To evaluate the performance of the eXtensible Bremen Aerosol/cloud and surfacE
parameters Retrieval (XBAER) algorithm, presented in the Part 1 companion paper to this paper, we apply the XBAER algorithm to the Sea and Land
Surface Temperature Radiometer (SLSTR) instrument on board Sentinel-3. Snow
properties – snow grain size (SGS), snow particle shape (SPS) and specific
surface area (SSA) – are derived under cloud-free conditions. XBAER-derived
snow properties are compared to other existing satellite products and
validated by ground-based and aircraft measurements. The atmospheric
correction is performed on SLSTR for cloud-free scenarios using Modern-Era
Retrospective Analysis for Research and Applications (MERRA) aerosol optical
thickness (AOT) and the aerosol typing strategy according to the standard XBAER
algorithm. The optimal SGS and SPS are estimated iteratively utilizing a
look-up-table (LUT) approach, minimizing the difference between
SLSTR-observed and SCIATRAN-simulated surface directional reflectances at
0.55 and 1.6
The comparison with aircraft measurements, during the Polar Airborne
Measurements and Arctic Regional Climate Model Simulation Project
(PAMARCMiP) campaign held in March 2018, also shows good agreement (with
Change in snow properties is both a consequence and a driver of climate change (Barnett et al., 2005). Snow cover and snow season, especially in the Northern Hemisphere, are reported by different models to decrease due to climate change (Liston and Hiemstra, 2011). The reduction in snow cover leads to a change in the surface energy budget (Cohen and Rind, 1991; Henderson et al., 2018), a reduction in Asian summer rainfall (Liu and Yanai, 2002; Zhang et al., 2019), a loss of Arctic plant species (Phoenix, 2018), and other impacts on societies and ecosystems (Bokhorst et al., 2016). Snow may influence the climate through both direct and indirect feedbacks (Lemke et al., 2007). The direct feedback is the snow–albedo feedback, and the indirect feedbacks involve atmospheric circulation. The snow–albedo feedback describes the mechanism by which melting snow (the absence of snow cover), caused by global warming, reflects less solar radiation and further enhances the warming (Thackeray and Fletcher, 2016). The snow indirect feedbacks describe the impact of snow property change on monsoonal and annual atmospheric circulation (Lemke et al., 2007; Gastineau et al., 2017). However, the snow cover may be declining even faster than thought due to large uncertainties in how models describe the snow feedback mechanisms (Flanner et al., 2011). The uncertainties in describing the snow feedback mechanisms are largely introduced by the uncertainties in knowledge of snow properties (Hansen et al., 1984; Groot Zwaaftink et al., 2011; Sarangi et al., 2019). Snow properties depend on snow age, moisture, and surrounding temperatures (LaChapelle, 1969; Sokratov and Kazakov, 2012).
Model simulations and field-based measurements provide valuable information
of snow properties (e.g., snow grain size (SGS), snow particle shape (SPS),
specific surface area (SSA)) for the understanding of changing snow and its
corresponding impact on climate change. Satellite observations offer another
effective way to derive those snow properties on a large scale with a high
quality (e.g., Painter et al., 2003, 2009; Stamnes et al., 2007; Lyapustin et
al., 2009; Wiebe et al., 2013). The similarities and differences in the
required snow parameters and their accuracy between the snow remote sensing
community and other communities (e.g., field measurement community) are
discussed in detail in the Part 1 companion paper (Mei et al., 2021a). In
this paper, SGS (effective radius) is defined as
