• No results found

Application of Flow Cytometry for Slow Sand Filters

N/A
N/A
Protected

Academic year: 2021

Share "Application of Flow Cytometry for Slow Sand Filters"

Copied!
70
0
0

Loading.... (view fulltext now)

Full text

(1)

Linköping University | Department of Physics, Chemistry and Biology Master thesis, 30 hp | Educational Program: Physics, Chemistry and Biology Spring term 2019 | LITH-IFM-x-EX—19/3649–SE

Application of Flow Cytometry for

Slow Sand Filters

Amanda Helstad

Examinator: Carl-Fredrik Mandenius

Supervisor: Robert Gustavsson

External Supervisor: Catherine Paul

Industrial Supervisor: Sandy Chan

2019-06-12

(2)
(3)

ABSTRACT

This project investigated the bacteria in water entering and leaving the slow sand filters at Ringsjö Water Works using flow cytometry. The purpose was to explore the possibility of utilising flow cytometry as a monitoring method for optimising water production using slow sand filters. Data describing the bacterial community in water was collected over seven weeks and analysed with FlowJo, flow cytometric image comparison and Minitab. The total cell count, intact cell count and the percentage of high nucleic acid bacteria were analysed. These parameters were highly dependent on scraping events, water entering the filters and season. The results indicated that flow cytometry has great potential for use as a monitoring method, although more data should be collected to establish expected trends and secure baseline values for routine comparisons.

(4)
(5)

CONTENTS

Abstract i Figures vii Tables ix Abbreviations xi 1 Introduction 1

1.1 Purpose of the Study . . . 2

1.2 Expected Impact of Study . . . 2

1.3 Objectives of the Work . . . 3

2 Theory 5 2.1 Scientific Background . . . 5

2.1.1 Slow Sand Filter . . . 5

2.1.2 Flow Cytometer . . . 7

2.1.3 Cytometric Histogram Image Comparison . . . 11

2.2 Methodology . . . 11

(6)

iv CONTENTS 3.1 Materials . . . 13 3.2 Methods . . . 14 3.2.1 Sampling . . . 14 3.2.2 Flow Cytometry . . . 15 3.2.3 Data Analysis . . . 17 4 Results 19 4.1 Bacterial Removal by SSFs . . . 19 4.2 Baselines . . . 21 4.3 Seasonality . . . 23

4.4 Factors Affecting the Outgoing Water . . . 25

4.4.1 Scraping Event . . . 25

4.4.2 Ingoing Water . . . 26

4.4.3 Season . . . 26

4.5 CHIC . . . 28

5 Discussion 31 5.1 Effects of SSFs Detected with FCM . . . 31

5.2 Baselines . . . 32 5.3 Impact of Season . . . 33 5.4 Impact of Scraping . . . 34 5.5 Future Prospects . . . 35 5.6 Summary of Discussion . . . 35 6 Conclusions 37 7 Acknowledgment 39 References 41 Articles . . . 41

(7)

CONTENTS v Books . . . 42 Webpages . . . 43 A Planning Report 44 B Technical Data 52 B.1 Sampling Schedule . . . 52 B.2 Screening Results . . . 54 B.3 Ingoing and Outgoing Water . . . 56

(8)
(9)

LIST OF FIGURES

1.1 Ringsjö Waterworks . . . 2

2.1 Slow Sand Filter . . . 6

2.2 Flow Cytometry Function . . . 8

2.3 SYBR Green I and PI . . . 8

2.4 Effects of SYBR Green I and PI . . . 9

2.5 Excitation and emission maxima for SYBR Green I and PI . . . 10

2.6 Fingerprint . . . 10

3.1 Samples for one SSF . . . 14

3.2 Sampling Outgoing Water . . . 15

3.3 Sampling Ingoing Water . . . 15

3.4 Plate Layout . . . 17

4.1 Removal Effect of the SSFs . . . 20

4.2 Baselines . . . 22

4.3 Seasonality . . . 24

4.4 Scraping Event . . . 25

(10)

viii List of Figures

4.6 Interval Plots for TCC, ICC and % HNA Against the Factor Week. . 27

4.7 Interval Plots for TCC, ICC and % HNA Against the Factor Season. 28 4.8 NMDS Plot from CHIC Analysis . . . 29

5.1 Scraping Patterns . . . 34

A.1 Slow Sand Filter . . . 46

A.2 Ringsjö Waterworks . . . 48

A.3 Samples for one SSF . . . 48

A.4 Plate Design . . . 49

B.1 Screening Source Water . . . 54

B.2 Screening Source Water . . . 55

B.3 Ingoing Water . . . 56

(11)

LIST OF TABLES

4.1 Filters identified above or below any baseline. . . 23

A.1 Milestone list. . . 50

A.2 Gantt chart. . . 51

B.1 Technical lab training week. Screening results. . . 52

B.2 First winter sampling week - 1W (Feb 11 - Feb 15). . . 52

B.3 Second winter sampling week - 2W (Feb 18 - Feb 20). . . 52

B.4 Third winter sampling week - 3W (Feb 27 - Mar 1). . . 52

B.5 Fourth winter sampling week - 4W (Mar 4 - Mar 7). . . 53

B.6 First spring sampling week - 1S (Apr 1 - Apr 5). . . 53

B.7 Second spring sampling week - 2S (Apr 8 - Apr 10). . . 53

(12)
(13)

ABBREVIATIONS

CHIC Cytometric histogram image comparison F CM Flow cytometry

HP C Heterotrophic plate count HN A High nucleic acid

ICC Intact cell count LN A Low nucleic acid

N M DS Nonmetric multidimensional scaling

P I Propidium iodide

SSF Slow sand filter

SY SYBR Green I

T CC Total cell count T OC Total organic carbon

(14)
(15)

CHAPTER

1

INTRODUCTION

A treatment for water that has been utilised for over 200 years and is still common at today’s modern drinking water treatment plants is the slow sand filter (SSF) (Huis-man and Wood, 1974). This type of filter belongs to the category biofilter which can remove both particles and pathogens with the help of microbial communities growing in the sand. However, this process is still a “black box” regarding the understanding of how the SSF works and what parameters are important for function (Chan, 2018). The microbial method required by law for controlling the safety and quality of drink-ing water today is the heterotrophic plate count (HPC). This is an established method where the bacteria present in a 1 ml sample of water are grown on an agar plate. If less than 100 colonies (< 10 coliforms and no E. coli (SLVFS 2001:30. Livsmedelsver-ket, 2006)) have grown on the agar plate, after a certain incubation time, the water is supposed to be drinkable. This requires days of preparation before a final result can be determined and can only detect less than 1 % of the total bacteria (Chan, 2018).

Several studies have shown that flow cytometry (FCM) can be a fast and reliable method that can also be used to determine and describe the microbial safety and quality of drinking water (Hammes, Berney, et al., 2008; Berney et al., 2008; Van Nevel et al., 2017). FCM with cytometric histogram image comparison (CHIC) was investigated at Ringsjö Waterworks where the influent and effluent water was sampled from three different SSFs (Chan et al., 2018). The filters differed in the composition of sand, where the first of them was an established filter with old sand which acted as a reference. The second contained cleaned sand from an old filter mixed with new sand, and the last one had completely new sand. After a scraping event, where the top layer of biofilm (schmutzdecke) was removed, the sampling started. With FCM and CHIC, differences between the three filters’ bacterial profiles could be distinguished. The established filter was stable, while the two new filters could not keep the performance and were sensitive to the scraping event. These results inspired further investigation of the possibilities for FCM for SSF monitoring.

(16)

2 Chapter 1. Introduction

1.1

Purpose of the Study

The purpose was to expand and continue the work described in Chan, et al. 2018, to develop FCM as a method for routine monitoring of SSFs. This project expanded the work described in this first publication to investigate all 18 filters at Ringsjö Waterworks with FCM.

