Browsing by Subject "zero-inflated model"
Now showing items 1-2 of 2
-
(2024)Fecal indicator bacteria (FIB) concentrations are used as proxies for hygienic quality of ambient waters. The most utilized FIB are Escherichia coli and intestinal Enterococci. The laboratory methods used to reliably determine these concentrations are time-consuming and do not provide immediate information on the hygienic quality of recreational waters. In this thesis I developed regression models for FIB concentrations using physicochemical in situ water quality measurements as explanatory variables. An accurate and reliable model would enable real-time and continuous determination of FIB concentrations, thereby enhancing the monitoring of water hygienic quality. Using a flow-through system, two spatially comprehensive physicochemical datasets were collected from the sea area of Helsinki Metropolitan area in September 2022 and May 2023. Alongside, total of 56 water samples were collected for analyzing concentrations of E. coli and intestinal Enterococci. For these datasets, I fitted multiple linear regression models, other generalized linear models, zero-inflated models, and hurdle models between the measured parameters and FIB concentrations. Among the measured variables, the fluorescent dissolved organic matter proved to be the superior indicator of declining hygienic quality of water. Other significant explanatory variables included concentrations of phycocyanin and CO2, as well as turbidity. Produced models for intestinal Enterococci were generally more accurate compared to those for E. coli. The preferred model for E. coli proved to be negative binomial regression model, whereas zero-inflated negative binomial model was the optimal model for intestinal Enterococci, due to the substantial proportion of zeros in the intestinal Enterococci dataset. Applying the selected models to the entire physicochemical dataset generated elevated FIB concentration estimations in areas that are subject to FIB loading, based on previous FIB monitoring results throughout the study area. However, thorough validation and further development of presented models is essential before engaging them in FIB monitoring practices.
-
(2024)Fecal indicator bacteria (FIB) concentrations are used as proxies for hygienic quality of ambient waters. The most utilized FIB are Escherichia coli and intestinal Enterococci. The laboratory methods used to reliably determine these concentrations are time-consuming and do not provide immediate information on the hygienic quality of recreational waters. In this thesis I developed regression models for FIB concentrations using physicochemical in situ water quality measurements as explanatory variables. An accurate and reliable model would enable real-time and continuous determination of FIB concentrations, thereby enhancing the monitoring of water hygienic quality. Using a flow-through system, two spatially comprehensive physicochemical datasets were collected from the sea area of Helsinki Metropolitan area in September 2022 and May 2023. Alongside, total of 56 water samples were collected for analyzing concentrations of E. coli and intestinal Enterococci. For these datasets, I fitted multiple linear regression models, other generalized linear models, zero-inflated models, and hurdle models between the measured parameters and FIB concentrations. Among the measured variables, the fluorescent dissolved organic matter proved to be the superior indicator of declining hygienic quality of water. Other significant explanatory variables included concentrations of phycocyanin and CO2, as well as turbidity. Produced models for intestinal Enterococci were generally more accurate compared to those for E. coli. The preferred model for E. coli proved to be negative binomial regression model, whereas zero-inflated negative binomial model was the optimal model for intestinal Enterococci, due to the substantial proportion of zeros in the intestinal Enterococci dataset. Applying the selected models to the entire physicochemical dataset generated elevated FIB concentration estimations in areas that are subject to FIB loading, based on previous FIB monitoring results throughout the study area. However, thorough validation and further development of presented models is essential before engaging them in FIB monitoring practices.
Now showing items 1-2 of 2