Researchers at the Wrocław University of Environmental and Life Sciences (UPWr) have published a paper in Satellite Navigation this month describing a method that uses deep learning to turn low-resolution global navigation satellite system (GNSS)-based atmospheric readings into sharper 3D maps of humidity.
The work focuses on water vapor, which plays a key role in thunderstorms, flash floods and hurricanes. The researchers say satellites have struggled to capture humidity in enough detail to help warn of fast-changing weather.
To improve the data, the team used a super-resolution generative adversarial network (SRGAN), a type of AI commonly used to make grainy images clearer. In this case, the model was trained on global weather data and powered by NVIDIA GPUs. The system “upscaled” low-resolution readings from navigation satellites into high-resolution humidity maps with fewer errors.
According to the source text, the method reduced forecast errors by 62% in Poland and 52% in California, even in rainy conditions. The researchers said the AI produced sharp gradients that matched what ground instruments observed, unlike older methods that blurred details.
The paper also includes explainable AI tools, using Grad-CAM and SHAP to show where the model focused when making decisions. The visualizations showed attention on storm-prone areas, including Poland’s western borders and California’s coastal mountains.
Saeid Haji-Aghajany, Assistant Professor at Wrocław University of Environmental and Life Sciences, said: “High-resolution, reliable humidity data is the missing link in forecasting the kind of weather that disrupts lives. Our approach doesn’t just sharpen GNSS tomography — it also shows us how the model makes its decisions. That transparency is critical for building trust as AI enters weather forecasting.”
The researchers say sharper humidity fields could be fed into physics-based or AI-driven weather models to improve forecasts and help catch sudden downpours or flash floods earlier.
Source: blogs.nvidia.com.
Companies can share verified announcements through Newz9’s international press release submission page.