Different retrieval algorithms to derive SGS have been developed for different instruments. The Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and Thematic Mapper (TM) on board Landsat are pioneer instruments used for the retrieval of SGS (Hyvarinen and Lammasniemi, 1987; Li et al., 2001). Painter et al. (2003, 2009) retrieved SGS using AVIRIS and Moderate Resolution Imaging Spectroradiometer (MODIS) data, exploring the information from both visible and near-infrared spectral channels. There are several available satellite SGS products for MODIS (Klein and Stroeve, 2002; Painter et al., 2009; Rittger et al., 2013) and its successor, the Visible Infrared Imaging Radiometer Suite (VIIRS) (Key et al., 2013). For instance, the MODIS Snow-Covered Area and Grain size (MODSCAG) product is created utilizing a spectral mixture analysis method based on the prescribed endmember. The endmember is a spectrum library for snow, vegetation, rock, and soil (Painter et al., 2009). The MODSCAG algorithm can provide the snow cover fraction and snow albedo besides SGS on a pixel base. Topographic effects in MODSCAG are not considered, and the MODSCAG product tends to overestimate SGS (Mary et al., 2013). Other retrieval algorithms have also been designed for and tested on the MODIS instrument (Stamnes et al., 2007; Aoki et al., 2007; Hori et al., 2007). Jin et al. (2008) retrieved SGS over the Antarctic continent using MODIS data based on an atmosphere–snow coupling radiative transfer model. Lyapustin et al. (2009) proposed a fast retrieval algorithm for SGS at a 1 km spatial resolution using MODIS observations. The algorithm is based on an analytical asymptotic radiative transfer model. Negi and Kokhanovsky (2011) proposed the use of the asymptotic radiative transfer (ART) theory to retrieve SGS. The retrieved snow albedo and grain size from Negi and Kokhanovsky (2011) were validated and showed good accuracy for clean and dry snow. However, potential problems have been reported for dirty snow (e.g., soot and/or dust contamination). The Snow Grain Size and Pollution (SGSP) algorithm retrieves SGS and pollution amount based on a snow model (Zege et al., 1998), without a priori assumptions about SPS (Zege et al., 2011). The SGSP algorithm has been validated using in situ measurements over central Antarctica, and an underestimation of SGSP-derived SGS was reported under a large solar zenith angle (Zege et al., 2011; Carlsen et al., 2017). The algorithm is currently implemented for the MODIS instrument and provides operational daily snow products (Wiebe et al., 2011). New instruments such as Hyperion on board Earth Observing-1 (EO-1) and OLCI have also been used to derive SGS (Zhao et al., 2013; Kokhanovsky et al., 2019). The algorithm proposed by Kokhanovsky et al. (2019) is conceptually based on an analytical ART model, which estimates snow reflectance by the given SGS and ice absorption (Kokhaovksy et al., 2018). The snow grains in the ART model are described as a fractal.
Snow particle shape is a fundamental parameter needed to describe snow
properties (Räisänen et al., 2017). The SPS keeps relatively stable
before falling on the ground under cold and dry conditions, while it has
large variabilities under warm and wet conditions (Dang et al., 2016).
A few attempts have been proposed to retrieve SSA from spaceborne
observations. The retrieval of SSA is actually performed based on the
pre-retrieved SGS with an assumption of a known SPS. Mary et al. (2013) retrieved SSA over mountain regions using MODIS data, assuming a
spherical ice crystal shape. The algorithm performs a topographic correction
for the surface reflectance to achieve a better retrieval accuracy. The
overall difference, compared to field measurements, is 9.4 m
This paper, as in the Part 1 companion paper, applies the eXtensible Bremen Aerosol/cloud and surfacE
parameters Retrieval (XBAER) algorithm to the
Sea and Land Surface Temperature Radiometer (SLSTR) on board Sentinel-3 to
derive SGS, SPS and SSA. The general concept is to use the channels, which
are sensitive to SGS and SPS, simultaneously. The channels used in XBAER
algorithms are 0.55 and 1.6
As mentioned in the Part 1 companion paper, the nine SPSs of Yang et al. (2013) used in the XBAER algorithm are proven to be a new option to describe the ice crystal local optical properties for the snow community (e.g., Saito et al., 2019; Pohl et al., 2020; Mei et al., 2021b), and we would also like to emphasize several more points to avoid misunderstandings between different scientific communities.
