Development of an OMI AI data assimilation scheme for aerosol modeling over bright 2 surfaces—a step toward direct radiance assimilation in the UV spectrum

Abstract


57
Currently, the main satellite inputs for operational aerosol modeling are AOD products 58 derived from passive-based polar orbiting imagers, such as the Moderate Resolution Imaging 59 Spectroradiometer (MODIS), the Visible Infrared Imaging Radiometer Suite (VIIRS), and the 60 Advance Very High Resolution Radiometer (AVHRR). Experimentation is proceeding with the 61 use of products from the multi-angle imaging spectroradiometer (MISR) (e.g., Lynch et al., 2016; 62 Randles et al. 2017;Buchard et al. 2017) and from geostationary instruments such as Himawari 63 and Geostationary Operational Environmental Satellite (GOES). A major advantage with such 64 passive-based satellite sensors is that the AOD is retrieved with high spatial and temporal to merge with AOD or solar radiance assimilation to influence aerosol loading, height and 93 absorption (e.g., VIIRS+OMPS product; such as Lee et al. 2015). Details of the developed OMI 94 AI assimilation system are presented in the paper, which is organized as follows: Data sets used 95 in the study are summarized in Section 2; Section 3 discusses the components of the AI-DA 96 system. Section 4 provides an evaluation of the developed system; and Section 5 contains a 97 summary discussion. System (NAAPS; Lynch et al., 2016), which was the first operational global aerosol mass transport 104 model available to the community. The assimilation system is based on spatial and temporal 105 variations of aerosol particles from NAAPS (Zhang et al., 2006;, and the Vector LInearized 106 Discrete Ordinate Radiative Transfer (VLIDORT; Spurr, 2006) code is used to construct a forward 107 model for the AI-DA system. (1) 121 Unbiased, noise-reduced, quality-assured AI data are necessary for AI data assimilation.

122
This is especially important for OMI observations, due to this particular sensor suffering from the 123 well-referenced "row anomalies" issues (Torres et al., 2018).  (Zhang et al., 2006;Hyer et al., 2011;Shi et al., 2014) are assimilated into 163 NAAPS through the Naval Research Laboratory Atmospheric Variation Data Assimilation 164 System-AOD system (NAVDAS-AOD; e.g., Zhang et al., 2008;Zhang et al., 2011;Zhang et al., 165 2014 process within the tropics (Joyce et al., 2004). The usage of CMORPH avoids the ubiquitous 170 precipitation bias that exists in all global atmospheric models (e.g. Dai, 2006) and is proven to 171 improve aerosol wet deposition, therefore yielding better AOD (Xian et al., 2009). The reanalysis 172 agrees reasonably well with AERONET data on a global scale (Lynch et al., 2016) and also 173 reproduces AOD trends that are in a good agreement with satellite based analysis (e.g., Zhang and 174 Reid, 2010;Hsu et al., 2012). In this study, we use a free running version of NAAPS reanalysis v1 175 without AOD assimilation to provide aerosol fields every 6 hours at 1x1 (Latitude/Longitude) 176 resolution. 4-vectors with respect to any atmospheric or surface property (Spurr, 2006). The model uses the VLIDORT model may be found in a recent review paper (Spurr and Christi, 2019, and 191 references to VLIDORT therein).

192
VLIDORT is used to simulate the AI in this study. Simulations at 354 and 388 nm are

214
To construct an AI-DA system, a forward model is needed to simulate AI using aerosol 215 concentrations from NAAPS. In this study, the forward model is built around the VLIDORT 216 model, following a similar method to that suggested in Buchard et al. (2015). Here VLIDORT is 217 configured to compute OMI radiances and Jacobians as functions of the observational conditions 218 at 354 and 388 nm, using geolocation information from OMI data such as satellite zenith, solar 219 zenith and relative azimuth angles, as well as ancillary OMI data (surface albedos at 354 and 388 220 nm).

221
To convert from NAAPS mass-loading concentrations to aerosol extinction and scattering 222 profiles, we require aerosol optical properties for the four species at 354 and 388 nm, which are 223 summarized in (GEOS-5) model (e.g. Colarco et al., 2014;Buchard et al., 2015). Note that the study period is . (2)
Here, Iaer354(ρ354) is the calculated radiance at 354 nm using NAAPS aerosol fields as well as the 253 OMI-reported surface albedo at 354 nm (ρ354). Iray354(R388 ʹ ) is the calculated radiance assuming a 254 Rayleigh atmosphere and the derived value of R388ʹ as surface albedo (Buchard et al., 2015).

255
The forward model-simulated OMI AI values are inter-compared with OMI AI values as 256 shown in Figure 1 for the study region. A total of one month (01-31 July 2007) of NAAPS 257 reanalysis data and OMI AI data were used. Note that OMI AI data over both cloud-free and 258 cloudy skies were used. Since surface albedos included in the OMI data represent reflectivities 259 under clear-sky situations, the albedo under cloudy sky is then computed Here, ρclr and fc are the clear sky surface albedo (e.g. ρ354 or ρ388) and the cloud fraction, both  Here, 1 and 2 are given respectively by Equations (7) and (8) Based on these equations, radiance Jacobians with respect to aerosol particles, K354,nk and K388,nk, 295 are computed at 354 and 388 nm, respectively, using OMI-reported surface albedo values (ρ354 296 and ρ388), followed by a calculation of the albedo Jacobian 354 ( 388 ) at 354 nm.

