The New MISR Research Aerosol Retrieval Algorithm: A Multi-Angle, Multi-Spectral, Bounded-Variable Least Squares Retrieval of Aerosol Particle Properties over Both Land and Water
- 1Earth Science Division, NASA Goddard Space Flight Center, Greenbelt, 20771, USA
- 2Science Systems and Applications Inc., Lanham, 20706, USA
- 3Department of Meteorology and Atmospheric Science, The Pennsylvania State University, State College, 16802, USA
- 4University of Maryland, College Park, MD, USA
- 1Earth Science Division, NASA Goddard Space Flight Center, Greenbelt, 20771, USA
- 2Science Systems and Applications Inc., Lanham, 20706, USA
- 3Department of Meteorology and Atmospheric Science, The Pennsylvania State University, State College, 16802, USA
- 4University of Maryland, College Park, MD, USA
Abstract. Launched in December 1999, NASA’s Multi-angle Imaging SpectroRadiometer (MISR) has given researchers the ability to observe the Earth from nine different views for the last 22 years. Among the many advancements that have since resulted from the launch of MISR is progress in the retrieval of aerosols from passive space-based remote-sensing. The MISR operational standard aerosol retrieval algorithm (SA) has been refined several times over the last twenty years, resulting in significant improvements to spatial resolution (now 4.4 km) and aerosol particle properties. However, the MISR SA still suffers from large biases in retrieved aerosol optical depth (AOD) as aerosol loading increases. Here, we present a new MISR research aerosol retrieval algorithm (RA) that utilizes over-land surface reflectance data from the Multi-Angle Implementation of Atmospheric Correction (MAIAC) to address these biases. This new over-land/over-water algorithm produces a self-consistent aerosol/surface retrieval when aerosol loading is low (AOD < 1); this is combined with a prescribed surface algorithm using a bounded-variable least squares solver when aerosol loading is elevated (AOD > 2). The two algorithms (prescribed + retrieved surface) are then merged as part of our combined-surface retrieval algorithm. Results are compared with AErosol RObotic NETwork (AERONET) validation sun-photometer direct-sun + almucantar inversion retrievals.
Over-land, with AERONET AOD (550 nm) direct-sun observations as the standard, the root-mean squared error (RMSE) of the MISR RA combined retrieval (n = 9680) is ~0.09, with a correlation coefficient (r) of ~0.93 and expected error of (0.225*[MISR AOD] + 0.025). For MISR RA-retrieved AOD > 0.5 (n = 565), we report Ångström exponent (ANG) RMSE of ~0.36, with a correlation coefficient of ~0.85. Retrievals of ANG and aerosol particle properties such as fine-mode fraction (FMF) and single-scattering albedo (SSA) improve as retrieved AOD increases. For AOD > 1.5 (n = 45), FMF RMSE is < 0.09 with correlation > 0.95, and SSA RMSE is < 0.02 with a correlation coefficient > 0.80.
Over-water, comparing AERONET AOD to the MISR RA combined retrieval (n = 4590), MISR RA RMSE is ~0.06 and r is ~0.94, with an expected error of (0.20*[MISR AOD] + 0.01). ANG sensitivity is excellent when MISR RA reported AOD > 0.5 (n = 211), with a RMSE of 0.30 and r = 0.88. Due to a lack of coincidences with AOD > 1 (n = 20), our conclusions about MISR RA high-AOD particle property retrievals over water are less robust (FMF RMSE = 0.12 and r = 0.96, whereas SSA RMSE = 0.022 and r = 0.32).
It is clear from the results presented that the new MISR RA has excellent sensitivity to aerosol particle properties (including FMF and SSA) when retrieved AOD exceeds 1–1.5, with qualitative sensitivity to aerosol type at lower AOD, while also eliminating the AOD bias found in the MISR SA at higher AODs. These results also demonstrate the advantage of using a prescribed surface when aerosol loading is elevated.
- Preprint
(12640 KB) -
Supplement
(1981 KB) - BibTeX
- EndNote
James A. Limbacher et al.
Status: final response (author comments only)
-
RC1: 'Comment on amt-2022-95', Stefan Kinne, 07 Apr 2022
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-95/amt-2022-95-RC1-supplement.pdf
-
RC2: 'Comment on amt-2022-95', Stefan Kinne, 07 Apr 2022
CORRECTION to REVIEW
in my review I also provided the MACv3 aerosol top down aerosol type approach.
Unfortunately there is an errot in table 1 as the (midvis) SSA for large mineral dust
is (not constant) but a function of size (the table lists the same values 0.962 value)Table 1 correction: for re= 1,5, 2.5, 4.0, 6.5 and 10 the mid-visible dust SSA values are 0.962, 0.931, 0.918, .882 and .840
for the same imaginary part (here 0.0011) ... that is very important to understand my comments of the review -
RC3: 'Comment on amt-2022-95', Meng Gao, 10 Apr 2022
The manuscripted by Limbacher et al provide a thorough and interesting study on the impact of surface reflectance on aerosol retrievals using a new MISR research algorithm. Analyses are conducted by using four years of MISR data over both land and water. Details in the aerosol model updates and optimization algorithms are provided, with improvement quantified by comparing quality ensured AROENT data and MISR retrieval results. The MISR algorithm has been well optimized for aerosol retrievals. The new research algorithm further demonstrates the most current capability of aerosol retrievals using multi-angle measurements.
