Implications of surface flooding on airborne thickness measurements of snow on sea ice

Snow thickness observations from airborne snow radars, such as the NASA’s Operation IceBridge (OIB) mission, have recently been used in altimeter-derived sea ice thickness estimates, as well as for model parameterization. A number of validation studies comparing airborne and in situ snow thickness measurements have been conducted in the 15 western Arctic Ocean, demonstrating the utility of the airborne data. However, there have been no validation studies in the Atlantic sector of the Arctic. Recent observations in this region suggest a significant and predominant shift towards a snowice regime, caused by deep snow on thin sea ice. During the Norwegian young sea ICE expedition (N-ICE2015) in the area north of Svalbard, a validation study was conducted on March 19, 2015, during which ground truth data were collected during an OIB overflight. Snow and ice thickness measurements were obtained across a two dimensional (2-D) 400 m × 60 20 m grid. Additional snow and ice thickness measurements collected in situ from adjacent ice floes helped to place the measurements obtained at the gridded survey field site into a more regional context. Widespread negative freeboards and flooding of the snow pack were observed during the N-ICE2015 expedition, due to the general situation of thick snow on relatively thin sea ice. These conditions caused brine wicking and saturation into the basal snow layers, causing more diffuse scattering and influenced the airborne radar signal to detect the radar main scattering horizon well above the snow/sea ice 25 interface, resulting in a subsequent underestimation of total snow thickness, if only radar-based information is used. The average airborne snow thickness was 0.16 m thinner than that measured in situ at the 2-D survey field. Regional data within 10 km of the 2-D survey field suggested however a smaller deviation between average airborne and in situ snow thickness, a 0.06 m underestimate in snow thickness by the airborne radar, which is close to the resolution limit of the OIB snow radar system. Our results also show a broad snow thickness distribution, indicating a large spatial variability in snow across the 30 region. Differences between the airborne snow radar and in situ measurements fell within the standard deviation of the in situ data (0.15 – 0.18 m). Our results suggest that, with frequent flooding of the snow-ice interface in specific regions of the Arctic in the future, it may result in an underestimate of snow thickness or an overestimate of ice freeboard, measured https://doi.org/10.5194/tc-2020-168 Preprint. Discussion started: 3 August 2020 c © Author(s) 2020. CC BY 4.0 License.

from radar altimetry, thereby affecting the accuracy of sea ice thickness estimates.

