A neural network radiative transfer model approach applied to the Tropospheric Monitoring Instrument aerosol height algorithm

Nanda, Swadhin; Graaf, M. de; Veefkind, J. P.; Linden, Mark ter; Sneep, Maarten; Haan, Johan de; Levelt, P. F.

doi:10.5194/amt-12-6619-2019

Cited by 29 publications

(29 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The RTM in this case is a neural network model that has learnt parts of a full physics RTM derived from de Haan et al (1987) called Determining Instrument Specifications and Analyzing Methods for Atmospheric Retrieval (DISAMAR; described in Sect. 3 of Nanda et al, 2019), which is 3 orders of magnitude faster than DISAMAR. In short, the atmosphere is simplified by DISAMAR in order to reduce computational burden, and the neural network forward model is implemented for a further performance boost in an operational environment; for instance, DISAMAR ignores rotational Raman scattering even though the literature has shown that the oxygen A-band ring effects are sensitive to ALH (Vasilkov et al, 2013;Wagner et al, 2010).…”

Section: Tropomi Alhmentioning

confidence: 90%

“…The consequence of the many assumptions in the model (described in Sect. 2.2 of Nanda et al, 2019) result in a large χ 2 (of the order of 1 × 10 4 to 1 × 10 7 ), with larger χ 2 representing a larger departure between the model and the observation. There are several reasons for these departures, the more important ones being the presence of undetected clouds in the scene, incorrect surface reflectance information and multiple aerosol layers.…”

Section: Tropomi Alhmentioning

confidence: 99%

“…Several passive retrieval strategies that are either currently in their operational phase or are upcoming remote sensing missions utilise the interaction of incoming solar radiation with the aerosol particles to retrieve height information. Some notable mentions of missions that retrieve ALH are the Multi-angle Imaging SpectroRadiometer (MISR) on board the NASA Terra satellite (Nelson et al, 2013), which measures aerosol height using geometric optics; the Deep Space Climate Observatory (DSCOVR) mission with its Earth Polychromatic Imaging Camera (EPIC) (Xu et al, 2017(Xu et al, , 2019; the Ozone Monitoring Instrument (OMI) on board the NASA Aura satellite (Chimot et al, 2017(Chimot et al, , 2018Choi et al, 2019); and finally the TROPOspheric Monitoring Instrument (TROPOMI) instrument on board the Sentinel-5 Precursor satellite (Veefkind et al, 2012). Xu et al (2017Xu et al ( , 2019 are the first studies to demonstrate that the diurnal cycle of aerosol height is retrievable.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A first comparison of TROPOMI aerosol layer height (ALH) to CALIOP data

et al. 2020

Self Cite

View full text Add to dashboard Cite

Abstract. The TROPOspheric Monitoring Instrument (TROPOMI) level-2 aerosol layer height (ALH) product has now been released to the general public. This product is retrieved using TROPOMI's measurements of the oxygen A-band, radiative transfer model (RTM) calculations augmented by neural networks and an iterative optimal estimation technique. The TROPOMI ALH product will deliver ALH estimates over cloud-free scenes over the ocean and land that contain aerosols above a certain threshold of the measured UV aerosol index (UVAI) in the ultraviolet region. This paper provides background for the ALH product and explores its quality by comparing ALH estimates to similar quantities derived from spaceborne lidars observing the same scene. The spaceborne lidar chosen for this study is the Cloud-Aerosol LIdar with Orthogonal Polarization (CALIOP) on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) mission, which flies in formation with NASA's A-train constellation since 2006 and is a proven source of data for studying ALHs. The influence of the surface and clouds is discussed, and the aspects of the TROPOMI ALH algorithm that will require future development efforts are highlighted. A case-by-case analysis of the data from the four selected cases (mostly around the Saharan region with approximately 800 co-located TROPOMI pixels and CALIOP profiles in June and December 2018) shows that ALHs retrieved from TROPOMI using the operational Sentinel-5 Precursor Level-2 ALH algorithm is lower than CALIOP aerosol extinction heights by approximately 0.5 km. Looking at data beyond these cases, it is clear that there is a significant difference when it comes to retrievals over land, where these differences can easily go over 1 km on average.

show abstract

Section: Tropomi Alhmentioning

confidence: 90%

Section: Tropomi Alhmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A first comparison of TROPOMI aerosol layer height (ALH) to CALIOP data

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…Aerosols are small liquid or solid particles suspended in the air. Aerosols have a direct effect on climate because they absorb and scatter solar and terrestrial radiation (e.g., Boucher, 2015;Penner et al, 2001). In terms of radiative properties, two types of aerosols can be distinguished: absorbing and scattering aerosols.…”

