Tastes and Odours in Reservoirs Research Report 73

Research Report 73 Tastes and Odours in Reservoirs

Peter Hobson1, Caroline Fazekas1, Jenny House1, Rob Daly1, Tim Kildea1, Steve Giglio1, Michael Burch1, Tsair-Fuh Lin2, Yan-Min Chen2

1Australian Water Quality Centre, South Australian Water Corporation, GPO Box 1751, , South , 5001 2Department of Environmental Engineering, National Cheng Kung University, 1 University Road, Tainan City, Taiwan 70101 (ROC)

Research Report No 73

TASTES AND ODOURS IN RESERVOIRS

DISCLAIMER

The Cooperative Research Centre for Water Quality and Treatment officially ended October 2008, and has been succeeded by Water Quality Research Australia Limited (WQRA), a company funded by the Australian water industry.

WQRA and individual contributors are not responsible for the outcomes of any actions taken on the basis of information in this research report, nor for any errors and omissions.

WQRA and individual contributors disclaim all and any liability to any person in respect of anything, and the consequences of anything, done or omitted to be done by a person in reliance upon the whole or any part of this research report.

This research report does not purport to be a comprehensive statement and analysis of its subject matter, and if further expert advice is required the services of a competent professional should be sought.

0 Water Quality Research Australia Limited 2010

Location: WQRA Head Office Level 3, 250 Square, Adelaide SA 5000

Postal Address: GPO BOX 1751, Adelaide SA 5001

For more information about WQRA visit the website www.wqra.com.au

Tastes and Odours in Reservoirs

Research Report 73

ISBN 18766 16997

MARCH 2010

2 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

FOREWORD

Research Report Title: Tastes and Odours in Reservoirs

Research Officers: Peter Hobson Peter Baker Greg Ingleton Paul Hackney Tim Kildea Kathy Cinque

Project Leader: Michael Burch

Research Nodes: Australian Water Quality Centre South East Queensland Water Corporation Ltd Water Corporation Hunter Water Corporation

CRC for Water Quality and Treatment Project No. 2.0.2.2.2.0 – Tastes and Odours in Reservoirs

3 TASTES AND ODOURS IN RESERVOIRS

EXECUTIVE SUMMARY

A range of organisms that grow naturally in source water are known to cause taste and odour (T&O) problems in drinking water supplies. The scientific literature and practical experience indicates that these can include cyanobacteria and algae, actinomycete bacteria, fungi and other aquatic microbiota. The organisms can be free-living in the water column (i.e. planktonic) or grow as benthic populations on the sediments; and depending upon the source water, they can be seasonally abundant. This project aimed to investigate the occurrence of these different organisms in Australian reservoirs and to better understand their relative potential to produce significant tastes and odours which affect water quality.

The background to this study was that in 2005 a survey of 37 drinking water providers in Australia, representing 5 million customers, found that 78% of the providers had experienced problems with earthy/musty odours and half had not positively identified the problem compounds and were not carrying out routine chemical analysis. Interestingly, only five respondents reported a definite link between algal numbers and T&Os, indicating a need for further investigation. Further feedback from a range of water quality managers in the Australian water industry assisted in scoping the project.

The project set out to characterise the occurrence of T&Os produced by benthic cyanobacteria and actinomycetes in a range of Australian reservoirs. These micro-organisms have been implicated as a possible cause of unexplained T&O incidents during the last few years where planktonic cyanobacteria were absent at the time in drinking water reservoirs in , Victoria, New South Wales, Queensland and Western Australia.

The first component of the project involved a field survey of reservoirs in four states to characterise T&O organisms. This study then developed a conceptual model and then a mass balance model that incorporated processes for production and loss of odour metabolites in a reservoir. The model led to specific investigations and experiments that were designed to calibrate the rates of the component processes identified in the model.

While there is considerable experience with the management of planktonic cyanobacteria there is very little knowledge or experience with the control and management of benthic cyanobacteria. The study evaluated the use of desiccation or ‘drying out’ by changing water level as a possible control technique. A discussion on management of benthic cyanobacteria is provided in this report, however given that knowledge on the ecology of these organisms is relatively limited it is difficult to develop a comprehensive management plan at this stage.

Components of the project are summarised below.

