The two principal quantities calculated by the OMI NO2 PGEs are the total vertical column density and the tropospheric column density of atmospheric nitrogen dioxide above each location viewed by OMI, at the time of overpass. At this time, the number of experiments that directly measure NO2—stratospheric, tropospheric, or total—from the ground or from aircraft is somewhat limited, but includes data collected during a few field campaigns, as well as data collected from a geographically distributed network of fixed measuring sites. A number of these measurements can be exploited to at least qualitatively, and in many cases semi-quantitatively, assess the quality of the OMIretrieved measurements. This document summarizes what has been learned thus far from ground-based measurements collocated with OMI measurements of NO2. This document is a supplement to the Readme File included as part of the first public release of the OMI NO2 data product (OMNO2). The Readme File may be found at http://toms.gsfc.nasa.gov/omi/no2/OMNO2_readme.pdf. Factors Affecting OMI NO2 Data Quality The OMI NO2 algorithm uses a spectral fitting technique (Differential Optical Absorption Spectroscopy, DOAS) to estimate the NO2 slant column density (SCD), which is the total NO2 density along the optical path (i.e., along the solar beam from the top of the atmosphere to the visible surface—cloud or ground—and then along the instrument’s line-of-sight, back to the top of the atmosphere). The rest of the algorithm is concerned with the computation of the air mass factor (AMF), which is the number used to convert from the SCD into the vertical column density (VCD), and with the separation of the NO2 column into stratospheric and tropospheric components. The determination of the SCD is affected by various types of noise in OMI’s detectors. Particular problems arise from the fact that the spectral fitting is applied to the ratio of the (earthshine) radiance and (solar) irradiance spectra, which are not measured simultaneously. In normal operations, the irradiance is measured once per day, and a particular solar measurement is then applied to the subsequent 14-15 orbits. It has been found that the patterns of “hot” pixels, and pixels affected by random telegraph signal (RTS) differ between the measurements of radiance and irradiance, resulting in significant noise in the ratios. This has led, most prominently, to the presence of “stripes” in most of the data products, where values retrieved at some of the 60 cross-track positions will be persistently higher or lower than the values at adjacent positions. Research groups working with different derived OMI products have taken different approaches to attempt to suppress these stripes. The NO2 algorithm uses a scheme that adjusts the means of the SCDs for each position, along a single orbital track, to lie on a smooth curve. This post-hoc correction has resulted in the suppression of most of the apparent striping in the NO2 (total and tropospheric), but at the expense of the possible introduction of an unknown bias. In addition, an anomaly occurred on 2006 February 28, which has taken the fold mirror, which switches between earthshine-view and solar-view modes, and since that time no new irradiance measurements have been made. Subsequent processing has proceeded using the last-measured irradiance spectrum, and some degradation has been seen in the retrieved NO2 data, including an increase in the apparent striping, even after the post-hoc correction. The calculation of the AMF rests on certain assumptions, for example, concerning the overall shape of the NO2 vertical profile. It also rest on some other data sets, such as the OMI-derived cloud fraction and cloud-top pressure (currently using the oxygen dimer algorithm product), and the surface albedo (currently using the GOME-derived albedo climatology of Koelemeijer). Finally, the calculation of the AMF is affected by a process that classifies an OMI FoV as having a significant tropospheric pollution component, or not. This classification is performed using an algorithm that constructs an “unpolluted field,” and compares the initial estimate of the VCD to the unpolluted field. To date, this algorithm has only been evaluated against a model-based data set. Though this algorithm is expected to be quite good, it has yet to be critically evaluated using real-world data. Groundand Aircraft-based Measurements of NO2 Ground-based measurements, particularly in regions of moderate-to-intense industrial activity, are subject to variation due to both spatial inhomogeneity of boundary layer NO2, which can show up as an azimuthal dependence of the measured NO2, and timeevolution of the NO2 concentration due to continual variations in NO2 sources and sinks, wind patterns, and insolation. Such variations