Assessments of Data-Driven Deep Learning Models on One-Month Predictions of Pan-Arctic Sea Ice Thickness

In recent years,deep learning methods have gradually been applied to prediction tasks related to Arctic sea ice concentration,but relatively little research has been conducted for larger spatial and temporal scales,mainly due to the limited time coverage of observations and reanalysis data.Meanwhile,deep learning predictions of sea ice thickness(SIT)have yet to receive ample attention.In this study,two data-driven deep learning(DL)models are built based on the ConvLSTM and fully convolutional U-net(FC-Unet)algorithms and trained using CMIP6 historical simulations for transfer learning and fine-tuned using reanalysis/observations.These models enable monthly predictions of Arctic SIT without considering the complex physical processes involved.Through comprehensive assessments of prediction skills by season and region,the results suggest that using a broader set of CMIP6 data for transfer learning,as well as incorporating multiple climate variables as predictors,contribute to better prediction results,although both DL models can effectively predict the spatiotemporal features of SIT anomalies.Regarding the predicted SIT anomalies of the FC-Unet model,the spatial correlations with reanalysis reach an average level of 89%over all months,while the temporal anomaly correlation coefficients are close to unity in most cases.The models also demonstrate robust performances in predicting SIT and SIE during extreme events.The effectiveness and reliability of the proposed deep transfer learning models in predicting Arctic SIT can facilitate more accurate pan-Arctic predictions,aiding climate change research and real-time business applications.

An improved algorithm for retrieving thin sea ice thickness in the Arctic Ocean from SMOS and SMAP L-band radiometer data

The aim of this study was to develop an improved thin sea ice thickness(SIT)retrieval algorithm in the Arctic Ocean from the Soil Moisture Ocean Salinity and Soil Moisture Active Passive L-band radiometer data.This SIT retrieval algorithm was trained using the simulated SIT from the cumulative freezing degree days model during the freeze-up period over five carefully selected regions in the Beaufort,Chukchi,East Siberian,Laptev and Kara seas and utilized the microwave polarization ratio(PR)at incidence angle of 40°.The improvements of the proposed retrieval algorithm include the correction for the sea ice concentration impact,reliable reference SIT data over different representative regions of the Arctic Ocean and the utilization of microwave polarization ratio that is independent of ice temperature.The relationship between the SIT and PR was found to be almost stable across the five selected regions.The SIT retrievals were then compared to other two existing algorithms(i.e.,UH_SIT from the University of Hamburg and UB_SIT from the University of Bremen)and validated against independent SIT data obtained from moored upward looking sonars(ULS)and airborne electromagnetic(EM)induction sensors.The results suggest that the proposed algorithm could achieve comparable accuracies to UH_SIT and UB_SIT with root mean square error(RMSE)being about 0.20 m when validating using ULS SIT data and outperformed the UH_SIT and UB_SIT with RMSE being about 0.21 m when validatng using EM SIT data.The proposed algorithm can be used for thin sea ice thickness(<1.0 m)estimation in the Arctic Ocean and requires less auxiliary data in the SIT retrieval procedure which makes its implementation more practical.

The Arctic Sea Ice Thickness Change in CMIP6’s Historical Simulations

This study assesses sea ice thickness(SIT)from the historical run of the Coupled Model Inter-comparison Project Phase 6(CMIP6).The SIT reanalysis from the Pan-Arctic Ice Ocean Modeling and Assimilation System(PIOMAS)product is chosen as the validation reference data.Results show that most models can adequately reproduce the climatological mean,seasonal cycle,and long-term trend of Arctic Ocean SIT during 1979-2014,but significant inter-model spread exists.Differences in simulated SIT patterns among the CMIP6 models may be related to model resolution and sea ice model components.By comparing the climatological mean and trend for SIT among all models,the Arctic SIT change in different seas during 1979-2014 is evaluated.Under the scenario of historical radiative forcing,the Arctic SIT will probably exponentially decay at-18%(10 yr)-1 and plausibly reach its minimum(equilibrium)of 0.47 m since the 2070s.