Survey of Australian Reservoirs

A survey was carried out of major drinking water supply reservoirs in four states: Hope Valley and Happy Valley Reservoirs (South Australia), North Pine Dam (Queensland), Grahamstown Dam (New South Wales) and Yan Yean Reservoir (Victoria). The aim was to characterise the benthic community (cyanobacteria and actinomycetes) and determine their relative contribution to odour production in these reservoirs. The initial focus of investigation was on benthic cyanobacteria, however, actinomycetes have recently been identified in the literature as possible sources of T&Os in water supplies and were also included.

In addition, historical data collected from both Hope Valley and Happy Valley Reservoirs by SA Water since 1997 were analysed to determine relationships between cyanobacteria and odour episodes. Both planktonic Anabaena and the benthic cyanobacterium Phormidium were shown to contribute significantly to the total geosmin content in Hope Valley and possibly . Total geosmin present in water entering treatment plant on some occasions came entirely from Phormidium. These results showed the important role that historical data can provide in identifying sources of T&Os in reservoirs.

A detailed spatial survey was undertaken at these two reservoirs in spring and summer to gain a better understanding about the ecology of both benthic cyanobacteria and actinomycetes. The survey

4 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

consisted of ten transects equally spaced around the perimeter where samples were collected at various depths down to 2 m. The aim was to assess the variation both within and between locations. The density of benthic cyanobacteria, Oscillatoria sp., Phormidium sp. and Pseudanabaena sp. varied significantly from site to site at both Hope Valley and Happy Valley Reservoirs. These organisms were found at all depths at some sites and were completely absent from other sites. This suggested variations in micro-environments and habitats within the reservoirs at relatively closely spaced locations. Growth was also found to vary seasonally with growth at some sites in spring but not in summer and vice versa. There was a significant increase (p<0.05) in numbers of Oscillatoria sp. with increasing depth in both spring and summer for Happy Valley Reservoir and summer for Hope Valley Reservoir. This suggested that Oscillatoria sp. preferred lower irradiance and possibly lower temperatures which are both found in deeper waters. Such observations of differences in growth preferences are important for developing a monitoring program for particular reservoirs.

Actinomycetes producing both geosmin and MIB were isolated from Hope Valley and Happy Valley Reservoirs. These were all members of the genus Streptomyces. All but one isolate from a total of 92 produced geosmin and approximately 40% of isolates from both reservoirs produced both geosmin and MIB in varying ratios. Correspondingly the remaining 60% of isolates produced geosmin only. The presence of T&O producing actinomycetes in these reservoirs confirms that they could be contributing to the reservoir odour load but further work is required to determine their significance. An issue with the technique of culture isolation used in this study is that it is not possible to determine whether the isolates that grew from reservoir samples were from actively growing cells or from dormant spores and it is therefore not possible to determine whether they were producing geosmin. The relationship between actively growing actinomycetes and geosmin production requires further study.

The benthic cyanobacterium, Phormidium sp., was identified in the sediments of Grahamstown Reservoir at a single depth at one site only. Large areas of the reservoir supported a dense coverage of torpedo grass (Panicum repens) which grew down to a depth of approximately 1 m and which could retard the growth of benthic cyanobacteria by shading and light limitation. However, an epiphytic community consisting of Pseudanabaena sp. and Anabaena sp. was found attached to the macrophytes indicating that they provide a substrate to support cyanobacterial growth. Actinomycetes were detected in the sediments at all 10 sites surveyed, but not at all depths, and included both Streptomyces sp. and Actinomycetale sp. Both species were found to produce only geosmin.

The water levels at both North Pine Dam and Yan Yean Reservoir were historically low at the time of the study, which was a result of drought conditions. As a consequence sediments around the margins consisted of fine silt representing material which would normally be found in deeper sections of the reservoirs. This sediment was devoid of benthic cyanobacteria or macrophytes apart from a single finding of Phormidium sp. at one site in North Pine Dam. The numbers of Phormidium sp. at this site were high however, and under appropriate conditions its distribution could be expected to be more widespread. Actinomycetes were not found at either North Pine Dam or Yan Yean Reservoir.