SSFs are elegantly simple and environmentally friendly. Only a small amount of energy is required for the maintenance, no chemicals need to be added and almost no waste is generated (Huisman and Wood, 1974). This water treatment technique is therefore valuable to explore further.

1.2

Expected Impact of Study

The study was performed together with Sydvatten at their drinking water treatment plant Ringsjö Waterworks. Sydvatten produces drinking water to almost one million people every day in the south of Sweden (Johansson et al., 2015). It is therefore of high importance that the water is safe to drink and can be delivered at all times. The SSFs at Ringsjö Waterworks are one of six treatment steps (Figure 1.1) and constitutes an approved microbiological barrier. A drinking water treatment plant needs at least two independent microbiological barriers when the source water origi-nates from surface water, which is the case for Ringsjö Waterworks (SLVFS 2001:30. Livsmedelsverket, 2006).The SSFs are one of those barriers and play therefore an important role when it comes to the safety of the drinking water.

Figure (1.1): Ringsjö Waterworks

Overview of the water treatment process at Ringsjö Waterworks (Johansson, 2019).

Today there is only one type of established method for controlling the microbiolog-ical parameters at drinking water treatment plants, namely HPC. This is a limited method due to that it can only grow bacteria that survives at 22 and/or 37 ℃ on a certain agar medium. The method can qualitatively detect E. coli and coliforms, which are indicator organisms, but can only detect a small fraction (< 1 %) of the total bacteria (Bartram et al., 2003). This means that a lot of information is lost, and it is easy to miss the overall picture of the microbiome. This study explored for that reason the possibility of utilising flow cytometry as a complement for the HPC method. Flow cytometry is more efficient and detects all the bacteria in a sample and has therefore a lot of potential.

(17)

1.3. Objectives of the Work 3

generated from FCM was proposed in the work of Besmer et al., 2014. The same type of baseline was made for the results in this Master Thesis and is a suggestion for an eventual future control system for the SSFs at Ringsjö Waterworks to make it easier to sustain well-functioning SSFs.

This study can also contribute with a scientific value for SSFs, as for the current situation there are no official data regarding the variability of all SSFs within one treatment plant. With enough data it could also be possible to steer the output and to optimise an SSF more easily. This study can also contribute with a better understanding of the SSFs at Ringsjö Waterworks. Subsequently, it will increase the safety of the water and make the production more efficient. FCM could also be complementing HPC, and possibly in the future, even replace it.

1.3

Objectives of the Work

The main goal of this project was to decide if it would be possible to set a baseline for a well-functioning SSF at Ringsjö Waterworks with FCM and CHIC. With a baseline, FCM and CHIC could be recommended as a method for monitoring SSFs, with deviations from the baseline as a potential early warning system. It could be able to indicate changes in the process and predict eventual contamination of the water at the drinking water treatment plant (Besmer et al., 2014).

Possible factors that could affect the FCM results were also investigated. This contributed to a better understanding of what can influence SSF results and is an important component for developing a suitable control system. One of the main challenges for this study was to investigate what parameters that would be the most useful to look at in a drinking water treatment perspective for SSFs.

(18)
(19)

CHAPTER

2

THEORY

This chapter contains a deeper background about SSFs, FCM and CHIC, which were the most central subjects in this project. The second part in this chapter will present an overview of the methodology.

2.1

Scientific Background

2.1.1

Slow Sand Filter

The design of an SSF is quite simple, although the function and the microbial ecosystem in the sand, including bacteria, algae, fungi and other microorganisms, is complex. The SSF has four important components and layers. The top layer is the water that will go through the filter, also referred to as the influent. It contributes with a hydraulic pressure and protects the next layer, the top layer of the sand, also known as schmutzdecke. This, and all additional layers, contain biofilm which has been produced by the microorganisms in the sand as well as both organic and inorganic material. Below the schmutzdecke comes the sand bed and is around one meter deep. In the final layer, before the water goes to the outlet, there is supporting gravel. This layer makes sure the sand stays in the filter and lets the water drain easily (Zheng and Dunets, 2012). Figure 2.1 presents the design of an SSF.

(20)

6 Chapter 2. Theory

Figure (2.1): Slow Sand Filter

Design of a standard SSF, modified from (Huisman and Wood, 1974).

An SSF contributes with both a physical filtration by the sand grains and a biological filtration by the microbes living in biofilm that grows on the sand. Pathogens that enter the SSF will not be able to compete with the microorganisms that thrive in the biofilm and will be removed (Zheng and Dunets, 2012). Biofilm is a microbiological community that is attached to a surface and is built up by a polymeric matrix created by the hosts. This way of living for a microorganism is beneficial in several approaches. It provides protection and nutrients and makes it possible for the cells to communicate with each other (Neagu et al., 2017).

SSFs can eliminate up to 99 % of enteric bacteria (bacteria found in the intestines) (Haig et al., 2015) and are effective in removing pathogens due to a few different mechanisms. One of them is the competition for nutrients and organic carbon in the schmutzdecke. E.g., E. coli can survive at low nutrient concentrations, but have difficulties with competing bacteria due to their subordinate kinetic properties (Vital et al., 2012). If the pathogens would survive the first layer, the organic matter will be lesser deeper down in the sand and the cells will eventually starve. Another mechanism is the microbiological predators in the top layer. It has been discovered that protozoan and viruses reduces E. coli, while for coliforms it is mostly algae and fungi that have a reducing effect (Haig et al., 2015). Microbes in the SSFs can also produce various toxins that are poisonous for the pathogens and makes the environment even more hostile for them (Ranjan and Prem, 2018).

Biofilters do not necessarily reduce the number of bacteria after a filtration but they are capable of changing the waters’ microbiome (Chan, 2018). They can also make the water biologically stable by removing nutrients, which is important for the drinking water distribution systems, to reduce the risk of bacteria regrowth (Oh et al., 2018).

It has been suggested that a well-functioning biofilter increases the percentage of low nucleic acid (LNA) bacteria in the effluent water (Lautenschlager et al., 2014). There are speculations that LNA bacteria are more reliant on biofilm for their sur-vival than high nucleic acid (HNA) bacteria. An increase of LNA bacteria in the effluent water could therefore be sign of a diverse and efficient transformation of the

(21)

2.1. Scientific Background 7

microbiome during passage through the SSF. It is also proposed that a high evenness in the effluent microbiome indicate a well-functioning SSF. A high evenness implies a high stress resistance against environmental factors and bacterial infiltrators (Chan, 2018).

When an SSF has been operating for some time, the growth of the biofilm in the schmutzdecke will start clogging the filter, reducing the flow of water through the SSF. The resistance for the water will eventually be too high to sustain a normal production flow. When this happens, the filter needs to be scraped (the operation procedure described below is the current procedure at Ringsjö Waterworks). The filter is emptied from water, and a few centimetres of the top layer on the sand is removed. After the scraping the filter is slowly filled again from below. When the water appears from below on the surface of the sand bed again, the filling of water can continue from above the filter. This type of event has shown to affect the performance of biofilters. The time that it takes for a filter to recover and to mature can vary from 6 hours to 12 weeks (Cullen and Letterman, 1985). It is still not known exactly how the development of maturation works, although, many are convinced that the top layer of biofilm is important for the functionality of a biofilter (Gimbel et al., 2006). In this project, also two other types of events occurred during the time of sampling. For one filter all the sand was exchanged, and another filter was ploughed where the top 30 centimetres was ploughed and mixed.

Factors that are important for the microbial ecosystem in the SSF are the amount of nutrients, oxygen and temperature in the water. Concentrations of organic matter in the ingoing water can limit the growth of a biofilm (Ranjan and Prem, 2018). Low temperatures can slow down the metabolism of all microorganisms in the sand. This decreases the degradation of the organic matter and SSFs have therefore a reduced effect when the temperature is low (Gimbel et al., 2006), while also requiring less scraping maintenance. The total organic carbon (TOC) and temperature of the water were therefore important factors for this investigation of the function of the SSFs at Ringsjö Waterworks.