This paper is structured as follows: instrument characteristics of SLSTR and the field-based measurements and aircraft measurements used for validation are described in Sect. 2. Section 3 describes the method including cloud screening, atmospheric correction and the flowchart of the XBAER algorithm. Some selected data products and comparisons with MODIS products and field-based measurements are shown in Sect. 4. The comparison with the recent campaign measurement is presented in Sect. 5. A discussion to illustrate a time series of the retrieval results is shown in Sect. 6. The conclusions are given in Sect. 7.
After the loss of Environmental Satellite (Envisat) on 12 April 2012, the
European Space Agency (ESA) launched Sentinel-3A and Sentinel-3B in February 2016 and April 2018, respectively. As the successor of Advanced Along-Track
Scanning Radiometer (AATSR) on board Envisat, Sentinel satellites take the
SLSTR instrument. The SLSTR instrument has similar characteristics to AATSR (see Table 1 for details). The instrument has nine
spectral bands in the visible and infrared spectral range. It also has
dual-view observation capability with swath widths of 1420 and 750 km for
nadir and oblique directions, respectively. The SLSTR and AATSR dual-view
observations of the Earth's surface make surface bidirectional reflectance
distribution function (BRDF) effect estimation possible, which is widely
used to retrieve both surface and atmospheric geophysical parameters (Popp
et al., 2016). Besides the heritage of AATSR, some new features (wider
swath, new spectral bands and higher spectral resolution for certain bands)
have been included in SLSTR instrument
(
Instrument characteristics of AATSR and SLSTR.
The validation of satellite-derived snow properties is challenging due to (i) limited available field-based measurements and (ii) the difficulties of spatial–temporal collocation between satellite observations and field-based measurements because of cloud coverage. This paper focuses on the Sentinel-3a satellite for the periods of February 2016 (launch month of Sentinel-3a) and December 2020. The field-based measurements from both permanent sites and campaign sites for the focal time period are collected. Figure 1 shows the geographic distribution of the validation sites. The site names used in this paper are listed near each site. Since XBAER retrieves SGS, SPS and SSA simultaneously, the SnowEx campaign, which provides the three parameters as well, will be introduced in detail first.
Geographic distribution of the validation sites. The colors represent the type of each site, and the site name used in this paper is indicated near each site.
The National Aeronautics and Space Administration (NASA) established a terrestrial hydrology program (SnowEx mission) in order
to better quantify the amount of water stored in snow-covered regions (Kim
et al., 2017). The measurements for the first year (2016–2017) were
carried out during February 2017 (between 8 and 25 February 2017) at Grand Mesa and the Senator Beck Basin in Colorado (hereafter referred to
as SnowEx17) (see Fig. 2a) (Elder et al., 2018). Grand Mesa is a forest
region covered by relatively homogeneous snow cover with an area size
similar to airborne instrument swath widths (Brucker et al., 2017) (see Fig. 2c). The Senator Beck Basin site has complex topography and is covered by
snow. The campaign used more than 30 remote sensing instruments, and most of
the instruments are from the NASA except some instruments such as ESA's radar (Kim et al., 2017).
The snow pit measurements provide information on snow grain size and
type/shape, stratigraphy profiles, and temperatures with certain information
about surface conditions (e.g., snow roughness) (Rutter et al., 2018). The
SnowEx17 campaign provides seven different shapes (new snow, rounds, facets,
mixed forms, melt–freeze, crust and ice lens). Table 2 lists both the
SnowEx17-measured snow grain shapes and SPSs defined in Yang et al. (2013).
The SPSs defined by
Photos taken during the SnowEx campaign.