Production and Loss of Tastes and Odours in Reservoirs

This study set out to measure the production of odour metabolites by natural populations of both planktonic and benthic cyanobacteria. It also aimed to determine the rates and relative significance of loss processes including biodegradation, volatilisation and photodegradation in reservoirs.

Potential production of geosmin by planktonic cyanobacteria was estimated by the determination of the cell content or ‘cell quota’ of the compound in a range of natural populations of Anabaena circinalis. Samples of natural blooms of A. circinalis in Myponga, Happy Valley and Mount Bold Reservoirs in South Australia and Yan Yean Reservoir in Victoria had a geosmin content ranging from approximately 0.02 to 0.05 pg/cell. These cell quotas can be combined with cyanobacterial cell counts to estimate the potential range of odour yield for natural populations and provide an estimate of the likely water treatment challenge.

The process by which T&Os are released from benthic cyanobacterial mats is poorly understood. A novel set of experiments using specially designed chambers measured the release rate or “flux” of geosmin from benthic mats in the reservoir. These flux experiments were done in the Hope Valley Reservoir in November and December 2006. This is historically a period of high growth of Phormidium sp. in this reservoir. Production rates ranged from 0.0067 to 0.0938 ng/cm2/h in

5 TASTES AND ODOURS IN RESERVOIRS

December and November respectively. The difference was directly related to the cell density of Phormidium sp. in the mats at the time of the experiments. The loss of geosmin from the experimental columns was also correspondingly greater in November than in December with rates of 3.74 and 1.06 ng/L/h respectively.

The study demonstrated that the rate of odour metabolite release can be measured in situ and is dependent upon the density of cyanobacterial cells within the mat. The density of benthic cyanobacteria per unit area can be estimated at individual locations by cell counts, and this can be used to derive a reservoir odour production rate or yield. The largest source of error in deriving the total reservoir odour yield is likely to be the estimation of total benthic coverage given that surveys indicated that the distribution and density of benthic mat is highly variable throughout a reservoir.

This study found that loss processes such as biodegradation and volatilisation have a major effect upon the concentration of T&Os in a reservoir. Biodegradation was found to be temperature dependent with optimum rates occurring at 20-25°C. Biodegradation followed a first-order reaction rate and based upon this the half-life of geosmin was approximately 1 day at 20°C. This means that it would take approximately 24 hours for geosmin to reach 50% of its original concentration through biodegradation alone. Volatilisation was also found to be a significant loss factor with the rate directly related to wind speed, whereas photodegradation is not likely to be an important loss mechanism in natural water. The project incorporated all of these processes into a conceptual and mass balance model.

Taste and Odour Model

An important outcome of this study was the development and validation of a model characterising the mass balance of T&O compounds in a reservoir. The model is summarised in the figure below. The model incorporated major variables of the odourants in inflow and outflow for the reservoir, production of T&Os by planktonic and benthic cyanobacteria, their degradation by photo-oxidation and biodegradation, and loss by volatilisation. The model critically assessed the relative importance of geosmin and 2-methylisoborneol (MIB) in both a Taiwanese and an Australian reservoir. Rate constants were developed for the production of geosmin and MIB by planktonic and benthic cyanobacteria and loss by photo-oxidation, biodegradation, and adsorption onto suspended solids for use in the model. Volatilisation and biodegradation were found to be the most important mechanisms governing loss of geosmin and MIB from the two reservoirs. Wind speed was shown to be an important variable influencing volatilisation i.e. high winds lead to increased loss of geosmin and MIB by volatilisation. Adsorption of geosmin and MIB onto suspended solids was estimated to be relatively low; however, further work is needed in this area to assess adsorption onto reservoir sediments.

Further development of this model could allow for the incorporation of the process rate constants into a deterministic reservoir process-based model to then do comprehensive forecasting and to use the information for management. These models allow for simulation of spatial and temporal relationships between physical, biological and chemical dynamics in water bodies, over single events or seasonal or longer timescales. The application of such a model to a reservoir could enable operators to predict taste and odour episodes caused by planktonic and benthic cyanobacteria.

6 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Figure 1 Representation of the major production and loss factors for odours that have been incorporated into a mass balance model for odours in reservoirs. Inputs are representented in green and losses are in red.

