<p>Earth observation from the satellite images has great importance, since it enables detection of such vital facts such as pollution, environmental degradation and climate-related changes. Hyperspectral images (HSIs), providing rich spectral information about the surface, serve as a tool for this purpose. This paper presents an efficient algorithm for mucilage detection from HSIs, aiming the large-scale marine pollution case appeared in the Sea of Marmara, Türkiye, in the spring of 2021. To this end, a discriminative and structure-regularized joint collaborative representation with Tikhonov-regularization (DSRJCRT) classifier is proposed. While the discriminative constraint enhances the class separability among the limited training samples, the structure-regularization term encourages the representation vector to assign higher coefficients to the class that best matches the test sample. The closed-form solution of DSRJCRT is solved for each homogeneous region extracted by a superpixel segmentation method, and then the class-label matrix is obtained. The proposed DSRJCRT is realized on the two hyperspectral mucilage datasets, called Bursa and Istanbul, acquired by the PRISMA satellite over the Sea of Marmara. The experimental results reveal that the proposed DSRJCRT provides higher accuracy than the recent studies on these datasets in the literature.</p>