The measurements over Greenland are obtained by the EastGRIP team over
75.63
The SSA measurements at Nunavut, northern Canada (69.20
The SPS and SSA measurements around Inuvik, Northwest Territories of Canada
(68.73
The SSA measurements above the French Alps (45.04
The SGS measurements were obtained over Nagaoka, Japan (37.41
The SGS measurements were obtained over Xinjiang province during a different
period (Chen et al., 2020); the dataset around the site (44.146
The SSA measurements at Dome C (75
Snow grain type (shape) provided by Yang et al. (2013), in situ
measurements in the SnowEx campaign and by
During the Polar Airborne Measurements and Arctic Regional Climate Model
Simulation Project (PAMARCMiP) campaign held in March and April 2018,
ground-based and airborne observations of surface, cloud and aerosol
properties were performed near the Villum Research Station (North
Greenland). One of the most important objectives of the PAMARCMiP 2018
campaign was to quantify the physical and optical properties of snow, sea
ice and the atmosphere (Egerer et al., 2019; Nakoudi et al., 2020). Airborne
spectral irradiance measurements by the Spectral Modular Airborne Radiation
Measurement System (SMART) on board the
The algorithm synergistically uses SLSTR and OLCI data to identify clouds
over the snow surface. The criteria for cloud screening over snow using
SLSTR and OLCI measurements can be found in Istomina et al. (2010) and Mei
et al. (2017), respectively. Short summaries of Istomina et al. (2010) and
Mei et al. (2017) are presented below, and more details can be found in the
original publications. The algorithm proposed by Istomina et al. (2010) for
the SLSTR instrument utilizes spectral behavior differences at SLSTR visible
and thermal infrared channels, and this algorithm was updated later by
Jafariserajehlou et al. (2019). Relative thresholds are determined based on
radiative transfer simulations under various atmospheric and surface
conditions. The method proposed by Mei et al. (2017b) for the OLCI instrument
uses different cloud characteristics: cloud brightness, cloud height and
cloud homogeneity. The TOA reflectance at 0.412
Due to the low atmospheric aerosol loading over the Arctic snow-covered
regions (e.g., Greenland), atmospheric correction using path radiance
representation (Chandrasekhar, 1950; Kaufman et al., 1997) can provide
accurate estimation of surface reflection even under relatively large SZAs
(Lyapustin, 1999). The TOA reflectance at selected channels (0.55 and 1.6
The theoretical background of the retrieval algorithm is given in Sect. 4 of the companion paper. The XBAER algorithm consists of three stages to derive SGS, SPS and SSA: (1) derivation of SGSs for each predefined SPS, (2) selection of the optimal SGS and SPS pairs for each scenario, and (3) calculation of SSA for each retrieved SGS and SPS. This section describes some implementation details such as the selection of the first guess for the retrieval parameters and the flowchart of the algorithm.
A reasonable first-guess value for the iteration process can significantly
reduce the computation time, which is important for retrievals of
atmospheric and surface properties over large geographic and temporal scales
with different instrument spatial resolutions. The first guess of SGS in the
XBAER algorithm is obtained employing the semi-analytical snow reflectance
model (Kokhanovsky and Zege, 2004; Kokhanovsky et al., 2018). Details of
using this model to derive SGS can be found in Lyapustin et al. (2009). Due
to the different band settings in MODIS and SLSTR (SLSTR has no 2.1
Figure 3 shows the flowchart of how XBAER derives SGS, SPS and SSA. The flowchart includes pre-processing of cloud screening using the synergy of OLCI and SLSTR and the atmospheric correction using MERRA providing AOT and a weakly absorbing aerosol type. The SGS and SPS are obtained using the LUT-based minimization routine. SSA is then calculated using the retrieved SGS and SPS.
Flowchart of the XBAER retrieval algorithm.
Greenland is the largest ice-covered land mass in the Northern Hemisphere
and the biggest cryospheric contributor to the global sea-level rise (Ryan
et al., 2019). XBAER-derived SGS, SPS and SSA over Greenland enable a good
understanding of the retrieval accuracy with a large and representative
geographic scale. Kokhanovsky et al. (2019) reported that July is an
optimal month to analyze satellite-derived snow properties over Greenland
because Greenland has a strong snow particle metamorphism process (SPMP) due
to higher temperatures in July (Nakamura et al., 2001). The SPMP, affected
strongly by temperature, is a dominant factor for the variabilities in SGS,
SPS and SSA (LaChapelle, 1969; Sokratov and Kazakov, 2012; Saito et al.,
2019). Snow particle size increases dramatically and the ice crystal
particles are compacted in the strong SPMP (Aoki et al., 1999; Nakamura et
al., 2001; Ishimoto et al., 2018).
