Abstract
Marine monitoring constitutes one of the main thematic areas of the Sentinel mission. The Sentinel 3 OLCI (S3) sensor provides satellite data for services relevant to the ocean and land. While the spatial resolution of S3 images (300 m) is suitable for most marine applications, there are some applications such as floating debris detection, suspended mater estimation, etc., that require higher resolution. To fulfill this requirement this study applies an unmixing-based data fusion technique on S3 and BRDF-corrected Sentinel 2 (S2) images and evaluates the fused data by calculating the correlation coefficient and the spectral angle distance (SAD) indexes. Then, it explores the increased monitoring capabilities of the fused image by applying improved chlorophyll-a (Chl-a) and total suspended matter (TSM) algorithms, developed for satellite data. The fused image presents spectral similarity to S3 data and spatial similarity to S2 image. Consequently, the products provided by the fused image have much better resolution than those of S3 image, which enables detailed estimations of Chl-a and TSM concentrations. However, the dynamic nature of the marine environment that results in the formation of time-varying patterns at sea surface, in relation to the time lag between S2 and S3 image acquisitions may locally affect the accuracy of the products in the neighborhood of these patterns. This study exploits the effective elimination of directional reflectance effects in S2 ocean images, interprets the fused image and the generated ocean products, and points out the constraints regarding the synergy of Sentinel optical data for ocean areas.
Highlights
S ENTINEL 3 OLCI products aim at supporting a) marine safety, where the quality of the ocean is observed, b) marine resources, where resources in the marine environment are detected and studied, c) marine and coastal environment, where it is observed how these environments are affected by natural processes or human intervention, and d) sea ice, where the behavior of frozen ocean water is examined
Image fusion techniques emerged from multispectral (MS) pansharpening techniques where MS images are fused with a panchromatic band (PAN) in order to improve their spatial resolution
The difference in azimuth angles between neighboring detectors vary from 25 to 100 degrees resulting in significant directional reflectance effects over non-Lambertian surfaces, which are described by the bidirectional reflectance distribution function (BRDF)
Summary
S ENTINEL 3 OLCI products aim at supporting a) marine safety, where the quality of the ocean is observed, b) marine resources, where resources in the marine environment are detected and studied, c) marine and coastal environment, where it is observed how these environments are affected by natural processes or human intervention, and d) sea ice, where the behavior of frozen ocean water is examined. After a conventional unmixing-based fusion and a preliminary fused image reconstruction by the optimized band image dictionary and reconstruction coefficients, a Spectral-Spatial Quality (SSQ) index was used for guiding the adversarial selection process. According to Linear Mixing Model (LMM), each pixel is a linear combination of numerous independent signals and spectral unmixing aims to analyze the composition of these mixed pixels into their endmembers and their corresponding abundances. Though, based on the studies in [8], [19] endmembers can be extracted from the hyperspectral image and spectrally downsampled to the resolution of the multispectral image. N-FINDR is based on the assumption that pure pixel endmembers exist in the image
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
