The Ozone Mapping and Profiler Suite Limb Profiler (OMPS/LP) has been flying
on the Suomi National Polar-orbiting Partnership (S-NPP) satellite since
October 2011. It is designed to produce ozone and aerosol vertical profiles
at

Accurate estimation of stratospheric aerosol is important because aerosols in the stratosphere have an important influence on climate variability through their contribution to direct radiative forcing, although the magnitude of this term is still uncertain (Ridley et al., 2014). Aerosols also play an important role in the chemical and dynamic processes related to ozone destruction in the stratosphere. Therefore, long-term measurement of the distribution of aerosols is necessary for a better understanding of stratospheric processes.

The Ozone Mapping and Profiler Suite Limb Profiler (OMPS/LP) is one of three OMPS instruments on board the Suomi National Polar-orbiting Partnership (S-NPP) satellite (Flynn et al., 2007). S-NPP was launched in October 2011, into a sun-synchronous polar orbit. The local time of the ascending node of the S-NPP orbit is 13:30. The LP instrument collects limb-scattered radiance data and solar irradiance data on a 2-D charge-coupled device (CCD) array over a wide spectral range (290–1000 nm) and a wide vertical range (0–80 km) through three parallel vertical slits. These spectra are primarily used to retrieve vertical profiles of ozone (Rault and Loughman, 2013; Kramarova et al., 2018), aerosol extinction coefficient (Loughman at al., 2018; Chen et al., 2018) and cloud-top height (Chen et al., 2016). More details about the OMPS/LP instrument design and capabilities are provided in Jaross et al. (2014).

Instruments that measure scattered radiation need to assume some form of aerosol size distribution (ASD) to convert their measured information into aerosol extinction. These instruments include limb-scattered instruments such as the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) (von Savigny et al., 2015; Malinina et al., 2018), Optical Spectrograph and Infrared Imaging System (OSIRIS) (Bourassa et al., 2008, 2012; Rieger et al., 2014, 2018), OMPS/LP (Loughman at al., 2018; Chen et al., 2018), and space- and ground-based lidars such as Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) (Winker et al., 2009). By contrast, instruments that employ solar, lunar and stellar occultation techniques such as SAGE II (Chu et al., 1989), SAGE III (Thomason et al., 2010) and Global Ozone Monitoring by Occultation of Stars (GOMOS) (Bertaux et al., 2010) can derive extinction directly from their transmission measurements without assuming an ASD.

In this study, we determine a new ASD by calculating a fit to results produced by the Community Aerosol and Radiation Model for Atmospheres (CARMA; Colarco et al., 2003, 2014) in order to improve the accuracy of aerosol extinction profiles retrieved from OMPS/LP measurements. The revised ASD is used in the new V1.5 OMPS/LP aerosol extinction retrieval algorithm, which demonstrates better performance in internal validation tests (e.g., absolute difference and spectral dependence of calculated radiances vs. measurements) compared to the V1.0 OMPS/LP algorithm. We also validate the revised ASD through comparisons to independent satellite retrievals of aerosol extinction from Stratospheric Aerosol and Gas Experiment on the International Space Station (SAGE III/ISS) solar occultation measurements. This work extends the previous results shown in Chen et al. (2018) to provide improved validation of the LP V1.5 aerosol extinction product.

The original LP aerosol extinction retrieval algorithm described in Rault and
Loughman (2013) uses radiance data measured at multiple wavelengths in the
visible and near-infrared spectral region. The updated V1.0 algorithm
described in detail by Loughman et al. (2018) is based on Mie theory, using
radiances from one wavelength at 675 nm. We briefly review the design of
this algorithm here. The aerosol extinction profiles are retrieved from
limb-scatter observations using the aerosol scattering index (ASI) as the
measurement vector. The ASI is the fractional difference between a given
radiance and the calculated radiance assuming a pure Rayleigh atmosphere
bounded by a Lambertian surface. This quantity is roughly proportional to
aerosol extinction, as described in Loughman et al. (2018), and is defined at
wavelength

Size distributions used in several recent limb scattering aerosol extinction retrieval algorithms.