2.1.2

Flow Cytometer

Flow cytometry can contribute to a better understanding of different aquatic ecosys-tems (Besmer et al., 2014). In this section, the information regarding SSF function and the variables that can be obtained with FCM will be presented.

FCM is an instrument that lets a stream of a water sample pass a laser beam in a small pipe that only allows one cell at the time, so-called hydrodynamic focusing. When the laser strikes a passing cell or particle, the light will either scatter or be absorbed. Fluorochromes that can stain a cell’s DNA absorb the laser light. The absorbed light will later be emitted as fluorescent light. These two types of light, scattered or fluorescent, can be registered and differentiated by the machine. The fluorescent light from the stained cells makes it possible to distinguish them from abiotic particles. It is therefore feasible to make precise cell counts with FCM (Hammes and Egli, 2010). Figure 2.2 illustrates a stained cell and the laser light in

(22)

8 Chapter 2. Theory

an FCM and what wavelengths that are emitted from the cell.

Figure (2.2): Flow Cytometry Function

Schematic figure for the function of an FCM. The illustration visualises the different lights that can emit or scatter from a cell (Hammes and Egli, 2010).

Two different fluorochromes were used to stain the bacteria in this project. The fluorochrome SYBR Green I stains the DNA of both intact and non-intact cells and made it possible to measure the total cell count (TCC) (Hammes and Egli, 2010). The second fluorochrome, propidium iodide (PI), is used indirectly to assess the membrane integrity of the cells. PI is a large and positively charged molecule that interacts with DNA and RNA but cannot pass an intact bacterial membrane. Hence, this fluorochrome can only stain bacteria with a damaged membrane (non-intact bacteria). A schematic picture of how the two different fluorochromes stain bacteria is illustrated in Figure 2.3. With the combination of these two stains, it is possible to determine the intact cell count (ICC) (Gatza et al., 2013).

Figure (2.3): SYBR Green I and PI

A schematic picture of how the SYBR Green I and PI interact with intact and non-intact bacteria. (Gatza et al., 2013).

To illustrate the staining effects, an example of a result from a source water sample is presented in Figure 2.4. In this figure there are two dot plots, where each dot

(23)

2.1. Scientific Background 9

represents detection of an event (cell or particle) with a certain combination of green (x-axis) and red (y-axis) fluorescence. The left plot is a sample stained with only SYBR Green I that determines TCC. The right plot represents the same sample stained with both SYBR Green I and PI. In this sample the non-intact bacteria will emit more red fluorescence compared to the intact bacteria because they have absorbed PI as well. The non-intact cells will therefore be moving higher up in the dot plot due to the increased red fluorescence. In the plots there is a customised gate called “bacteria”. The dots inside this gate will be regarded as bacteria and the dots outside the gate will be treated as background noise. When comparing the two plots, it is noticeable that a lot more bacteria in the right plot are now outside the gate compared to the left plot. The non-intact bacteria are now considered as background noise in the right sample and it becomes possible to calculate the ICC in the sample (Gatza et al., 2013).

Figure (2.4): Effects of SYBR Green I and PI

The graphs are representing the same sample of source water but with two different types of staining. The graph to the left was stained with SYBR Green I and determined TCC. The graph to the right was stained with SYBR Green I and PI and determined the ICC.

(x-axis: green fluorescence, y-axis: red fluorescence.)

Samples with SYBR Green I and samples with SYBR Green I and PI can be run with the same excitation laser (488 nm) in the FCM (Biosciences, 2019b). This is because these different fluorochromes excite at the same wavelength but emit at different wavelengths. This means that their fluorescent light will not overlap each other (Biosciences, 2019a) (Figure 2.5).

(24)

10 Chapter 2. Theory

Figure (2.5): Excitation and emission maxima for SYBR Green I and PI

The blue lines in both graphs represents the wavelengths where SYBR Green I (top graph) and PI (bottom graph) can be excited (note that these wavelengths are overlapping). The green and red area represents the wavelengths where SYBR Green I

and PI emit their fluorescence, respectively (not overlapping) (Biosciences, 2019a).

A histogram presenting the green fluorescence intensity (x-axis) and cell count (y-axis) plotted for each sample generates an individual “fingerprint” (Figure 2.6) (Gatza et al., 2013). These histograms can be divided into two regions, where all bacteria to the right of the value 2 × 104 a.u. will be recognised as HNA cells, and

all bacteria to the left as LNA cells (Prest et al., 2013).

Figure (2.6): Fingerprint

The fingerprint for a source water sample stained with SYBR Green I. (x-axis: green fluorescence, y-axis: cell count.)

(25)

2.2. Methodology 11

2.1.3

Cytometric Histogram Image Comparison

The complexity of the results for the microbial communities that are measured with the FCM can be difficult to interpret. The results also become highly person de-pendent when the gate is made manually (see Chapter 2.1.2). To overcome this problem, the dot plots retrieved from the FCM are compared with each other using a program called Cytometric Histogram Image Comparison (CHIC). This was the method applied in Chan et al. 2018 for the comparison of the FCM data describing three different SSFs. The version flowCHIC was used for this project where the script was provided in R. This program is an automated evaluation procedure and has a gate set already in the script, which makes this comparison method less per-son dependent. The program can therefore identify small changes in the microbial communities represented by the distribution of events in the raw data files (Koch et al., 2013).

2.2

Methodology

The methodology for this project was divided into the following steps:

1. Studying slow sand filters and flow cytometry, summarising the project in a planning report

2. Technical lab training 3. Winter sampling and FCM 4. Data analysis

5. Summarising the first results into a half-time report 6. Spring sampling and FCM

7. Data analysis

8. Writing of final report

The sampling was performed the same day as the samples were run in the FCM. Samples were taken of ingoing and outgoing water for all filters.

A first week of technical lab training was performed before the experiments for all filters started. For the lab training, samples from the source water and for one SSF (14) were sampled throughout a day to investigate if the results could be different dependent on the time of the day the sampling was performed (from 7:00 AM to 01:00 PM). The results were similar to each other, and the time of sampling for the experiments was decided to not be of big importance (results can be found in Appendix B.2).

(26)

12 Chapter 2. Theory

The experiments were performed in two different time periods. The first period was called Winter sampling (four weeks, February 11 - March 7) and the second period was called Spring sampling (three weeks, April 1 – April 17). (A more detailed schedule for the sampling is available in Appendix B.1) With this arrangement it was possible to detect changes in the results with the season.

The raw data from the FCM was extracted as Flow Cytometry Standard (FCS) files. These files were processed in the software FlowJo calculating the results for TCC, ICC and % HNA content bacteria and converted into Excel-files (Goetz and Hammerbeck, 2018). The FCS files were also analysed with the software flowCHIC. Statistical analysis of TCC, ICC and % HNA content bacteria together with different factors was performed in the program Minitab.

(27)

CHAPTER

3

MATERIALS AND METHODS

This chapter presents the materials and the methods that have been used in this project. The methods are divided into three subsections: sampling, flow cytometry and data analysis.