Ice thickness,internal layers,and surface and subglacial topography in the vicinity of Chinese Antarctic Taishan station in Princess Elizabeth Land,East Antarctica

We present the results of two ground-based radio-echo-sounding（RES） and GPS surveys performed in the vicinity of new Chinese Taishan station,Princess Elizabeth Land,East Antarctica,obtained in two austral summers during CHINARE 21（2004/2005） and CHINARE 29（2012/2013）.The radar surveys measured ice thickness and internal layers using 60- and 150-MHz radar systems,and GPS measurements showed smooth surface slopes around the station with altitudes of 2607-2636 m above sea level（a.s.l.）.Radar profiles indicate an average ice thickness of 1900 m,with a maximum of 1949 m and a minimum of 1856 m,within a square area measuring approximately 2 km × 2 km in the vicinity of the station.The ice thickness beneath the station site is 1870 m.The subglacial landscape beneath the station is quiet sharp and ranges from 662 to 770 m a.s.l.,revealing part of a mountainous topography.The ice volume in the grid is estimated to be 7.6 km^3.Along a 60-MHz radar profile with a length of 17.6 km at the region covering the station site,some disturbed internal layers are identified and traced;the geometry of internal layers within the englacial stratigraphy may imply a complex depositional process in the area.

Distribution of ice thickness and subglacial topography of the "Chinese Wall" around Kunlun Station,East Antarctica

As fundamental parameters of the Antarctic Ice Sheet,ice thickness and subglacial topography are critical factors for studying the basal conditions and mass balance in Antarctica.During CHINARE 24（the 24 th Chinese National Antarctic Research Expedition,2007/08）,the research team used a deep ice-penetrating radar system to measure the ice thickness and subglacial topography of the ＂Chinese Wall＂ around Kunlun Station,East Antarctica.Preliminary results show that the ice thickness varies mostly from 1600 m to 2800 m along the ＂Chinese Wall＂,with the thickest ice being 3444 m,and the thinnest ice 1255 m.The average bedrock elevation is 1722 m,while the minimum is just 604 m.Compared with the northern side of the ice divide,the ice thickness is a little greater and the subglacial topography lower on the southern side,which is also characterized by four deep valleys.We found no basal freeze-on ice in the Gamburtsev Subglacial Mountains area,subglacial lakes,or water bodies along the ＂Chinese Wall＂.Ice thickness and subglacial topography data extracted from the Bedmap 2 database along the ＂Chinese Wall＂ are consistent with our results,but their resolution and accuracy are very limited in areas where the bedrock fluctuates intensely.The distribution of ice thickness and subglacial topography detected by ice-penetrating radar clarifies the features of the ice sheet in this ＂inaccessible＂ region.These results will help to advance the study of ice sheet dynamics and the determination of future locations of the GSM＇s geological and deep ice core drilling sites in the Dome A region.

An Arctic sea ice thickness variability revealed from satellite altimetric measurements

A modified algorithm taking into account the first year（FY） and multiyear（MY） ice densities is used to derive a sea ice thickness from freeboard measurements acquired by satellite altimetry ICESat（2003–2008）. Estimates agree with various independent in situ measurements within 0.21 m. Both the fall and winter campaigns see a dramatic extent retreat of thicker MY ice that survives at least one summer melting season. There were strong seasonal and interannual variabilities with regard to the mean thickness. Seasonal increases of 0.53 m for FY the ice and 0.29 m for the MY ice between the autumn and the winter ICESat campaigns, roughly 4–5 month separation, were found. Interannually, the significant MY ice thickness declines over the consecutive four ICESat winter campaigns（2005–2008） leads to a pronounced thickness drop of 0.8 m in MY sea ice zones. No clear trend was identified from the averaged thickness of thinner, FY ice that emerges in autumn and winter and melts in summer. Uncertainty estimates for our calculated thickness, caused by the standard deviations of multiple input parameters including freeboard, ice density, snow density, snow depth, show large errors more than 0.5 m in thicker MY ice zones and relatively small standard deviations under 0.5 m elsewhere. Moreover, a sensitivity analysis is implemented to determine the separate impact on the thickness estimate in the dependence of an individual input variable as mentioned above. The results show systematic bias of the estimated ice thickness appears to be mainly caused by the variations of freeboard as well as the ice density whereas the snow density and depth brings about relatively insignificant errors.