Assuming that optically thin conditions prevail, the radiance sensitivity is
approximately proportional to the change in aerosol extinction

The primary change introduced for the LP V1.5 aerosol retrieval algorithm is the revised particle size distribution described in Sect. 3. Other changes with less impact on the retrieved extinction values include the use of vector radiative transfer calculations and the implementation of intra-orbit tangent height adjustments as described by Moy et al. (2017). In addition, the V1.0 retrievals only allowed a factor-of-2 change in extinction for each iteration and executed three iterations, rather than the larger values (factor-of-5 change, four iterations) given in Loughman et al. (2018). Based on inspection of test results, we revised those parameters for the V1.5 algorithm to allow a factor-of-3 change in extinction for each iteration and four iterations of the retrieval.

Retrieval of aerosol extinction profiles from limb scattering measurements
requires the specification of an aerosol size distribution (ASD) to
represent the microphysical properties of the aerosol particles. Different
functional forms can be selected to represent the ASD. The V1.0 LP aerosol
algorithm retrieves extinction profiles by assuming a bimodal lognormal size
distribution (BD):

The main motivation for using a bimodal size distribution arose from the
desire to make comparisons with the existing in situ optical particle
counter (OPC) data set, which generally features a bimodal size distribution
at the altitudes where the stratospheric aerosol extinction is greatest
(Deshler et al., 2003). However, specifying the five independent parameters (two
mode radii, two mode widths and the coarse-mode fraction) needed to define
this more complex distribution can be challenging. Most OPC measurements have
no independent information at radii less than 0.1

To select an alternate ASD for use in LP retrievals, we have used results
from CARMA. CARMA is a sectional aerosol and cloud microphysics model that has been used to
study a wide variety of problems in planetary atmospheres (Toon et al., 1979,
1988; Turco et al., 1979; Bardeen et al., 2008; Colarco et al., 2003, 2014;
English et al., 2011, 2012; Yu et al., 2015). The CARMA model is coupled here
to the NASA Goddard Earth Observing System (GEOS) Earth system model, a
three-dimensional atmospheric general circulation model, as described in
Colarco et al. (2014), and provides simulated aerosol distributions over a
full range of latitude and longitude, altitude, and season. Colarco et
al. (2014) describes how CARMA was implemented initially for dust and sea
salt. The usage in GEOS for sulfate aerosols is a relatively new capability,
with the sulfur chemistry mechanism and aerosol microphysics as in English et
al. (2011) and as described and evaluated in Aquila et al. (2018). The
particle size distribution is represented by 22 size bins covering a wide
range of radii from 0.000267 to 2.79

The aerosol optical properties can be directly calculated based on a
radiative transfer model for each bin of the CARMA size distribution. However, the Mie
calculation in the current LP aerosol code requires an analytic aerosol mode,
rather than bin data. In this study, we use an analytical model of aerosol
particle size distribution which deals with the ASD as a mean of size
spectrum to accurately fit a cumulative distribution function (CDF) on the
binned data using Deshler's method (Deshler et al., 2003). We have chosen the
gamma distribution (GD; e.g., Chylek et al., 1992) to describe the size
distribution of aerosols for OMPS/LP retrievals (Chen et al., 2018). This
function is described in Eq. (6).

We created a subset of CARMA results that is approximately consistent with
the Deshler et al. (2003) long-term measurements by averaging
June–July–August model results to create a climatology, then extracting
aerosol size distribution values for the approximate location
(41

Comparison between gamma size distribution (GD), bimodal lognormal
size distribution (BD) and unimodal normal distribution (UD).