Introduction 35
Snow and sea ice thickness in a changing Arctic climate system are the matter of many recent studies (e.g. Webster et al. 2018), since the snow layer on top of the frozen ocean generates several contradictory effects on the polar climate. On the one hand, in winter, snow acts as an insulator between the relatively warm ocean and the cold atmosphere and hinders the heat exchange between ocean and atmosphere, reducing the sea ice growth rate (Sturm, 2002;Perovich, 2003). On the other hand, in spring and summer, with its high optical albedo in the range of 0.7-0.85, compared to about 0.6 for bare sea ice 40 Maykut, 1977, Perovich, 1996), Snow reflects short-wave radiation and prevents the underlying ice from melting. In addition, snow can be a positive contributor to the sea ice mass balance since snow can transform to snow-ice Merkouriadi et al., 2017a) and superimposed ice (Eicken et al., 2004;Wang et al., 2015).
Besides the importance of snow from a radiative and mass balance perspective, knowledge of snow thickness on sea ice is also required for the accurate retrieval of sea ice thickness from satellite altimetry. The method relies on the assumption that 45 sea ice floating in the ocean is in hydrostatic equilibrium, and sea ice thickness can be calculated by using observations of either ice-freeboard (from radar altimeters) or snow-freeboard (from laser altimeter) and assumptions about the respective densities of snow, ice and water. Ice-and snow-freeboard describe the distances above the local sea level to the snow/sea ice or air/snow interface, respectively. The error budget of the derived ice thickness from laser altimetry is dominated by uncertainties of snow thickness and ice and snow densities, and as well as uncertainties due to remaining 50 errors in the sea surface height (Giles et al., 2007;Kern et al., 2015;Skourup et al., 2017).
Thus, accurate knowledge of snow thickness on sea ice would be helpful to reduce the error in the sea ice thickness calculations and is important for quantifying climatological processes in Polar Regions. The Operation IceBridge (OIB) airborne campaigns (Koenig et al., 2010), which began in 2009, measure snow thickness and surface elevation with an ultra-wideband snow radar (e.g., Yan et al., 2017) and an airborne topographic mapper (ATM) laser altimetry system 55 , respectively. With this constellation of sensors, both the air/snow and the snow/sea ice interface can be detected with the snow radar (e.g., Newman et al., 2014), and the surface elevation can be mapped with the ATM (e.g., Farrell et al., 2012). Hence, the OIB data are a valuable source for validating satellite remote sensing sea ice products as well as for model parameterization. Furthermore, the comparison of airborne OIB data with in situ field measurements is necessary to understand the processes affecting radar penetration into snow covered sea ice and the impact of the snow load 60 on the snow/sea ice interface.
techniques (Kwok, 2014;King et al., 2015). However, in the northern hemisphere, all evaluation studies (except those connecting to satellite data) have thus far focused on snow in the Canada Basin, in the central Arctic Ocean, or only in peripheral sub regions of the Arctic. To our knowledge, no OIB validation study has been conducted in the Atlantic sector of the Arctic. In recent years, a significant change towards thinner ice with thicker snow cover (Renner et al., 2014;Rösel et al., 2018) has been observed in this region, caused by an increase in intense storm events and associated precipitation in 70 this area (Woods and Caballero, 2016;Graham et al., 2017;Rinke et al., 2017). In addition, previous studies indicate that radar signal penetration through the snow pack might be lower under certain geophysical snow-ice conditions in this area (Gerland et al., 2013;King et al., 2018;Nandan et al., 2020) and also in the Weddell Sea ice in the Antarctic (Kwok and Kacimi, 2018). Snow and ice conditions in this region differ to those in the Canada Basin and central Arctic (e.g., Webster et al., 2014;, and they have been found to induce substantial negative freeboards with subsequent flooding of the snow 75 pack, more akin to the conditions in the seasonal ice pack of the Southern Ocean (Massom et al., 2001). This may have an impact on remote sensing methods of snow and ice thickness estimation, which have so far only been validated for more typical Arctic conditions.
In this paper, we present in situ observations of sea ice and snow thickness, and snow and ice characteristics from the N-ICE2015 expedition, alongside near-coincident airborne measurements acquired on 19 March 2015 during an OIB 80 overflight. We calculate ice freeboard values from a variety of sensors to investigate the prevalence of negative freeboards and flooding at the snow/sea ice interface. We investigate the impact of flooded snow layers on the airborne radar observations. Utilizing a combination of methodologies, we assess sea ice thickness conditions in the region. We discuss our results in the context of satellite-derived ice thickness and consider the impact of flooding on estimating thickness in regions with thin sea ice and deep snow, such as in the Southern Ocean or in the Atlantic sector of the Arctic Ocean. 85 2 Data and Methods

Study Area
Field observations for this study were acquired during the Norwegian young sea ICE expedition (N-  (Granskog et al., 90 2016(Granskog et al., 90 , 2018. In this study, we focus on sea ice and snow related observations from the drift of Floe 2, covering a time period from 24 February to 19 March 2015. Data from the OIB overflight employed in this study was collected on 19 March 2015 at 82 • 29'N and 22 • 37'E above the drifting sea ice floe. https://doi.org/10.5194/tc-2020-168 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.

