Ground-based MAX-DOAS observations of NO2 and H2CO at Kinshasa and comparisons with TROPOMI observations
Abstract. We present a database of MAX-DOAS (Multi-AXis Differential Optical Absorption Spectroscopy) ground-based observations of NO2 and H2CO performed for the first time in the city of Kinshasa. These measurements were conducted between November 2019 and July 2021 and processed using the standardized inversion tools developed in the ESA FRM4DOAS (Fiducial Reference Measurements for Ground-Based DOAS Air-Quality Observations) project. The retrieved geophysical quantities are used to validate column observations from the TROPOspheric Monitoring Instrument (TROPOMI) in Kinshasa. In the validation, we experiment three different comparison cases of increasing complexity. In the first case, a direct comparison between MAX-DOAS observations (average +/- 60 minutes around overpass) and TROPOMI shows an underestimation of TROPOMI with a median bias of -40 % (s=0.26 and R=0.41) for NO2 and -26 % (s=0.24 and R=0.28) for H2CO. The second case takes into account the different vertical sensitivities of the two instruments and the apriori profile. We note a slight decrease of the biases and a strong improvement of the linear regression parameter, about -35 % (s=0.72 and R=0.74) for NO2 and 1 % (s=1.01 and R=0.66) for H2CO. The third case, which is considered more realistic than the first two, builds on the second case by considering also the direction of sight of the MAX-DOAS. For this third case, we find a bias of -2 % (s= 1.09; R= 0.59) for NO2 and 13 % (s= 1.51; R= 0.60) for H2CO. Those results indicate a large impact of the vertical sensitivity and horizontal heterogeneity in this validation process at this site. In order to evaluate the capability of the GEOS-Chem model in this region, we performed the comparisons between TROPOMI and the simulations made for 2020. We found a bias of 16 % (s= 0.42 and R = 0.80) for NO2 and bais of 61 % (s= 0.05 and R = 0.24) for H2CO.
Rodriguez Yombo Phaka et al.
Status: final response (author comments only)
- RC1: 'Comment on amt-2022-327', Anonymous Referee #1, 16 Feb 2023
- RC2: 'Comment on amt-2022-327', Anonymous Referee #2, 17 Feb 2023
Rodriguez Yombo Phaka et al.
Rodriguez Yombo Phaka et al.
Viewed (geographical distribution)
This paper discusses the comparison of MAX-DOAS and TROPOMI observations of NO2 and H2CO at Kinshasa. I found two aspects of the paper particularly interesting: the location of the measurements, in a very data-sparse but interesting and relevant region, and the three cases presented, which nicely show the impact of vertical profiles and line of sight on the quantitative comparisons. The paper is generally well written, with a good set of references and a good introduction.
But, in general I found the comparisons with GEOS-Chem not very revealing/useful. The resolution of the model is very low, 2x2.5 degree, and a direct comparison with the MAX-DOAS is not even presented (maybe because of this). The model-TROPOMI comparison is limited to one location. A real model evaluation with TROPOMI would involve assessments over a larger area, addressing aspects of the emissions (choice inventory) and sectors (fires, transport, industry) contributing emissions, and evaluation of other model processes (transport, chemistry, deposition). Drawing conclusions from a time series at one location is not really possible. A reasonal comparison is shown for Kinshasa, which is pleasing but may be coincidental. One aspect which is of interest for this paper is the NO2/H2CO profile from GEOS-Chem, and comparisons with MAX-DOAS and TROPOMI a-priori/a-posteriori profiles. This could be extended and structured differently.
Because of this I would be in favour of publication of this work, after my comments below have been dealt with. In particular, the comparison with GEOS-Chem could be shortened and could be given a different focus.
- Abstract: The word "bias" should be used in a more balanced way to my opinion. A "bias" normally points to one of the two datasets, taking the other as reference (assuming it to be more accurate). On line 10 "shows an underestimation of TROPOMI with a median bias of -40% (s=0.26 and R=0.41) for NO2 and -26% (s=0.24 and R=0.28) for H2CO". The reader will conclude from this that TROPOMI is biased low. But the other cases show that the difference is influenced by the way the analysis is done. So, I conclude that the -40%/-26% are not so much to be attributed to TROPOMI, but also reflect the comparison approach. An alternative formulation could be "MAXDOAS is biased high by +40/+26%", which sounds like a very different conclusion. I would suggest to use a more neutral "mean difference between MAXDOAS and TROPOMI" instead of "the bias of TROPOMI" throughout the paper.
- Abstract, line 16: "We found a bias of 16% (s= 0.42 and R = 0.80) for NO2 and bais of 61% (s= 0.05 and R = 0.24) for H2CO". Is the model or TROPOMI higher in this case?
- Abstract, line 16: "bais"
- Fig. 1. Do these yellow lines correspond to the MAX-DOAS viewing direction? It may be useful to indicate the viewing (azimuth) direction in Fig. 12/13 as a line or arrow. Maybe Figs 12+13 could be brought to the beginning of the paper, e.g. after Fig. 1. The spatial distribution of NO2/H2CO is useful as background information before reading the rest of the paper. Would be nice to see MODIS AOD as well.
- line 107: "only MMF data selected for their consistency with corresponding MAPA results are retained." What does that mean? Which dataset(s) is (are) submitted?
- line 113: "we only considered MMF due to inconsistencies in the MAPA aerosol retrievals for our Kinshasa spectra." Please explain the "inconsistencies".
- line 113: It would be valuable to see the results for both retrieval approaches and to know how much the MAX-DOAS results (tropospheric columns) differ between the two (e.g summarise the findings of the papers cited in line 112-113). Would it be possible to present MAPA results?
