Related Experiment Video
Updated: Mar 29, 2026
Early Detection of Cyanobacterial Blooms and Associated Cyanotoxins using Fast Detection Strategy
Published on: February 25, 2021
A comprehensive analysis of retrieving optically inactive indicators from multi-level remote sensing product(s) in
Abdul Majed Sajib1, Md Galal Uddin1, Agnieszka I Olbert1
1Department of Civil Engineering, School of Engineering, College of Science and Engineering, University of Galway, Ireland; Ryan Institute, University of Galway, Ireland; Eco-HydroInformatics Research Group (EHIRG), School of Engineering, College of Science and Engineering, University of Galway, Ireland.
Researchers developed 2101 models to retrieve dissolved oxygen (DOX) from Irish waters using remote sensing (RS) data. Model quantity did not improve performance, highlighting challenges in using machine learning (ML)/artificial intelligence (AI) for optically inactive water quality indicators.
Area of Science:
- Oceanography
- Environmental Science
- Data Science
Background:
- Dissolved oxygen (DOX) is a critical water quality (WQ) indicator.
- Retrieving DOX using remote sensing (RS) is challenging due to its optically inactive nature.
- Copernicus Marine Services products offer potential for WQ monitoring.
Purpose of the Study:
- To retrieve dissolved oxygen (DOX) from Irish transitional and coastal waters using Copernicus Marine Services products.
- To develop and validate numerous machine learning (ML)/artificial intelligence (AI) and statistical models for DOX retrieval.
- To assess the performance of Sentinel-3 OLCI (S3-OLCI) and Multi-sensor (MS) datasets for DOX monitoring.
Main Methods:
- Development and validation of 2101 ML/AI (supervised learning, ensembles) and statistical models.
- Utilized multi-level Sentinel-3 OLCI (S3-OLCI) and Multi-sensor (MS) remote sensing (RS) datasets.
- Incorporated in-situ and modelled DOX datasets for model training and independent validation.
Main Results:
- Supervised ML/AI models showed high training accuracy but poor generalizability on independent validation datasets.
- Sentinel-3 OLCI (S3-OLCI) data demonstrated superior performance over Multi-sensor (MS) data with lower uncertainty.
- Spatio-temporal analysis revealed highest DOX inshore/semi-enclosed bays and lowest offshore.
Conclusions:
- Model performance is dictated by methodological characteristics, not quantity.
- Retrieving optically inactive water quality indicators like DOX using RS and ML/AI presents significant validation challenges.
- Findings support mapping baseline oxygen conditions and advancing ML/AI for WQ indicator retrieval.
