Multi-source satellite remote sensing of harmful algal blooms in the East China Sea: cross-sensor comparison of GOCI, Sentinel-2 and Sentinel-3
- 1College of Ocean Science and Technology, Shandong University of Science and Technology, Qingdao, Shandong, China
- 2Key Laboratory of Marine Ecological Monitoring and Restoration Technologies, East China Sea Ecological Center, Ministry of Natural Resources, Shanghai, China
Abstract
IntroductionHarmful algal blooms (HABs) represent recurring coastal ecological hazards that threaten aquaculture, marine ecosystems and near‑shore environmental safety. Satellite remote sensing provides essential observational capacity for HAB monitoring; nevertheless, systematic quantitative inter‑comparisons between geostationary GOCI‑II and polar‑orbiting Sentinel‑2 MSI / Sentinel‑3 OLCI for identical bloom events remain limited. This study addresses the complementary strengths and limitations of these three ocean‑colour sensors for East China Sea HAB observation.MethodsWe selected two well‑documented HAB episodes occurring in the East China Sea during 2021 (30 April and 22 September). GOCI‑II, Sentinel‑2 MSI and Sentinel‑3 OLCI datasets were processed using sensor‑matched atmospheric correction schemes (standard GOCI‑II ocean‑colour processing, ACOLITE‑DSF for Sentinel‑2, C2RCC for Sentinel‑3 OLCI). Six spectral indices including NDCI, RBD, NRTI, RAB, PDI and DI were applied for bloom detection. Spatial‑degradation experiments at 20 m, 300 m and 500 m were implemented to separate geometric scale‑related effects on bloom area and boundary representation. Near‑simultaneous overpass comparisons and hourly GOCI‑II time‑series tracking were performed to quantify inter‑sensor discrepancies and intra‑day bloom dynamics.ResultsUnder ~10‑minute near‑coincident overpass on 30 April 2021, GOCI‑II mapped a bloom area of 901.52 km² versus 571.52 km² from Sentinel‑2B; the difference originated jointly from spatial resolution, spectral configuration, atmospheric correction and detection thresholds, while bloom advection exerted minor influence. On 22 September 2021, hourly GOCI‑II observations captured a complete intra‑day bloom cycle: bloom optical‑anomaly area varied from 37.30 km² up to a peak of 760.20 km² and declined to 107.67 km², accompanied by a 45.74 km centroid displacement, a dynamic that single polar‑orbiting snapshots only partially captured. Spatial degradation analysis demonstrated that boundary‑length loss exceeded 76 – 85 % at coarser resolutions, whereas total‑area deviations ranged from −14.8 % to +11.0 %, highlighting Sentinel‑2’s unique value for fine‑scale near‑shore boundary delineation. Even after resolution normalisation, substantial spatial mismatch persisted among sensor bloom products (IoU as low as 0.014).DiscussionIndividual sensors exhibit distinct trade‑offs: GOCI‑II delivers high‑frequency temporal screening, Sentinel‑2 provides fine‑scale spatial detail, and Sentinel‑3 OLCI supplies rich spectral information for regional contextual analysis. Accordingly, we propose a conditional three‑stage multi‑source synergistic workflow: GOCI‑II for rapid hourly screening and drift tracking, Sentinel‑2 MSI for priority‑zone fine‑scale inspection, and Sentinel‑3 OLCI for regional spectral validation and cross‑sensor consistency checking. Discrepancies across sensors stem from combined effects of spectral band configuration, atmospheric correction, index thresholds and sampling timing. Limitations include only two analysed bloom events and absence of dense pixel‑level in‑situ ground truth. Further multi‑event validation with harmonised atmospheric correction is required to generalise this conditional monitoring template for operational HAB early‑warning in optically‑complex East Asian coastal waters.
