Although the Advanced Topographic Laser Altimeter System (ATLAS) onboard the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) was primarily designed for glacier and sea-ice measurement, it can also be applied to monitor lake surface height (LSH). However, its performance in monitoring lakes/reservoirs has rarely been assessed. Here, we report an accuracy evaluation of the ICESat-2 laser altimetry data over 30 reservoirs in China using gauge data. To show its characteristics in large-scale lake monitoring, we also applied an advanced radar altimeter SARAL (Satellite for ARgos and ALtika) and the first laser altimeter ICESat (Ice, Cloud and land Elevation Satellite) to investigate all lakes and reservoirs (>10 km²) in China. We found that the ICESat-2 has a greatly improved altimetric capability, and the relative altimetric error was 0.06 m, while the relative altimetric error was 0.25 m for SARAL. Compared with SARAL and ICESat data, ICESat-2 data had the lowest measurement uncertainty (the standard deviation of along-track heights; 0.02 m vs. 0.17 m and 0.07 m), the greatest temporal frequency (3.43 vs. 1.35 and 1.48 times per year), and the second greatest lake coverage (636 vs. 814 and 311 lakes). The precise LSH profiles derived from the ICESat-2 data showed that most lakes (90% of 636 lakes) had a quasi-horizontal LSH profile (measurement uncertainty <0.05 m), and special methods are needed for mountainous lakes or shallow lakes to extract precise LSHs.

Lake volume variation is closely related to climate change and human activities, which can be monitored by multi-source remote-sensing data from space. Although there are usually two routine ways to construct the lake volume by the digital elevation model (DEM) or satellite altimetric data combined with the lake area, rarely has a comparison been made between the two methods. Therefore, we conducted a comparison between the two methods in Texas for 14 lakes with abundant validation data. First, we constructed the lake hypsometric curve by five commonly applied DEMs (SRTM, ASTER, ALOS, GMTED2010, and NED) or satellite altimetric products combined with the gauge lake area. Second, the lake volume was estimated by combining the hypsometric curve with the gauge lake area time series. Finally, the estimation error has been quantitatively calculated. The results show that the relative lake volume estimation error (rVSD) of the altimetric data (4%) is only 10–18% of that of the DEMs (22–41%), and the DEM with the highest resolution (NED) has the least rVSD with an average of 22%. Therefore, for large-scale lake monitoring, we suggest the application of satellite altimetric data with the lake area to estimate the lake volume of large lakes, and the application of high-resolution DEM with the lake area to calculate the lake volume of small lakes that are gapped by satellite altimetric data.

Automatic modulation classification (AMC) is a basic technology in intelligent wireless communication systems. It is important for tasks such as spectrum monitoring, cognitive radio, and secure communications. In recent years, deep learning methods have made great progress in AMC. However, mainstream methods still face two key problems. First, they often use time-frequency images instead of raw signals. This causes loss of key modulation features and reduces adaptability to different communication conditions. Second, most methods rely on supervised learning. This needs a large amount of labeled data, which is hard to get in real-world environments. To solve these problems, we propose a self-supervised learning framework called RIS-MAE. RIS-MAE uses masked autoencoders to learn signal features from unlabeled data. It takes raw IQ sequences as input. By applying random masking and reconstruction, it captures important time-domain features such as amplitude, phase, etc. This helps the model learn useful and transferable representations. RIS-MAE is tested on four datasets. The results show that it performs better than existing methods in few-shot and cross-domain tasks. Notably, it achieves high classification accuracy on previously unseen datasets with only a small number of fine-tuning samples, confirming its generalization ability and potential for real-world deployment.