DETECTION AND TRACKING OF MUCILAGE PHENOMENON IN THE SEA OF MARMARA BY REMOTE SENSING IMAGES

Basit Öğe Kaydı
dc.contributor.authorKucuk, Sefa
dc.contributor.authorAbaci, Bahri
dc.contributor.authorDede, Murat
dc.contributor.authorYuksel, Seniha Esen
dc.contributor.authorYilmaz, Mete
dc.coverage.doi10.26650/B/LSB21LSB37.2024.023
dc.date.accessioned2026-02-08T15:15:53Z
dc.date.available2026-02-08T15:15:53Z
dc.date.issued2025
dc.departmentBursa Teknik Üniversitesi
dc.description.abstractMarine mucilage is a collection of mucus-like organic matter released by marine microorganisms. Intense mucilage formation in the sea prevents fisheries, maritime, and tourism activities, reduces oxygen levels, and adversely affects biodiversity. The traditional method of detecting mucilage involves taking samples from the sea and analyzing them in a laboratory. However, detecting mucilage with these standard methods is laborious since it can spread over kilometers. On the other hand, several satellites in orbit regularly collect data from the Earth's surface, making it possible to monitor the presence of mucilage through satellite data analysis. Therefore, using both traditional and deep learning algorithms, we utilized PRISMA hyperspectral and Sentinel-2 multispectral data to detect mucilage in its early stages. Sentinel-2A has four 10m fine bands and six 20m coarse bands. To benefit from all bands of Sentinel 2A, the spectral bands must have the same spatial resolution. Although the Sentinel-2A does not have a panchromatic band, the spatial resolution of the 20m bands has been increased to 10m employing its four fine bands as a panchromatic band. We aim to identify or construct a suitable panchromatic band for coarse bands using seven of the existing pansharpening techniques to enhance the spatial resolution of 20m bands to 10m. After preprocessing, we comprehensively compare four different methods, namely Linear regression, Random forest, U-Net, and Vescovi index, on two datasets. On the multispectral dataset, we correctly detect 87.8% of the mucilage formations with the U-Net model and achieve the area under the curve (AUC) score of 0.977. However, the Random forest model has outperformed the other methods, identifying 89.8% of the mucilage formations on the hyperspectral dataset. Experimental results on satellite data with multiple resolutions, bands, different days, and times indicate that detecting mucilage from satellite data with high accuracy and without massive effort is possible.
dc.description.sponsorshipTUBITAK 1001 Project [121G085]; Science Academy, Turkey under the BAGEP 2023 Awards
dc.description.sponsorshipThis study has been funded by TUBITAK 1001 Project No: 121G085 titled Mucilage Detection and Tracking from Multi-Band, Multi-Resolution, and Multi-Satellite Data. Assoc. Prof. Seniha Esen Yuksel would also like to acknowledge the support of the Science Academy, Turkey under the BAGEP 2023 Awards.
dc.identifier.doi10.26650/B/LSB21LSB37.2024.023.014
dc.identifier.endpage542
dc.identifier.isbn978-605-07-1633-7
dc.identifier.startpage517
dc.identifier.urihttps://doi.org/10.26650/B/LSB21LSB37.2024.023.014
dc.identifier.urihttps://hdl.handle.net/20.500.12885/6014
dc.identifier.wosWOS:001628446800015
dc.identifier.wosqualityN/A
dc.indekslendigikaynakWeb of Science
dc.language.isoen
dc.publisherIstanbul Univ Press, Istanbul Univ Rectorate
dc.relation.ispartofEcological Changes in The Sea of Marmara
dc.relation.publicationcategoryKitap Bölümü - Uluslararası
dc.rightsinfo:eu-repo/semantics/closedAccess
dc.snmzWOS_KA_20260207
dc.subjectHyperspectral
dc.subjectMucilage
dc.subjectMultispectral
dc.subjectPRISMA
dc.subjectSentinel-2
dc.titleDETECTION AND TRACKING OF MUCILAGE PHENOMENON IN THE SEA OF MARMARA BY REMOTE SENSING IMAGES
dc.typeBook Chapter

Dosyalar

Koleksiyon

WoS İndeksli Yayınlar Koleksiyonu