Specifically, the main motivation of this work is the observation of large biases in retrieved aerosol optical depth (AOD) as aerosol loading increases (>1). To resolve this issue, the authors proposed to use the surface reflectance data from the Multi-Angle Implementation of Atmospheric Correction (MAIAC). A combined algorithm is developed with surface properties directly retrieved for low AOD (<1), and a prescribed surface reflectance from MAIAC for large AOD (>2), and a linear combination of the two surface options are used for 1<AOD<2. By comparing with AERONET product and the MISR research algorithm product, the AOD uncertainties are well quantified as: ± (0.225*[MISR AOD] + 0.025) over land, and ±(0.20*[MISR AOD] + 0.01) over water.
This study provides useful experiences and techniques in exploiting aerosol and surface information from multi-angle measurement. Please find my suggestive comments for the authors to consider.
Main comments:
Most of my questions are related to how the surface reflectance are treated and how they impact retrieval results:
- Since there is a larger number of retrieval parameters when using directly retrieval surface properties, it makes sense that there could be large uncertainties. But it is still not clear to me why this leads to a negative bias of AOD as clearly shown in Fig 2(b).
- Page 4, line 25 “The fact that this bias correction was not sufficient to remove the AOD bias seen in the prescribed surface retrieval over-land (especially at AODs < 0.20) indicates that a camera-by-camera correction should probably be used in the future.” Since the MAIAC reflectance has been corrected according to MISR retrieval results at low AOD (Page 4, since line 19), do you suggest the angular shape is still different between the MAIAC and retrieved surface reflectance? Is this part of the reason to have the bias in AOD retrievals?
- Since the authors have done retrieval using both retrieved and prescribed surface reflectance, it could be useful to compare the angular/spectral shape of these surface reflectance to understand exactly where the difference are. Specifically:
- What are the retrieved surface reflectance difference under low and high AOD? How do they compare with the prescribed surface reflectance?
- How does the surface reflectance (retrieved and prescribed) impact aerosol property retrievals differently? Currently only AOD are discussed which shows clear bias over land, it would be interesting to understand how the surface reflectance impacts other properties, such as SSA, FMF etc.
- Page 25, Fig 6, MISR retrieved surface case seem work good over water comparing with the prescribed ocean surface. Does the prescribed ocean surface derived in the same way from MAIAC as discussed for land? Do you have the same correction coefficients applied for the surface reflectance over water? I am curious why there is less AOD bias over ocean than over land.
Minor comments:
- Page 3, line 29: “SSA spectral slope (“Brown Smoke” AOD fraction)”. Are they the same here?
- Page 4, line 5, “applies a spectrally invariant angular-shape-similarity assumption to derive 5 the surface reflectance (over land)”. This is probably explained in later discussions, but do you assume that the same land surface reflectance at different angles and wavelengths?
- Page 4, line 7, “whereas the other algorithm prescribes the surface reflectance for both land and water from other sources”, specify the sources or add reference?
- Page 4, line 11, “We then correct these TOA reflectances for the following: gas absorption, out-of-band light, stray-light from instrumental artifacts, flat-fielding, and temporal calibration trends”. Do you have an estimated accuracy after all those correction in the measurement?
- Page 4, Line 21/22, “surface reflectance”, are they defined in the same way in Eq (1) using ETOA (or EBOA)?
- Page 5, Line 7, “10m wind-speed”. What 10m mean here? The wind speed is retrieved, right?
- Page 5, Line 10, “appropriate solar/viewing geometry”, do you consider spherical shell effect of the atmosphere?
- Page 6, Table 1, it would clear to explain BrS and BlS in the caption.
- Page 6, Line 7, how to do you define “non-sphericity” by mixing two coarse modes?
- Page 8, line 16, “(2)” and “(3)” seem not used for referred later on?
- Page 8, Line 20, cost function seems not normalized by the total number of measurements (N)? The current definition seems agree with a Chi square function which will have the most probable value at N. Is this the case here?
- Page 9, Line 15 “and MAIAC retrieved surface reflectance error (which should be much larger for the MISR 70Ë-viewing cameras than for the near-nadir cameras). Does this relate to earth spherical shell effect too?
- Page 10, line 15/18, “set the result to 0”, so you are finding both A* and Lc to minimize the cost function, right? (I appreciate the authors provide details in the optimization approach (eg. Sec 2.1.2). The optimization are represented by a system of linear equations, which seems work well for this algorithm. )
- Page 11, line 19, “an additional 9 pieces of information”, do you mean the total parameters for land surface are 9+4=13?
