Landsat-Sentinel-2 Algal Bloom Mapping Using Vision Transformers: Model Description, Implementation, and Examples

2026-06-15 · Source: Computer Vision and Pattern Recognition · Field: Science & Research — Environmental Science & Earth Systems, Artificial Intelligence & Machine Learning · Depth: Advanced, quick

Summary

A study introduces the first successful implementation of vision transformer-based coastal algal bloom mapping using 30-m Landsat-8/9 and Sentinel-2 A/B/C images. These missions provide frequent, spatially detailed observations every 2-3 days, crucial for detecting fragmented bloom structures. Deep learning offers a data-driven alternative to traditional bio-optical methods, overcoming spectral limitations. The research generated a globally distributed bloom patch dataset and compared four transformer architectures against a convolutional baseline. All deep learning models effectively detected floating bloom areas, with 8-65% omission and commission errors. Notably, the Swin Transformer outperformed traditional spectral-index approaches under cloud and glint stress, avoiding false positives. Higher spatial resolution also proved beneficial over MODIS-derived products for fragmented blooms.

Key takeaway

For environmental scientists and remote sensing specialists monitoring dynamic coastal environments, adopting vision transformer-based methods for algal bloom detection is highly recommended. This approach provides consistent, high-resolution mapping using Landsat-Sentinel-2 data, effectively mitigating false positives from cloud and glint interference that challenge traditional spectral indices. Consider integrating these deep learning models to enhance the accuracy and reliability of your bloom monitoring programs.

Key insights

Vision Transformers enable robust, high-resolution algal bloom mapping from harmonized Landsat-Sentinel-2 imagery, outperforming traditional methods under challenging conditions.

Principles

• Deep learning overcomes spectral limitations in aquatic remote sensing.
• Transformers excel in fine-scale bloom detection.
• Higher spatial resolution improves fragmented bloom detection.

Method

Generated a globally distributed bloom patch dataset. Compared four transformer architectures against a convolutional baseline for 30-m Landsat-Sentinel-2 image classification under various optical water types and atmospheric conditions.

In practice

• Utilize Swin Transformer for cloud/glint-affected areas.
• Integrate Landsat-Sentinel-2 for detailed bloom monitoring.
• Apply deep learning for aquatic remote sensing.

Topics

• Vision Transformers
• Algal Blooms
• Remote Sensing
• Landsat-Sentinel-2
• Deep Learning
• Environmental Monitoring

Best for: Computer Vision Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• HLS-GPT: A Generative Pretrained Transformer (GPT) for Continental-Scale NASA Harmonized Landsat and Sentinel-2 (HLS) Reflectance Reconstruction Across All Bands on Arbitrary Dates — HLS-GPT reconstructs continental-scale Landsat and Sentinel-2 reflectance time series for all bands and dates using a hierarchical Transformer.
• Explainable Flood Segmentation on Sentinel-1 SAR Imagery: A Comparative Study of CNN and Transformer Architectures — SegFormer-b2 excels in SAR flood segmentation, outperforming CNNs on specific datasets, with distinct explainability characteristics.
• LatticeVision: Image to Image Networks for Modeling Non-Stationary Spatial Data — Image-to-image networks globally estimate non-stationary spatial parameters faster and more accurately than local methods.
• ATT-CR: Adaptive Triangular Transformer for Cloud Removal — ATT-CR uses triangular attention and feature gating to efficiently remove clouds from remote sensing images, improving accuracy.
• SWNet: A Cross-Spectral Network for Camouflaged Weed Detection — Cross-spectral fusion and edge refinement are crucial for accurate camouflaged weed detection in agriculture.
• LALE: Lightweight-Transformer Architecture for Land-Cover Estimation — LALE efficiently segments remote sensing imagery by combining ConvMixers for local detail and transformers for global context in a bifurcated encoder.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.