Sea ice thickness analyses for the Bohai Sea using MODIS thermal infrared imagery

Level ice thickness distribution pattern in the Bohai Sea in the winter of 2009-2010 was investigated in this paper using MODIS night-time thermal infrared imagery. The cloud cover in the imagery was masked out manually. Level ice thickness was calculated using MODIS ice surface temperature and an ice surface heat balance equation. Weather forcing data was from the European Centre for Medium-Range Weather Forecasts （ECMWF） analyses. The retrieved ice thickness agreed reasonable well with in situ observations from two off-shore oil platforms. The overall bias and the root mean square error of the MODIS ice thickness are -1.4 cm and 3.9 cm, respectively. The MODIS results under cold conditions （air temperature 〈 -10~C） also agree with the estimated ice growth from Lebedev and Zubov models. The MODIS ice thickness is sensitive to the changes of the sea ice and air temperature, in particular when the sea ice is relatively thin. It is less sensitive to the wind speed. Our method is feasible for the Bohai Sea operational ice thickness analyses during cold freezing seasons.

Practical Model of Sea Ice Thickness of Bohai Sea Based on MODIS Data

Sea ice thickness is one of the most important input parameters in the studies on sea ice disaster prevention and mitigation. It is also the most important content in remote sensing monitoring of sea ice. In this study, a practical model of sea ice thickness(PMSIT) was proposed based on the Moderate Resolution Imaging Spectroradiometer(MODIS) data. In the proposed model, the MODIS data of the first band were used to estimate sea ice thickness and the difference between the second-band reflectance and the fifth-band reflectance in the MODIS data was calculated to obtain the difference attenuation index(DAI) of each pixel. The obtained DAI was used to estimate the integrated attenuation coefficient of the first band of the MODIS at the pixel level. Then the model was used to estimate sea ice thickness in the Bohai Sea with the MODIS data and then validated with the actual sea ice survey data. The validation results showed that the proposed model and corresponding parameterization scheme could largely avoid the estimation error of sea ice thickness caused by the spatial and temporal heterogeneity of sea ice extinction and allowed the error of 18.7% compared with the measured sea ice thickness.

Towards a semi-empirical model of the sea ice thickness based on hyperspectral remote sensing in the Bohai Sea

Sea ice thickness is one of the most important input parameters for the prevention and mitigation of sea ice disasters and the prediction of local sea environments and climates. Estimating the sea ice thickness is currently the most important issue in the study of sea ice remote sensing. With the Bohai Sea as the study area, a semiempirical model of the sea ice thickness（SEMSIT） that can be used to estimate the thickness of first-year ice based on existing water depth estimation models and hyperspectral remote sensing data according to an optical radiative transfer process in sea ice is proposed. In the model, the absorption and scattering properties of sea ice in different bands（spectral dimension information） are utilized. An integrated attenuation coefficient at the pixel level is estimated using the height of the reflectance peak at 1 088 nm. In addition, the surface reflectance of sea ice at the pixel level is estimated using the 1 550–1 750 nm band reflectance. The model is used to estimate the sea ice thickness with Hyperion images. The first validation results suggest that the proposed model and parameterization scheme can effectively reduce the estimation error associated with the sea ice thickness that is caused by temporal and spatial heterogeneities in the integrated attenuation coefficient and sea ice surface. A practical semi-empirical model and parameterization scheme that may be feasible for the sea ice thickness estimation using hyperspectral remote sensing data are potentially provided.

Fhe application of electromagnetic-induction on the measurement of sea ice thickness in the Antarctic

As an important component of the cryosphere, sea ice is very sensitive to climate change. The study of sea ice physics needs accurate sea ice thickness. This paper presents an electromagnetic induction （EM） technique which can be used to measure the sea ice thickness distribution efficiently and its successful application in the Antarctic Neila Fjord. Based on the electrical properties of sea ice and seawater and the application of electromagnetic field theory, this technique can accurately detect the distance between the EM instrument and the ice/water interface to measure the sea ice thickness. Analyzing the apparent conductivity data obtained by the electromagnetic induction technique and drill-hole measurements at same location allows the construction of a transform equation for the apparent conductivity and sea ice thickness. The verification of the calculated sea ice thickness using this equation indicates that the electromagnetic induction technique is able to determine reliable sea ice thickness with an average relative error of only 5.5%. The ice thickness profiles show the sea ice distribution in Neila Fjord is basically level with a thickness of 0.8 - 1.4 m.

