Tingting Liu

Recently, rapid warming scenarios along with intensified climate extremes were observed in the Antarctic Peninsula (AP). Acting as a key indicator of changing climate, snowline refers to the lower limit where snow remains on the glacier surface at the end of melting season. The snowline altitude (SLA) also approximately represents the equilibrium line altitude (ELA) which reflects glacier surface mass balance (SMB) dynamics. Unfortunately, unlike continuous SLA monitoring in the High Mountain Asia and circum-Arctic regions, there is a lack of precise SLA information due to non-available methods in the AP. In this study, a novel workflow, namely SLAMU, which mainly consists of a stacking machine learning (ML) model and a U-Net network enhanced by CBAM modules was proposed. The SLAMU workflow, for the first time, utilized both optical and SAR satellite imagery to identify snowlines. The stacking ML model achieved higher classification accuracy and Kappa coefficients (0.936, 0.903) than the random forest model (0.884, 0.833), thereby generating a more representative training label dataset. Based on visually interpreted snowline, SLAMU results exhibit more robust consistency than results either from ML models or published algorithms. The comparisons illustrated that SLAMU could nicely capture minor differences between snow and ice by fully grasping various glacier landcover characteristics, therefore precisely estimating SLA. A long time-series SLA dataset in the AP based on the SLAMU workflow can be expected, which further serves as a baseline for better quantitatively investigating glacier SMB under climate change.

Arctic sea-ice surface temperature (IST) is an important environmental and climatic parameter. Currently, wide-swath sea-ice surface temperature products have a spatial resolution of approximately 1000 m. The Medium Resolution Spectral Imager (MERSI-I) offers a thermal infrared channel with a wide-swath width of 2900 km and a high spatial resolution of 250 m. In this study, we developed an applicable single-channel algorithm to retrieve ISTs from MERSI-I data. The algorithm accounts for the following challenges: (1) the wide range of incidence angle; (2) the unstable snow-covered ice surface; (3) the variation in atmospheric water vapor content; and (4) the unique spectral response function of MERSI-I. We reduced the impact of using a constant emissivity on the IST retrieval accuracy by simulating the directional emissivity. Different ice surface types were used in the simulation, and we recommend the sun crust type as the most suitable for IST retrieval. We estimated the real-time water vapor content using a band ratio method from the MERSI-I near-infrared data. The results show that the retrieved IST was lower than the buoy measurements, with a mean bias and root-mean-square error (RMSE) of −1.928 K and 2.616 K. The retrieved IST is higher than the IceBridge measurements, with a mean bias and RMSE of 1.056 K and 1.760 K. Compared with the original algorithm, the developed algorithm has higher accuracy and reliability. The sensitivity analysis shows that the atmospheric water vapor content with an error of 20% may lead to an IST retrieval error of less than 1.01 K.

Ice surface temperature (IST) plays a fundamental role in the Antarctic ice sheet/shelf study. However, the production of spatially and temporally continuous Antarctic IST products remains a challenge. We proposed an instantaneous IST retrieval framework that can generate the spatially and temporally continuous Antarctic IST using Advanced Microwave Scanning Radiometer 2 (AMSR2) data. To generate a temporally continuous IST product, we developed an innovative scheme, which was based on the acquisition times difference between input and output data. We considered the impact of terrain and sensor observation state. The corresponding parameters were used as the auxiliary variables to improve the model accuracy. We trained and validated nine machine learning models using the generated sample set. The Light Gradient Boosting Machine (LightGBM) model presents the best performance, and the root mean square error (RMSE) of the LightGBM model is only half of that of the typical linear models. The RMSE of the LightGBM model decreased with training sample set size and stabilized at 1.67 K. Further validation using muti-source data showed that the IST retrieved using the LightGBM model has RMSEs of 1.39–2.32 K (relative to IST from Landsat-8) and 3.7–5.9 K (relative to IST from Baseline Surface Radiation Network data). Compared to the commonly used ERA5 IST data, the retrieved IST in this study has higher accuracy. We retrieved Antarctic IST from 2013 to 2020. Antarctic IST decreased continuously from 2013 to 2015. After 2015, Antarctic IST increased with large fluctuations.