Figure 1b compares the derived differential size distributions from the three
fits, which are plotted as

Limb scattering measurements from other satellite instruments have also been
used to retrieve stratospheric aerosol extinction profiles, with their own
choices of particle size distribution. Bourassa et al. (2012) used a unimodal
size distribution based on Deshler et al. (2003) OPC data to retrieve aerosol
extinction from OSIRIS radiance data at 750 nm. Rieger et al. (2014) not only used
the same function and initial parameters but also investigated the addition
of 1530 nm radiance data to enable the simultaneous retrieval of extinction
and mode radius. Rieger et al. (2018) evaluated the errors associated with
both unimodal and bimodal size distributions in the OSIRIS and SCIAMACHY
retrieval algorithms. Von Savigny et al. (2015) used a unimodal size
distribution based on different OPC data (Deshler, 2008) to retrieve aerosol
extinction from SCIAMACHY radiance data at 750 nm, although the radiance
data were not normalized with 470 nm radiance data. Malinina et al. (2018)
used an alternate approach with SCIAMACHY data in which radiances at seven
wavelengths between 750 and 1530 nm were included and the number density
profile was held constant. This allowed the simultaneous retrieval of mode
radius and distribution width for cloud-free observations at tropical
latitudes (20

Phase functions at the 675 nm wavelength derived from the aerosol
size distributions listed in Table 1, including OMPS V1.0 (green), GD (blue),
OSIRIS (red) and SCIAMACHY (black). The Rayleigh phase function is also shown
as a dashed line for reference. The V1.0

We have selected the gamma size distribution derived from CARMA results in
this work to assess the impact of ASD on stratospheric aerosol extinction
profile retrieval from OMPS/LP limb measurements. The two fitted parameters
(

Figure 3a shows the impact on the gamma distribution

It is important to point out that OMPS/LP measurements cover a wide range of
scattering angles with a well-defined latitude dependence. Figure 4 shows the
variation of

This figure shows how simulated phase function and retrieved
extinction change when

Variation of scattering angle (

Scatter diagrams of retrieved aerosol extinctions for the V1.5 ASD
(blue) and the V1.0 ASD (green) at 25.5 km

In Sect. 3, we described the creation of the gamma aerosol size distribution
model derived from CARMA results. To understand the quality of the present
aerosol size distribution and to estimate the uncertainty associated with the
retrieved aerosol extinction, we first perform the aerosol retrieval code
runs for conditions without a significant volcanic eruption. This provides a
baseline situation. To evaluate the performance of the presented aerosol size
distribution, aerosol extinction profiles were retrieved from OMPS/LP
measurements before and after the Calbuco volcano eruption to see if the
volcanic eruption can be captured by the new model. This eruption occurred in
Chile (41.3

Figure 5 shows scatter diagrams of retrieved aerosol extinctions at 20.5 and
25.5 km as a function of latitude for the V1.5 ASD (blue) and from V1.0 ASD
(green), as well as their ratios (

Scatterplots of extinction ratio (V1.5

Zonal mean of ASI residuals (ASI

Zonal mean of ratio ASI(745 nm)

Figure 5e and f show significantly more variability in extinction ratio of

While the altitude normalization used to construct the ASI measurement vector
in Eq. (1) reduces the effect of DUR in the LP aerosol extinction profile
retrieval considerably, there are second-order effects present that make ASI
sensitive to

In Fig. 7, ASI residuals (difference between the measured ASI and the calculated ASI) from V1.0 and V1.5 retrievals at 20.5 km are plotted as a function of latitude for wavelengths not used in the LP aerosol retrieval (352, 430, 508, 600, 745, 869 nm) for the V1.5 test processing of the 1-year data set. Residuals at the retrieval wavelength (675 nm) are not shown because they are very close to zero for both cases. The residuals produced by the V1.5 ASD are closer to zero than the V1.0 residuals for all wavelengths, indicating that the gamma function ASD more effectively represents the OMPS/LP measurements.

Figure 8 shows the ratio of ASI(745 nm)

SAGE III/ISS solar occultation coverage compared to OMPS/LP coverage. Red: sunrises; black: sunsets; light blue: OMPS/LP.