- Page 11, line 26, what is the ‘prescribed surface AOD’, are they also provided by MAIAC?
- Page 15, line 12, Do you remove the measurements at particular cameras if the inputs are not ‘good’?
- Page 15, Line 14, cost function < 1, check the normalization of the cost function as mentioned previously.
- Page 15, line 30, “A larger 2nd derivative corresponds to a steeper minimum in our cost function with respect to AOD; we use 10 as a lower bound here in quality flag 6 as this tends to mask out some lower quality results (mostly clouds)”. How do you determine the threshold? Since the derivatives are available, can the authors compute the uncertainties using error propagation, which can provide a more meaningful criteria?
- Page 17, line 8, a prognostic error is introduced here, but not well explained. Some information seems scattered in the Fig 3 captain and discussion from later sections. It would be useful to explain early how the error is computed. Another question: what dataset bins are used to compute the 68th percentiles? Are these bins with respect to AOD, reflectance, uncertainty?
- Fig 3(b): 2% of reflectance?
- Page 22, Fig 5: It seems the MISR algorithm have the flexibility to deal with different aerosol types (therefore different refractive index). For large AOD at Fig 5 (bottom row), the data are peaked at either small or large FMF, which results in better SSA and non-sphericity agreement with AERONT. But for small AOD, there are many intermediate FMF values. If I recall correctly, AERONET retrieval algorithm assumes the same refractive indices for both fine and coarse mode. So the AERONET product should have better representation for fine or coarse mode dominated cases. Does this partially explain what we observe in Fig 5 here?
-
RC4: 'Comment on amt-2022-95', Anonymous Referee #3, 11 Apr 2022
#####GENERAL COMMENTS#####
This manuscript by Limbacher et al. present a new MISR research algorithm (RA) for retrieving aerosol over land and water surface. To address the issues of large biases of high aerosol loading in MISR operational standard aerosol algorithm (SA), the proposed RA utilized and combined 2 schemes: (i) retrieved surface; (ii) prescribed surface from MODIS/MAIAC product. If the prescribed surface algorithm reported AOD<1, then the results from retrieved surface algorithm will be used; if the prescribed surface algorithm reported AOD>2, the results will adopt from prescribed surface algorithm; while if the 1<AOD<2, the results will be merged from 2 algorithms. In general, the methodology is sound. The validation with AERONET suggest a good quality for AOD, ANG as well as FMF, SSA and non-sphericity both over land and water. Overall, I think this paper is well-structured and clearly written, I recommend this paper to be published in AMT after some minor comments have been addressed.
(i) One interesting part however missing in the current manuscript is the direct comparison with MISR operational SA product. I would suggest to add at least some demonstrations of this part to show the evolution.
(ii) In the validation section, the authors evaluate the fine mode fraction with AERONET almucantar inversion product. Why not to use AERONET SDA FMF, which definitely will provide more coincidences?
#####SPECIFIC COMMENTS#####
Page 4 Line 15: How the temporally interpolation is done? Meanwhile, how do you deal with the differences of MISR and MODIS wavelengths?
Page 4 Line 20: MISR’s 36 channels? This should be a mistake. Do you mean 4 wls x9 angles?
Page 5 Line 21: Surface reflectance correction? This is not clear to me.
You correct your retrieved surface reflectance? If yes, how it can help to remove AOD bias?
or you correct measured TOA reflectance?
Page 6 Line 15: How do you derive ANG from your algorithm, this is not clear in the text?
Page 14 Line 24: ANG at 550 nm?
Page 15 Line 17: So the NDVI<0.1 is not retrieved over land, right? Or you still retrieve it but not pass with high quality flag.
Page 20 Line 1: it's not clear from Section 2. How the ANG is derived from the algorithm? Only AOD at 550 nm is mentioned.
Tables 4 and 6: it looks like incorrect for AOD blocks. 0.2<AOD<0.5 not 0.2>AOD>0.5?
-
RC5: 'Comment on amt-2022-95', Alexei Lyapustin, 13 Apr 2022
This is a very good study describing research algorithm development for MISR. The standard MISR over-land retrieval has a long-standing problem of underestimating AOD at high AOD because the EOF algorithm fails when the surface contrast disappears at high AOD. This development uses prescribed MAIAC BRDF dataset over land (similar over ocean) to significantly improve the RA aerosol characterization at high AOD.
I recommend publication after the authors address my specific mostly editorial comments which I provide in the annotated manuscript. A minor re-structuring would also benefit this paper improving readability and understanding.
Alexei.
James A. Limbacher et al.
James A. Limbacher et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
319 | 89 | 17 | 425 | 17 | 6 | 8 |
- HTML: 319
- PDF: 89
- XML: 17
- Total: 425
- Supplement: 17
- BibTeX: 6
- EndNote: 8
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1