Figure 4 shows an example of the XBAER-derived SGS on 28 July 2017 from SLSTR,
XBAER first guess, and its comparison with the same scenario from the MODSCAG
product (Painter et al., 2009). Here we chose MODIS on Aqua rather than
MODIS on Terra to avoid the impact of instrument degradation of MODIS on Terra
(Lyapustin et al., 2014). The visualization of XBAER-derived SGS is shown to
be between 10 and 500
Figure 5 shows XBAER-retrieved SGS, SPS and SSA for 28 July 2017. Since there
are no available products of SPS and SSA from MODSCAG, it is a great
challenge to make a similar comparison to that in the case of SGS. Fortunately,
campaign-based and laboratory investigations provide valuable information on
typical snow shapes at different times and locations with a wide range of
atmospheric conditions. According to Kikuchi et al. (2013), the typical SPSs
in the polar regions include column crystal (e.g., solid column, bullet-type
crystal) with SGSs of about 50
The geographic distribution of SSA is somehow anti-correlated with the
geographic distribution of SGS, due to the definition of SSA. Most SSAs fall
into the range of 10–40 m
A comparison of the MODSCAG SGS
XBAER-derived SGS, SPS and SSA over Greenland for the same scenario as in Fig. 4.
In this section, we will quantitatively validate XBAER-derived snow properties with field-measured data and aircraft measurements.
In order to have a quantitative evaluation of XBAER-derived SGS, SPS and SSA, we have collocated the SLSTR observations with recent campaign measurements provided by SnowEx17 and SnowEx20, as described in Sect. 2. Due to overpass time and cloud cover, only limited match-ups between XBAER retrievals and SnowEx17 and SnowEx20 measurements have been obtained. No match-up is obtained for SnowEx20.
Table 3 summarizes match-up information. The first three columns in Table 3
show the observation times and locations (longitude and latitude). The
fourth and fifth columns indicate the cloud conditions. Cloud conditions in
Table 3 are given in three categories: cloud-free snow, cloud-contaminated
snow and cloud-covered snow. These three categories are classified by the
XBAER cloud identification results (see Sect. 3.1) and are illustrated by
the RGB composition figures, covering the SnowEx campaign area, as presented
in Fig. 6. An optically thin cloud over a melting snow layer, a thick cloud
over snow and snow scenarios are presented in Fig. 6a, b and c,
respectively. The cloud optical thickness (COT), estimated using the
independent XBAER cloud retrieval algorithm, as presented in Mei et al. (2018), is
Information of match-ups between SnowEx and SLSTR during February 2017.
Zoom of the RGB composition figures (created using ESA official SLSTR software SNAP) for the three selected dates presented in Table 3. The yellow point indicates the SnowEx instrument position.
Even though the synergistical use of SLSTR and OLCI provides valuable
information for separating cloud and snow, the identification of an optically
thin cloud above a snow layer is a great challenge due to the similar
wavelength dependence of snow and cloud reflectance, especially between snow
and ice cloud (Mei et al., 2020b). The identification of the cloud from an
underlying snow layer in XBAER relies mainly on the O
The RGB composition
Table 4 summarizes the comparison between XBAER retrieval results, the MODSCAG
product and SnowEx17 campaign measurements. The first three columns in
Table 4 are the same as those of Table 3, showing the observation time and locations
(longitude and latitude). The second three columns are the SnowEx17-measured
SGS. Since the SnowEx17 provides the SGS profile up to a 1 m depth, the
minimum (SnowEx_min), average (SnowEx_avg)
and maximum (SnowEx_max) values of SGS are listed in Table 3.
The last two columns are MODSCAG- and XBAER-derived SGS. For the four
cloud-filter-passed match-ups, XBAER-derived SGS shows good agreement with
SnowEx17 measurements, especially for the 22 February. The average
absolute difference is less than 10
An underestimation is found for the first match-up on the 9 February. This is explained by the cirrus cloud contamination as presented
in Fig. 11. According to an independent XBAER cloud retrieval (Mei et al.,
2018), the COT is
The comparison between SnowEx SGS measurements, XBAER- and MODSCAG-retrieved SGS during February 2017.