Control of Benthic Cyanobacteria by Desiccation

Desiccation or ‘drying out’ was investigated as a possible method for the control of benthic cyanobacteria using mats collected from Happy Valley Reservoir, South Australia. This involved an air-drying trial to assess cell viability over a period of approximately 7 weeks. The results showed that some cells within a benthic mat could remain viable following long periods of desiccation. Around 20% of the population of Phormidium was still viable after 46 days exposure and it is expected that around 5% would still be viable after 90 days i.e. 3 months. It is very possible that this proportion of healthy cells could provide a seed population for re-growth of benthic mats when they are re-wet.

Management of Benthic Cyanobacteria

While there is considerable experience with the management of planktonic cyanobacteria there is very little knowledge on the management of benthic cyanobacterial problems. A discussion on the options for management of benthic cyanobacteria is provided in this report.

The occurrence of T&Os in a reservoir in the absence of known planktonic producers is the most direct indicator of a potential benthic source. There are some general reservoir features that are likely indicators of the risk of benthic cyanobacteria growing in a reservoir.

The distribution of benthic cyanobacteria is restricted by the depth of light penetration. Shallow reservoirs, especially those with high water transparency, will have greater area available for benthic cyanobacterial growth than deep reservoirs. Shallow reservoirs with gently sloping margins as opposed to deep, steep sides will have higher sediment surface to water volume ratios, resulting in less potential for dilution of the odours. The proximity of the benthic mats to the water off-take is also an important consideration. As a general guide, benthic cyanobacteria need about 1% of the surface irradiance to grow, however this may be lower depending upon the species or type. The area of the reservoir potentially available for benthic cyanobacterial growth can be calculated from the extinction co-efficient of the water and the bathymetry of the reservoir.

A review of historical data from the reservoir can provide significant information about taste and odour issues associated with benthic cyanobacteria. Benthic cyanobacteria can become dislodged from the sediment surface and fragments can float into the water column and can then be detected in routine

7 TASTES AND ODOURS IN RESERVOIRS water samples. The relative concentration of total and dissolved odour compounds can help to identify the source. High levels of dissolved or extracellular odour compounds can suggest a source other than planktonic cyanobacteria.

Reservoirs at high risk of benthic cyanobacterial growth and any reservoir which has a history of unexplained T&O problems should be surveyed to determine the abundance and species composition of benthic cyanobacteria and these should be compared with known T&O producers. Samples should be analysed using appropriate analytical methods for T&O compounds. Visual observations of the sediment surface can provide very useful information on the distribution of benthic cyanobacteria.

The aim of a survey is to obtain a representative assessment of numbers and distribution of benthic cyanobacteria. Samples should be collected from a number of transects around the perimeter of a reservoir. Particular attention should be paid to shallow protected bays and any areas where benthic mats have been observed. Samples at varying depths may need to be collected down to approximately 5 metres, although this will depend upon light attenuation in the water body.

Methods for control of benthic cyanobacteria are limited and have varying success. This study evaluated the control of benthic cyanobacteria by desiccation or drying-out through lowering of reservoir levels and showed that the organisms are quite hardy and resistant to desiccation. The limitation to doing this is having operational flexibility so that the level can remain down for an extended period.

Reduction of catchment-derived nutrients may decrease cyanobacterial growth potential but it is likely to be less effective on benthic forms.

Physical removal by raking or towing drag lines across the sediment surface can be effective to scour the surface and disrupt and dislodge the benthic mat communities. This technique has not been widely used in Australia and is reported from North America and Europe. The technique is only applicable for reservoirs of a certain type, size and bathymetry, and may require repeated application over the growing season. This practice is also likely to generate turbidity and possibly localised spikes in odour compounds in the water above the sediment which may be an issue depending upon the proximity of the supply offtake.

The application of algicides such as copper sulphate has been shown to be effective but this may not be acceptable for environmental reasons in local circumstances. In relation to the specific treatment of benthic cyanobacteria, copper sulphate has been applied as large crystals directly to the areas of benthic mat with the aim that the crystals sink to the sediment and dissolve adjacent to the target organisms to enhance the toxic effect of the copper. The application should be targeted to the area where the mats are found and this is therefore dependent upon having good survey information. Dose rates can be adjusted to treat only the affected areas and not the entire reservoir. It is important to ensure that any chemical that is applied as an algicide is a registered (i.e. approved) product for that purpose. It should be noted that the T&O producing genus Phormidium is recognised as being more tolerant to copper sulphate than common nuisance planktonic species of Microcystis and Anabaena.