undoubtedly account for some of the differences that are seen between the OMI-derived and ground-based measured NO2 amounts, and must be considered when making such comparisons. Most current ground-based measurements of NO2 rely on the absorption of solar irradiance by NO2 molecules. The Brewer Method [Cede and Herman], for example, measures the NO2 absorption from the direct solar beam using a Brewer spectrophotometer. These measurements may only be made during clear-sky conditions (cloud and aerosol optical thickness less than 1.5). The air mass factors used to transform the measured SCD to the VCD are relatively simple: They are geometric AMFs, modified to account for the absorption by stratospheric ozone. Besides the simplicity of the AMF calculation and the simplicity of the direct-sun measurement, the Brewer method also has the advantage of requiring a reasonably short measurement time (4 minutes, routinely repeated every half-hour while the solar zenith angle is less than 80°), so measurements made around the OMI overpass time (~13:45 local time) may be isolated for comparison. The intrinsic variability of the individual measurements, arising ultimately from the fact that the NO2 optical thickness is small at the wavelengths used, means that a large number of measurements must be made and averaged together to ensure precision. This, of course, means that the measurements are taken over a period of time long enough that the actual atmospheric composition may change, in addition to the optical geometry. Nonetheless, Cede and Herman have collected a large base of data using an instrument located in Greenbelt, Maryland (currently the only station making NO2 measurements using a Brewer), and Cede et al. have made preliminary comparisons with the OMI total column NO2 product. The results of those comparisons are presented and discussed in a subsequent section of this document. Another instrument, the SAOZ (Système d’Analyse par Observations Zénithales), has a lengthy record of NO2 measurements, made with a network of instruments, primarily located in the polar latitudes. These instruments measure the zenith-sky radiation (from 400 to 600 nm) during sunrise and sunset (solar zenith angles between 86° and 91°), and perform a DOAS fit to retrieve both O3 and NO2. The measurement technique— more specifically, the optical geometry—makes it far more sensitive to NO2 in the stratosphere than the troposphere, but these instruments are mostly stationed in locations far from sources of anthropogenic NO2. While the fact that there are a number of SAOZ instruments distributed over the Earth, regularly measuring NO2, is an advantage, the fact that these instruments only measure NO2 at sunrise and sunset, while OMI measures in the middle of the day, means that the measurements are never actually collocated. This is a particular problem, since stratospheric NO2 varies greatly over the course of the day. Ionov et al. have attempted to get around this by using a chemical transport and photochemistry model to adjust the OMI-measured mid-day measurements to the values that obtained at the time of the SAOZ measurement (they chose to compare to the sunrise measurements). The results of the comparison based on these adjusted data are presented in a later section of this document. Actually, because of the orbital inclination of the EOS-AURA satellite (97°), the high-latitude, northern hemisphere sites tend to have their OMI overpasses earlier in the day than the 13:45 equator crossing time, so the adjustment of the OMI-measured stratospheric NO2 to the sunrise value is not as great is it is in the tropics, or the southern hemisphere stations. One possible further complication of the SAOZ measurement technique is that, measuring at sunrise and sunset, one is measuring at the very times when the stratospheric NO2 is changing the most rapidly, due to photochemical destruction. The Multi-axial DOAS (MAX-DOAS) instruments are designed to measure the vertical profile of NO2 and O3 in the troposphere. The DANDELIONS (Dutch Aerosol and Nitrogen Dioxide Experiments for vaLIdation of OMI aNd SCIAMACHY) campaign, organized by KNMI (Koninklijk Nederlands Meteorologisch Instituut/Royal Dutch Meteorological Institute) during May and June 2005, used three MAX-DOAS instruments to measure NO2. Preliminary results have been presented by Brinksma et al. This campaign was conducted in Cabaw, the Netherlands, and measured a number of air quality-related quantities, but with a primary focus on NO2. One objective of this field campaign was to intercompare individual MAX-DOAS instruments (and a newer “miniMAX-DOAS” model). However, the MAX-DOAS results were also compared with collocated OMI tropospheric NO2 measurements. We now turn to the results of the validation studies that have been made so far.