Published Paper Details:

Ocean contamination represents one of the most pressing ecological challenges confronting contemporary civilization, with severe implications for aquatic ecosystems, public health outcomes, and global atmospheric systems. This research introduces an integrated framework for continuous identification and surveillance of oceanic water quality degradation by synthesizing cutting-edge sensing technologies (IoT-enabled monitoring networks), advanced geospatial observation platforms (satellite-based imagery acquisition), and computational intelligence methods (deep learning and machine learning). The integrated architecture collects environmental parameters--including pH measurements, water clarity indicators, oxygen saturation levels, organic matter decomposition rates, salinity concentrations, trace element distributions, and petroleum hydrocarbon residues--via networked sensor platforms and subsurface monitoring equipment deployed throughout maritime and coastal territories. Contemporary imaging data from Sentinel series and Landsat satellites undergoes transformation through neural network-based image recognition systems, specifically Convolutional Neural Networks, to identify surface-level water quality issues including crude oil contamination plumes, harmful algal proliferation events, and anthropogenic plastic particle aggregation. Computational classification techniques including Random Forest decision systems, Support Vector classification, and recurrent neural architectures (LSTM) facilitate categorization of aquatic condition states and anticipatory modeling of future contamination patterns. The proposed system incorporates a geospatial visualization interface displaying instantaneous surveillance outputs, contaminant distribution visualization, and instantaneous warning broadcasts to governmental environmental entities. Real-world validation utilizing authentic measurement datasets demonstrates the Convolutional Neural Network module achieves 94.2% classification accuracy for satellite-derived oil spill identification, whereas the Random Forest model generates an F1-score of 0.91 for categorical water assessment. This contribution strives to deliver an implementable, precise, and financially viable contamination surveillance mechanism for facilitating legal compliance and ecosystem restoration programs.