Ground-based measurements
The ice station on Floe 2 was set up on an aggregation of different ice types: refrozen leads, first year ice (FYI), and second 95 year ice (SYI). Modal sea ice thickness at the field station was 0.3 m, 0.9 m, and 1.7 m, for refrozen leads, FYI, and SYI, respectively . Snow thickness was on average 0.56 ± 0.17 m on FYI and SYI , while on refrozen leads it was approximately 0.02 m, likely redistributed from blowing snow. For this study, a 400 m × 60 m survey field was established. Red flag poles, with black snow-filled trash bags, marked the outline making it visible from air (see Figure 1). On this two dimensional (2-D) survey field, short after the overflights, snow thickness measurements (ℎ , 100 N=1121) were obtained with a GPS snow probe (SP) from Snow-Hydro (Fairbanks, AK, USA). Total snow and ice thickness measurements (ℎ , N=1046) were obtained using the EM31 electromagnetic device (Geonics Ltd., Mississauga, Ontario, Canada) within the 2-D survey field following a "snake line" sampling pattern with 5 m spacing (see Figure 1). The snow probe is a thin pole with a sliding disk, 0.2 m in diameter. The pole penetrates the snow pack to the snow/sea ice interface, while the disk rests on the snow surface. Inside the pole a magnetic device measures the distance between the disk 105 and the lower tip of the pole providing the snow thickness (Sturm and Holmgren, 1999;. Each measurement is timetagged and geolocated, and recorded on a data logger. The accuracy of the measurements over sea ice may vary between ±1-3 mm (Marshall et al., 2006;Sturm and Holmgren, 2018) and the footprint is the size of the disk (i.e., 0.2 m). Snow thickness measurements were made approximately every 5 m following the snake line sampling pattern within the 2-D survey field (see Figure 1). A person dragging the EM31 instrument on a plastic sledge followed the snow probe sampler. The EM31 110 measurements were sampled with a frequency of 2 Hz. The footprint size of the EM31 ranges from 3 m to 5 m (e.g., Haas et al., 1997), depending on the ice and snow thickness. The accuracy of the EM31 measurements is approximately ±0.1 m (Haas et al., 2009) for level ice, but higher for rough and deformed ice. For comparison to the EM31 data and to collect direct measurements, we drilled 10 equispaced holes with a 2" auger and measured ice and snow thickness, and ice freeboard with a thickness gauge from Kovacs Enterprises (Roseburg, OR, USA). The gauge is a specific tape measure with a foldable metal 115 weight at the bottom that can be deployed through the drill-holes. The accuracy of the readings is estimated to be ±0.01 m.
In addition, a snow pit was dug in the vicinity of the 2-D survey field (Merkouriadi et al., 2017c), and an ice core was obtained to measure ice salinity, temperature, and density . The core was extracted with a 0.09 m diameter ice corer from Kovacs Enterprises (Roseburg, OR, USA). During the N-ICE2015 expedition a set of long, and independent, transects with combined EM31 and snow thickness measurements (N=5053) were obtained within a maximum radius of 5 km 120 around the ship. These were performed to characterize the spatial variability of snow and ice thickness in the area surrounding the main ice camp. Further details can be found in Rösel et al. (2018). We use the results of the independent snow transects from Floe 2 to provide the regional context for the measurements obtained in the smaller 2-D survey field.

Airborne measurements
The OIB aircraft surveyed the 2-D survey field three times (see Figure 2) on 19 March 2015. First a surveillance overflight 125 occurred at 15:28 UTC. A second and third pass directly over the 2-D survey field occurred at 15:37 UTC and 15:43 UTC, respectively. Because the first pass did not adequately intersect the 2-D survey field, we focus our analysis on measurements obtained during passes 2 and 3 of the aircraft. Although the ice floe drifted during the airborne survey, the alignment of transects 2 and 3 were such that they directly intersected the 2-D survey field on both passes. For sea ice studies, the aircraft was equipped with an ATM laser altimeter system , an ultra-wideband frequency modulated continuous 130 waveform (FMCW) snow radar system (e.g., Yan et al., 2017) and a digital camera system (DMS) that provides highresolution (0.1 m) geolocated visible-band images of the snow surface (Dominguez, 2018) allowing for visual interpretation of sea ice conditions in the vicinity of the 2-D survey field (see Figure 1). indicated by black stars) located on Floe 2. The three airborne transects across the field are indicated. During the survey, the ice floe drifted south at an approximate drift speed of 0.15 ms -1 .