A number of alternative methods for the control of cyanobacteria, both chemical and non-chemical, are available in the market place but most have not had rigorous and scientifically valid testing with evaluations published in the peer-reviewed scientific literature. It is crucial that such testing be undertaken so an appropriate alternative to copper sulphate can be identified that will be effective not only for planktonic but also for benthic cyanobacteria.

Actinomycetes have emerged in a number of places around the world as a possible contributor to T&Os in water supplies. This study has identified them as potential sources of both geosmin and MIB in a range of Australian reservoirs but the significance of their contribution to the total T&O load is unclear at this time. There is insufficient quantitative information to incorporate actinomycetes into the model developed as part of this project, however the study confirmed their presence in several drinking water reservoirs in Australia and their production of geosmin and/or MIB. This is an important first step in determining their significance as a T&O contributor and further research is required in this area.

8 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Recommendations

1. Sampling for Benthic Cyanobacteria: Standard techniques are required for sampling and quantitative biomass determination. The sediment sampling and cell counting techniques developed here require further validation with regard to precision and reproducibility with a range of natural populations and sediment types.

2. Monitoring for Benthic Cyanobacteria: Research is required into rapid monitoring techniques for assessment of number and types of benthic micro-organisms. Some options that are worthy of investigation include in situ fluorescence techniques using fluoroprobes applied to mat samples; and molecular techniques involving specific quantitative PCR for geosmin producers.

3. Actinomycetes: Techniques are required for quantitative assessment of numbers and viability of Actinomycetes in environmental samples to determine their significance and potential to contribute T&Os in reservoirs.

4. Modelling Tastes and Odours in Reservoirs: The processes indentified for production and loss of T&Os in reservoirs and presented in a mass balance model should be incorporated into a reservoir process-based model that simulates physical, chemical and biological dynamics which could then be used for forecasting benthic growth and odour dynamics over time and predicting odour episodes to enable use of the information for management.

5. Control of Benthic Cyanobacteria: Research is required to develop control methods for benthic cyanobacteria. This should include susceptibility of local species to algicides, and into specific methods for delivery of chemicals to target organisms.

9 TASTES AND ODOURS IN RESERVOIRS

CONTENTS

Foreword ...... 3 Executive Summary ...... 4 Contents ...... 10 List of Figures ...... 12 List of Tables ...... 15 Abbreviations ...... 16 1 Introduction ...... 18 2 Survey of Reservoirs ...... 23 2.1 Introduction ...... 23 2.2 Aim ...... 23 2.3 Materials and Methods ...... 23 2.3.1 Sediment Analysis ...... 26 2.3.2 Statistical analysis ...... 27 2.3.3 Variation to Sampling Protocol ...... 27 2.4 Results and Discussion ...... 27 2.4.1 Historical Data from Hope Valley and Happy Valley Reservoirs - South Australia ...... 27 2.4.2 Survey of Hope Valley and Happy Valley Reservoirs – South Australia ...... 31 2.4.3 Grahamstown Dam – New South Wales ...... 42 2.4.4 North Pine Dam – Queensland ...... 45 2.4.5 Yan Yean Reservoir – Victoria...... 47 2.5 Conclusions ...... 47 3 Production and Loss of Tastes and Odours from Reservoirs ...... 50 3.1 Photodegradation ...... 50 3.1.1 Introduction ...... 50 3.1.2 Aim ...... 50 3.1.3 Materials and Methods ...... 50 3.1.4 Results and Discussion ...... 51 3.2 Biodegradation ...... 55 3.2.1 Introduction ...... 55 3.2.2 Aim ...... 55 3.2.3 Materials and Methods ...... 55 3.2.4 Results and Discussion ...... 56 3.3 Volatilisation ...... 62 3.4Geosmin Production by Planktonic Cyanobacteria ...... 63 3.4.1 Introduction ...... 63 3.4.2 Aim ...... 63 3.4.3 Materials and Methods ...... 63 3.4.4 Results and Discussion ...... 64 3.5 Geosmin Production by Benthic Cyanobacteria ...... 65 3.5.1 Introduction ...... 65 3.5.2 Aim ...... 65 3.5.3 Materials and Methods ...... 65 3.5.4 Results and Discussion ...... 67 3.6 Conclusions ...... 71

