Employing a Probabilistic Neural Network for Classifying Cyprus Coastal Eutrophication Status

¹ Department of Electrical Engineering, Computer Engineering and Informatics, Cyprus University of Technology, 3036 Limassol, Cyprus
² Department of Fisheries and Marine Research, Ministry of Agriculture, Rural Development and the Environment of the Republic of Cyprus, 2033 Nicosia, Cyprus

^* Corresponding author: e.hadjisolomou@cut.ac.cy

Abstract

Good coastal water quality is important for human well-being but also for marine organisms. The European Water Framework Directive (2000/60/EC) has established threshold values for regional seas, with Cyprus collaborating with Greece to assess conditions and set common chlorophyll-a (chl-a) thresholds. In the Levantine Basin, known for its oligotrophic waters, chl-a levels categorize water quality: under 0.1 (μg/l) indicates high quality, 0.1 to 0.4 (μg/l) indicates good quality, and over 0.4 (μg/l) indicates moderate quality. A study developed a Probabilistic Neural Network (PNN) to classify coastal water quality based on factors such as dissolved nitrogen (DIN), ortho-phosphates (PO43−), salinity, dissolved oxygen (DO), pH, electrical conductivity (EC), and water temperature (WT). Over a 20-year monitoring period (2000-2020), the PNN demonstrated impressive accuracy, achieving 98.1% overall classification accuracy and a macro-averaged F1-score of 97.9%. This model serves as an effective tool for environmental management, capable of accurately predicting the water quality status of the Cypriot coastline based on various measurements, thus contributing to better understanding and preservation of coastal ecosystems.