Table 5 shows the same match-up information as in Table 4 but for SPS. We would like to highlight again that the SPSs proposed by Yang et al. (2013) are used for the radiative transfer calculation. From a single-ice-crystal point of view, those shapes are very unlikely to occur exactly in reality. This is similar to the issue in field measurements. In field-based measurements, a spherical-shape assumption is widely used (e.g., the calculation of SSA from SGS); however, a pure spherical shape is also very unlikely to occur in natural snow. To have a reasonable comparison between satellite-derived SPS and field-measured SPS, the quantitative information of “roundish” or “irregular” shapes from both satellite and field measurement communities may be an option. Under this comparison strategy, a “droxtal” shape derived from satellite observation is somehow identical with a “spherical shape” in field measurement.
The second and third columns in Table 5 show SnowEx17-measured and XBAER-derived SPS. The abbreviations of the SPS are listed in Table 2. The fourth–sixth columns are the temperature, wetness of snow and the comments provided by campaign participants, respectively. Previous publications show that ice cloud and fresh snow are best described by aggregate of 8 columns (Platnick et al., 2017; Järvinen et al., 2018). Both 9 and 11 February are retrieved to be aggregate of 8 columns because both of them are affected by ice cloud. The first sample on 22 February is reported to be aggregate of 8 columns and the observation of SnowEx17 is fresh snow. The SPS of the second sample on 22 February is “facet” while XBAER says droxtal, indicating possible linkage between XBAER-derived droxtal and field-measured facet. It is interesting to compare the SPS for the third sample on 22 February. The SPSs are round and aggregate of 8 columns for the SnowEx17 measurement and XBAER retrieval, respectively. The atmospheric condition is reported to be “windy”, and the snow layer is wind-affected and not very well banded ice crystal. The ice crystal shape in blowing snow is likely to be irregular and aggregated (Lawson et al., 2006; Fang and Pomeroy, 2009; Beck et al., 2018), which is strongly affected by the near-surface processes (Beck et al., 2018). Snow grains may also become rounded due to sublimation in blowing snow (Domine et al., 2009). The wind blowing snow may be well-represented optically by an “aggregate-of-8-columns” shape, as retrieved by XBAER.
The comparison between SnowEx snow grain shape and XBAER-retrieved SGP during February 2017.
Table 6 shows the comparison of SSA. For the three cloud-free samples, the
difference in XBAER-derived SSA and SnowEx17-measured SSA is 2.7 m
The comparison between SnowEx SSA and XBAER-retrieved SSA during February 2017.
The above validation for the retrieval of SGS, SPS and SSA using the XBAER algorithm, although with limited samples, indicates the consistency of the sensitivity study from the Part 1 companion paper and the retrieval results in Part 2, as presented in this section.
For comprehensive validation, we have analyzed the rest of the sites
besides the SnowEx site. The comparison is performed based on the daily mean
observation following the method from Wiebe et al. (2011). We have
restricted the SGS in the range of 0–300
The potential linkage between XBAER-derived SPS and field-measured SPS is
also presented in Fig. 8. This is named SPS similarity in this
paper. The SPS similarity is defined as the ratio of the match-up number
for a given SPS pair (XBAER-retrieved SGS from Yang et al., 2013, field-measured
Validation of XBAER-derived SGS, SPS and SSA. The upper panels show
the scatterplots for SGS and SSA, while the lower panel shows the
relationship of SPS between XBAER and
Figures 9 and 10 show the time series of SGS and SSA over each site. We
can see that sites Greenland and Antarctica provide most of the match-ups.
Both SGS and SSA show good agreement between XBAER-derived and field-measured values over these two sites. For SGS, the correlation coefficients
are 0.85 and 0.89 and the RMSEs are 14 and 9
Time series of XBAER-derived and field-measured SGS for each site. The
match-ups for SGS are distinguished by the AOT values. The correlation
coefficient (
Time series of XBAER-derived and field-measured SSA for each site.
