Coastal water quality prediction based on machine learning with feature interpretation and spatio-temporal analysis. (September 2022)

• Record Type: Journal Article
• Authors: Grbčić, Luka; Družeta, Siniša; Mauša, Goran; Lipić, Tomislav; Lušić, Darija Vukić; Alvir, Marta; Lučin, Ivana; Sikirica, Ante; Davidović, Davor; Travaš, Vanja; Kalafatovic, Daniela; Pikelj, Kristina; Fajković, Hana; Holjević, Toni; Kranjčević, Lado
• Is Part Of: Environmental modelling & software. Volume 155(2022)
• Publication Date: 2022-09
• Subjects: Coastal water quality -- Machine learning -- Shap -- Catboost -- Fecal indicator bacteria; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2022.105458
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Coastal water quality management is a public health concern, as water of poor quality can potentially harbor dangerous pathogens. In this study, we employ routine monitoring data of E s c h e r i c h i a C o l i and enterococci across 15 beaches in the city of Rijeka, Croatia, to build machine learning models for predicting E . C o l i and enterococci based on environmental features. Cross-validation analysis showed that the Catboost algorithm performed best with R 2 values of 0.71 and 0.69 for predicting E . C o l i and enterococci, respectively, compared to other evaluated algorithms. SHapley Additive exPlanations technique showed that salinity is the most important feature for forecasting both E . C o l i and enterococci levels. Furthermore, for low water quality sites, the spatial predictive models achieved R 2 values of 0.85 and 0.83, while the temporal models achieved R 2 values of 0.74 and 0.67. The temporal model achieved moderate R 2 values of 0.44 and 0.46 at a site with high water quality. Graphical abstract: Highlights: Coastal water FIB measurements were used to train machine learning models. Machine learning models trained with Catboost showed peak performance. Strong R 2 scores (0.85, 0.74) were achieved for spatial and temporal FIB prediction. SHapley Additive exPlanations (SHAP) was used to analyze FIB environmental influences. SHAP analysis showed salinity to be a major FIB stressor.