Estimation of Design Sea Ice Thickness with Maximum Entropy Distribution by Particle Swarm Optimization Method

The maximum entropy distribution, which consists of various recognized theoretical distributions, is a better curve to estimate the design thickness of sea ice. Method of moment and empirical curve fitting method are common-used parameter estimation methods for maximum entropy distribution. In this study, we propose to use the particle swarm optimization method as a new parameter estimation method for the maximum entropy distribution, which has the advantage to avoid deviation introduced by simplifications made in other methods. We conducted a case study to fit the hindcasted thickness of the sea ice in the Liaodong Bay of Bohai Sea using these three parameter-estimation methods for the maximum entropy distribution. All methods implemented in this study pass the K-S tests at 0.05 significant level. In terms of the average sum of deviation squares, the empirical curve fitting method provides the best fit for the original data, while the method of moment provides the worst. Among all three methods, the particle swarm optimization method predicts the largest thickness of the sea ice for a same return period. As a result, we recommend using the particle swarm optimization method for the maximum entropy distribution for offshore structures mainly influenced by the sea ice in winter, but using the empirical curve fitting method to reduce the cost in the design of temporary and economic buildings.

Regional characteristics of sea ice thickness in Canadian shelf and Arctic Archipelago measured by Ground Penetrating Radar

Ground Penetrating Radar （GPR） measurements of sea ice thickness including undeformed ice and ridged ice were carried out in the central north Canadian Archipelago in spring 2010. Results have shown a significant spatial heterogeneity of sea ice thickness across the shelf. The undeformed multi-year fast ice of （2.05±0.09） m thick was investigated southern inshore zone of Borden island located at middle of the observational section, which was the observed maximum thickness in the field work. The less thick sea ice was sampled across a flaw lead with the thicknesses of （1.05±0.11） m for the pack ice and （1.24±0.13） m for the fast ice. At the northernmost spot of the section, the undeformed multi-year pack ice was （1.54±0.22） m thick with a ridged ice of 2.5 to 3 m, comparing to the multi-year fast ice with the thickness of （1.67±0.16） m at the southernmost station in the Prince Gustaf Adolf Sea.

Ice thickness distribution and volume estimation of Burqin Glacier No.18 in the Chinese Altay Mountains

Information on the thickness distribution and volume of glacier ice is highly important for glaciological applications;however,detailed measurements of the ice thickness of many glaciers in the Chinese Altay Mountains remain lacking.Burqin Glacier No.18 is a northeast-orientated cirque glacier located on the southern side of the Altay Mountains.This study used PulseEKKO®PRO 100A enhancement ground-penetrating radar(GPR)to survey the ice thickness and volume of Burqin Glacier No.18 in summer 2018.Together with GPR surveying,spatial distributed profiles of the GPR measurements were concurrently surveyed using the real-time kinematic(RTK)global navigation satellite system(GNSS,Unistrong E650).Besides,we used QuickBird,WorldView-2,and Landsat TM to delineate accurate boundary of the glacier for undertaking estimation of glacier ice volume.GPR measurements revealed that the basal topography of profile B1-B2 was flat,the basal topography of profile C1-C2 presented a V-type form,and the basal topography of profile D1-D2 had a typical U-type topographic feature because the bedrock near the central elevation of the glacier was relatively flat.The longitudinal profile A1-A2 showed a ladder-like distribution.Glacier ice was thin at the terminus and its thickness increased gradually from the elevation of approximately 2620 m a.s.l.along the main axis of the glacier tongue with an average value of 80(±1)m.The average ice thickness of the glacier was determined as 27(±2)m and its total ice volume was estimated at 0.031(±0.002)km3.Interpretation of remote sensing images indicated that during 1989–2016,the glacier area reduced from 1.30 to 1.17 km2(reduction of 0.37%/a)and the glacier terminus retreated at the rate of 8.48 m/a.The mean ice thickness of Burqin Glacier No.18 was less than that of the majority of other observed glaciers in China,especially those in the Qilian Mountains and Central Chinese Tianshan Mountains;this is probably attributable to differences in glacier type and climatic setting.

