Extracting Value From Complex High-Frequency Multivariate Water Quality Data: Exploring Routinely Collected Operational Data

Abstract

Drinking water treatment works are increasingly placed under external stressors including climatic variability, land use and management, and pollution incidents. Routine high-frequency water quality monitoring is an integral part of operational control and is used to inform the treatment process and support the identification of risks. However, in order to improve decision making using the complex, time-series of water quality data that are generated (and typically archived), there must be distinction between basic sensor errors, artefacts of system design and management, and process driven patterns. This paper explores these complex data in order to support synthesis of uncleaned (or raw), high-frequency data; extracting information value from routine catchment wide monitoring. The data are presented in a form that enhances the capability and capacity to utilise existing complex data; improves understanding of complex surface water systems; and helps facilitate data driven models to investigate and forecast the dynamics between water quality determinands during hard-to-treat spate (or rainfall-runoff) events.

Beaver dams attenuate flow: A multi‐site study

“…The systematic extraction of rainfall‐runoff events and corresponding metrics was undertaken using a semi‐automated rules‐based approach for the identification and pairing of rainfall and flow geometries from sub‐hourly observations (Ashe et al, 2019; Deasy et al, 2009; Glendell et al, 2014; Ladson et al, 2013; Luscombe, 2014; Puttock et al, 2017) summarized in Figure 2. Data were sub‐sampled at 15 min intervals and pre‐processed for quality control (Ashe et al, 2019). The automated systematic approach for flow event extraction is sensitive to low flow variability in the discharge time series.…”