The match-ups for SGS are distinguished by the AOT values. The correlation
coefficient (
The optical snow grain size over Arctic sea ice was derived from airborne
SMART measurements as described in Sect. 2.3. Figure 11a shows the
retrieved grain size along the flight track (black-encircled area) taken on
26 March 2018 between 12:00 and 14:00 UTC north of Greenland. During this period
of cloudless conditions, a Sentinel-3 overpass (12:29 UTC) delivered SGS data
based on the XBAER algorithm as displayed in the background of this map with
a 1 km spatial resolution. In general, lower SGSs were observed by both methods
in the vicinity of Greenland, while in particular in the northeast region
of the map (dashed red circle in Fig. 11a) SGS values of up to 350
The SGS retrieval based on the algorithm suggested by Zege et al. (2011) and
Carlsen et al. (2017) gives the optical radius of the snow grains such that
the SSA can be derived applying Eq. (A1) from the companion paper. The map of
the SSA (Fig. 11c) reflects a similar pattern to that observed for the SGS,
showing an inverse behavior to that depicted in Fig. 11a. On average, XBAER (mean SSA
24
Since XBAER is also designed to support the MOSAiC campaign on an Arctic-wide scale (Mei et al., 2020c), it is important to have an overview of how snow properties look on an Arctic-wide scale for the existing campaign. Figure 12 shows the SGS, SPS and SSA geographic distribution over the whole Arctic for 26 March 2018. Northern Greenland, North America and central Russia show large snow particles, especially over North America. And the SPS shows more diversity in lower latitudes compared to the central Arctic, indicating a stronger SPMP. An aggregated shape such as aggregate of 8 columns is the dominant shape in the central Arctic, while column is one of the dominant shapes in lower latitudes. SSA shows large values in the lower-latitude Arctic (northern Canada, southern Greenland, western Norway, southern Finland, northern Russia), while the values are smaller in the central Arctic.
The distribution of XBAER-derived SGS, SPS and SSA over the whole Arctic for 26 March 2018.
XBAER-derived SGS, SPS and SSA over Greenland during 27–30 July 2017.
Wind direction (referenced to the north) and wind speed (unit: m/s) over Greenland during 27–30 July 2017 (data from ECMWF).
The above analysis shows the promising quality of XBAER-derived SGS, SPS and
SSA results. The XBAER-retrieved SGS, SPS and SSA can be used to understand
the change in snow properties temporally. Even though the snow metamorphism
depends on the environmental conditions, Aoki et al. (2000) and Saito et al. (2019) pointed out that a 4 d timescale is a reasonable time span to see
the temporal change in snow properties. Figure 13 shows XBAER-derived SGS
(upper panels), SPS (middle panels) and SSA (lower panels) over Greenland
during 27–30 July 2017. Large variability for SGS, SPS and SSA can be
seen during these 4 d, indicating the impacts of snow metamorphism on
the snow properties. Figure 13 shows the snow melting process in both western and
northeastern parts of Greenland, especially on 28 July. The strong
melting in July over Greenland has also been reported by Lyapustin et al. (2009). The SPS over the southeastern part of Greenland becomes smaller during those
4 d. No snowfall was reported according to the relevant Polar Portal report
(
SGS, SPS and SSA are three important parameters to describe snow properties. SGS, SPS and SSA all play important roles in the changes in snow albedo/reflectance and further impact the atmospheric and energy exchange processes. A better knowledge of SGS, SPS and SSA can provide more accurate information to describe the impact of snow on Arctic amplification processes. The information about SGS, SPS and SSA may also be used to explore new applications to understand atmospheric conditions (e.g., aerosol loading). Although some previous attempts (e.g., Lyapustin et al., 2009) show the capabilities of using passive remote sensing to derive SGS over a large scale, no publications have been found that derive SGS, SPS and SSA simultaneously. This is the first paper, to the best of our knowledge, attempting to retrieve SGS, SPS and SSA using passive remote sensing observations.
