Integration of Sentinel-1 and Sentinel-2 Data for Land Cover Mapping Using W-Net

Area of Science:

• Remote Sensing
• Earth Observation
• Geospatial Analysis

Background:

• Sentinel 2 data is crucial for land cover mapping but is limited by cloud cover.
• Existing methods struggle with data gaps caused by persistent cloud cover.
• Accurate land cover classification is essential for environmental monitoring and management.

Purpose of the Study:

• To develop a new approach for fusing Sentinel 1 (S1) and Sentinel 2 (S2) data for enhanced land cover mapping.
• To overcome the limitations of Sentinel 2 data caused by cloud cover by integrating Sentinel 1 data.
• To improve the accuracy and efficiency of land cover classification, particularly for distinguishing rice, water, and bare soil.

Main Methods:

• A novel multi-temporal W-Net approach is proposed for segmenting Interferometric Wide swath mode (IW) Sentinel-1 data.
• Sentinel 1 data is used to generate Sentinel 2-like segmentation maps to fill data gaps.
• The method utilizes multi-temporal Sentinel-1 data from ascending/descending orbits for improved discrimination of land cover types.

Main Results:

• The proposed multi-temporal W-Net approach significantly improves segmentation accuracy by 0.18 and F1-score by 0.25 compared to single-date methods.
• The fusion of Sentinel 1 and Sentinel 2 data effectively addresses cloud cover issues, enabling continuous land cover mapping.
• Performance gains were observed in classical segmentation metrics and computational time in the Albufera National Park.

Conclusions:

• The developed W-Net based solution offers a robust and efficient method for land cover mapping by integrating Sentinel 1 and Sentinel 2 data.
• This approach enhances the reliability of land cover classification in areas prone to cloud cover.
• The study demonstrates the potential of multi-temporal Sentinel-1 data for improving the accuracy and reducing the computational cost of land cover mapping.