Satellite remote-sensing capability to assess tropospheric column ratios of formaldehyde and nitrogen dioxide: case study during the LISTOS 2018 field campaign
Abstract. Satellite retrievals of tropospheric column formaldehyde (HCHO) and nitrogen dioxide (NO2) are frequently used to investigate the sensitivity of ozone (O3) production to concentrations and emissions of nitrogen oxides (NOx) and volatile organic carbon compounds (VOCs). Space-based remote-sensing information of chemical proxies for NOx (i.e., NO2) and VOCs (i.e., HCHO), in particular the ratios of tropospheric column HCHO and NO2 (FNRs), provide insight into the non-linear relationship of O3 formation in the lower troposphere. Ultraviolet–visible (UV/VIS) satellite spectrometers such as the Ozone Monitoring Instrument (OMI) and TROPOspheric Monitoring Instrument (TROPOMI) are capable of providing FNR information with high spatiotemporal coverage, yet a recent study suggested that the biases and noise of satellite retrievals are the largest source of uncertainty for applying satellite-derived FNRs to better understand O3 production sensitivities. To quantify, and inter-compare, the uncertainties in two of the most commonly-applied satellite sensors to investigate O3 production sensitivities, we evaluated OMI and TROPOMI retrievals of NO2 and HCHO tropospheric columns, and resulting FNRs, using Geostationary Trace gas and Aerosol Sensor Optimization (GeoTASO) and GEO-CAPE Airborne Simulator (GCAS) airborne remote-sensing data taken during the Long Island Sound Tropospheric Ozone Study 2018 (LISTOS 2018).
Compared to suborbital remote-sensing observations of tropospheric column NO2 and HCHO, the accuracy of OMI (using both the National Aeronautics and Space Administration (NASA) version 4 and the Quality Assurance for Essential Climate Variables (QA4ECV) retrieval algorithms) and TROPOMI were magnitude-dependent with high biases (i.e., satellite tropospheric columns > suborbital tropospheric columns) in clean/background environments and a tendency towards a low bias (i.e., satellite tropospheric columns < suborbital tropospheric columns) in moderate to polluted regions. Campaign-averaged NO2 median biases for OMI, using both the NASA and QA4ECV algorithms, were similar at 0.4±4.1 × 1015 molecules cm-2 (6.3 %) and 0.4±4.5 × 1015 molecules cm-2 (6.8 %), respectively. TROPOMI retrievals of NO2 had a campaign-averaged median bias of -0.3±3.7 × 1015 molecules cm-2 (-4.8 %) and 0.3±3.3 × 1015 molecules cm-2 (5.8 %) when averaged at finer (0.05° × 0.05°) and coarser (0.15° × 0.15°) spatial resolution. The three satellite products (NASA OMI, QA4ECV OMI, and TROPOMI) differed more when evaluating tropospheric column HCHO retrievals. Noise in the HCHO retrievals, likely due to low signal-to-noise ratios and the fact the UV/VIS measurement sensitivity at shorter wavelengths used in HCHO retrievals are low in the troposphere, resulted in low correlations and high oscillation/variability in bias (bias standard deviation) in all three satellite products, with campaign-averaged median biases of 5.1±7.8 × 1015 molecules cm-2 (38.7 %), 2.3±8.9 × 1015 molecules cm-2 (17.3 %), 1.9±6.7 × 1015 molecules cm-2 (12.9 %), and 2.9±4.9 × 1015 molecules cm-2 (23.1 %) for NASA OMI, QA4ECV OMI, and TROPOMI at finer and coarser spatial resolution, respectively. Spatially-averaging TROPOMI tropospheric column HCHO, along with NO2 and FNRs, to coarser resolutions similar to OMI native pixel size proved to reduce the bias standard deviation of the retrieval data. While large median biases, and enhanced variability in bias, were derived for HCHO, errors in both NO2 and HCHO tropospheric columns tended to offset as all three satellite products compared well to observed FNRs with campaign-averaged median biases from NASA OMI, QA4ECV OMI, and TROPOMI of 0.4±3.8 (11.0 %), -0.2±3.3 (-5.4 %), and 0.4±2.3 (13.0 %), respectively. While satellite-derived FNRs had minimal campaign-averaged median biases, the statistical analysis shows that all satellite FNR values still had large bias standard deviation due to unresolved errors in satellite retrievals of HCHO. This result is important as accurate retrievals (minimal median biases) of FNRs from satellites do not suggest the accuracy of the underlying proxy species. The reduction in noise in satellite retrievals of HCHO with additional calibration and improved sensor design and/or improved a priori information of the vertical profiles of HCHO in the troposphere to avoid the impact of the low measurement sensitivity in the shorter UV/VIS wavelengths used to retrieve HCHO is critical for reducing unresolved biases in satellite retrievals of FNRs. Furthermore, this work demonstrates the large impact of a) a priori vertical profiles of NO2 and HCHO for calculations of Air Mass Factors in tropospheric column trace gas retrievals in both OMI and TROPOMI, b) spatiotemporal averaging to increase signal-to-noise, and c) different retrieval algorithms on retrieval errors. Finally, the novel diurnal information of tropospheric FNRs that is expected to be provided by the upcoming NASA geostationary sensor Tropospheric Emissions: Monitoring of Pollution (TEMPO) is investigated and compared to low earth orbiting sensors currently applied to investigate tropospheric FNRs.
Matthew S. Johnson et al.
Status: final response (author comments only)
RC1: 'Comment on amt-2022-237', Anonymous Referee #1, 01 Nov 2022
- AC1: 'Reply on RC1', Matthew S. Johnson, 06 Feb 2023
RC2: 'Comment on amt-2022-237', Anonymous Referee #2, 08 Nov 2022
- AC2: 'Reply on RC2', Matthew S. Johnson, 06 Feb 2023
Matthew S. Johnson et al.
Matthew S. Johnson et al.
Viewed (geographical distribution)
This study evaluates OMI and TROPOMI retrievals of NO2, HCHO and FNR using aircraft measurements during the LISTOS campaign. The manuscript is well-written, and it is a good for for AMT. See my comments below.
Abstract: The abstract is lengthy. I’d suggest the authors shorten the abstract to include only the core findings of this work. For example, the first paragraph may belong to introduction.
Line 370: What are the quality flags for? Is this the same quality flag as for TROPOMI? If so, why do you choose different thresholds? Better to include references here.
Table 2: I’d suggest include an estimate of the error, such as normalized mean standard errors. NMB doesn’t tell much about the precision of the retrievals.
Line 530: Maybe you could have a figure of the mean biases of HCHO to show where OMI or TROPOMI HCHO is biased high?
Line 600: I’m not sure if we could call this as a ‘high pollution’ day because ozone was actually low on this day. This could very well be a cold day when the lifetime of NO2 is long, and the the photolysis is low. I’m not sure how much value there is to evaluate FNR on this day. It’d be more interesting to add another day with both high ozone and high NO2.
Line 700: It is interesting to see that improved the a priori from CMAQ does not improve the retrieval performance of OMI. The authors attribute this to coarse resolution of OMI. Could this be due to the coarse resolution of cloud and surface albedo data used in the retrieval?
Line 835: While the low mean biases of FNR is low, the standard deviation is very large. The R sure is also low for FNR. Thus I don’t think the errors of HCHO and NO2 could cancel out. The errors in HCHO and NO2 can offset only if the errors are correlated. I’d suggest the authors make a scatter plot for errors of HCHO versus NO2, and see if they are correlated.