Modeling ice thickness distribution and storage volume of glaciers in Chandra Basin,western Himalayas

This article reports modeled ice thickness distribution and total ice volume of the 65 selected glaciers(>0.5 km^2)of Chandra basin,located in the Western Himalayas.This is a first-of-its-kind study that gives detailed insights about the current ice thickness distribution at an individual glacier level in the Western Himalayas.The estimates are obtained using an optimally parameterized Glab Top2_IITB[Glacier Bed Topography Indian Institute of Technology Bombay(IITB)version]model with highresolution Digital Elevation Model(DEM)as an input.The total estimated volume of all the 65 selected glaciers is about 55.32 km^3 covering a total area of about 591.03 km^2.Using hypsometric analysis,it is found that the maximum amount of ice volume,i.e.,about 12.79 km^3,is currently residing at the elevation range of 5200–5400 m a.s.l.Ice thickness estimates obtained in the current study are compared with the ensemble estimates obtained in the Global Glacier Thickness Initiative(G2TI)project for three large glaciers,namely,Bada Shigri,Samudra Tapu,and Gepang Gath glaciers.The obtained results indicate that the difference between both the studies is marginal in terms of mean ice thickness and maximum ice thickness estimates except Samudra Tapu glacier.Moreover,the uncertainty of the estimated glacier ice volume from this study is about±15%whereas,from the G2TI project,it is about 25%.The main reasons for the difference could be the quality of the inputs used,model structure,model parameterization as well as the time stamp of the input used.The obtained results from this study indicate that the use of appropriate shape factor and better DEM would result in more reliable glacier ice thickness estimates even by using a simple slopedependent model like Glab Top2_IITB.

Assessment and application of electromagnetic induction method to measure Arctic sea ice thickness

The electromagnetic induction method is widely used to measure sea ice thickness. Based on the electrical properties of sea ice and seawater, the method measures the apparent conductivity, which represents the conductivity of the half-space, and calculates the thickness of the sea ice. During the fourth Chinese National Arctic Research Expedition in summer 2010, an integrated electromagnetic induction system was set up on the icebreaker R/V XUE LONG to measure sea ice thickness along the ship＇s tracks to the north of the Chukchi Sea. The conductivities of sea ice, seawater, and brine were measured and a simple forward model was used to explain the effect of changes in those conductivities on the apparent conductivity over a horizontal layered structure. The results of this analysis indicated that when using the electromagnetic induction method to measure sea ice thickness, the conductivity of sea ice could be neglected and the conductivity of seawater could be treated as a constant. The ice distribution results derived from the electromagnetic induction method showed that the typical sea ice thickness was 160 cm and 90 cm during the outbound and the return legs of the voyage, respectively.

Measurements of sea ice thickness and its subice morphology analysis using ice-penetration radar in the Arctic Ocean

Based on radar penetrating measurements and analysis of sea ice in the Arctic Ocean, The potential of radar wave to measure sea ice thickness and map the morphology of the underside of sea ice is investigated. The results indicate that the radar wave can penetrate Arctic summer sea ice of over 6 meters thick; and the propagation velocity of the radar wave in sea ice is in the range of 0.142 m·ns -1 to 0.154 m·ns -1 . The radar images display the roughness and micro-relief variation of sea ice bottom surface. These features are closely related to sea ice types, which show that radar survey may be used to identify and classify ice types. Since radar images can simultaneously display the linear profile features of both the upper surface and the underside of sea ice, we use these images to quantify their actual linear length discrepancy. A new length factor is suggested in relation to the actual linear length discrepancy in linear profiles of sea ice, which may be useful in further study of the area difference between the upper surface and bottom surface of sea ice.

Determination of ice thickness,subice topography and ice vol-ume at Glacier No.1 in Tien Shan,China by ground penetrating radar

We describe a radio-echo sounding (RES) survey for the determination of ice thickness, subglacial topography and ice volume of Glacier No. 1 , in Tien Shan, China, using ground-penetrating radar (GPR). Radar data were collected with 100-MHz antennas that were spaced at 4 m with a step size of 8 m. The images produced from radar survey clearly show the continuity of bedrock echoes and the undulation features of the bedrock surface. Radar results show that the maximum ice thickness of Glacier No. 1 is 133 m, the thickness of the east branch of Glacier No. 1 averages at 58. 77 m while that of the west branch of Glacier No. 1 averages at 44. 84 m. Calculation on ice volume indicates that the ice volume of the east branch of Glacier No. 1 is 51. 87 × 106 m3 and that of the west branch of Glacier No. 1 is 20. 21 × 106 m3. The amplitude of the undulation of the bedrock surface topography revealed by radar profiles is larger than that of the glacier surface topography, indicating that the surface relief does not directly depend on that of the bedrock undulation in Glacier No. 1 , in Tien Shan.

