Killian Gleeson, S. Husband, John Gaffney, J. Boxall
{"title":"Algorithms to mimic human interpretation of turbidity events from drinking water distribution systems","authors":"Killian Gleeson, S. Husband, John Gaffney, J. Boxall","doi":"10.2166/hydro.2023.159","DOIUrl":null,"url":null,"abstract":"Deriving insight from the increasing volume of water quality time series data from drinking water distribution systems is complex and is usually situation- and individual-specific. This research used crowd-sourcing exercises involving groups of domain experts to identify features of interest within turbidity time series data from operational systems. The resulting labels provide insight and a novel benchmark against which algorithmic approaches to mimic the human interpretation could be evaluated. Reflection on the results of the labelling exercises resulted in the proposal of a turbidity event scale consisting of advisory <2 NTU, alert 2 < NTU < 4, and alarm >4 NTU levels to inform utility response. Automation was designed to enable event detection within these categories. A time-based averaging approach, calculating averages based on data at the same time of day, was found to be most effective for identifying low-level (<2 NTU) events. Simple flat-line event detection was sufficient to identify higher-level alert and alarm events. The automation of event detection and categorisation presented here provides the opportunity to gain actionable insight to safeguard drinking water quality from aging infrastructure.","PeriodicalId":507813,"journal":{"name":"Journal of Hydroinformatics","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2023-11-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Hydroinformatics","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.2166/hydro.2023.159","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Deriving insight from the increasing volume of water quality time series data from drinking water distribution systems is complex and is usually situation- and individual-specific. This research used crowd-sourcing exercises involving groups of domain experts to identify features of interest within turbidity time series data from operational systems. The resulting labels provide insight and a novel benchmark against which algorithmic approaches to mimic the human interpretation could be evaluated. Reflection on the results of the labelling exercises resulted in the proposal of a turbidity event scale consisting of advisory <2 NTU, alert 2 < NTU < 4, and alarm >4 NTU levels to inform utility response. Automation was designed to enable event detection within these categories. A time-based averaging approach, calculating averages based on data at the same time of day, was found to be most effective for identifying low-level (<2 NTU) events. Simple flat-line event detection was sufficient to identify higher-level alert and alarm events. The automation of event detection and categorisation presented here provides the opportunity to gain actionable insight to safeguard drinking water quality from aging infrastructure.