Sea ice leads in the Arctic Ocean and Antarctic Ocean are of great significance to polar ecology, climate change, and ship navigation. While rule-based remote-sensing classification methods, such as the threshold method, have been widely used for surveying and mapping polar sea ice leads, they have difficulty in overcoming the problems of noise and poor generalization. In this article, we presented an automated, deep learning sea ice open lead detection network in low wind speed conditions, entropy-weighted network (EW-Net). In EW-Net, a dense block was introduced into a U-Net baseline network to strengthen feature propagation, and an entropy sampling structure was designed to improve the pertinence of the network to leads and alleviate the influence of noise. Furthermore, an entropy-weighted feature fusion block was designed to better restore features. EW-Net was trained on the Sentinel-1 synthetic aperture radar (SAR) dataset in order to obtain lead maps of different polar regions from Sentinel-1 images. The results showed that the proposed EW-Net model was more effective and more generalizable for polar sea ice lead detection than the threshold method and the deep learning methods U-Net and DeepLab V3 Plus, with an overall accuracy (OA) exceeding 0.97. In addition, EW-Net showed advantages in terms of outstanding generalization from cross-sensor image monitoring of leads. The method proposed in this article could be beneficial for providing accurate maps of polar sea ice leads in open water and contribute to further research on polar science.

Building segmentation is a classical and challenging task in high-resolution remote sensing imagery. This approach has achieved remarkable performance based on a fully convolutional network with adequate pixel-wise annotations. However, due to differences in sensor technology as well as appearance in different regions, datasets gathered from these various sources are quite distinct, and dense annotations for a particular area are not always available. Thus, directly applying a segmentation model trained on one dataset (source domain) to another unseen dataset (target domain) usually results in a drop in performance, called the domain gap. In this article, we propose a weakly-supervised domain adaptation method using adversarial entropy for building segmentation to address this problem. First, we use an adversarial entropy strategy to decrease the entropy and improve the prediction certainty for target images, causing the distributions between the source and target domains to become closer to each other. Second, we propose a simple and effective self-training strategy for the target domain that produces high-confidence predictions using pseudolabels. We use a series of thresholds to generate the pseudolabels without introducing extra parameters. This strategy effectively enhances the discriminability of the target domain and further minimizes the distribution discrepancy between the two domains. Experiments on cross-domain aerial datasets have demonstrated the effectiveness and superiority of our proposed method when compared to other state-of-the-art methods.

The COVID-19 outbreak is under control in China. Mobility interventions, including both the Wuhan lockdown and travel restrictions in other cities, have been undertaken in China to mitigate the epidemic. However, the impact of mobility restrictions in cites outside Wuhan has not been systematically analyzed. Here we ascertain the relationships between all mobility patterns and the epidemic trajectory in Chinese cities outside Hubei Province, and we estimate the impact of local travel restrictions. We estimate local inter-city travel bans averted 22.4% (95% PI: 16.8–27.9%) more infections in the two weeks after the Wuhan lockdown, while local intra-city travel prevented 32.5% (95% PI: 18.9–46.1%) more infections in the third and fourth weeks. More synchronized implementation of mobility interventions would further decrease the number of confirmed cases in the first two weeks by 15.7% (95% PI:15.4–16.0%). This study shows synchronized travel restrictions across cities can be effective in COVID-19 control.

Wetlands are one of the most important ecosystems on the Earth and play a critical role in regulating regional climate, preventing floods, and reducing flood severity. However, it is difficult to detect wetland changes in multitemporal Landsat 8 OLI satellite images due to the mixed composition of vegetation, soil, and water. The main objective of this study is to quantify change to wetland cover by an image-to-image comparison change detection method based on the image fusion of multitemporal images. Spectral distortion is regarded as candidate change information, which is generated by the spectral and spatial differences between multitemporal images during the process of image cross-fusion. Meanwhile, the normalized difference vegetation index (NDVI) and normalized difference water index (NDWI) were extracted from the cross-fused image as a normalized index image to enhance and increase the information about vegetation and water. Then, the modified iteratively reweighted multivariate alteration detection (IR-MAD) is applied to the generally fused images and normalized difference index images, providing a good evaluation of spectral distortion. The experimental results show that the proposed method performed better to reduce the detection errors due to the complicated areas under different ground types, especially in cultivated areas and forests. Moreover, the proposed method was tested and quantitatively assessed and achieved an overall accuracy of 96.67% and 93.06% for the interannual and seasonal datasets, respectively. Our method can be a tool to monitor changes in wetlands and provide effective technical support for wetland conservation.

ICESat-2 altimeter data could be used to estimate sea ice freeboard and thickness values with a higher measuring accuracy than that achievable with data provided by previous altimeter satellites. This study developed an improved algorithm considering variable lead proportions based on the lowest point method to derive the sea surface height for the retrieval of Antarctic sea ice freeboard and thickness values from ICESat-2 ATL-07 data. We first collocated ICESat-2 tracks to corresponding Sentinel-1 SAR images and calculated lead (seawater) proportions along each track to estimate the sea surface height in the Antarctic Ocean. Then, the Antarctic sea ice freeboard and thickness were estimated based on a local sea surface height reference and a static equilibrium equation. Finally, we assessed the accuracy of our improved algorithm and ICESat-2 data product in the retrieval of the Antarctic sea ice thickness by comparing the calculated values to ship-based observational sea ice thickness values acquired during the 35th Chinese Antarctic Research Expedition (CHINARE-35). The results indicate that the Antarctic sea ice freeboard estimated with the improved lowest point method was slightly larger than that estimated with the ICESat-2 data product algorithm. The root mean squared error (RMSE) of the improved lowest point method was 35 cm with the CHINARE-35 measured sea ice thickness, which was smaller than that determined with the ICESat-2 data product algorithm (65 cm). Our improved algorithm could provide more accurate data on the Antarctic sea ice freeboard and thickness, thus supporting Antarctic sea ice monitoring and the evaluation of its change under global effects.

Thermodynamic parameters play a crucial role in determining polar sea ice thickness (SIT); however, modeling their relationship is difficult due to the complexity of the influencing mechanisms. In this study, we propose a self-attention convolutional neural network (SAC-Net), which aims to model the relationship between thermodynamic parameters and SIT more parsimoniously, allowing us to estimate SIT directly from these parameters. SAC-Net uses a fully convolutional network as a baseline model to detect the spatial information of the thermodynamic parameters. Furthermore, a self-attention block is introduced to enhance the correlation among features. SAC-Net was trained on a dataset of SIT observations and thermodynamic data from the 2012–2019 freeze-up period, including surface upward sensible heat flux, surface upward latent heat flux, 2 m temperature, skin temperature, and surface snow temperature. The results show that our neural network model outperforms two thermodynamic-based SIT products in terms of accuracy and can provide reliable estimates of SIT. This study demonstrates the potential of the neural network to provide accurate and automated predictions of Arctic winter SIT from thermodynamic data, and, thus, the network can be used to support decision-making in certain fields, such as polar shipping, environmental protection, and climate science.