Related
- Spatio-Temporal Analysis of Urban Expansion in Awka and its Environs, Nigeria Using Sentinel-2 Satellite Imagery and Google Earth Engine (2021 to 2026)
- Spatial and temporal modelling of pastoral value in the Eastern Alps using Sentinel-2 and satellite embeddings
- Forecasting the moisture content of live forest fuels with a transformer machine learning model and Sentinel-2 satellite data
- Recent Changes in Lake Chad’s Surface Water Extent: A Remote Sensing Assessment Using Sentinel-2 Land Cover Data (2017-2025)
- Decadal dynamics of surface water extent and quality in Fujian Province, China: An integrated optical-SAR remote sensing analysis using Landsat and Sentinel-1 time series (2015–2024)
BibTeX
@article{zhang2026multisource,
title = {{Multi-source satellite remote sensing of harmful algal blooms in the East China Sea: cross-sensor comparison of GOCI, Sentinel-2 and Sentinel-3}},
author = {Chengxin Zhang and Chengxin Zhang and Honghua Shi},
journal = {Frontiers in Marine Science},
year = {2026},
month = oct,
doi = {10.3389/fmars.2026.1966777},
url = {https://doi.org/10.3389/fmars.2026.1966777},
note = {CC BY},
}
RIS
TY - JOUR
TI - Multi-source satellite remote sensing of harmful algal blooms in the East China Sea: cross-sensor comparison of GOCI, Sentinel-2 and Sentinel-3
AU - Chengxin Zhang
AU - Chengxin Zhang
AU - Honghua Shi
PY - 2026
DA - 2026/10/26
JO - Frontiers in Marine Science
DO - 10.3389/fmars.2026.1966777
UR - https://doi.org/10.3389/fmars.2026.1966777
AB - IntroductionHarmful algal blooms (HABs) represent recurring coastal ecological hazards that threaten aquaculture, marine ecosystems and near‑shore environmental safety. Satellite remote sensing provides essential observational capacity for HAB monitoring; nevertheless, systematic quantitative inter‑comparisons between geostationary GOCI‑II and polar‑orbiting Sentinel‑2 MSI / Sentinel‑3 OLCI for identical bloom events remain limited. This study addresses the complementary strengths and limitations of these three ocean‑colour sensors for East China Sea HAB observation.MethodsWe selected two well‑documented HAB episodes occurring in the East China Sea during 2021 (30 April and 22 September). GOCI‑II, Sentinel‑2 MSI and Sentinel‑3 OLCI datasets were processed using sensor‑matched atmospheric correction schemes (standard GOCI‑II ocean‑colour processing, ACOLITE‑DSF for Sentinel‑2, C2RCC for Sentinel‑3 OLCI). Six spectral indices including NDCI, RBD, NRTI, RAB, PDI and DI were applied for bloom detection. Spatial‑degradation experiments at 20 m, 300 m and 500 m were implemented to separate geometric scale‑related effects on bloom area and boundary representation. Near‑simultaneous overpass comparisons and hourly GOCI‑II time‑series tracking were performed to quantify inter‑sensor discrepancies and intra‑day bloom dynamics.ResultsUnder ~10‑minute near‑coincident overpass on 30 April 2021, GOCI‑II mapped a bloom area of 901.52 km² versus 571.52 km² from Sentinel‑2B; the difference originated jointly from spatial resolution, spectral configuration, atmospheric correction and detection thresholds, while bloom advection exerted minor influence. On 22 September 2021, hourly GOCI‑II observations captured a complete intra‑day bloom cycle: bloom optical‑anomaly area varied from 37.30 km² up to a peak of 760.20 km² and declined to 107.67 km², accompanied by a 45.74 km centroid displacement, a dynamic that single polar‑orbiting snapshots only partially captured. Spatial degradation analysis demonstrated that boundary‑length loss exceeded 76 – 85 % at coarser resolutions, whereas total‑area deviations ranged from −14.8 % to +11.0 %, highlighting Sentinel‑2’s unique value for fine‑scale near‑shore boundary delineation. Even after resolution normalisation, substantial spatial mismatch persisted among sensor bloom products (IoU as low as 0.014).DiscussionIndividual sensors exhibit distinct trade‑offs: GOCI‑II delivers high‑frequency temporal screening, Sentinel‑2 provides fine‑scale spatial detail, and Sentinel‑3 OLCI supplies rich spectral information for regional contextual analysis. Accordingly, we propose a conditional three‑stage multi‑source synergistic workflow: GOCI‑II for rapid hourly screening and drift tracking, Sentinel‑2 MSI for priority‑zone fine‑scale inspection, and Sentinel‑3 OLCI for regional spectral validation and cross‑sensor consistency checking. Discrepancies across sensors stem from combined effects of spectral band configuration, atmospheric correction, index thresholds and sampling timing. Limitations include only two analysed bloom events and absence of dense pixel‑level in‑situ ground truth. Further multi‑event validation with harmonised atmospheric correction is required to generalise this conditional monitoring template for operational HAB early‑warning in optically‑complex East Asian coastal waters.
SN - 2296-7745
ER -