- line 115: "both algorithms." Does this refer to NO2 vs H2CO, or MMF vs MAPA? But MAPA is not used?
- line 116: "monthly climatology" Why not use the actual meteorological variables from for instance the ERA-5 reanalysis? Is the retrieval sensitive to meteorology (temperature)?
- line 124: Please add a comment on the a-priori error used: how much is the retrieval constrained by the a-priori?
- Sec 2.3: For FRM4DOAS you cite the ATBD. Likewise it would be useful to include a reference to he TROPOMI ATBDs.
- line 129: What is this "S5P-PAL" product? Please explain in one or two sentences.
- line 132: "Only pixels within a radius of 20km around the observation site" Why 20 km?
- Sec 2.4: Apart from Marais, it would be useful to add a few key references for this global GEOS-Chem (version 12). Are there other relevant studies done over Africa with the model?
- Sec 2: I would expect a section on the intercomparison approach, dealing with aspects like profile shape, horizontal gradients, line of sight, collocation and meaning (use) of the circles in Figs 12 and 13. Instead, the cases are discussed in section 3.2.
- line 168: I was wondering how much the biomass burning season is contributing to AOD in comparison to local (dust, transportation, industry, household) contributions? Could you summarise what is known from e.g. the inventories.
- Fig. 5. "The error bars represent the standard deviation." The standard deviation of what? Is it an error bar or a measure of the spread of the values?
- line 184: "with some delay, " What would be a typical delay during daytime?
- line 187: Looking at figure 12 this first case does not seem very useful. There is clearly a strong gradient and a lot of clean area is included in the average which is not observed by the MAX-DOAS instrument. I could imagine that for H2CO a larger circle may be needed because of the larger noise level compared to NO2.
- line 195: Please provide the details. What are the units of the MAX-DOAS profile (molecules/cm^3?). How is the interpolation done? Does the interpolation conserve the column amount? What is the collection of MAX-DOAS observations from which the median is computed? Why a median instead of a mean, and does it matter?
- Case 3: Is this done in the same way as described in Dimitropoulou et al? I was wondering if the method could be visualised? For instance for one day / one overpass, showing the region like in Fig 12/13, the azimuthal viewing line and TROPOMI pixels selected. Are weights applied to the TROPOMI observations? Are pixels close to the MAXDOAS more important?
- line 203: "a coincidence test is performed" What are the criteria? What is the "surface point" of a pixel?
- line 213: Looking at the figure it seems that the retrieval is producing negative concentrations in several cases. Do you apply a clipping, or are negatives used as is?
- line 216: "motivate the application of the transformation". According to the optimal estimation theory of Rodgers averaging kernels are to be used in profile comparisons. So this would be the main motivation, rather than an observed difference in profile shape. The difference in profile shapes indicates that case 1 and 2 may differ substantially.
- Fig.6. In May-September the profiles for H2CO look quite different. Which one of the two would be more realistic? You mentioned before that the sensitivity of the MAX-DOAS rapidly decreases at 2km and above. This is also indicated by the small spread around 2 km altitude. So maybe the ground-based observation is not sensitive enough to capture elevated layers? Could this explain part of the difference ground-satellite? Please comment.
- Fig.7. What kind of regression method is used? Does it account for satellite and ground-based retrieval errors? Please indicate in the caption that this is a case-3 comparison.
- Fig.7: "TROPOMI error bars are standard deviations. " For the monthly values I assume this is the spread in the individual column amounts used in the average. But what is shown for the daily points? Is it now the retrieval error, or still a spread in values?
- line 277: "As for NO2, the results of the third case are shown in Figure 8" .. for H2CO.
- line 278: "The dynamic range of MAX-DOAS measurements is small compared to that of NO2 ". Does this refer to the blue-green error bars?
"..because of the different points filtered .." this is unclear to me.
- line 282: "reduced number of TROPOMI measurements" How many measurements are used on average for NO2 and H2CO?
- Fig. 9: It would be nice if the a-priori profiles from TROPOMI and MAX-DOAS could be added in this figure as well. Like in Fig. 6 it would be good to see the season-averaged profile shape. Maybe Fig 6 and Fig 9 could be combined?
- Why is GEOS-Chem only compared with TROPOMI and not with the MAX-DOAS? Of course the resolution of the model is very low compared to TROPOMI, leading to large mismatches in the air masses probed.
- line 331: "The general underestimation of TROPOMI compared to MAX-DOAS ". It would be good to mention that you refer to the difference between case 2 and case 1 here. The "best" comparison, case3, does not show a prominent underestimation.
- line 344: "Additional uncertainties comes from clouds and aerosols". How are clouds treated/filtered in the retrieval of the MAX-DOAS? This information is not provided in the paper.
- line 363: Why refer to the raw comparison here? The best comparison is presented in Figs 10 and 11.
- Section 4.2. See my general comment above. The discussion cites a few studies and possible shortcomings. But there is not enough data to draw conclusions concerning GEOS-Chem (or TROPOMI) apart from a general reasonable agreement and similar seasonality.
- line 404: I have the feeling that the noise in individual TROPOMI H2CO retrievals is an important reason for a poor correlation with GEOS-Chem and/or MAX-DOAS. This role of the retrieval noise could be discussed in more detail.
- line 406: "The present comparisons have shown the importance of correcting the initial TROPOMI products with the profile measured over the observation site and taking into account the horizontal variability of the studied molecules." Could you generalise this finding and formulate recommendations for other sites and satellite-ground remote sensing comparisons in general? Large gradients near cities are common, and the case 3 comparison
approach could be a general recommendation for future validation work. How should previous comparisons (e.g. Verhoelst et al, Vigouroux et al) be interpreted?