CryoSat-SMOS Merged Sea Ice Thickness

[vox_mundi]

First Assimilation of CryoSat-2 Summer Observations Provides Accurate Estimates of Arctic Sea Ice Thickness
https://phys.org/news/2023-11-assimilation-cryosat-summer-accurate-arctic.html



Scientists have improved a data assimilation system for better estimating Arctic summer sea ice thickness (SIT) by assimilating satellite-based summer SIT and ice concentration data with an incremental analysis update (IAU) approach. Their study shows promising results for the improved estimations of Arctic SIT by assimilating the latest breakthrough of satellite-retrieved SIT for summer in the Arctic.

Their work was published in the journal Ocean-Land-Atmosphere Research.

"To date, there are no studies focusing on the assimilation of satellite-based SIT observations during summer. In this study, we assimilate the latest biweekly summer SIT and daily sea ice concentration data from satellite observations into a coupled ice-ocean model to improve the model estimates," said Dr. Chao Min at Sun Yat-sen University, and Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai), China; and the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Germany.

Prior to this study, satellite-derived summer SIT had not been assimilated into numerical models since SIT observations were limited in summer. Recent progresses on SIT observations based on CryoSat-2, a satellite dedicated to the study of ice, have provided scientists with reliable year-round sea ice data. With CryoSat-2 summer SIT, the team delved into the proper approach to assimilate CryoSat-2 data. Their study marks the first successful instance in which satellite-based summer SIT data has been assimilated into a coupled ice-ocean model.

Directly assimilating the rather infrequent CryoSat-2 observations would introduce discontinuities in the development of sea ice volume and thickness estimates. The team overcame this challenge by implementing an incremental analysis update (IAU) approach in their data assimilation system. The IAU method provides a gradual development of the sea ice fields over time while allowing the assimilation of infrequent summer SIT data, which are only available on the biweekly basis, in conjunction with daily sea ice concentration data.

The team's efforts have led to improved estimates of Arctic SIT, which exhibit stronger correlations with independent SIT observations when compared to a sea ice reanalysis (CMST) that does not incorporate CryoSat-2 summer observations. By combining CryoSat-2 summer observations and model dynamics, the team obtained significant improvements in sea ice estimates.

They noted significant improvements in the SIT field in the areas where the sea ice is roughest and experiences strong deformation, such as around the Fram Strait and the northern coast of the Canadian Arctic Archipelago and Greenland.

... "A continuous long-term ice thickness record with a finer temporal-spatial resolution that assimilates both the year-round sea ice concentration and thickness will be reconstructed in the future. In addition, with the improved sea ice initial states, our data assimilation system has the potential to improve sea ice forecasts, particularly in summer," said Prof. Qinghua Yang, a professor at the School of Atmospheric Sciences, Sun Yat-sen University, and Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai), China.

Chao Min et al, Improving Arctic sea-ice thickness estimates with the assimilation of CryoSat-2 summer observations, Ocean-Land-Atmosphere Research (2023)
https://spj.science.org/doi/10.34133/olar.0025

[Niall Dollard]

Thanks Vox Mundi.

Just for the casual observer, please note the charts in the previous post refer to the summer of 2016.

The article mentions that "The overestimation of the ice thickness by CMST is corrected, particularly in the Fram Strait and in the Arctic Ocean on the northern coast of the Canadian Arctic Archipelago and Greenland. These are regions where the sea ice experiences stronger deformation and the ice surface is roughest".

I have often suspected that some of the thickness charts (Hycom?) have a tendency to have very thick ice right up close to the CAA and north coast of Greenland. Perhaps then it is the ice formation/pressure ridges in this area that is making the ice appear thicker. 

[Steven]

It would be a game-changer if accurate year-round sea ice thickness estimates from CryoSat would become publicly available, even when it's updated only twice per month.  But we don't seem to be at that point yet. 

The study shows that the raw CryoSat thickness estimates in summer tend to underestimate the real observed sea ice thickness. 