The new algorithm is designed within the framework of the XBAER algorithm. The XBAER algorithm has been applied to derive SGS, SPS and SSA using the newly launched SLSTR instrument on board the Sentinel-3 satellite. The cloud screening is performed with a synergistical technique using both OLCI and SLSTR measurements. The synergistical cloud screening in XBAER is easily implementable and effectively runnable on a global scale, with a high quality, which enables a cloud-contamination-minimized SGS, SPS and SSA retrieval using passive remote sensing.
Besides the cloud screening, another pre-process is the atmospheric
correction. Aerosols have a non-ignorable impact on the retrieval of SGS,
SPS and SSA, even over the Arctic regions, where aerosol loading is small
(AOT at 0.55
The SGS, SPS and SSA retrieval algorithm is based on the publication by Yang
et al. (2013), in which a database of optical properties for nine typical ice
crystal shapes are provided. Previous publications show that this database
can be used to retrieve ice crystal properties in both ice cloud and snow
layers (e.g., Järvinen et al., 2018; Saito et al., 2019). The algorithm is
a LUT-based approach, in which the minimization is achieved by the
comparison between atmosphere-corrected TOA reflectance at 0.55 and 1.6
The comparison between XBAER-derived SGS, SPS and SSA shows good agreement
with the SnowEx17 campaign measurements. The average absolute and relative
difference between XBAER-derived SGS and SnowEx17-measured SGS is about 10
Although the presented version of the XBAER retrieval algorithm shows promising results, we see at least four possibilities for improving its accuracy. Potential cloud contamination may still occur according to the analysis, exploiting the time-series technique, as described in Jafariserajehlou et al. (2019). Currently only a single-ice-crystal shape is used in the retrieval; the mixture of different ice crystal shapes, i.e., the snow grain habit mixture model (e.g., Saito et al., 2019), will be tested in further work. Another potential improvement may be linked to the use of polydisperse ice crystals (e.g., gamma distribution). The potential impacts of the vertical structure of SGS and SPS also need to be investigated in the future.
XBAER-derived SGS, SPS and SSA will be used to support the analysis of the MOSAiC expedition and other campaign-based measurements (Jäkel et al., 2021).
The data over the Antarctica site are provided by Ghislain Picard. The data
over the Greenland site are provided by Hans Christian Steen-Larsen. The
data over the Canada-Alex site are provided by Alexandre Langlois. The
data over the Canada-Josh site are provided by Joshua King. The data over the
China site are provided by Tao Che. The data over the Japan site are available
at
LM and VR conceptualized the study, LM implemented the code and processed the data. LM, VR EJ, XC analyzed the data. LM prepared the manuscript with contribution from all co-authors. LM, VR, MV and JPB polished the whole manuscript.
The authors declare that they have no conflict of interest.
This research was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project ID 268020496 – TRR 172. The comments by Alexandre Langlois, Ghislain Picard and the anonymous reviewer helped to improve the quality of the manuscript significantly. The authors highly appreciate the effort from Adam Povey (University of Oxford) to help to deal with the huge number of SLSTR L1 data. The authors would like to thank Knut von Salzen from Environment and Climate Change Canada for valuable discussion. We thank the support from Lisa Booker from the National Snow and Ice Data Center, Boulder, in understanding the SnowEx17 campaign data. We thank Alexander Kokhanovsky from Vitrociset, Darmstadt, Germany, and Jason E. Box from the Geological Survey of Denmark and Greenland (GEUS) for valuable discussion. We thank Salguero Jaime for providing the MODSCAG snow products. The MODIS snow product data are provided by MODSCAG team, and SLSTR–OLCI data are provided by ESA. The Zoom meeting and valuable discussion with Joshua King are highly appreciated.
This research has been supported by the Deutsche Forschungsgemeinschaft (grant no. 268020496 – TRR 172).The article processing charges for this open-access publication were covered by the University of Bremen.
This paper was edited by Alexandre Langlois and reviewed by Ghislain Picard and one anonymous referee.