3.1

Materials

Chemical and biological materials required for the experiments in this project:

• Water samples from the SSFs, containing bacteria • Dimethyl sulfoxide (DMSO)

• SYBR Green I

– Staining solution SY: 5 µl SYBR Green I (10 000 x) and 495 µl DMSO • Propidium iodine (PI)

– Staining solution SYPI: 105 µl SY staining solution (100 x) and 21 µl PI solution (1 mg/ml)

• Deionized water

Solutions required to run the FCM, BD Accuri™ C6 Plus, that was used in this project:

• Sheath solution (BD™) • Detergent solution (BD™)

(28)

14 Chapter 3. Materials and Methods

• FACS Clean solution (BD™)

For the quality control of the FCM a test sample with beads (BD™) was run. Other material that was required for the experiments in this project are listed below:

• Falcon tubes (15 ml) • Eppendorf tubes (3 ml)

• Pipettes (2-20, 20-200 and 200-1 000µl) • Matching sterile pipette tips to the pipettes • Sterile gloves (protecting the water samples)

3.2

Methods

3.2.1

Sampling

Each SSF was sampled every sampling week (in total seven weeks). Both ingoing and outgoing water were sampled, stained with either SY or SYPI, in triplicate to be able to investigate the effect of the filters on the microbial community in the water (Figure 3.1).

Figure (3.1): Samples for one SSF

Visualisation for the number of samples that will be run for each SSF each week of sampling in the FCM. (IN: influent water, EFF: effluent water, SYPI and SY: staining

solutions.)

Both ingoing and outgoing water were sampled with sterile falcon tubes. The out-going water was sampled from taps underground below the filters (Figure 3.2). The

(29)

3.2. Methods 15

taps were sterilised by flaming and the water was sampled a few minutes after the taps had been open, making sure that the stagnation water was not sampled.

Figure (3.2): Sampling Outgoing Water

Picture of a tap for a filter where outgoing water was sampled.

The ingoing water was sampled by immersing a falcon tube into the water of the filter with the help of a sampling holder (Figure 3.3). The sampling was performed in the middle of the long side of each filter. Between each filter the sampling holder was sterilised with ethanol (70 %).

Figure (3.3): Sampling Ingoing Water

Picture of the ingoing water sampling, showing the custom designed sample holder with 15 ml Falcon tube.

3.2.2

Flow Cytometry

When the samples had been collected, a specific volume of water was pipetted into Eppendorf tubes for staining. The samples were stained with the two different staining solutions. (Water samples stained with SY: total volume: 500 µl, final concentration of SYBR Green I: 1 x. Water samples stained with SYPI: total volume

(30)

16 Chapter 3. Materials and Methods

500 µl, final concentration of SYBR Green I: 1 x, PI: 3 µM.) Negative controls were prepared with deionized water and were treated the same as the samples stained with SY. Water samples with stains in the Eppendorf tubes were vortexed and then incubated for 15 minutes at 36 ℃. After incubation, the samples were vortexed again and loaded into the FCM (Gillespie et al., 2014). The staining solutions are light sensitive and were protected from light as much as possible during the laboratory work with aluminium foil. A quality control test for the FCM was always performed before any water analysis tests. The setup for the samples and the negative controls in the FCM was according to Prest et al., 2013:

• Run limit: 50 µl

• Fluidics: medium (flow rate: 35 µl, core size: 16 µm) • Threshold: FL1-H less than 500

• Agitate plate: 1 cycle every 1 well(s) • SIP rinse settings: none

Between each sample there was either a negative control or a well of deionized water. This setup minimises the risk for contamination between the different samples. The setup in the FCM for the samples with only deionized water was:

• Run limits: 2 min or 200 µl

• Fluidics: fast (flow rate: 66 µl, core size: 22 µm) • Threshold: FSC-H less than 8 000

The waste from the FCM was autoclaved before it was discarded.

Three different water samples could be measured in one run with the FCM. Each sample was probed in two triplicates, stained with SY or SYPI respectively (Figure 3.4). In total, to measure all 18 SSFs in one week required running 12 plates.

(31)

3.2. Methods 17

Figure (3.4): Plate Layout

The layout for three samples in one plate. (DI: deionized water, N: negative control, S1-3: sample 1-3.) In each plate there was always two negative controls. Blue circles, deionized water or negative control, green circles, stained with SY, pink circles stained with SY and

PI.

3.2.3

Data Analysis

The raw data from FCM was handled with the software FlowJo V10. The gate “bacteria” was set to exclude background noise in the results and so fingerprints could be generated. The bars for deciding percentages of HNA and LNA were also set here based on the values reported by Prest et al., 2013. The generated TCC, ICC and % HNA content bacteria results were extracted to Excel-files. In Excel the means and standard deviations were calculated, and graphs designed. The negative controls were not to exceed TCC-values larger than the standard deviation within the triplicates of the same water sample.

The statistical analysis one-way ANOVA and Pearson correlation were performed using the software Minitab. The one-way ANOVA analysis is presented with interval plots in the results section, which display the mean for each group together with a 95 % confidence interval of the mean. This function was used to visualise statis-tically significant differences and trends between groups and factors. The Pearson correlation analysis was used to investigate correlations between factors and results (Newton, 2014).

CHIC is the tool that was needed to compare the dot plots and identified relevant alterations of the microbiological communities. The software R was required to utilise CHIC (Koch et al., 2013). The script for R for this analysis was already available (S. Chan, personal communication). The version called flowCHIC of this method was used. For the first step, the dot plot images of the raw data were generated for each water sample, using packages flowCore and ggplot2. The second step was to exclude the background noise in the dot plots to only compare relevant data, and for the last step, the similarities and differences between the images were calculated in a distance matrix form (64-pixel resolution) (Schumann et al., 2015). The distance matrix makes comparisons between every sample and calculates a number that describes how similar (close to zero) or different (higher numbers)

(32)

18 Chapter 3. Materials and Methods

all the samples are to each other. This matrix was later visualised with a non-metric multidimensional scaling (NMDS) graph. The longer the distance between samples in this graph, the more different they are to each other (Digby and Kempton, 2012).

(33)

CHAPTER

4

RESULTS

As FCM analysis generates large and abundant data sets, only the most relevant results are presented here, in five subsections. Bacterial removal by the SSFs visu-alised with the three different descriptive parameters. A proposition of how a baseline could look like in the future when more data has been added is presented, including identifying filters that were outside the baseline in this study period and could thus be regarded as abnormally-behaving filters. The FCM data and trends are presented in the context of temperature and TOC concentrations, which are factors that de-pend on the season and could have affected the bacteria in the water. Factors that could have influenced the outgoing water from the filters are described and finally, a comparison of the FCM profiles using flowCHIC is presented.

4.1

Bacterial Removal by SSFs

Changes in the TCC, ICC and % HNA content bacteria as the water passed through an SSF can be presented as percentage of removal (Figure 4.1). A negative value on the y-axis represents an increase of TCC/ICC/% HNA content bacteria in the outgoing water as the water has passed the SSF, while a positive value represents a decrease. For all parameters increased the removal effect over the sampling weeks.

(34)

20 Chapter 4. Results

Figure (4.1): Removal Effect of the SSFs

The percentage removal of TCC (A), ICC (B) and % HNA content bacteria (C) (y-axis) are presented for each SSF (x-axis). Positive values represent a lower value in the outgoing water compared to the ingoing water, and negative values represent a higher value in the outgoing water compared to the ingoing water. The colour scale visualises the

(35)

4.2. Baselines 21

4.2

Baselines

Baselines were created based on the results of TCC, ICC and % HNA content bac-teria in the effluent water (Figure 4.2). The baselines were generated by taking the mean and standard deviation from all results for all filters and weeks for each vari-able respectively excluding outliers (for TCC: values higher than 260 000 TCC/ml, for ICC: values higher than 210 000 ICC/ml, and for % HNA content bacteria: val-ues higher than 25 %). During the first two weeks the outliers for the TCC and ICC parameters can be found below the baseline. In these two weeks, the values for these parameters were also overall low compared to the other weeks. Over the later weeks, these parameters increased in value and outliers can only be found above these base-lines. The filters that were identified as outliers were filters that had been scraped, ploughed or had an exchange of sand recently. For the baseline with % HNA content bacteria the outliers were more spread out over the weeks and are both below and above the baseline. There was no identified pattern for this parameter.

