Coastal water quality prediction based on machine learning with feature interpretation and spatio-temporal analysis

• DOI: 10.48550/arxiv.2107.03230
• Publication Date: 2021-01-01
• Authors: Luka Grbčić; Siniša Družeta; Goran Mauša; Tomislav Lipić; Darija Vukić Lušić; Marta Alvir; Ivana Lučin; Ante Sikirica; Davor Davidović; Vanja Travaš; Daniela Kalafatović; Kristina Pikelj; Hana Fajković; Toni Holjević; Lado Kranjčević
• Keywords: Predictive modelling

Coastal water quality management is a public health concern, as poor coastal can harbor pathogens that are dangerous to human health. Tourism-oriented countries need actively monitor the condition of at tourist popular sites during summer season. In this study, routine monitoring data $Escherichia\ Coli$ and enterococci across 15 beaches in city Rijeka, Croatia, were used build machine learning models for predicting their levels based on environmental parameters well investigate relationships with stressors. Gradient Boosting (Catboost, Xgboost), Random Forests, Support Vector Regression Artificial Neural Networks trained measurements from all sampling predict $E.\ values features. The evaluation stability generalizability 10-fold cross validation analysis models, showed Catboost algorithm performed best R$^2$ 0.71 0.68 enterococci, respectively, compared other evaluated ML algorithms including Xgboost, Networks. We also use SHapley Additive exPlanations technique identify interpret which features have most predictive power. results show site salinity measured important feature forecasting both levels. Finally, spatial temporal accuracy examined lowest quality. $E. achieved strong 0.85 0.83, while 0.74 0.67. model moderate 0.44 0.46 high