10 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

4. Taste and Odour Model ...... 73 4.1 Introduction ...... 73 4.2 Aim ...... 73 4.3 Model Development ...... 73 4.4 Site Description ...... 74 4.4.1 Hope Valley Reservoir ...... 74 4.4.2 Tai-Lake Reservoir ...... 74 4.5 Estimation of Model Parameters ...... 75 4.5.1 Emission of Geosmin from Benthic Mats ...... 75 4.5.2 Planktonic Generation of Geosmin ...... 75 4.5.3 Mass in the Influent and Effluent of the Reservoir ...... 75 4.5.4 Rate of Biodegradation ...... 76 4.5.5 Rate of Photodegradation ...... 76 4.5.6 Rate of Volatilisation ...... 76 4.5.7 Rate of Adsorption ...... 77 4.6 Assessing and Modelling the Importance of Loss Terms for Odourants ...... 77 4.7 Effect of Environmental Parameters on Loss Terms ...... 78 4.8 Conclusion ...... 79 5 Control of Benthic Cyanobacteria by Desiccation ...... 81 5.1 Introduction ...... 81 5.2 Aim ...... 81 5.3 Materials and Method ...... 81 5.4 Results and Discussion ...... 82 5.5 Conclusions ...... 83 6 Management of Benthic Cyanobacteria ...... 84 7 Summary and Conclusions ...... 87 8 References ...... 92

11 TASTES AND ODOURS IN RESERVOIRS

LIST OF FIGURES

Figure 1 Representation of the major production and loss factors for odours that have been incorporated into a mass balance model for odours in reservoirs. Inputs are representented in green and losses are in red...... 7 Figure 2.1 Schematic of the reservoir transects showing sampling depths...... 23 Figure 2.2 0.4m x 0.4m quadrat divided into 16 squares and 25 intersections. Each intersection was used to calculate benthic mat coverage (see text)...... 24 Figure 2.3 Pictures of Benthic Corer...... 25 Figure 2.4 A. Insertion of the plunger into the corer pushes the sediment plug to the top of the column and enables collection of the benthic layer sample. B. Cross section of core showing benthic mat layer at the surface of sediment...... 26 Figure 2.5 Data collected by SA Water from Happy Valley Reservoir since November 1997 showing A: Anabaena circinalis numbers in water at Location 1, B: Phormidium numbers in water at Location 1 (primary sampling site for reservoir), C: Total geosmin concentration of water entering water treatment plant (WTP) from reservoir, D: Estimated potential geosmin concentration of water entering the WTP (calculated by combining cell counts in A with measured geosmin concentration of 0.041 pg/cell), E: Extracellular geosmin concentration of water entering WTP from reservoir, and F: Intracellular geosmin concentration in water entering WTP from reservoir. Columns  and  represent periods of interest which will be discussed in the report...... 29 Figure 2.6 Data collected by SA Water from Hope Valley Reservoir since November 1997 showing A: Anabaena circinalis numbers in water at Location 9, B: Phormidium numbers in water at Location 9 (primary sampling site for reservoir), C: Total geosmin concentration of water entering the water treatment plant (WTP) from reservoir, D: Estimated potential geosmin concentration of water entering the WTP (calculated by combining cell counts in A with measured geosmin concentration of 0.041 pg/cell), E: Extracellular geosmin concentration of water entering WTP from reservoir, and F: Intracellular geosmin concentration in water entering WTP from reservoir. Columns, ,  and  represent periods of interest which will be discussed in the report...... 30 Figure 2.7 Distribution of Actinomycetes in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 33 Figure 2.8 Distribution of Oscillatoria sp. in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 34 Figure 2.9 Distribution of Phormidium sp. in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 35 Figure 2.10 Distribution of Pseudanabaena sp. in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 36 Figure 2.11 Distribution of Actinomycetes in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 37 Figure 2.12 Distribution of Oscillatoria sp. in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 38 Figure 2.13 Distribution of Phormidium sp. in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 39 Figure 2.14 Distribution of Pseudanabaena sp. in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites ...... 40 Figure 2.15 Changes in water level over a five month period July-Dec 2007: A. Hope Valley Reservoir and B. Happy Valley Reservoir, South Australia. Dates of survey sampling for benthic communities are indicated...... 41 Figure 2.16 Map of Grahamstown Dam showing all 10 sampling locations ...... 43 Figure 2.17 Location 4 at Grahamstown Dam showing dense growth of Panicum repens (torpedo grass) growing out into the reservoir to a depth of approximately 1 m ...... 44 Figure 2.18 Location 1 at Grahamstown Dam showing Eleocharis (spike rushes) growing out into the reservoir to a depth of approximately 1 m ...... 44 Figure 2.19 Map of North Pine dam showing the sampling locations ...... 45

