Restoration of missing ocean color data in high-latitude oceans using neural network model
Satellite ocean color remote sensing plays a crucial role in monitoring marine environment at both regional and global scales. However, due to the reduced accuracy of atmospheric correction models under large solar zenith angles (≥70°), publicly available satellite ocean color products lack valid da...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-04-01
|
| Series: | Big Earth Data |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/20964471.2025.2474655 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Satellite ocean color remote sensing plays a crucial role in monitoring marine environment at both regional and global scales. However, due to the reduced accuracy of atmospheric correction models under large solar zenith angles (≥70°), publicly available satellite ocean color products lack valid datasets for high-latitude oceans (≥50°S or ≥50°N) during winter. Based on a neural network atmospheric correction model designed for high solar zenith angle observation environments (which used a Rayleigh scattering lookup table generated by PCOART-SA to compute Rayleigh scattering radiance and a neural network model to invert remote sensing reflectance from Rayleigh-corrected radiance), this study has established a monthly ocean color product dataset for high-latitude oceans, named NN-LAT50, covering the period from 2003 to 2020. We validated the accuracy of the ocean color products in NN-LAT50 dataset using multiple in situ datasets, and the results indicated that NN-LAT50 had more reliable and accurate retrievals compared to the NASA released ocean color products in high latitude oceans. Furthermore, during autumn and winter, coverage of the NN-LAT50 dataset far exceeds that of products released by NASA. For instance, during the winter in the Southern Hemisphere, the coverage rates are 3.02% for MODIS/Aqua, 21.59% for VIIRS, and 1.74% for OLCI, while the NN-LAT50 dataset maintains a coverage rate of 38.64%. This study is the first to establish a long-term (2003–2020) ocean color product dataset covering high-latitude oceans during winter, which can significantly enhance the observation of ecological changes in polar and subpolar oceans. |
|---|---|
| ISSN: | 2096-4471 2574-5417 |