Time series of individual extinctions at 20.5 km observed by LP
V1.5 (blue), V1.0 (green), SAGE sunrises (red) and SAGE sunsets (black) for
six different

Comparison of zonal mean profiles of collocated SAGE III/ISS (red
solid lines), LP Version 1 (green solid lines) and LP Version 1.5 (blue solid
lines) aerosol extinction profiles for six different

Relative differences of the mean aerosol extinction profiles between
675 nm OMPS/LP and 676 nm SAGE III/ISS. Difference

Correlation plot of SAGE III/ISS vs. OMPS/LP V1.5 (blue) and SAGE
III/ISS vs. OMPS/LP V1.0 (green) zonal mean aerosol extinctions in
10

The Stratospheric Aerosol and Gas Experiment on the International Space
Station (SAGE III/ISS) developed by NASA Langley Research Center (LaRC) was
launched to the International Space Station in February 2017. SAGE III/ISS
provides limb occultation measurements of aerosols and gases in the
stratosphere and upper troposphere (Chu and Veiga, 1998). The SAGE series of
occultation measurements have been extensively evaluated and compared with
other space-based instruments and have been found to have relatively high
precision and accuracy (Bourassa et al., 2012). A general description of the
solar occultation measurement technique is provided by McCormick et
al. (1979). The ISS travels in a low Earth orbit at an altitude of
330–435 km at an inclination of 51.6

Figure 10 shows time series of OMPS/LP and SAGE III/ISS extinctions for six
10

Further indication of the level of agreement between OMPS/LP and SAGE III/ISS
is provided by comparing the zonal average profiles. For this comparison, a
relatively broad collocation requirement of

Figure 11 shows the zonal mean extinction profiles between 15 and 30 km
altitude in

Figure 12 shows relative differences between the mean LP and SAGE profiles
using the same latitude bins shown in Fig. 11. The absolute value of the
relative differences between LP V1.5 and SAGE III is generally

Figure 13 shows a scatterplot of individual zonal mean extinction values
from each data set between 20.5 and 25.5 km, selected for collocation within
10

This paper describes the derivation of a revised aerosol size distribution function to retrieve aerosol extinction profiles from OMPS/LP limb scattering radiance measurements. We use results from the CARMA microphysical model as a basis for the revised ASD to take advantage of CARMA's large range of particle size information. We find that using an ASD based on a gamma function fit (designated V1.5) requires fewer free parameters than our previous choice of a bimodal lognormal ASD (designated V1.0) and is more consistent with the CARMA particle size results. Evaluation of LP observed radiances is complicated by the measurement geometry (typically backward scattering in the SH, forward scattering in the NH) and the corresponding variation in phase function, as well as variations in scene reflectivity. The V1.5 ASD improves the performance of radiance-based retrieval algorithm internal validation tests, including reducing the magnitude of residuals between calculated and measured radiance and spectral dependence.

We also evaluated our revised ASD by comparing V1.5 retrieved extinction
profiles to SAGE III measurements during June–December 2017. Relative
differences between collocated zonal mean profiles are less than 10 %
between 19 and 29 km, with increased differences below 18 km. Regression fits
to all data between 20 and 25 km show a better correlation coefficient between
SAGE III data and LP retrievals with the V1.5 ASD (

The OMPS LP Version 1.5 aerosol data product will be
available through the NASA Goddard Earth Sciences Data and Information
Services Center (GES DISC) at

One of the longest and most comprehensive records of local stratospheric
aerosol conditions comes from the University of Wyoming's optical particle
counters (OPCs) carried on weather balloons at Laramie, Wyoming, USA
(41

An example of this situation is illustrated in Fig. A1, which shows four
bimodal lognormal distribution fits to the same OPC data at 20 km altitude,
all having similar Ångström exponents of approximately 2.4, but each
with a different values of coarse-mode fraction

All four fits to the OPC data are equally good, but they differ from each other
significantly in the radius range between 0.01 and 0.1

Four BD fits to OPC data measured at Laramie, Wyoming, on 12 April 2010 at 20 km.

Estimated bimodal lognormal cumulative distributions

Aerosol phase functions at 675 nm as a function of single scattering angle for the four ASDs listed in Table A1.

The authors declare that they have no conflict of interest.

We thank the OMPS/LP team at NASA Goddard and Science Systems and Applications, Inc. (SSAI) for help in producing the data used in this study. We also would like to thank Tong Zhu for her technical support. Zhong Chen and Matthew DeLand were supported by NASA contract NNG17HP01C. Edited by: Piet Stammes Reviewed by: three anonymous referees