(36)

22 Chapter 4. Results

Figure (4.2): Baselines

A: baseline for TCC, B: baseline for ICC and C: baseline for % HNA content bacteria. The mean is represented as the middle line in the graphs and the lines below and above

are representing the mean ± two standard deviations. Each dot represents the value obtained for that parameter in the effluent water from one SSF in the week indicated on

(37)

4.3. Seasonality 23

By comparing individual values of effluent water relative to the defined baselines, filters that were above or below any baseline any week were identified (Table 4.1.). Effluent water with TCC outside of baseline values tended to also have ICC outside of baseline values, while outliers with respect to % HNA content bacteria were less clearly related to noncompliance in the other parameters.

Table (4.1): Filters identified above or below any baseline.

Week 1 2 3 4 8 9 10 TCC Above 13 2 17, 13,3 16, 13 13, 4,1 Below 1 14 ICC Above 13 2 17, 13 16, 13 13, 4 Below 14 % HNA Above 18, 14,13 18, 9,3, 2, 1 18, 14,13 13 13 Below 10 16, 15 10, 7

4.3

Seasonality

TCC and % HNA content bacteria for both ingoing and outgoing water from all filters during the whole sampling campaign were compared to variation in TOC and temperature (Figure 4.3). An attempt of visualising how the season changes the prerequisites for the SSFs and how the results could depend on them (Ranjan and Prem, 2018; Gimbel et al., 2006). TCC-values increased for every week for both the ingoing and outgoing water to the filters and this correlated to the increasing temperature. The TCC-values increased drastically for the ingoing water after the second week during winter sampling and this event occurred at the same time as for the culmination of TOC. As for the % HNA content bacteria, there has been a slight decreasing trend throughout the weeks.

(38)

24 Chapter 4. Results

Figure (4.3): Seasonality

A: % HNA content bacteria against TCC/ml (logarithmic scale) for all filters in all weeks. The first weeks are coloured blue and the last weeks are coloured red. The ingoing

water is represented with crosses and the outgoing water with dots. (This graph is divided into two graphs in Appendix B.3, one for the ingoing water and one for the outgoing water.) B: dynamics in possible water quality parameters TOC (green line) and

temperature (yellow and orange) during the weeks of sampling that may have had an impact on TCC and % HNA content bacteria in the effluent water.

(39)

4.4. Factors Affecting the Outgoing Water 25

4.4

Factors Affecting the Outgoing Water

Scraping event, ingoing water and season were three different factors that were further investigated and are presented in this subsection of the results.

4.4.1

Scraping Event

Six different filters were sampled 12 h, 24 h and 36 h after a scraping event (Figure 4.4). The graph visualises the dynamic changes of the factors TCC and % HNA content bacteria short after a scraping event. A pattern found for every filter was a decrease of TCC between 12 and 36 h after scraping. For the % HNA content bacteria the filters are more individual and does not have an obvious trend.

Figure (4.4): Scraping Event

Filters were sampled 12 h, 24 h and 36 h after a scraping event. The results are presented with TCC/ml on the x-axis with a logarithmic scale (note: reversed axis relative to other

Figures in this report) and % HNA content bacteria on the y-axis.

In order to see if TCC in the outgoing water was correlated to the number of days since the last scraping, all results from all 18 filters and 7 weeks were divided into four time groups. Each time group represented an interval of days after a scraping event. Time group 1: less or equal to 3 days, time group 2: 4-15 days, time group 3: 16-99 days and time group 4: more or equal to 100 days. The responses TCC and % HNA content bacteria were tested statistically against the factor "Time Group" with one-way ANOVA (Figure 4.5).

The intervals of the means for the different groups for the response TCC are not overlapping each other and have a decreasing trend of TCC for increasing days after

(40)

26 Chapter 4. Results

a scraping event. The means of the groups are statistically different from each other (p-value < 0.05, one-way ANOVA). The intervals of the means for the different groups for the response % HNA content bacteria are overlapping each other and do not have a clear trend with increasing days after a scraping event. Although, the means of the groups are statistically different from each other (p-value < 0.05, one-way ANOVA).

Figure (4.5): Interval Plots for TCC and % HNA

The Figure presents interval plots for the responses TCC (A) and % HNA content bacteria (B).

TCC and Time Group had a strong negative relationship (-0.660, p-value < 0.05, Pearson correlation). This means that as time passes following a scraping event the TCC is decreasing, in contrast to the % HNA content bacteria which did not have a statistically significant relationship to time (0.007, p-value > 0.05, Pearson correlation).

4.4.2

Ingoing Water

TCC in the outgoing water had a strong positive correlation to the TCC in the ingoing water (0.779, p-value < 0.05, Pearson correlation), meaning that if the TCC would increase in the ingoing water, it is therefore very likely that the TCC in the outgoing water will increase as well.

The % HNA content bacteria in the outgoing and ingoing water was also positively correlated (0.170, p-value < 0.05, Pearson correlation) and an increase of % HNA content bacteria in the ingoing water will increase the % HNA content bacteria in the outgoing water.

4.4.3

Season

To investigate if the season had any influence on TCC, ICC and % HNA content bacteria in the effluent water, the results were divided into weeks of sampling (Figure 4.6). The means of the results each week were statistically different from each other for every response (p-value < 0.05, one-way ANOVA). Both TCC and ICC

(41)

4.4. Factors Affecting the Outgoing Water 27

had increasing trends from winter to spring, while % HNA content bacteria had oscillating results and did not have a noticeable trend.

Figure (4.6): Interval Plots for TCC, ICC and % HNA Against the Factor Week.

Interval plots with the responses TCC (A), ICC (B) and % HNA content bacteria (C) in the effluent water against the sampling week. The four first weeks (1-4) are results from

winter sampling and the three last weeks (5-7) are results from spring sampling.

To more directly compare changes with seasons, the results for TCC, ICC and % HNA content bacteria were collapsed into the two seasons: winter and spring (Figure 4.7). The means of the results each season were statistically different from each other for every response (p-value < 0.05, one-way ANOVA). For these results the trends became much clearer. Especially for the % HNA content bacteria, where a decreasing trend could be visualised.

(42)

28 Chapter 4. Results

Figure (4.7): Interval Plots for TCC, ICC and % HNA Against the Factor Season.

Interval plots with the changes in TCC (A), ICC (B) and % HNA content bacteria (C) in the effluent water against the season (season 1 is winter and season 2 is spring).

4.5

CHIC

The analysis with CHIC investigated the two first winter weeks (blue) and the two first spring weeks (green) for both ingoing (no fill) and outgoing water (filled) (Figure 4.8). The plot shows that the ingoing water (top right) is different from the outgoing water (bottom left). For the winter weeks the ingoing and outgoing water had longer distances to each other compared to the spring weeks. This means that the ingoing and outgoing water were more similar to each other during spring. There is a cluster in the upper left corner in the plot which only has samples of filters that have been scraped 12-36 h before sampling.

(43)

4.5. CHIC 29

Figure (4.8): NMDS Plot from CHIC Analysis

The NMDS plot generated from flowCHIC. A visualisation of the distance matrix, describing the differences between all the samples for the first two winter weeks (blue)

(44)
(45)

CHAPTER

5

DISCUSSION

This chapter will discuss and summarise the results for this project. The first sub-section will present the effects of the SSFs that could be seen with FCM, followed by an evaluation of the baselines that were proposed in the results. The factors season and scraping were further analysed and discussed and this chapter ends with the future prospects and a summary.

5.1

Effects of SSFs Detected with FCM

The effect of the SSFs on the bacterial content of the outgoing water were visualised by plotting the percentage of removal for TCC, ICC and % HNA content bacteria (Figure 4.1). The results suggested that the behaviour of the filters was dynamic and changed over the weeks. The effects seemed dependent on the season which effected both the ingoing and outgoing water. Higher TOC values and higher temperatures must have increased the microbiological activity and therefore also contributed to a higher removal effect of all parameters for the filters (Gimbel et al., 2006).

