Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Evaluating flow cytometry for investigating slow sand filter function in drinking water treatment

Westberg, Agnes LU (2026) In TVVR 5000 VVRM05 20252
Division of Water Resources Engineering
Abstract
Quality assurance is crucial for producers of drinking water, and several steps are taken to ensure a safe and aesthetically acceptable drinking water. Monitoring the microbiological quality of drinking water, both finished and within the drinking water treatment plants, is done through cultivation-based techniques. These techniques can take up several days before results can be seen, and quicker results would allow drinking water producers to act on potential quality issues before the water reaches the distribution system.

Flow cytometry offers quick and high-throughput results, measuring total and intact cell counts along with other parameters depending on which reagents are used. It also allows for mapping the cells detected to... (More)
Quality assurance is crucial for producers of drinking water, and several steps are taken to ensure a safe and aesthetically acceptable drinking water. Monitoring the microbiological quality of drinking water, both finished and within the drinking water treatment plants, is done through cultivation-based techniques. These techniques can take up several days before results can be seen, and quicker results would allow drinking water producers to act on potential quality issues before the water reaches the distribution system.

Flow cytometry offers quick and high-throughput results, measuring total and intact cell counts along with other parameters depending on which reagents are used. It also allows for mapping the cells detected to two-dimensional histograms based on, for example, fluorescence. This project studies the possibility of relating traditional cultivation-based techniques to results from flow cytometric measurements on water effluent of slow sand filters, first through exploratory data analysis and then with machine learning. Machine learning was used to find patterns in the fingerprints to determine if heterotrophic plate counts or coliform content were above or below a set threshold.

Data was provided by three different drinking water treatment plants: Norsborg’s drinking water treatment plant (Stockholm vatten och avfall), Ringsjöverket (Sydvatten) and Berggården (Tekniska verken). Correlations were found between flow cytometric data and heterotrophic plate counts at Norsborg’s drinking water treatment plant. The machine learning models also showed promise at determining if the heterotrophic plate count would be above or below 20 colony forming units/mL by analysing the fingerprints. Correlations were also found between total cell counts and head loss in the filters at Norsborg’s drinking water treatment plant and Ringsjöverket. No relationships between flow cytometric data and cultivation-based techniques were observed at Berggården or Ringsjöverket, and the machine learning models were less effective at these sites. (Less)
Popular Abstract (Swedish)
Flödescytometri som verktyg i dricksvattenproduktionens kvalitetssäkrande arbete

Dricksvattenproducenter har flera metoder för att undersöka och kontrollera vattnets kvalitet. Den mikrobiella kvaliteten kan ta allt ifrån 18h till sju dygn att fastställa. Flödescytometri är en metod som har föreslagits som ett potentiellt alternativ eller tillägg till dessa metoder, och som erbjuder korta analystider och hög potential för parallella analyser. I det här projektet har vi undersökt hur flödescytometriska mätningar kan jämföras med mer traditionella metoder.

Dricksvatten måste vara rent och säkert för att kunna distribueras till konsumenter. Detta kontrollerar dricksvattenproducenter genom olika kemiska och mikrobiella mätningar. Resultat... (More)
Flödescytometri som verktyg i dricksvattenproduktionens kvalitetssäkrande arbete

Dricksvattenproducenter har flera metoder för att undersöka och kontrollera vattnets kvalitet. Den mikrobiella kvaliteten kan ta allt ifrån 18h till sju dygn att fastställa. Flödescytometri är en metod som har föreslagits som ett potentiellt alternativ eller tillägg till dessa metoder, och som erbjuder korta analystider och hög potential för parallella analyser. I det här projektet har vi undersökt hur flödescytometriska mätningar kan jämföras med mer traditionella metoder.