The authors of the study use a modified model, which they call ANA, which incorporates CryoSat thickness.  It gives decent estimates of observed sea ice thickness estimates in summer north of Greenland (from the IceBird aerial mission) and Beaufort Sea (from buoy data), but works poorly in the Laptev Sea.
 


It still looks promising, and hopefully we'll hear more about this soon. 

Also interesting:

While the new CryoSat-2 summer SIT [Sea Ice Thickness] data represent a considerable improvement in satellite monitoring of Arctic sea ice, the SIT uncertainties are relatively larger in summer than in winter [18]. These SIT uncertainties include uncertainties in the radar freeboard estimates introduced by the interaction of CryoSat-2 radar waves with snow overlying the sea ice [67]. This is because the ability of radar waves to penetrate through snow to the ice surface varies depending on the salinity and roughness of the snow as well as changing snow properties in response to air temperatures and wind speed [67,68]. More importantly, for summer datasets, the interaction between melt ponds on the snow surface and CryoSat-2 radar waves is very poorly understood [31].

[Steven]

Meanwhile it looks like conditions have been favorable for sea ice growth over the last week, maybe due to the relatively low temperatures over the Arctic Ocean.  CryoSat-Smos total volume is now 6th lowest on record.  But the differences between 2nd and 6th place are small, and weather forecasts suggest that warmer air will move over the Arctic Ocean in a few days. 

Regionally, the low volume for Beaufort Sea and especially CAA stands out.  Baffin and Hudson Bay are also low.  While the entire Atlantic edge (Laptev, Kara, Barents and Greenland Sea) has relatively high volume.

https://sites.google.com/view/arctic-sea-ice/home/cryosat-regional

[Steven]

There has been a major slowdown of the CryoSat-SMOS volume growth in the last 7 days, probably due to the unfavorable weather conditions.  Total volume is now 2nd lowest on record for the date.

https://sites.google.com/view/arctic-sea-ice/home/cryosat-regional

[uniquorn]

Dispersion reducing average thickness north of Greenland?
cs2smos merged sea ice thickness with lead ice concentration overlaid at 60%
oct21-dec9

https://go.nasa.gov/3t67dGy

[uniquorn]

cs2smos today still showing ice thinning north of Lincoln Sea, nov27-dec10

[uniquorn]

update on thickest ice north of Lincoln/Ellesmere, oct21-dec19
same scale as above

[uniquorn]

thickness update with leads overlay at 60% (16MB)
Still waiting for thickest ice to thicken. Dartmouth2022#6 further west still hasn't thickened from 96cm yet.

[HapHazard]

Is this normal? I don't recollect this thickening-slowness before, but admittedly my memory is often lacking.

Either way, thanks for your tireless efforts, as always, uni.

[uniquorn]

There is no normal 

Worth re-reading this paper and comparing their results to this year's buoys.

Changes in the annual sea ice freeze–thaw cycle in the Arctic Ocean from 2001 to 2018
Long Lin, Ruibo Lei, Mario Hoppmann, Donald K. Perovich, and Hailun He
https://tc.copernicus.org/articles/16/4779/2022/

(Bottom Freeze Onset, Surface Freeze Onset)

The time delay between the BFO and SFO ranged from 38 to 115 d, with a mean of 82 ± 26 d.

Generally, the lower surface air temperature, thinner sea ice, and thinner snow cover are related to the earlier basal ice growth and vice versa, suggesting the time lag between the BFO and SFO can be significantly attributed to the ice column cooling efficiency. Without considering the SAT, His also has a significant relationship with the time lag (R=0.79, p<0.05) because the SAT does not show significant differences between the buoys. These results imply a negative feedback; i.e., thinner snow and ice favor earlier basal freeze-up in the following winter. Since sea ice thickness and snow depth of each IMB vary in a wide range, that is the most likely explanation why the BFO exhibits a much larger variability.