The two first weeks had low values for all parameters for both the ingoing and outgo-ing water and the removal effect was low, while for the latter weeks the parameters increased for both ingoing and outgoing water, but the removal effect increased dra-matically. Although, there was a higher TCC in the outgoing water the latter weeks even though there was a higher removal effect.

The ingoing water has been more unstable and have had a higher variation of TCC, ICC and % HNA content bacteria compared to the outgoing water. The ingoing water was even divided into two clusters in Figure 4.3 (two first winter weeks vs. the rest of the weeks). (The clusters can be seen more clearly in Appendix B.3.) The outgoing water has stayed in the same cluster and have had a more evenness in the results compared to the ingoing water. This is also a trait that could be an indicator for well-functioning SSFs (Chan, 2018).

(46)

32 Chapter 5. Discussion

For a closer look on the percentage removal, the filters appear to be quite individual. Two filters that were standing out for both TCC and ICC ratios were filters 5 and 13. These two filters have never, during these weeks of sampling, had a reducing effect. These filters had also maintenance problems at the drinking water treatment plant. E.g. filter 13 had to exchange all its sand following the maintenance schedule and operational problems. Filter 8 and 18 are also standing out but in the opposite way, they have almost every week had a reducing effect. These performances could possibly be explained by their effect on % HNA content bacteria discussed below. When observing the percentage removal for % HNA content bacteria, the trend for an increasing percentage ratio after every week was similar to the factors TCC and ICC. However, for this factor there were other filters that were standing out. Filter 9, 14 and 17 had always an increase of % HNA content bacteria in the outgoing water, and filter 8 and 18 had for almost every week an increase. What can be reasons for these behaviours? One interesting effect is that filter 8 and 18 had a strong reducing effect of TCC and ICC, but almost always an increase of % HNA content bacteria. A hypothesis for this phenomenon is that these filters are better at reducing LNA bacteria than HNA bacteria. In the work of Chan, 2018, it was suggested that a well-functioning SSF with established biofilm is more likely to have effluent water with an increase of LNA bacteria. The results for filter 8 and 18 could therefore be signs of bacteria leakage and that these filters have not fully matured or have another composition of biofilm that leads to another microbiological composition of the effluent water. Another interesting observation was that filter 8, 14, 17 and 18 have relatively young sand in their filters, they have either had an exchange of all the sand or are newly built. This strengthens the hypothesis that they could have another type of biofilm composition and are therefore not behaving like the other filters at the drinking water treatment plant.

5.2

Baselines

The three different parameters TCC, ICC and % HNA content bacteria were pro-posed as possible baselines measurements for SSF monitoring. The baselines with TCC and ICC had similar results when it came to filters found below and above the baseline. The filters that were above both these baselines had been scraped between 12-48 hours before sampling, had recently (1-13 days) had a complete change of sand (filter 13), or had been, not only scraped, but ploughed 5 days before sampling (filter 1). Filters that were found below the baseline had not been scraped for a while (28 and 137 days). The baseline with TCC recognised a few more filters as outliers than the baseline with ICC did. These two baselines seem to have a simple explanation for the outliers, namely the amount of days between scraping and sampling.

The baseline with % HNA content bacteria was more complex. The filters that were found above and below the baseline had a mixed variation of days between scraping and sampling and cannot be an explanation for the outcome. The % HNA content bacteria in the outgoing water has also varied a lot from week to week (Figure 4.6). Only a small trend of a reduced amount of % HNA content bacteria in the outgoing

(47)

5.3. Impact of Season 33

water was found when comparing the seasons (Figure 4.7). It seemed that the % HNA content bacteria in the outgoing water is quite individual for each filter, and there was no factor identified in this study that could explain the different results and why some filters are outliers. Although, a high % HNA content bacteria could possibly be an indication for leakage of bacteria (discussed in chapter 5.1). If this would be true, this baseline could act as a warning system for malfunctioning filters. Filter 8 and 18 that could have a leakage of bacteria, have not until today showed any signs of malfunctioning. However, these filters would be interesting to follow more closely to see if their behaviour will change and become more like the other filters or behave more irregular. Could this possibly be an early warning sign, which we do not know the outcome of yet? This idea would however require a lot more research and data before it could be applied. A baseline that could work in a more near future is with the parameter TCC. It detected more outliers than the baseline with ICC, probably because TCC is a broader parameter and becomes therefore more detailed and sensitive. There is also a better understanding of the TCC baseline compared to the baseline with % HNA content bacteria. It could be developed as an assessment for determining when a filter has matured after scraping, ploughing or exchange of all sand and when it would be ready for production. This type of baseline with FCM results has therefore a big potential for improving on the current microbiological monitoring technology at Ringsjö Waterworks (Besmer et al., 2014).

5.3

Impact of Season

Before the third week of sampling there was a culmination of TOC in the source water that seemed to have affected the ingoing and outgoing water from the filters dramatically. The major increase of TOC is dependent on events that happens at the lake where the water comes from, Lake Bolmen in Småland. A TOC increase can be a result of wind, heavy rain, temperature variations or other weather conditions that are different from season to season. After this event the TCC and ICC values increased greatly, but the removal effect increased as well.

When comparing all the weeks of sampling (Figure 4.6), the % HNA content bacteria showed a great variation for each week and this pattern is difficult to explain. When the data was divided into two seasons (Figure 4.7), the % HNA content bacteria in the effluent water had a slight decreasing trend. It was proposed in the work Chan, 2018 that % HNA content bacteria decreases after a biofiltration. The first weeks of sampling the results showed the opposite in this project, the % HNA content bacteria increased after the SSFs. However, the % HNA content bacteria had a decreasing trend (Figure 4.7) in this project, and the experiments in Chan, 2018 were made during summer. This can give an indication that the removal of % HNA content bacteria could be strongly dependent on the season.

The results in the CHIC analysis also indicated that the season has a big impact on the results. The spring season seemed to have a higher removal effect than the winter season (discussed in 5.1), but when looking at the NMDS plot with CHIC there was a higher difference between the ingoing and outgoing water in the winter season

(48)

34 Chapter 5. Discussion

compared to the spring season. In winter, the filters could therefore be playing another role than during spring and have another type of effect. It is difficult to try to understand these different types of results with only the data available from this project. This type of analysis will be interesting to use more in the future to investigate the differences between the seasons.

Due to the effects the season had on all factors, a baseline should be made for every season to not be misleading.

5.4

Impact of Scraping

A scraping event had different effects on different filters. This was presented in Figure 4.4 where filters that had been sampled 12 h, 24 h and 36 h after a scraping event showed different patterns in the recovery. Something that all filters had in common was a decrease of TCC over time. However, % HNA content bacteria values are more individual. There were three different patterns that were recognised for the maturing of the filters (Figure 5.1).

Figure (5.1): Scraping Patterns

Presentation of the three different scraping patterns.

The first pattern was recognised for filter 12 and 17. These two filters have had a normal amount of time (3-4 months) between scrapings. These two could be representing a normal maturing development of a filter. The filters 1 and 3 had pattern number 2, and these filters have had short times between each scraping (1-2 months) because of too high resistance in the filters. With such frequent scraping, these filters could possibly represent unstable filters, with values of % HNA content bacteria going up and down. This is however just a speculation, because filter 16 has had a normal amount of time between the scrapings like filter 12 and 17 but had also pattern 2. Albeit, filter 16 started with lower values of % HNA content bacteria already after the 12 h sampling. The third pattern was recognised for filter 2 and has also had a normal amount of time between each scraping. However, this filter had uniquely high TCC values compared to the other filters and patterns. Filter 2 had a complete change of sand in 2017 and could maybe be a part of an explanation for the extreme values. Other factors could also have affected the scrapings, such as sampling week, temperature and TOC. It is still difficult to say

(49)

5.6. Summary of Discussion 35

why the filters have had these types of scraping patterns and require more data for a further understanding about this event.

