September 5, 2025Open Access

A Novel Approach for Reliable Classification of Marine Low Cloud Morphologies with Vision–Language Models

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Analysis reveals vision-language models classify cloud morphologies accurately, suggesting benefits for satellite imagery and climate models.

The vision-language model achieves an overall accuracy of 0.84 for mesoscale cloud types, enhancing cloud classification reliability.
K-fold cross-validation shows that the VLM performs robustly under limited training sizes, making it suitable for diverse cloud datasets.
For marine stratocumulus types, the model attains 0.86 accuracy, significantly outperforming traditional image-only models under similar conditions.
These findings suggest that vision-language models could reshape how satellite remote sensing is applied in climate science.

Erfani et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Analysis reveals vision-language models classify cloud morphologies accurately, suggesting benefits for satellite imagery and climate models.