The rapid development of artificial intelligence (AI) and deep learning has revolutionized the field of data analysis in recent years, including signal data acquired by remote sensors. Light Detection and Ranging (LiDAR) technology is widely used in atmospheric research for measuring various atmospheric parameters. Wind measurement using LiDAR data has traditionally relied on the spectral centroid (SC) algorithm. However, this approach has limitations in handling LiDAR data, particularly in low signal-to-noise ratio (SNR) regions. To overcome these limitations, this study leverages the capabilities of customized deep-learning techniques to achieve accurate wind profile reconstruction. The study uses datasets obtained from the European Centre for Medium Weather Forecasting (ECMWF) Reanalysis v5 (ERA5) and the mobile Incoherent Doppler LiDAR (ICDL) system constructed by the University of Science and Technology of China. We present a simulation-based approach for generating wind profiles from the statistical data and the associated theoretical calculations. Whereafter, our team constructed a convolutional neural network (CNN) model based on the U-Net architecture to replace the SC algorithm for LiDAR data post-processing. The CNN-generated results are evaluated and compared with the SC results and the ERA5 data. This study highlights the potential of deep learning-based techniques in atmospheric research and their ability to provide more accurate and reliable results.

Stratospheric airship navigation requires accurate wind field measurements at a ~20 km altitude, where low pressure and density limit the effectiveness of conventional wind sensors. To address this, we present a compact direct-detection Rayleigh Doppler wind lidar based on the molecular double-edge technique. The system utilizes a 532 nm fiber-coupled pulsed laser (0.5 W, 5 ns) and a fixed-cavity dual-channel Fabry–Perot etalon as the frequency discriminator. A liquid crystal variable retarder (LCVR) combined with a polarization beam splitter (PBS) enables non-mechanical, high-speed beam switching between two orthogonal line-of-sight (LOS) directions for horizontal wind measurement. Systematic tests are performed in controlled wind fields within Mie-dominated and Rayleigh-dominated regimes. The lidar effectively captures the sharp radial velocity profiles at wind speeds up to 7.6 m/s. Comparative experiments with a reference anemometer show that the system delivers reliable performance at 0.48 m range resolution, with measurement uncertainty below 0.34 m/s. With its compact, lightweight, and high-precision design, the developed lidar demonstrates reliable wind measurement capability under laboratory conditions, indicating its potential for future deployment on stratospheric airships.

Large language models (LLMs) offer new possibilities for natural-language interaction with geospatial analysis systems, but their use in remote sensing instrument data analysis remains limited by weak execution control, poor reproducibility, and limited integration with domain-specific computation. The paper presents an agent for Incoherent Doppler wind LiDAR (ICDL) data analysis, named ICDL-Agent, a tool-augmented LLM framework for remote sensing instrument workflows. The system maps conversational user requests to executable analysis pipelines for wind retrieval, uncertainty estimation, visualization, and higher-level diagnostics through structured planning over a registry of domain-specific tools. To improve execution reliability, the system combines schema-constrained workflow generation, shared-state reuse of intermediate scientific products, and validation with bounded repair. In addition to supporting routine LiDAR processing, the framework can generate new tools when required and adapt to related analytical tasks through domain-aware guidance and procedural documentation. We evaluate the system on multiple atmospheric wind-observation datasets in China and show that it faithfully reproduces the refined Doppler wind-retrieval pipeline, achieving representative R²/MAE values of 0.52/3.73 m/s against ERA5 and 0.80/2.31 m/s against radiosonde observations, while supporting downstream analyses such as profile comparison, climatological interpretation, and gravity-wave diagnostics. More broadly, this study demonstrates how constrained LLM orchestration can support LiDAR researchers, remote-sensing instrument teams, and geospatial analysts seeking transparent, reproducible, and automated scientific data-processing workflows.