Reconstruction of the three-dimensional dissolved oxygen and its spatio-temporal variations in the Mediterranean Sea using machine learning
Journal of Environmental Sciences · 2025-11 · 157 · 710-728
DOI: 10.1016/j.jes.2025.01.010 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Environmental Sciences, 2025-11. Verfasst von Guangsheng Liu und 5 weiteren Autor:innen. Beteiligte Einrichtung: Remote Sensing Information and Digital Earth Center, School of Computer Science and Technology, Qingdao University, Qingdao 266071, China; Aerospace …. Hauptthemen laut MEDLINE: Oxygen, Environmental Monitoring, Machine Learning. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 76 Literaturangaben. Gefördert von: Qingdao University, Forschungszentrum Dynamische Systeme, Otto-von-Guericke-Universität Magdeburg, Natural Science Foundation of Shandong Province.
Abstract (Originalsprache)
Oceanic dissolved oxygen (DO) concentration is crucial for assessing the status of marine ecosystems. Against the backdrop of global warming, DO shows a general decrease, posing a threat to the health of marine ecosystems. Therefore, there is an urgent need to develop advanced tools to characterize the spatio-temporal variations of three-dimensional (3D) DO. To address this challenge, this study introduces the Light Gradient Boosting Machine (LightGBM), combining satellite remote sensing and reanalysis data with Biogeochemical Argo data to accurately reconstruct the 3D DO structure in the Mediterranean Sea from 2010 to 2022. Various environmental parameters are incorporated as inputs, including spatio-temporal features, meteorological characteristics, and ocean color properties. The LightGBM model demonstrates excellent performance on the testing dataset with R2 of 0.958. The modeled DO agrees better with in-situ measurements than products from numerical models. Using the Shapley Additive exPlanations method, the contributions of input features are assessed. Sea surface temperatures provide a correlation with DO at the sea surface, while spatial coordinates supplement the view of the ocean interior. Based on the reconstructed 3D DO structure, we identify an oxygen minimum zone in the western Mediterranean that expands continuously, reaching depths of approximately 300-800 m. The western Mediterranean exhibits a significant declining trend. This study enhances marine environmental evidence by proposing a precise and cost-effective approach for reconstructing 3D DO, thereby offering insights into the dynamics of DO variations under changing climatic conditions.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Remote Sensing Information and Digital Earth Center, School of Computer Science and Technology, Qingdao University, Qingdao 266071, China; Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 1…
- Remote Sensing Information and Digital Earth Center, School of Computer Science and Technology, Qingdao University, Qingdao 266071, China.
- Department of Biological and Agricultural Engineering, University of California, Davis 95616, USA.
Förderung
- Qingdao University
- Forschungszentrum Dynamische Systeme, Otto-von-Guericke-Universität Magdeburg
- Natural Science Foundation of Shandong Province
Zitieren
APA
Liu, G., Yu, X., Zhang, J., Wang, X., Xu, N., & Ali, S. (2025). Reconstruction of the three-dimensional dissolved oxygen and its spatio-temporal variations in the Mediterranean Sea using machine learning. Journal of Environmental Sciences, 157, 710-728. https://doi.org/10.1016/j.jes.2025.01.010
Vancouver
Liu G, Yu X, Zhang J, Wang X, Xu N, Ali S. Reconstruction of the three-dimensional dissolved oxygen and its spatio-temporal variations in the Mediterranean Sea using machine learning. Journal of Environmental Sciences. 2025;157:710-728. doi:10.1016/j.jes.2025.01.010
BibTeX
@article{Liu2025,
author = {Liu, Guangsheng and Yu, Xiang and Zhang, Jiahua and Wang, Xiaopeng and Xu, Nuo and Ali, Shawkat},
title = {Reconstruction of the three-dimensional dissolved oxygen and its spatio-temporal variations in the Mediterranean Sea using machine learning},
journal = {Journal of Environmental Sciences},
year = {2025},
volume = {157},
pages = {710--728},
doi = {10.1016/j.jes.2025.01.010}
}
EUVIMED