Methodology 155
We number the variables measured with the different observation techniques as follows: 1) derived from in situ measurements with EM31 and SP; 2) derived from in situ measurements from drill holes; 3) derived from OIB measurements (ATM and snow radar); 4) derived from a combination of in situ and OIB measurements.

Drift correction
In order to obtain spatial coincidence between the in situ and airborne measurements of snow thickness, freeboard and sea ice 160 thickness, and the positions of all measurements were first corrected to mitigate the impact of the drifting sea ice during the experiment. As a reference, we determined the position of the four corners of the 2-D survey field using the DMS imagery collected during the second and third OIB overpasses. We extracted the precise latitude and longitude from the geo-referenced DMS imagery and by comparing the differences for each corner marker between the two overpasses, we were able to deduce that the ice floe was drifting south at a speed of 0.15 ms -1 . To correct for the drift that occurred during the EM31 and SP 165 sampling of the 2-D survey field, we followed the procedure described in Rösel et al. (2018): The EM31 data was resampled onto the coordinates of the SP track, and a Gaussian filter was applied to the EM31 data. Afterwards, both the EM31 and the SP data were median-sampled on a 5 m regular grid to avoid oversampling (Geiger et al., 2015).

Density of sea water, ice and snow
In all calculations we used the following values: the density for sea water was = 1027 kgm -3 (Meyer et al., 2017), the 170 bulk density for the snow pack was = 328 kgm -3 (Merkouriadi et al., 2017b), and the bulk density for FYI was = 910 kgm -3 . All values are based on actual measurements during the N-ICE2015 expedition at floe 2.

In situ snow thickness, sea ice thickness, and freeboards
In Figure 3, the concept of isostatic equilibrium is shown for two cases: On the left side, the ratio of snow thickness (ℎ ) to sea ice thickness (ℎ ) is small, resulting in a positive sea ice freeboard (ℎ ). On the right side, the situation of the 'new' 175 Arctic as described in Rösel et al. (2018) is schematically presented: A thick snow layer ℎ is pushing a relatively thin sea ice ℎ layer below the ocean surface. The resulting sea ice freeboard ℎ becomes negative, and subsequently the sea ice surface is vulnerable for flooding. https://doi.org/10.5194/tc-2020-168 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.
To obtain sea ice thickness (ℎ ) from the in situ measurements, the resampled snow thickness measurements from SP (ℎ , N=1046) were subtracted from the total sea ice thickness from EM31 measurements (ℎ , N=1046): 180 185 Figure 3: The concept of isostatic equilibrium of sea ice: On the left, the ratio of snow thickness (ℎ ) to sea ice thickness (ℎ ) is small, sea ice freeboard (ℎ ) is positive. On the right, the ratio of ℎ to ℎ is high, ℎ is negative. Snow freeboard ℎ is the same in both cases.
which results in the freeboard (ℎ ) derived from the in situ (IS) measurements using obtained snow thickness and sea ice thickness information and densities given above: 195 For wet snow conditions, or a flooded state of the sea ice we refer to the studies of Zwally et al., 2008 andOzsoy-Cicek et al., 2013 where either ℎ is set equal to ℎ , or a slush layer is included in the calculations, respectively. In addition, we 200 gain in situ information from the drill-hole readings: Sea ice thickness (ℎ ), snow thickness (ℎ ),freeboard (ℎ ), and snow freeboard (ℎ ) The measurement accuracy of in situ freeboards ℎ and ℎ is estimated to be ±0.06 m; the accuracy of freeboard from the drill-hole measurements ℎ and ℎ is ±0.01 m .