The NMDS plot showed also that a scraping event had an impact on the ingoing and outgoing water for an SSF. Filters that had been scraped between 12-36 h before sampling had their own cluster in the upper left corner. CHIC could therefore also be a possible baseline for deciding when a filter can go back to production after scraping, just like the TCC baseline. Although, for now is the TCC baseline simpler to generate, but it is not known how much better or worse CHIC would be for the purpose.

5.5

Future Prospects

This Master Thesis was just the beginning of a bigger project of investigating the filters at Ringsjö Waterworks with FCM. The results look promising for continuing the investigation. This project has created a lot of data and new information about the filters. To establish FCM as a monitoring method it would require more data, especially for the baselines. With more data it will be easier to understand the results and what factors that could be the most important. In the future it would be likely that this method could help the production and make it possible to steer and optimise the filters.

The parameter TCC has a high potential to improve the current monitoring of SSF outgoing water. The TCC baseline could in a near future be a part of the control for maturing filters after a scraping, and it is able to do this much faster than the HPC method.

To investigate the factor % HNA and its possible correlation to leakage of bacteria, the filters 8 and 18 should be measured with not only FCM, but also the content of E. coli and coliforms in the outgoing water. This is to see if the hypothesis with a higher % HNA content bacteria in the outgoing water compared with the ingoing is a warning sign for leakage.

In this project, the time was to short to make a deeper analysis of the flowCHIC results. Although, for the generated results, CHIC seems to have a great potential of discovering differences among the filters in more detail. It could also visualise the impact of season in another way than the removal effect did and would therefore be a very interesting factor to follow with this analysis method.

5.6

Summary of Discussion

The FCM has opened a new world for visualising the complexity of microbiological communities in drinking water. The results that can be achieved by the FCM can feel overwhelming and it is now a challenge to make the results comprehensive in a drinking water treatment perspective. The results from this project has showed

(50)

36 Chapter 5. Discussion

that it is possible to distinguish the filters from each other and to find outliers. It is also possible to see the seasonal changes and how it effects the function of the filters. The filters could probably be more efficient with different settings for e.g. water flow or other settings for the treatment steps before the SSFs for the different seasons. This is one optimising step that could be developed with FCM.

It was surprising to see the strong increasing effect of % HNA content bacteria in the outgoing water the first two weeks, when the opposite effect was proposed in several other articles (Chan et al., 2018; Lautenschlager et al., 2014; Vital et al., 2012). A more reducing effect of % HNA content bacteria was though increasing throughout the weeks and the results for this parameter will probably conform more to the results from the other articles during summer season.

(51)

CHAPTER

6

CONCLUSIONS

The main conclusions for this master thesis project are listed below.

• A baseline with TCC values could be developed as an assessment for determin-ing when a filter has matured after a scrapdetermin-ing event and is ready for production. • With more research, a baseline with % HNA could be exploited for detecting

bacteria leakage in the filters.

• The removal effect of the SSFs are dependent on the season. The filters are more efficient at higher temperatures.

• The % HNA in the outgoing water seems to be strongly dependent on the season. High values during winter and low values during summer.

• The reduction of TCC for a filter is dependent on the time since a scraping event.

(52)
(53)

CHAPTER

7

ACKNOWLEDGMENT

I would like to thank my wonderful supervisors Sandy Chan and Catherine Paul who have guided me through this project. They have always been supportive and helpful, and I have learned so much from them. I am so grateful that I have had the opportunity to work with you.

I would also like to say thank you to all the colleagues at Sydvatten who have been welcoming me with open arms. It has been a joy to have had the chance to investigate your slow sand filters.

Thank you Robert Gustavsson and Carl-Fredrik Mandenius, tutor and examinator from Linköping University, for all the feedback and help I have received from you.

(54)
(55)

REFERENCES

Articles

Berney, Michael, Marius Vital, Iris Hülshoff, Hans-Ulrich Weilenmann, Thomas Egli, and Frederik Hammes (2008). ”Rapid, cultivation-independent assessment of mi-crobial viability in drinking water”. In: Water research 42.14, pp. 4010–4018. Besmer, Michael D, David G Weissbrodt, Bradley E Kratochvil, Jürg A Sigrist,

Mathias S Weyland, and Frederik Hammes (2014). ”The feasibility of automated online flow cytometry for in-situ monitoring of microbial dynamics in aquatic ecosystems”. In: Frontiers in microbiology 5, p. 265.

Chan, Sandy (2018). ”Processes governing the drinking water microbiome”. In: Chan, Sandy, Kristjan Pullerits, Janine Riechelmann, Kenneth M Persson, Peter

Rådström, and Catherine J Paul (2018). ”Monitoring biofilm function in new and matured full-scale slow sand filters using flow cytometric histogram image comparison (CHIC)”. In: Water research 138, pp. 27–36.

Cullen, Thomas R and Raymond D Letterman (1985). ”The effect of slow sand filter maintenance on water quality”. In: Journal-American Water Works Association 77.12, pp. 48–55.

Gatza, Erin, Frederik Hammes, and Emmanuelle Prest (2013). ”Rapid and accurate quantitation of bacteria in drinking water is essential to monitor, control, and optimize water treatment processes, and to illuminate the biology of low nutri-ent water systems. Historically, laboratories have relied on heterotrophic plate counts (HPCs) to monitor water quality, but this method is unreliable and time intensive.” In:

Gillespie, Simon, Patrick Lipphaus, James Green, Simon Parsons, Paul Weir, Kes Juskowiak, Bruce Jefferson, Peter Jarvis, and Andreas Nocker (2014). ”Assess-ing microbiological water quality in drink”Assess-ing water distribution systems with disinfectant residual using flow cytometry”. In: Water research 65, pp. 224–234.

(56)

42 References

Haig, Sarah-Jane, Melanie Schirmer, Rosalinda D’amore, Joseph Gibbs, Robert L Davies, Gavin Collins, and Christopher Quince (2015). ”Stable-isotope probing and metagenomics reveal predation by protozoa drives E. coli removal in slow sand filters”. In: The ISME journal 9.4, p. 797.

Hammes, Frederik, Michael Berney, Yingying Wang, Marius Vital, Oliver Köster, and Thomas Egli (2008). ”Flow-cytometric total bacterial cell counts as a de-scriptive microbiological parameter for drinking water treatment processes”. In: Water Research 42.1-2, pp. 269–277.

Hammes, Frederik and Thomas Egli (2010). ”Cytometric methods for measuring bacteria in water: advantages, pitfalls and applications”. In: Analytical and bio-analytical chemistry 397.3, pp. 1083–1095.

Huisman, Leendert and William E Wood (1974). ”Slow sand filtration”. In:

Koch, Christin, Ingo Fetzer, Hauke Harms, and Susann Müller (2013). ”CHIC—an automated approach for the detection of dynamic variations in complex microbial communities”. In: Cytometry Part A 83.6, pp. 561–567.

Lautenschlager, Karin, Chiachi Hwang, Fangqiong Ling, Wen-Tso Liu, Nico Boon, Oliver Köster, Thomas Egli, and Frederik Hammes (2014). ”Abundance and com-position of indigenous bacterial communities in a multi-step biofiltration-based drinking water treatment plant”. In: Water research 62, pp. 40–52.

Oh, Seungdae, Frederik Hammes, and Wen-Tso Liu (2018). ”Metagenomic character-ization of biofilter microbial communities in a full-scale drinking water treatment plant”. In: Water research 128, pp. 278–285.