The safe and efficient navigation of ships traversing the Arctic waters demands accurate information regarding sea ice concentrations. This study investigates the prediction of Arctic sea ice concentration using historical low-frequency sea ice concentration data, comparing the predictive effects of the SARIMA model and the XGBoost model. Moreover, the study examines the influence of the input data length on prediction accuracy when the XGBoost model is used for single-step predictions with high-frequency data and compares the forecasting performance of the XGBoost model for single-step and multi-step forecast. The experimental results demonstrate that the SARIMA model outperforms the XGBoost model in predicting low-frequency Arctic sea ice concentration data. Furthermore, the XGBoost model provides the most accurate predictions for single-step forecast with high-frequency data when the input data length is 50. The single-step forecast was more accurate than seven-step forecast when the input data length was 50 and the output data length was 7. Therefore, in practical applications, the appropriate model can be selected based on the specific sea ice prediction task and data situation, providing more reliable forecasting results for the planning and managing the Arctic shipping route.

A long-term time series of ice sheet surface elevation change (SEC) is an essential parameter to assess the impact of climate change. In this study, we used an updated plane-fitting least-squares regression strategy to generate a 30-year surface elevation time series for the Greenland Ice Sheet (GrIS) at monthly temporal resolution and 5×5 km grid spatial resolution using ERS-1 (European Remote Sensing), ERS-2, Envisat, and CryoSat-2 satellite radar altimeter observations obtained between August 1991 and December 2020. The ingenious corrections for intermission bias were applied using an updated plane-fitting least-squares regression strategy. Empirical orthogonal function (EOF) reconstruction was used to supplement the sparse monthly gridded data attributable to poor observations in the early years. Validation using both airborne laser altimeter observations and the European Space Agency GrIS Climate Change Initiative (CCI) product indicated that our merged surface elevation time series is reliable. The accuracy and dispersion of errors of SECs of our results were 19.3 % and 8.9 % higher, respectively, than those of CCI SECs and even 30.9 % and 19.0 % higher, respectively, in periods from 2006–2010 to 2010–2014. Further analysis showed that our merged time series could provide detailed insight into GrIS SEC on multiple temporal (up to 30 years) and spatial scales, thereby providing an opportunity to explore potential associations between ice sheet change and climatic forcing. The merged surface elevation time series data are available at https://doi.org/10.11888/Glacio.tpdc.271658 (Zhang et al., 2021).

Sea ice motion is an essential parameter when determining sea ice deformation, regional advection, and the outflow of ice from the Arctic Ocean. The Robeson Channel, which is located between Ellesmere Island and northwest Greenland, is a narrow but crucial channel for ice outflow. Only three Eulerian sea ice motion products derived from ocean/sea ice reanalysis are available: GLORYS12V1, PSY4V3, and TOPAZ4. In this study, we used Lagrangian ice motion in the Robeson Channel derived from Sentinel-1 images to assess GLORYS12V1, PSY4V3, and TOPAZ4. The influence of the presence of ice arches, and wind and tidal forcing on the accuracies of the reanalysis products was also investigated. The results show that the PSY4V3 product performs the best as it underestimates the motion the least, whereas TOPAZ4 grossly underestimates the motion. This is particularly true in regimes of free drift after the formation of the northern arch. In areas with slow ice motion or grounded ice floes, the GLORYS12V1 and TOPAZ4 products offer a better estimation. The spatial distribution of the deviation between the products and ice floe drift is also presented and shows a better agreement in the Robeson Channel compared to the packed ice regime north of the Robeson Channel.

Sea ice concentration (SIC) is an indispensable parameter for the study of polar sea ice. The existing methods can obtain accurate SICs for most situations, but they usually perform poorly in low SIC regions because of the spatial differences in the neighboring pixels induced by the discontinuity of the sea ice cover. In this letter, to cope with the difficulty of this problem, an improved SIC estimation method is proposed to retrieve SIC, focusing on low SIC regions. The proposed method introduces the spatial relationships into SIC estimation by employing a total variation (TV) regularizer. Moreover, nonnegative constrained least squares (NCLS) is used to derive the optimal solutions from the SIC estimation equation. Verification was conducted in low SIC regions (0%-50%) of the Antarctic utilizing ship-based in situ data and the Moderate Resolution Imaging Spectroradiometer (MODIS), and the results were compared with those of some of the mature methods. The results indicated that the proposed method can obtain a superior accuracy with a smaller root-mean-square error (RMSE) (6.0%-14.61%) than the other algorithms in low SIC regions. Furthermore, the proposed method can accurately estimate the SIC of both first-year ice and multiyear ice. The findings of this study confirm the need to consider the spatial relationships in the processing of SIC estimation.