Presentation is loading. Please wait.

Presentation is loading. Please wait.

Institute of Environmental Physics and Remote Sensing IUP/IFE-UB Physics/Electrical Engineering Department 1 Institute.

Similar presentations

Presentation on theme: "Institute of Environmental Physics and Remote Sensing IUP/IFE-UB Physics/Electrical Engineering Department 1 Institute."— Presentation transcript:

1 Institute of Environmental Physics and Remote Sensing IUP/IFE-UB Physics/Electrical Engineering Department 1 Institute of Environmental Physics and Institute of Remote Sensing University of Bremen Improving OMI NO 2 retrievals Andreas Richter, A. Hilboll, and J. P. Burrows OMI Science Team Meeting De Bilt, March 12, 2014 TRYING TO IMPROVE BETTER UNDERSTAND

2 2 Introduction OMI data provides excellent coverage and unprecedented spatial detail It will soon be followed by the TROPOMI / S5P instrument which has many similarities but promises even better performance  As preparation for S5P, IUP Bremen has started to look into OMI spectra

3 3 Choice of fitting window Starting point is transfer of GOME-2 settings to OMI data First check: how consistent are results from different fitting windows? –425 – 497 nm vs 425 – 450 and 405 – 465 nm => very good Second check: how consisitent are results with operational NASA product? –Very good correlation with 405 – 465 nm, offset of 1.5x10 15 molec cm -2 => will use GOME-2 settings

4 4 Choice of fitting window Why these specific settings? Good results in GOME-2 data –Little noise –Good sensitivity to tropospheric NO 2 The same window is used in our ground-based analysis => somewhat random choice

5 5 Spike removal In spite of the good signal to noise, OMI data in SAA region have large scatter Data can easily be flagged using RMS of fits, but can‘t we do better?

6 6 Spike removal Apply GOME-2 approach: –Run DOAS fit on original data –Identify “spikes“ using threshold of 10 x mean RMS –Re-run DOAS fit on reduced number of spectral points

7 7 Spike removal Application to data in SAA region strongly reduces the number of outliers compared to operational NASA product This approach ignores all flagging in lv1 data Data elsewhere remains nearly unchanged Is there an issue with potential for biases?

8 8 Spike removal The number of useful retrievals in the SAA region is strongly increased

9 9 Pacific background Although the NO 2 product is rather robust when using the mean solar background, some “stripes” are apparent They can be removed using destriping Use of earth-shine background leads to similar results => offset? No destriping Destriping Pacific background

10 10 AMF dependence on NO 2 column One of the main DOAS assumptions is, that the light path enhancement (AMF) for a trace gas is independent of its column amount For strong absorbers (O 3, SO 2, IR gases) this approximation is not good enough and the change of sensitivity with wavelength needs to be accounted in the fit ("modified“ DOAS) NO 2 is generally considered to be a weak absorber, but is that still true for very polluted scenarios?

11 11 AMF dependence on NO 2 column Up to a vertical column of about 1x10 16 molec cm -2, small dependence of AMF on NO 2 amount For larger columns, the AMF decreases and NO 2 absorption structures appear

12 12 AMF dependence on NO 2 column In relative units, the effect is –5% for a column of 5x10 16 molec cm -2 –5 – 12% for 1x10 17 molec cm -2 The effect is larger at smaller wavelengths (larger absorption cross- section, more Rayleigh scattering)

13 13 AMF dependence on NO 2 column Effect on DOAS fit on synthetic data is larger than on AMF alone as both smaller AMF and spectral structures reduce NO 2 columns Relative effect increases with SZA Effect is similar in typical fitting windows Even at a column of 5x10 16 molec cm 2, the error is > 10%

14 14 Excursion: GOME-2 NO 2 above China Monthly GOME-2 tropospheric NO 2 data are missing most of the large values These were removed by cloud filtering as aerosol was so thick that data were classified as partially cloudy No cloud screening

15 15 Is it only Aerosols? Even without cloud screening, there are data gaps over pollution hot spots on some days This is due to quality checking as these fits are poor No Chisq. screening

16 16 Why are the fits poorer at strong pollution? There are large and clearly structured residuals in fits over pollution hot spots This is not random noise! Comparison to NO 2 cross-sections shows that scaling of NO 2 should change over fitting window

17 17 Wavelength dependence of Air Mass Factor For constant albedo, AMF of NO 2 layer close to the surface increases with wavelength in a Rayleigh atmosphere For a surface layer, this can be a significant effect With radiative transfer modelling and a formal inversion, this should provide information on the altitude of the NO 2 About +/- 20%

18 18 Empirical Approach for NO 2 (λ) Take standard NO 2 x-section Scale to increase amplitude with wavelength Orthogonalise to leave NO 2 columns unchanged When introduced in the fit, large residuals are fixed