Prest, EI, F Hammes, Stefan Kötzsch, MCM Van Loosdrecht, and Johannes S Vrouwenvelder (2013). ”Monitoring microbiological changes in drinking water systems using a fast and reproducible flow cytometric method”. In: Water re-search 47.19, pp. 7131–7142.

Ranjan, Prem and Manjeet Prem (2018). ”Schmutzdecke-A Filtration Layer of Slow Sand Filter”. In: Int. J. Curr. Microbiol. App. Sci 7.7, pp. 637–645.

Schumann, J, C Koch, I Fetzer, and S Müller (2015). ”flowCHIC-Analyze flow cyto-metric data of complex microbial communities based on histogram images”. In: Bioconductor R. package version 1.8. 0.

SLVFS 2001:30. Livsmedelsverket (2006). ”Vägledning till Livsmedelsverkets föreskrifter om dricksvatten”. In:

Van Nevel, Sam, Benjamin Buysschaert, Karen De Roy, Bart De Gusseme, Lieven Clement, and Nico Boon (2017). ”Flow cytometry for immediate follow-up of drinking water networks after maintenance”. In: Water research 111, pp. 66–73. Vital, Marius, Frederik Hammes, and Thomas Egli (2012). ”Competition of Es-cherichia coli O157 with a drinking water bacterial community at low nutrient concentrations”. In: Water research 46.19, pp. 6279–6290.

Zheng, Youbin and Siobhan Dunets (2012). ”Slow sand filtration”. In: Greenhouse and nursery water treatment information system. Univ. of Guelph, Ontario, Canada, pp. 1–9.

(57)

Webpages 43

Books

Bartram, Jamie, JA Cotruvo, M Exner, C Fricker, and A Glasmacher (2003). Het-erotrophic plate counts and drinking-water safety. IWA publishing.

Digby, PGN and Rodney Alistair Kempton (2012). Multivariate analysis of ecological communities. Vol. 5. Springer Science & Business Media.

Gimbel, Rolf, Nigel Graham, and M Robin Collins (2006). Recent progress in slow sand and alternative biofiltration processes. IWA Publishing.

Goetz, Christine and Christopher Hammerbeck (2018). The Language of Flow Cy-tometry and Experimental Setup. Springer.

Neagu, Liliana, Doina Maria Cirstea, Carmen Curutiu, Magda Mihaela Mitache, Veronica Lazăr, and Mariana Carmen Chifiriuc (2017). Microbial biofilms from the aquatic ecosystems and water quality. Elsevier, pp. 621–642.

Newton, Isaac (2014). Minitab cookbook. Packt Publishing Ltd.

Webpages

Biosciences, BD (2019a). Absorption and emission spectra. url: http : / / www . bdbiosciences . com / us / applications / s / spectrumguidepage (visited on 04/26/2019).

— (2019b). BD Biosciences Fluorochrome/Laser Reference Poster. url: https : //www.bdbiosciences.com/documents/multicolor_fluorochrome_laser_ chart.pdf (visited on 04/26/2019).

Johansson, Jörgen (2019). Sydvatten – collaborating for public welfare. url: http: //sydvatten.se/wp-content/uploads/2016/02/Sydvatten-in-English.pdf (visited on 04/23/2019).

Johansson, Jörgen, Markus Holm, Justyna Berndtsson, Marie Nordkvist Persson, Kenneth M Persson, and Anna-Karin Wickström (2015). Skånes dricksvatten-försörjning i ett förändrat klimat. url: http://sydvatten.se/wp-content/ uploads / 2015 / 09 / skanes dricksvattenforsorjning i ett forandrat -klimat-lu-1.pdf(visited on 04/19/2019).

(58)

APPENDIX

A

PLANNING REPORT

Application of Flow Cytometry for

Slow Sand Filters

Amanda Helstad

Examinator: Carl-Fredrik Mandenius

Supervisor: Robert Gustavsson

External Supervisor: Catherine Paul

Industrial Supervisor: Sandy Chan

2019-02-12

(59)

45

Abstract

The main purpose of this project is to investigate variation in the microbiology between the slow sand filters at Ringsjö water works, using flow cytometery. This investigation can lead to a new standardised and efficient method for monitoring slow sand filters. It will also contribute to more scientific data for flow cytometry application on slow sand filters. In turn, it can be easier to control the filters which can give them a better industrial status than they have today. The hypothesis is that the slow sand filters will show similar results, which means that the water leaving the filters will all have similar microbial composition. If this is observed, it may be possible to set a standard baseline microbial water content for a functional slow sand filter. All slow sand filters will be sampled and analysed with a flow cytometer and statistical tools will be applied to investigate the variance. Due to the fact that slow sand filters are biological systems, they can vary considerably in microbial composition when external factors are changing, such as weather conditions. Therefore, the filters will be sampled in two different seasons, winter and spring.

Purpose of the Project

A treatment for drinking water that has been utilized for over 200 years and is still common at today’s modern drinking water treatment plants is the slow sand filter (SSF). This type of filter belongs to the category biofilter which can remove both particles and pathogens with the help of microbial communities growing in the sand. However, this process is still a “black box” when it comes to the understanding how it works and what parameters important for its function. (Chan, 2018, p.31-35).

The method for controlling the safety and quality of drinking water today that is required by law is the heterotrophic plate count (HPC). This is an established method where the bacteria present in a 1 ml sample of water are grown on an agar plate. If less than 100 colonies are counted on the agar plate, after a certain time, the water is supposed to be drinkable. This requires days of preparation before the final result can be determined (Chan, 2018, p.7).

A method that can investigate and control the safety and quality of water at a drinking water treatment plant that is fast and reliable, is therefore of big interest. Several studies have shown that flow cytometry possess these desirable characteristics (Hammes F et al. 2008, Barney M et al. 2008, Van Nevel S et al. 2017). A suggestion of using flow cytometry (FCM) with cytometric histogram image comparison (CHIC) was stated in the article Monitoring biofilm function in new and matured full-scale slow sand filters using flow cytometric image comparison (CHIC), Chan, Pullerits et al. 2018. The method was investigated at Ringsjö water works where they sampled influent and effluent water for three different slow sand filters. The filters differed in the composition of sand. The first was an established filter with older sand, the second had cleaned sand from an old filter and some new sand, while the last one had completely new sand. After a scraping event, where the top layer of biofilm (schmutzdecke) was removed, the sampling started. With FCM and CHIC, they could distinguish differences between the three filters’ bacterial profile. The established filter was stable, while the two new filters could not keep the performance and were therefore very sensitive for scraping (Chan, Pullerits et al. 2018). The design of a slow sand filter is shown in Figure A.1, where the different layers are illustrated.

References

Related documents

46 Konkreta exempel skulle kunna vara främjandeinsatser för affärsänglar/affärsängelnätverk, skapa arenor där aktörer från utbuds- och efterfrågesidan kan mötas eller

För att uppskatta den totala effekten av reformerna måste dock hänsyn tas till såväl samt- liga priseffekter som sammansättningseffekter, till följd av ökad försäljningsandel

The plots show the uncorrected velocity fields, where (a) is an evaporating droplet placed on a 60 °C sapphire plate, and (b) shows the velocity distribution inside a freezing

The main objective of this study was to redesign the base structure of a commercially available water filter (by Blöndal Water Group) in order to allow for an automated

From the simulation it appeared that initially, there was a large outgoing concentration of NH4N occurring during the first 15 days for the sand filter (0-10 mg/l) and first 25

Figure 33: Reduction of Mo and Ni during stress experiment with low pH compared to pH at normal operation and 15 minutes contact time in the cork filters and compared to the

Potential opening date: The facility will start functioning from May 8 with the first set of instruments (above table, instruments 1-4 &amp; 6-7). This is due to the time it takes

Industrial Emissions Directive, supplemented by horizontal legislation (e.g., Framework Directives on Waste and Water, Emissions Trading System, etc) and guidance on operating