Airborne snow thickness, sea ice thickness, and freeboards
The DMS images were used to identify the geographical coordinates of areas of open water (with little or no ice cover) within the large refrozen lead, located in the southwest of the 2-D survey field site and adjacent to it (see Figure 1). ATM elevation 210 measurements associated with these areas were averaged to estimate the local sea surface height (SSH). The SSH within the lead was then subtracted from all ATM elevations, to obtain the ATM snow freeboard ( ℎ ). Individual ATM measurements were resampled on the same 5-m regular grid as the in situ snow and ice measurements across the sea ice floe (see Figure 2). The snow radar echoes from passes 2 and 3 from the OIB survey also illustrate the presence of open water, refrozen leads and areas with deep snow cover on the N-ICE2015 ice floe (see Figure 4). We calculated snow thickness from 215 snow radar (ℎ ), following the methodology of Newman et al. (2014). Since the basal snow layers were saline in some locations, the snow/sea ice interface could not always be detected. Therefore, a running average at 25 m length-scale (equivalent to five snow radar measurements) was used to account for an observed diffuse snow-ice interface at the 2-D survey field site, possibly caused by a saline basal layer in the lower snow pack.

220
Ice freeboard (ℎ , ) and sea ice thickness (ℎ , ) derived from a combination of the airborne data measurements acquired over the 2-D survey field site with the in situ snow-probe data were calculated as follows: Here we introduce the term ℎ which is the difference between the ATM snow freeboard (ℎ ) and the snow radar snow thickness (ℎ ). ℎ is used to indicate the depth of the ice freeboard plus the snow-ice basal layer, if present.
Where ℎ is the ice freeboard, ℎ is the thickness of the snow-ice basal layer, and is any remaining errors due to the interface picking algorithms as applied to the snow radar echos (Figure 4).  To gain knowledge about the agreement in the snow thickness between the airborne and in situ observations on a more regional scale, we compared observations from snow radar within a 10 km radius around the position of R/V Lance with the average snow thickness conditions measured in situ during the drift of Floe 2 during the N-ICE2015 expedition within a 5 While the one-to-one comparison over the survey field can be considered as a direct validation study, the statistical regional comparison across the larger area can potentially be influenced by geophysical and thermodynamic processes such as ice dynamics, snow redistribution, snow metamorphism etc. that occurred during the entire drift duration of floe 2 (23 days) 270 where in situ data were acquired.
The drill-hole measurements made at the 2-D survey field site (N=10) gave the following averaged results: ice thickness ℎ = 1.39 ± 0.33m, snow thickness ℎ = 0.50 ± 0.18 m, freeboard (FB2) ℎ = 0.01 ± 0.07 m and snow freeboard ℎ = 0.50 ± 0.12 m (see Table 2 in Appendix). These values lie within the standard deviation of all measurements collected at the 2-D survey field site, i.e. our results demonstrate very good agreement across all observation methods. Three 275 out of the ten drill-holes were flooded, already before drilling.
https://doi.org/10.5194/tc-2020-168 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. probe measurements from ATM surface elevation (FB4, see Figure 6). Taking the ±0.06 m uncertainty into account, this results in a negative freeboard area fraction of 19% and 14% across the 2-D survey field site for FB1 and FB4, respectively (see Figure 7). An estimate of the ice freeboard plus the thickness of the snow-ice basal layer at the survey site that was impacted by brine wicking may be obtained by subtracting the snow radar measurements from the nearest ATM surface 295 elevation value, which results in an average ℎ , = 0.20 ± 0.10 m but varies across the site (Figure 7c) Table 1: Results of snow, sea ice, and freeboard measurements and calculations of the 2-D survey field site Finally, in Figure 8, we show the sea ice thickness distributions collected at the 2-D survey field site ℎ and ℎ , as well asfor the region surrounding the R/V Lance, ℎ . For comparison we include ℎ , calculated from a combination of 305 the ATM data and the in situ snow probe measurements. The in situ measurements show that the 2-D survey field site was situated on an ice floe 1.4 to 1.5 m thick (Figure 8). This compares to a slightly thinner overall ice cover regionally (1.09 ± 0.92m). The variability in sea ice thickness in the region surrounding the R/V Lance is about three times larger than that at the 2-D survey field site. We note that the average sea ice thickness ℎ = 1.90 m is biased high because the thickness equation (eqn. 6) does not take into account the two-layer snow set-up, with each snow layer having a different depth and 310 density. Our results suggest that the resulting bias is approximately 0.4 m, and this is consistent with the result shown in Figure 6, which indicates a bias of approximately 0.03 m in FB1, which arose for the same reason (i.e., eqn. 2 also does not account for the two-layer snow situation).

