Lake ice is a robust indicator of climate change. The availability of information contained in Moderate Resolution Imaging Spectroradiometer (MODIS) daily snow products from 2000 to 2017 could be greatly improved after cloud removal by gap filling. Thresholds based on open water pixel numbers are used to extract the freeze‐up start and break‐up end dates for 58 lakes on the Tibetan Plateau (TP), 18 lakes are also selected to extract the freeze‐up end and break‐up start dates. The lake ice durations are further calculated based on freeze‐up and break‐up dates. Lakes on the TP begin to freeze‐up in late October and all the lakes start the ice cover period in mid‐January of the following year. In late March, some lakes begin to break‐up, and all the lakes end the ice cover period in early July. Generally, the lakes in the northern Inner‐TP have earlier freeze‐up dates and later break‐up dates (i.e. longer ice cover durations) than those in the southern Inner‐TP. Over 17 years, the mean ice cover duration of 58 lakes is 157.78 days, 18 (31%) lakes have a mean extending rate of 1.11 d yr^‐1 and 40 (69%) lakes have a mean shortening rate of 0.80 d yr^‐1. Geographical location and climate conditions determine the spatial heterogeneity of the lake ice phenology, especially the ones of break‐up dates, while the physico‐chemical characteristics mainly affect the freeze‐up dates of the lake ice in this study. Ice cover duration is affected by both climatic and lake specific physico‐chemical factors, which can reflect the climatic and environmental change for lakes on the TP.

High Mountain Asia (HMA) glaciers are critical water reserves for montane regions, which are readily influenced by climate change. The glacier mass balance during 2000–2021 over HMA was estimated by comparing the elevations from ICESat-2 and the NASADEM. Radar penetration depth could be one of the intrinsic error sources in estimating glacier mass balance by using NASADEM. Therefore, we doubled elevation differences between the X-band Shuttle Radar Topography Missions (SRTMs) and NASADEM to estimate the potential error. The spatial characteristics of the altitude-dependent penetration depth can be detected in most sub-regions of HMA. Relatively deep penetrations in the Himalaya (2.3–3.7 m) and Hissar Alay (4.3 m) regions and small penetrations in the south-eastern HMA (1.0 m) were observed. The HMA region experienced a significant mass loss at a rate of −0.18 ± 0.12 m w.e. a⁻¹, in which the Hengduan Shan exhibited the highest mass loss of −0.62 ± 0.10 m w.e. a⁻¹, the West Kun Lun experienced a substantial mass gain of 0.23 ± 0.13 m w.e. a⁻¹, and the Karakoram showed a more or less balance. Our results are in agreement with previous studies that assessed the mass balance of HMA glaciers from different methods.

Daily snow observation data from 672 stations in China, particularly the 296 stations with over 10 mean snow cover days (SCDs) in a year during the period of 1952–2010, are used in this study. We first examine spatiotemporal variations and trends of SCDs, snow cover onset date (SCOD), and snow cover end date (SCED). We then investigate the relationships of SCDs with number of days with temperature below 0 °C (TBZD), mean air temperature (MAT), and Arctic Oscillation (AO) index. The results indicate that years with a positive anomaly of SCDs for the entire country include 1955, 1957, 1964, and 2010, and years with a negative anomaly of SCDs include 1953, 1965, 1999, 2002, and 2009. The reduced TBZD and increased MAT are the main reasons for the overall late SCOD and early SCED since 1952. This explains why only 12 % of the stations show significant shortening of SCDs, while 75 % of the stations show no significant change in the SCDs trends. Our analyses indicate that the distribution pattern and trends of SCDs in China are very complex and are not controlled by any single climate variable examined (i.e. TBZD, MAT, or AO), but a combination of multiple variables. It is found that the AO has the maximum impact on the shortening trends of SCDs in the Shandong peninsula, Changbai Mountains, Xiaoxingganling, and north Xinjiang, while the combined TBZD and MAT have the maximum impact on the shortening trends of SCDs in the Loess Plateau, Tibetan Plateau, and Northeast Plain.

Sea ice type is one of the most sensitive variables in Arctic ice monitoring and detailed information about it is essential for ice situation evaluation, vessel navigation, and climate prediction. Many machine-learning methods including deep learning can be employed for ice-type detection, and most classifiers tend to prefer different feature combinations. In order to find the optimal classifier-feature assembly (OCF) for sea ice classification, it is necessary to assess their performance differences. The objective of this letter is to make a recommendation for the OCF for sea ice classification using Cryosat-2 (CS-2) data. Six classifiers including convolutional neural network (CNN), Bayesian, K nearest-neighbor (KNN), support vector machine (SVM), random forest (RF), and back propagation neural network (BPNN) were studied. CS-2 altimeter data of November 2015 and May 2016 in the whole Arctic were used. The overall accuracy was estimated using multivalidation to evaluate the performances of individual classifiers with different feature combinations. Overall, RF achieved a mean accuracy of 89.15%, followed by Bayesian, SVM, and BPNN (~86%), outperforming the worst (CNN and KNN) by 7%. Trailing-edge width (TeW) and leading-edge width (LeW) were the most important features, and feature combination of TeW, LeW, Sigma0, maximum of the returned power waveform (MAX), and pulse peakiness (PP) was the best choice. RF with feature combination of TeW, LeW, Sigma0, MAX, and PP was finally selected as the OCF for sea ice classification and the results that demonstrated this method achieved a mean accuracy of 91.45%, which outperformed the other state-of-art methods by 9%.