et al., 2004), leading to discussions on the ability of retrieving air quality from space.

Satellites have the advantage of providing global coverage and long-term observations. However, important considerations must be made using satellite observations regarding, for example, cloud contamination, uncertainties in AOD retrievals, and sensor-specific data gaps (e.g., lack of AOD retrievals from the Moderate resolution Imaging Spectroradiometer (MODIS) sensor over bright surfaces for the dark-target aerosol retrieval algorithm). Additionally, AOD is a column-integrated observation and PM_2.5 concentration is a surface measurement, and estimates that use AOD as the only predictor of PM_2.5 can have unusually large uncertainties as a result. Several studies have shown the importance of considering the influence of other parameters on PM_2.5 space-based retrievals, such as the relative humidity or the altitude of the aerosol layer (e.g., Liu et al., 2005, Engel-Cox et al., 2006, van Donkelaar et al., 2006, Schaap et al., 2009, Gupta and Christopher, 2009b, Gupta and Christopher, 2009a, Crumeyrolle et al., 2014, Toth et al., 2014). For example, van Donkelaar et al., 2010, van Donkelaar et al., 2015 use a combination of AOD retrieved from satellites and the GEOS-Chem chemical transport model to represent the vertical placement of aerosol and aerosol optical properties to produce PM_2.5 estimates on a global scale.

Data assimilation tools combine the high temporal and spatial extent of a global model, constrained by available observations (e.g. AOD in the Modern Era Retrospective analysis for Research and Applications aerosol reanalysis (MERRAero)). Therefore, they can potentially provide a better characterization of PM_2.5 than either a model or observational network alone. Several studies (Li et al., 2013, Chen et al., 2014, Schwartz et al., 2014) have shown that assimilation of MODIS AOD and/or surface PM_2.5 observations within the WRF-Chem model improves simulated surface PM_2.5 relative to observations over the US. Saide et al. (2014) showed that the assimilation of AOD from planned geostationary satellites is expected to improve air quality forecasts when included in a current model that already assimilates MODIS AOD. Compared to observed relationships between AOD and PM_2.5, which cannot account for the influence of elevated aerosol layers on the column AOD, a data assimilation-equipped model includes simulation of the aerosol vertical profile, including transport of aerosols from remote and or elevated surfaces.

In this study, we assess the quality of the simulated surface PM_2.5 in MERRAero (see below) using ground-based measurements from the US Environmental Protection Agency (EPA) and Interagency Monitoring of PROtected Visual Environments (IMPROVE) networks over the continental United States. While MERRAero includes assimilation of bias-corrected AOD from the MODIS instrument on both Terra and Aqua satellites, it does not assimilate any data capable of directly constraining its aerosol vertical placement or composition, so it remains important to evaluate these aspects of the model. The accuracy of the aerosol reanalysis PM_2.5 diagnostic depends on the quality of AOD observed, the quality of the background forecast (speciation, size and vertical structure), the aerosol optical properties assumed to convert modeled aerosol mass to AOD, and the parameterization of error covariances in the aerosol data assimilation algorithm. Therefore, we use independent (non-assimilated) measurements from the AERONET network available over the United States to evaluate MERRAero AOD and airborne High Spectral Resolution Lidar (HSRL) measurements during air quality field campaigns over part of the US and CALIOP data to assess the quality of MERRAero aerosol vertical structure.

In Section 2, we summarize the GEOS-5 aerosol modeling and data assimilation system. Section 3 briefly describes all data products used in this study. In Section 4, MERRAero surface PM_2.5 is evaluated over the continental US using the data collected by the EPA and IMPROVE networks. In Section 5, we assess the quality of MERRAero AOD by doing comparisons to AERONET retrievals over the US, followed by the evaluation of the vertical distribution with lidar observations from CALIOP and from airborne HSRL instruments during the NASA DISCOVER-AQ (Deriving Information on Surface Conditions from COlumn and VERtically Resolved Observations Relevant to Air Quality) and SEAC⁴RS (Studies of Emissions and Atmospheric field Composition, Clouds and Climate Coupling by Regional Surveys) field campaigns. Concluding remarks appear in Section 6.