Discussion and Conclusions
The mean and modal snow thickness estimates derived from the snow radar were 0.15-0.16 m lower than the in situ snow probe measurements, obtained at the 2-D survey field site. Over a larger regional scale of 10 km radius about the R/V Lance location, snow thickness estimates derived from the snow radar underestimate in situ snow probe-derived thickness by 0.06 m, which is close to the measurement uncertainty of the snow radar system associated with its range resolution (Newman et al., 2014).
In radar altimetry, it is assumed that the radar signal penetrates completely through a dry snow pack and energy is reflected from the snow/sea ice interface, which represents the height of the sea ice-freeboard above local sea level. This assumption is 350 valid (Beaven et al., 1995) for a cold, dry and homogenous snow pack, typical of Arctic sea ice in winter. However, for snow packs exhibiting high moisture content or higher densities (e.g. due to ice lenses/crusts), radar signals undergo absorption within the snow volume (e.g., Kwok and Maksym, 2014;Ricker et al., 2015). Recent studies suggest reduced signal penetration into the snow pack, with a more diffuse snow/sea ice interface, on both Arctic and Antarctic sea ice (Gerland et al., 2013;Kwok and Kacimi, 2018), especially if the snow pack is saline (Nandan et al., 2020;Nandan et al., 2017) or very 355 deep with ice lenses present . In addition, deep snow pushes the ice surface below the water level, leading to negative freeboard and induces flooding and slushy, snow-ice formation in the basal layers of the snow pack, which, when measured with a radar altimeter system, can result in a dominant scattering horizon above the true snow/ice interface (Nandan et al., 2020), and hence an overestimation of ice freeboard and thus sea ice thickness, and an underestimate of total snow thickness. 360 On FYI, overlying snow also wicks brine upwards from the sea ice surface during freeze-up, producing saline snow layers, predominately observed in the bottom-most 0.06-0.08 m of the snow pack (Drinkwater and Crocker, 1988;Geldsetzer et al., 2009;Nandan et al., 2016Nandan et al., , 2017Nandan et al., , 2020. Sea ice salinity observations from the 1-m thick FYI floe, collected on 5 and 23 March 2015, show salinities of > 11 psu in the uppermost 0.10 m sea ice layers (Figure 10 in Appendix). Snow salinity observations from snow covers (0.26 m and 0.34 m thick) overlying the FYI floe indicate highly-saline, 0.10 m deep basal 365 layers in both snow covers, by up to 10 psu. Additionally, one-third of the drill-holes at the 2D field site indicated flooding of the snow pack, and negative freeboard, which induced the formation of highly saline and saturated slush in the basal snow layers. The presence of this slushy snow-ice layer at the 2D field site resulted in a challenging geophysical setting for the measurement of total snow and ice thickness using remote sensing techniques.
Previous studies (Barber et al., 1998;Barber and Nghiem, 1999;Nghiem et al., 1995;Geldsetzer et al., 2009;Nandan et al., 370 2017;Nandan et al., 2020) have reported the impact of saline snow on FYI, which alters the geophysical, thermodynamic, dielectric and radar scattering properties of the snow cover, thereby impacting radar signal penetration through the snow pack. Nandan et al. (2017) and showed that a saline snow cover on a FYI setting, induced by upward brine wicking from the sea ice surface, shifted the main radar scattering horizon away from the snow/sea ice interface by up to 0.07 m, and weakened the signal arising from the saline scattering horizon. In these studies, covering the Canadian (Nandan et al., 2017) 375 and the Atlantic (Nandan et al., 2020) sectors of the Arctic, the conditions at the survey field site included saline, wet, and deep snow which impacted the accuracy of the snow thickness derived from the snow radar signal. In the Nandan et al. (2020) case study from the N-ICE2015 experiment, they demonstrated significant overestimations in FYI thickness by up to 95% https://doi.org/10.5194/tc-2020-168 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. between simulated FYI thickness, snow radar and ATM-derived FYI thickness. They simulated the Ku-band radar scattering horizon from 0.36 m and 0.45 m deep snow on 0.69 m and 0.92 m thin ice, exhibiting negative freeboards by 0.04 m and 380 0.07 m, respectively. They found that the FYI thickness overestimations was a result of vertical shift in the radar scattering horizon, caused by brine wicking from the slushy snow layers, caused by negative freeboards.