19 19 Results Empirical Approach NO 2 (λ): GOME-2 The empirical NO 2 AMF proxy is found over the pollution hotspot in China It is not found at other locations where the NO 2 slant column is large There is some noise in the retrieval of the proxy

20 20 Results Empirical Approach NO 2 (λ): OMI As for GOME-2 data, the empirical NO 2 AMF proxy is found over the pollution hotspot in China There is more noise than in GOME-2 data Problems with row anomaly

21 21 Results Empirical Approach NO 2 (λ): OMI At very large NO 2 columns, the wavelength dependency can clearly be found in both fitting windows The retrieval improves when including the empirical parameter The fitting coefficient – relative to that of NO 2 – should contain information on the presence of BL NO 2

22 22 Is there more than China? GOME-2 Fit is improved by AMF proxy everywhere over pollution hotspots

23 23 Comparison to NO 2 columns: GOME-2 Overall pattern similar to NO 2 map Differences in distributions of maxima Artefacts over water noise

24 24 Impact of Clouds: GOME-2 On many days in winter, very large NO 2 slant columns are observed over Europe and the US The NO 2 AMF proxy picks up only very few of these signals This is linked to the fact that most of the events are related to cloudy scenes or snow on the surface, resulting in small wavelength dependence

25 25 Sensitivity Study for NO 2 (λ) / NO 2 Ratio of AMF proxy and NO 2 has strong dependence on NO 2 layer height Dependence on albedo is small between 3% and 7% Synthetic data: Rayleigh atmosphere Constant albedo NO 2 layer in different altitudes DOAS fit on spectra NO 2 temperature dependence corrected by using 2 NO 2 x-sections AMF proxy included Ratio of AMF proxy / NO 2 to normalise signal

26 26 Sensitivity Study NO 2 (λ) / NO 2 : SZA Effect varies with SZA; larger effect at larger SZA At large SZA, AMF proxy also found for high NO 2 Dependence on albedo is small between 3% and 7%

27 27 Sensitivity Study NO 2 (λ) / NO 2 : Bright Surfaces  multiple scattering over bright surfaces is stronger at shorter wavelengths  wavelength dependence of AMF is inverted Increasing albedo reduces effect as expected for reduced importance of Rayleigh scattering For large albedo (> 50%), negative fit factors are found for AMF proxy => wavelength dependence is inverted and only weakly dependent on altitude

28 28 NO 2 (λ) / NO 2 : Impact of Aerosols and Clouds Shielding Similar effect for both, AMF proxy and NO 2 => will cancel in ratio Ratio will give layer height for cloud free part of pixel Light path enhancement Light path enhancement in clouds / aerosols depends only weakly on wavelength Effect on NO 2 but no effect on AMF proxy Ratio will no longer be representative of NO 2 layer height

29 29 NO 2 (λ) / NO 2: Case Study Highveld: GOME-2 NO 2 plume from Highveld power plants can be tracked onto the ocean NO 2 SC values increase downwind of the source AMF Proxy also has higher values within the plume, but –Is more narrow –Has largest values at beginning of plume, not at the end of it

30 30 OMI NO 2 : The Indonesian Mystery Clear sea- land contrast Strange shadows Oszillation with lower values in the North and higher values in the South

31 31 OMI NO 2 : The Indonesian Mystery No cloud screening, simple stratospheric AMF Effect is also visible in pseudo-stratospheric columns (no surface a priori) IUP product seems less affected

32 32 OMI NO 2 : The Indonesian Mystery IUP product has Less “shadow“ structures Less NO 2 over the islands (basically no NO 2 over many of the smaller ones) Overall lower values => more detailed investigation needed!NASA IUP No cloud screening, simple stratospheric AMF

33 33 Summary We have joined the OMI NO 2 club Application of our GOME-2 settings to OMI proved successful –Good consistency between fitting windows but offset vs. Operational product –Spike removal works very well At large NO 2 values (> 5 x molec cm -2 ), AMF becomes a clear function of NO 2 column which needs to be corrected (> 10% effect) At large BL NO 2 values, the wavelength dependence of the NO 2 AMF becomes relevant in the fit in particular in the GOME-2 fitting window and can be used to derive some information on NO 2 layer height Over the Indonesian region, something strange is going on in the operational NO 2 product We‘re looking forward to digging deeper into OMI (and TROPOMI) NO 2 data! Funding by DLR Bonn under Contract 50EE1247

Download ppt "Institute of Environmental Physics and Remote Sensing IUP/IFE-UB Physics/Electrical Engineering Department 1 Institute."

Similar presentations

Ads by Google