%0 Journal Article %A Page, Rebecca M. %A Waldmann, Daniel %A Gahr, Achim %D 2017 %T CCWI2017: F113 'Online Water-Quality Monitoring based on Pattern Analysis' %U https://orda.shef.ac.uk/articles/journal_contribution/CCWI2017_F113_Online_Water-Quality_Monitoring_based_on_Pattern_Analysis_/5364505 %R 10.15131/shef.data.5364505.v1 %2 https://orda.shef.ac.uk/ndownloader/files/9219112 %K CCWI2017 %K Drinking water quality %K online early warning system %K multivariate pattern analysis %K Civil Engineering not elsewhere classified %X To date, drinking water quality monitoring frequently relies on a threshold-based approach coupled with occasional manual sampling for reference analysis and as evidence for legal requirements concerning the water quality. However, the increased availability of online measurements provides a good basis for an adaptive approach to high-resolution monitoring of water quality. In this case study, patterns in water quality of limestone springs were identified using multivariate analysis and artificial neural networks. Self-organizing maps were used to calculate system states based on six online parameters (spring discharge, turbidity, pH, el. conductivity and spectral absorption at 254 nm). A non-linear Sammon projection highlighted the relationship between the different system states, rendering a basis for the quantification of change occurring during the observation period in December 2015 - January 2016. The multivariate approach highlighted different phases during an event based on the relative location in a scatter plot and on the xy distance between two system states based on consecutive measurements. As this approach does not require the definition of thresholds and considers actual changes in system state, it is applicable to complex systems and adaptive management strategies. %I The University of Sheffield