Section: Introductionmentioning

confidence: 99%

Effects of clouds on the UV Absorbing Aerosol Index from TROPOMI

et al. 2020

Self Cite

View full text Add to dashboard Cite

Abstract. The ultraviolet (UV) Absorbing Aerosol Index (AAI) is widely used as an indicator for the presence of absorbing aerosols in the atmosphere. Here we consider the TROPOMI AAI based on the 340 nm/380 nm wavelength pair. We investigate the effects of clouds on the AAI observed at small and large scales. The large-scale effects are studied using an aggregate of TROPOMI measurements over an area mostly devoid of absorbing aerosols (Pacific Ocean). The study reveals that several structural features can be distinguished in the AAI, such as the cloud bow, viewing zenith angle dependence, sunglint, and a previously unexplained increase in AAI values at extreme viewing and solar geometries. We explain these features in terms of the bidirectional reflectance distribution function (BRDF) of the scene in combination with the different ratios of diffuse and direct illumination of the surface at 340 and 380 nm. To reduce the dependency on the BRDF and homogenize the AAI distribution across the orbit, we present three different AAI retrieval models: the traditional Lambertian scene model (LSM), a Lambertian cloud model (LCM), and a scattering cloud model (SCM). We perform a model study to assess the propagation of errors in auxiliary databases used in the cloud models. The three models are then applied to the same low-aerosol region. Results show that using the LCM and SCM gives on average a higher AAI than the LSM. Additionally, a more homogeneous distribution is retrieved across the orbit. At the small scale, related to the high spatial resolution of TROPOMI, strong local increases and decreases in AAI are observed in the presence of clouds. The BRDF effect presented here is a first step – more research is needed to explain the small-scale cloud effects on the AAI.

show abstract

“…For speed-up, LUTs have often been used in trace gas retrieval algorithms to serve as proxies for RT modeling or to perform corrections to on-line RT approximations. In recent years, applying neural network techniques and principal component analysis (PCA) to RT computational performance has received quite a lot of attention (e.g., Natraj et al, 2005;Spurr et al, 2013;Liu et al, 2016;Yang et al, 2016;Loyola et al, 2018;Nanda et al, 2019;Liu et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Radiative transfer acceleration based on the Principal Component Analysis and Look-Up Table of corrections: Optimization and application to UV ozone profile retrievals

Bak¹,

Liu²,

Spurr³

et al. 2020

Preprint

View full text Add to dashboard Cite

Abstract. In this work, we apply a principal component analysis (PCA)-based approach combined with look-up tables (LUTs) of corrections to accelerate the VLIDORT radiative transfer (RT) model used in the retrieval of ozone profiles from backscattered ultraviolet (UV) measurements by the Ozone Monitoring Instrument (OMI). The spectral binning scheme, which determines the accuracy and efficiency of the PCA–RT performance, is thoroughly optimized over the spectral range 265 to 360 nm with the assumption of a Rayleigh-scattering atmosphere above a Lambertian surface. The high level of accuracy (~ 0.03 %) is achieved from fast-PCA calculations of full radiances. In this approach, computationally expensive full multiple scattering (MS) calculations are limited to a small set of PCA-derived optical states, while fast single scattering and 2-stream multiple scattering calculations are performed, for every spectral point. The number of calls to the full MS model is only 51 in the application to OMI ozone profile retrievals with the fitting window of 270–330 nm where the RT model should be called at fine intervals (~ 0.03 nm with ~ 2000 wavelengths) to simulate OMI native measurements at 229 wavelengths (spectral resolution: 0.4–0.6 nm). We also developed a Look Up Table (LUT) to correct RT approximations performed using a scalar RT model with 4 streams (discrete ordinates) and 24 layers, thereby achieving the accuracy at the level attainable from simulations with a vector model with 12 streams and 72 layers; this speeds up the RT calculations by more than 2 orders of magnitude when ignoring other overhead. Overall, we speed up our OMI retrieval by a factor of 3.3 over the previous version, which has already been significantly sped up over line-by-line calculations due to various RT approximations. Improved treatments for RT approximation errors using LUT corrections improve spectral fitting (2–5 %) and hence retrieval errors, especially for tropospheric ozone by up to ~ 10 %; the remaining errors due to the forward model errors are within 5 % in the troposphere and 3 % in the stratosphere.

show abstract

A neural network radiative transfer model approach applied to the Tropospheric Monitoring Instrument aerosol height algorithm

Cited by 29 publications

References 36 publications

A first comparison of TROPOMI aerosol layer height (ALH) to CALIOP data

A first comparison of TROPOMI aerosol layer height (ALH) to CALIOP data

Effects of clouds on the UV Absorbing Aerosol Index from TROPOMI

Radiative transfer acceleration based on the Principal Component Analysis and Look-Up Table of corrections: Optimization and application to UV ozone profile retrievals

Contact Info

Product

Resources

About