It is also notable that the total basal growth shows a significant correlation (R=0.63, p<0.01) with the length of the basal freeze season (Fig. 7e). As investigated above with IMB observations, basal growth of thinner sea ice started earlier compared to thicker ice. In combination with the negative conductive feedback (i.e., thinner ice grows faster than thicker ice), thinner ice generally experienced a longer freezing season and a larger ice growth. Considering the thermal insulation effect of the snow cover, the initial equivalent ice thickness His (defined above) was used to identify the link between the initial ice thickness and the total ice growth. As shown in Fig. 7f, the total sea ice growth during the entire freezing season increased by 0.26 m with the initial His decreasing by 1 m. For all IMBs that experienced the complete melting or freezing seasons, the average ice melt was 0.56 m at the surface and 0.65 m at the ice bottom, while the average ice growth was 0.74 m. Thus, the average annual ice thickness budget derived from all IMB observations during 2000–2018 amounts to −0.47 m, which clearly confirms the ongoing decline of the Arctic sea ice thickness. One remarkably similar result was for example presented by Petty et al. (2020), who used the February–March ice thickness retrieved from the satellite altimeter measurement of ICE-Sat (Ice, Cloud, ad land Elevation Satellite) to show a decrease of ∼ 0.37 m or ∼ 20 % thinning across an inner Arctic Ocean domain from 2008 to 2019. We infer that the growth of multiyear sea ice in winter is not sufficient to compensate for the melt in summer, even though the negative conductive feedback enhances the ice growth during the freezing season.

4 Conclusions

In this study, we determined the timings of sea ice annual freeze–thaw cycles in the Beaufort Gyre and central Arctic Ocean using a multisource data approach. Our main focus was on the detailed analysis of observations obtained by a large number of ice mass balance buoys (IMBs), which we used as a reference to compare our results calculated from satellite passive microwave (PMW) data. Since the IMB dataset may be potentially biased towards thicker and older sea ice, we additionally supplemented this analysis with observations from upward-looking sonars (ULSs) on moorings.

Based on these three very different datasets, we calculate and intercompare four pairs of surface melt and freeze onsets and two pairs of basal melt and freeze onsets that we determined using distinct automated algorithms. Our results reveal that the PMW and SAT threshold methods can reliably capture the surface melt and freeze cycle when compared with reference IMB surface mass balance observations. The average BMOs were comparable in the central Arctic Ocean and approximately 17 d earlier than SMOs in the Beaufort Gyre. The average BFOs were lagging behind almost 3 months compared to the SFOs for the pan Arctic Ocean.

During the transition of the SMO, the topmost snow temperature increased to above melting, indicating the start of surface melt. Reanalysis data indicated that the SMO was primarily driven by longwave radiation rather than shortwave radiation. In contrast, the SFO is driven by seasonal decline of shortwave radiation. Synchronous ice and underlying ocean observations confirmed that the ice bottom melt began when oceanic heat flux surpassed the upward conductive heat flux at the sea ice bottom. The ice basal freeze-up delay relative to the surface can be attributed to the regulation of heat capacity of sea ice itself and the oceanic heat release from ocean mixed layer and subsurface layer. The ice cooling index determined by the near-surface air temperature, snow depth, and ice thickness shows a significant correlation with the temporal delay between BFO and SFO, with lower surface air temperature, thinner sea ice, and thinner snow cover favoring earlier onset of basal ice growth and vice versa.

In the Beaufort Gyre, both Lagrangian IMB observations and Eulerian ULS observations exhibit a trend towards earlier basal melt onset, which can be attributed to the earlier warming of the surface ocean. In contrast, there is a trend towards earlier onset of basal ice growth evident from the IMB observations, which is associated with the reduction of ice thickness of the multiyear ice. At the same time, we determined a trend towards delayed onset of basal ice growth in the ULS observations because of the frequent occurrence of ice-free summers in the southern Beaufort Gyre region in recent years.