Our study shows that saline snow conditions can lead to the observed underestimation of total snow thickness by the radar, due to a combination of factors including reflection from a scattering horizon in the snow pack that is above the main snow/ice interface, a diffuse scattering horizon within the snow volume, and potential errors in the height of the snow/ice 385 interface picked in individual snow radar echoes. Even though the snow radar underestimated mean snow thickness by between 0.06 m and 0.16 m across the 2-D survey field site and the regional survey, the radar was able to fully reproduce the snow thickness variability, when compared to in situ measurements (standard deviations of 0.16 m and 0.15 m for the survey field, respectively; see Figure 3). Thus, we can report that the airborne snow radar is capable of measuring meaningful snow thickness distributions even in challenging snow pack conditions. However, ambiguous radar signal penetration through 390 slushy layers (caused by sea ice flooding) and saline snow covers (caused by brine wicking from sea ice surface) may introduce a potential bias in accurate estimates of total snow thickness. For a radar altimeter, the main scattering horizon within the snow volume is not just a function of snow thickness, but also depends on the thermodynamic properties of the snow cover (i.e., snow temperature, density, salinity, wetness, roughness, and grain microstructure). Further research is required to understand the relationship between the main scattering horizon and variability in snow cover properties. 395 Biases caused by uneven penetration of radar altimeter signals within slushy and saline layers in the snow pack will also have implications on estimates of snow and sea ice thickness measurements from currently operational satellite-based radar altimeters such as SARAL AltiKa (Ka-band), CryoSat-2 and Sentinel-3A/B (Ku-band), and the ESA's upcoming Kuand Ka-band dual-frequency satellite radar altimeter mission CRISTAL. King et al. (2018) reported underestimation of sea ice thickness derived from CryoSat-2 data caused by negative freeboards in the same region as was investigated in our study. A 400 detailed quantification of the contributions to the error budget associated with freeboard retrieval from CryoSat-2 was made by the ESA CryoVal-SI project team and is described in Ricker et al. (2014) and Haas et al. (2016). To examine the impact of a deep snow pack with saline and/or flooded snow-ice interface, we show as an example the CryoSat-2 sea ice products provided by the NASA Goddard Space Flight Center (GSFC, Kurtz et al., 2014) for the region surrounding R/V Lance location in March 2015 (Figure 9). It is striking that for this region the CryoSat-2 sea ice freeboard is up to 0.3 m (Figure  405 9b), and the derived sea ice thickness is overestimated by over 1.0 m (Figure 9c), when compared with the in situ results reported in Rösel et al., 2018). The assumed snow thicknesses of 0.15 m and 0.37 m (derived from Warren et al., 1999;Kurtz et al., 2014, Figure 9a) are underestimated when compared to the observed in situ snow thickness, which averaged 0.55 m.
Sea ice parameters from the GSFC CryoSat-2 are derived with a waveform fitting procedure using an empirical waveform model (Sallila et al., 2019), which should account for snow geophysical properties. But, to our knowledge, currently, none of 410 the conventional CryoSat-2 retracker algorithms (Hendricks et al. 2010;Ricker et al., 2014) or the empirical models (Kurtz https://doi.org/10.5194/tc-2020-168 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. et al., 2014) include flooding in the snow pack as a consideration. This issue could cause misinterpretation of both, airborne and satellite radar altimeter signals especially in areas where flooding of the snow cover has been observed such as in the Atlantic sector of the Arctic, and also in parts of the Weddell and Bellingshausen Seas in the Southern Ocean which results in an overestimation of sea ice thickness. Our study suggests that this issue should to be considered and improved in the future. 415   https://doi.org/10.5194/tc-2020-168 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License.