A study on remote sensing models of sea ice thickness by microwave radiometry

Extrapolating from the propagation theories of electromagnetic waves in a layered medium, a three-layer medium model is deduced in this paper by using microwave radiometric remote sensing technology which is suitable to first-year sea ice condition of the northern part of China seas. Comparison with in situ data indicates that for microwave wavelength of 10 cm, the coherent model gives a quite good fit result for the thickness of sea ice less than 20 cm, and the incoherent model also works well for thickness within 20 to 40 cm. Based on three theoretical models, the inversion soft ware from microwave remote sensing data for calculating the thickness of sea ice can be set up. The relative complex dielectrical constants of different types of sea ice in the Liaodong Gulf calculated by using these theoretical models and measurement data are given in this paper. The extent of their values is (0. 5-4. 0)-j(0. 07~0. 19).

Retrieval of Antarctic sea ice freeboard and thickness from HY-2B satellite altimeter data

Antarctic sea ice is an important part of the Earth’s atmospheric system,and satellite remote sensing is an important technology for observing Antarctic sea ice.Whether Chinese Haiyang-2B(HY-2B)satellite altimeter data could be used to estimate sea ice freeboard and provide alternative Antarctic sea ice thickness information with a high precision and long time series,as other radar altimetry satellites can,needs further investigation.This paper proposed an algorithm to discriminate leads and then retrieve sea ice freeboard and thickness from HY-2B radar altimeter data.We first collected the Moderate-resolution Imaging Spectroradiometer ice surface temperature(IST)product from the National Aeronautics and Space Administration to extract leads from the Antarctic waters and verified their accuracy through Sentinel-1 Synthetic Aperture Radar images.Second,a surface classification decision tree was generated for HY-2B satellite altimeter measurements of the Antarctic waters to extract leads and calculate local sea surface heights.We then estimated the Antarctic sea ice freeboard and thickness based on local sea surface heights and the static equilibrium equation.Finally,the retrieved HY-2B Antarctic sea ice thickness was compared with the CryoSat-2 sea ice thickness and the Antarctic Sea Ice Processes and Climate(ASPeCt)ship-based observed sea ice thickness.The results indicate that our classification decision tree constructed for HY-2B satellite altimeter measurements was reasonable,and the root mean square error of the obtained sea ice thickness compared to the ship measurements was 0.62 m.The proposed sea ice thickness algorithm for the HY-2B radar satellite fills a gap in this application domain for the HY-series satellites and can be a complement to existing Antarctic sea ice thickness products;this algorithm could provide long-time-series and large-scale sea ice thickness data that contribute to research on global climate change.

Joint Probability Analysis and Prediction of Sea Ice Conditions in Liaodong Bay

Sea ice conditions in Liaodong Bay of China are often described by sea ice grades,which classify annual sea ice conditions based on the annual maximum sea ice thickness(AM-SIT)and annual maximum floating ice extent(AM-FIE).The joint probability distribution of AM-SIT and AM-FIE was established on the basis of their data pairs from 1949/1950 to 2019/2020 in Liaodong Bay.The joint intensity index of the sea ice condition in the current year is calculated,and the joint classification criteria of the sea ice grades in past years are established on the basis of the joint intensity index series.A comparison of the joint criteria with the 1973 and 2022 criteria revealed that the joint criteria of the sea ice grade match well,and the joint intensity index can be used to quantify the sea ice condition over the years.A time series analysis of the sea ice grades and the joint intensity index sequences based on the joint criteria are then performed.Results show a decreasing trend of the sea ice condition from 1949/1950 to 2019/2020,a mutation in 1990/1991,and a period of approximately 91 years of the sea ice condition.In addition,the Gray-Markov model(GMM)is applied to predict the joint sea ice grade and the joint intensity index of the sea ice condition series in future years,and the error between the results and the actual sea ice condition in 2020/2021 is small.