Note that, some limitation of our results should be considered. First, IMBs only collected one-dimension point measurements of mass balance and are representative of the special ice floe where they are deployed. As a result, the melt and freeze onsets of other ice categories such as ponded ice and ridged ice are out of our scope. Second, interior ice melt, surface pond, and false bottoms, as well as the unfrozen cavities within the rubble of ridges, greatly affected the energy budget consequently the basal melt and freeze (Shestov et al., 2018; Provost et al., 2019; Smith et al., 2022). But the effect for these different conditions could have was not considered in our study. Third, the majority of IMBs were deployed on multiyear undeformed ice (Planck et al., 2020), so the basal melt and freeze onsets of seasonal ice are underrepresented. Compared to multiyear ice, seasonal ice has higher bulk brine, resulting in a smaller specific heat capacity and latent heat of fusion (Tucker et al., 1987; Wang et al., 2020), as well as a higher permeability during the summer (Lei et al., 2022), thus affecting the sea ice basal melt and freeze processes. Finally, due to the limited vertical observation range of ocean profile automatic observation instruments, some special processes near the ice bottom, such as supercooling and false bottoms, were not characterized well.

Therefore, more intensive and elaborative ice mass balance observations of diverse ice types by IMB observations and other methods and simultaneous upper-ocean water properties observations in the future will vastly improve our capability to fully understand the ice–ocean system and the mass balance of sea ice in a changing Arctic.

[HapHazard]

Nice read. I think that freezing onset may well be influenced by Fram export, as well, at least in that area. Thicker ice thickens slower, IIRC, and a lot of thicker ice was flushed in that direction the past year. Not sure how much that influences the halocline & surface layer, either.

[uniquorn]

<>I think that freezing onset may well be influenced by Fram export<>

The thicker ice is probably still cooling from top to bottom, especially if there is snow. ('The ice basal freeze-up delay relative to the surface can be attributed to the regulation of heat capacity of sea ice itself and the oceanic heat release from ocean mixed layer and subsurface layer.')
Fram Export generates more leads north of Lincoln and that lowers the average ice thickness.

The time delay between the BottomFreezeOnset and SurfaceFreezeOnset ranged from 38 to 115 d, with a mean of 82 ± 26 d.

The freezing season thread started 95days ago so we may be a little over 'normal' north of Lincoln this year.

[uniquorn]

sit-leads overlaid onto ascat at 20% opacity, oct21-dec31.
closer look at thickest ice north of Lincoln Sea here

[uniquorn]

thickness update with leads overlaid at 60% opacity (12MB)

[uniquorn]

Important Notice: JRA-3Q will replace the present JRA55 reanalysis at the end of January 2024. Both L3 and L4 sea ice thickness products are affected.
Created by Xiangshan Tian-Kunze on Jan 23, 2024

JRA55 reanalysis is an auxiliary data set for SMOS sea ice thickness retrieval. The update of this near real-time JRA55 data will terminate at the end of January 2024, instead, the Japanese Reanalysis for Three Quarters of a Century (JRA-3Q) will be used thereafter. The impact of this transition on L3 SMOS and L4 CS2SMOS products will be analyzed in detail after the reprocessing of the complete data in April 2024. A preliminary comparison of JRA55 and JRA-3Q temperature and its impact on L3 SMOS sea ice thickness product is shown in Figure 1. There is temperature difference up to 5K in some regions in the Arctic, which is caused by the update of sea-ice and snow schemes in JRA-3Q.

Figure 1. Temperature difference between JRA-3Q and JRA55 in December 2023 (left) and the difference in SMOS sea ice thickness caused by the transition from JRA55 to JRA-3Q for the same month (right).

[uniquorn]

thickness update with leads overlaid at 60% opacity (19MB)
oct21-jan28

[uniquorn]

Comparison of HYCOM ESPC-D-V02 with CS2SMOS at end date oct22
gif and static