Forward model for Jacobians of AI
To check this analytic Jacobian calculation in Eqns. (6)- (8), we compute the aerosol AI 298 Jacobians using a finite difference (FD) method. Here, the derivative of AI as a function of aerosol 299 concentration of a species, k, in layer n, is computed using Here Cnk and Cnk' are the baseline and perturbed aerosol concentrations, respectively, and AI and 302 AI' are computed using Cnk and Cnk', respectively.  AIsmoke + AIdust. The same assumption holds for dust aerosols.

324
Given these two principles, the overall design concept for the OMI AI assimilation can be

348
The background error covariance matrix is constructed from modeled error variances and 349 error correlations, following the methodology in previous studies (Zhang et al., 2008;2011). The

353
Here, x and y are two given locations, and Rxy is the great circle distance. L is the averaged error 354 correlation length and is set to 200 km based on Zhang et al. (2008). Similarly, the vertical error 355 correlation between two pressure levels p1 and p2 is also based on the SOAR function, this time in (12)

358
Here, L is a unit-less number representing vertical correlation length and is set to 0.015.

359
The horizontal error variance is based on the RMS error of aerosol concentrations, which 360 is arbitrarily set to 100 µg/m 3 for near-surface dust aerosols (ground to 700 hPa  Africa, heavy aerosol plumes, as hinted at in NAAPS AOD from the natural run, cover larger 406 spatial areas than those inferred from OMI AI data. In comparison, NAAPS AOD patterns from 407 the OMI AI data assimilation cycle closely resemble aerosol patterns as suggested from OMI AI 408 data. Also shown in Figures 3e and 3f are the simulated AI using NAAPS data from the natural and OMI AI DA runs (data from Figures 3c and 3d) respectively. Clearly, with the use of NAAPS 410 data from the natural run, simulated OMI AI are overestimated in comparison with OMI AI data 411 (Figure 3b). Simulated AI patterns with the used of NAAPS data from the OMI AI DA run rather 412 closely resemble AI patterns from the OMI data, again, indicating the OMI AI DA system is 413 functioning reasonably as designed.

414
The performance of AI-DA is also evaluated using OMI AI for the whole study period, as 415 shown in Figure 4. These data are constructed using collocated OMI AI and NAAPS data respectively. Figure 4b is the spatial distribution of the simulated AI using NAAPS data from AI-  (Figure 4c). This is also seen from 424 Figure 4d, which is the difference between simulated AI from NAAPS AI-DA runs and OMI AI.

425
In contrast with the situation in Figure 4d, Figure 4h, which is the difference between simulated 426 AI from NAAPS natural runs and OMI AI, shows much larger differences in AI values.

427
While it is not too difficult to make the model mimic the AI product, proof of real skill lies 428 in any improvements to AOD calculations. To this end, the performance of OMI AI assimilation 429 was evaluated with the use of AERONET data. Figure 5a shows the inter-comparison of NAAPS As mentioned in Section 3, aerosol properties for non-smoke aerosol types were obtained 457 from the NASA GEOS-5 model (e.g. Colarco et al., 2014;Buchard et al., 2015). Yet, different  The OMI AI data assimilation system is a proxy for all-sky, all-band modeling system 504 radiance assimilation. It contains all the necessary components for such radiance assimilation,

517
In contrast with the assimilation of retrieved aerosol properties, both aerosol absorption 518 and scattering need to be accounted for when assimilating radiance or OMI AI in the UV spectrum.

519
This requires the inclusion of more dynamic aerosol optical properties into the data assimilation 520 process, and properties that vary with region and season. As noted already, even for biomass 521 burning aerosols over South Africa, lower single scattering albedo values were found at earlier 522 stages of burning seasons (e.g. Eck et al., 2013). A look-up- This study also suggests that NAAPS analyses with OMI AI data assimilation cannot out-560 perform NAAPS reanalyses data that were incorporated with MODIS and MISR AOD 561 assimilation, and validated against AERONET data. This is not surprising, as OMI AI is only a 562 proxy for the AOD and is sensitive to other factors such as surface albedo and aerosol vertical 563 distribution. Also, AERONET data are only available over cloud-free field of views, so the 564 performance of our OMI AI data assimilation system over cloudy regions has not been evaluated.

565
There are a number of issues arising from our study. For example, aerosol optical 566 properties are needed for the OMI AI-DA system -these have strong regional and temporal 567 signatures that need to be carefully quantified before applying them to the AI-DA on a global scale.

568
Also, OMI AI retrievals are rather noisy and contain known and unknown biases. Abnormally 569 high OMI AI values are found over mountain regions as well the polar regions. Sporadic high AI values are also known to occur, for reasons that are still not properly understood. Even though 571 quality assurance steps were proposed in this study, detailed analysis of OMI AI data are needed 572 for future implementation of OMI AI data assimilation for aerosol studies.

573
Lastly, AI values are derived from radiances and thus, the AI-DA system presented in the 574 study can be thought of as a radiance assimilation system for the UV spectrum. This is because 575 the AI-DA system contains all necessary components for radiance assimilation, based on a forward 576 model for calculating not only simulated satellite radiances, but also the aerosol-profile Jacobians 577 of these radiance, both quantities as functions of observation conditions. This study is among the 578 first attempts at radiance assimilation at the UV spectrum and indicates the future potential for 579 direct radiance assimilation at the UV and visible spectra for aerosol analyses and forecasts.