Dricksvatten måste vara rent och säkert för att kunna distribueras till konsumenter. Detta kontrollerar dricksvattenproducenter genom olika kemiska och mikrobiella mätningar. Resultat för de kemiska parametrarna kan oftast inhämtas inom en arbetsdag, medan de mikrobiella sällan kan kontrolleras förrän tidigast dagen efter provtagning. Eftersom dricksvatten inte kan återkallas efter att det nått distributionsnätet, är snabba resultat av stor vikt. Flödescytometri kan ge resultat för flera prover och mikrobiologiska parametrar mycket snabbt, ofta under en timma. Att kunna koppla flödescytometriska resultat till dricksvattnets kvalitet skulle vara värdefullt för dricksvattenproducenter, då det skulle göra kvalitetssäkring och felsökning mycket snabbare.

I det här projektet har vi undersökt om, och i så fall hur, de flödescytometriska parametrarna totalantal celler, antalet intakta celler och fluorescens-histogram kan kopplas till de mer traditionella analysmetoderna odlingsbara mikroorganismer (3-dygns) och antalet koliforma bakterier. Datan som vi undersökt kom från tre olika svenska vattenverk, ett i Stockholm, ett i Linköping och ett i Stehag. I projektet såg vi stor variation i vilka flödescytometriska parametrar som kunde kopplas till de mer traditionella metoderna, både i den mer utforskande delen av projektet, och i maskininlärningen.

Slutsatserna som vi drog av detta var att mer ingående studier behövs för att fastställa huruvida totalantal eller intakta celler korrelerar med odlingsbara mikroorganismer eller antalet koliforma bakterier. Resultaten från den förutsägande maskininlärningen var också diffusa, och mer studier behövs för att kunna bestämma tydligare samband. Vi drog även slutsatsen att flödescytometri kan vara ett kraftfullt verktyg i sig, från vilket mycket information kan hämtas, och att det kan vara begränsande att se det från samma perspektiv som de mer traditionella metoderna. (Less)
Please use this url to cite or link to this publication:
author
Westberg, Agnes LU
supervisor
organization
alternative title
Exploratory data analysis and implementation of machine learning on pre-existing datasets from three Swedish drinking water treatment plants
course
VVRM05 20252
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Flow cytometry, Machine learning, Slow sand filtration, fluorescence fingerprinting, drinking water production
publication/series
TVVR 5000
report number
TVVR-25/5019
ISSN
1101-9824
language
English
additional info
Examiner: Kenneth M. Persson
id
9224251
date added to LUP
2026-03-26 15:40:34
date last changed
2026-03-26 15:40:34
@misc{9224251,
  abstract     = {{Quality assurance is crucial for producers of drinking water, and several steps are taken to ensure a safe and aesthetically acceptable drinking water. Monitoring the microbiological quality of drinking water, both finished and within the drinking water treatment plants, is done through cultivation-based techniques. These techniques can take up several days before results can be seen, and quicker results would allow drinking water producers to act on potential quality issues before the water reaches the distribution system.

Flow cytometry offers quick and high-throughput results, measuring total and intact cell counts along with other parameters depending on which reagents are used. It also allows for mapping the cells detected to two-dimensional histograms based on, for example, fluorescence. This project studies the possibility of relating traditional cultivation-based techniques to results from flow cytometric measurements on water effluent of slow sand filters, first through exploratory data analysis and then with machine learning. Machine learning was used to find patterns in the fingerprints to determine if heterotrophic plate counts or coliform content were above or below a set threshold.

Data was provided by three different drinking water treatment plants: Norsborg’s drinking water treatment plant (Stockholm vatten och avfall), Ringsjöverket (Sydvatten) and Berggården (Tekniska verken). Correlations were found between flow cytometric data and heterotrophic plate counts at Norsborg’s drinking water treatment plant. The machine learning models also showed promise at determining if the heterotrophic plate count would be above or below 20 colony forming units/mL by analysing the fingerprints. Correlations were also found between total cell counts and head loss in the filters at Norsborg’s drinking water treatment plant and Ringsjöverket. No relationships between flow cytometric data and cultivation-based techniques were observed at Berggården or Ringsjöverket, and the machine learning models were less effective at these sites.}},
  author       = {{Westberg, Agnes}},
  issn         = {{1101-9824}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{TVVR 5000}},
  title        = {{Evaluating flow cytometry for investigating slow sand filter function in drinking water treatment}},
  year         = {{2026}},
}