12 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Figure 2.20 Location 1 at North Pine Dam near the wall showing the low level of water in the reservoir and fine silt sediment...... 46 Figure 2.21 Location 7 at North Pine Dam showing the low level of water in the reservoir and fine silt sediment (inset)...... 46 Figure 2.22 Map of Yan Yean Reservoir showing sampling locations ...... 48 Figure 2.23 Location 4 at Yan Yean Reservoir showing the low level of water in the reservoir and dried fine silt sediment...... 49 Figure 2.24 Location 8 at Yan Yean Reservoir showing the low level of water in the reservoir and silty sediment...... 49 Figure 3.1 Methylacrylate cuvettes used as sample exposure containers were attached to the acrylic sheet using silicon sealant ...... 51 Figure 3.2 Suspension of sample and control cuvettes immersed in an outdoor tank to a depth of 10 cm to investigate UV degradation of geosmin and MIB in water. A temperature logger was also installed to continuously monitor water temperature...... 51 Figure 3.3 Changes in geosmin concentration in high purity water after exposure to natural sunlight for 14 days ...... 52 Figure 3.4 Changes in MIB concentration in high purity water after exposure to natural sunlight for 14 days ...... 52 Figure 3.5 Temperature of water in the water-filled drum used in the experiment ...... 53 Figure 3.6 Transmission spectra of Ultra-pure water and water from Hope Valley and Happy Valley Reservoirs for the range 200-800nm through a 5 cm path length...... 54 Figure 3.7 Loss of geosmin in water from at different temperatures (10-25°C) over 48 hours. Samples were spiked with geosmin to reach 100 ng/L and sampled at various times up to 48 hours. Error bars are ± standard deviation...... 58 Figure 3.8 Loss of geosmin over 14 days in water from Blue Lake at 20°C. Samples were spiked with geosmin to reach 100 ng/L and sub-sampled at various times up to 14 days. Error bars are ± standard deviation...... 58 Figure 3.9 Loss of MIB in water from Myponga Reservoir at different temperatures (10-25°C) over 48 hours. Samples were spiked with MIB to reach 100 ng/L and sampled at various times up to 48 hours. Error bars are ± standard deviation...... 59 Figure 3.10 Loss of MIB over 14 days in water from Blue Lake at 20°C. Samples were spiked with MIB to reach 100 ng/L and sub-sampled at various times up to 14 days. Error bars are ± standard deviation...... 59 Figure 3.11 Cell numbers of Anabaena circinalis at a location the near water supply offtake (Location 1), mean cell number in the reservoir and total geosmin concentration in water at the reservoir outlet (BC tap) for three major blooms of Anabaena circinalis in Myponga Reservoir which occurred over the spring and summer of 2006/2007...... 60 Figure 3.12 Cell numbers of Anabaena circinalis at a location near water supply offtake (Location 1); the actual measured total geosmin concentration at the water supply offtake (BC tap); and expected geosmin concentration calculated assuming complete lysis of all A. circinalis cells and release of geosmin into water at the reservoir outlet for three major blooms of Anabaena circinalis in Myponga Reservoir over the spring and summer of 2006/2007...... 60 Figure 3.13 Cell numbers of Anabaena circinalis near dam wall (Location 1), intracellular and extracellular (dissolved) concentrations of geosmin measured at the reservoir water treatment plant (WTP) and the ratio of dissolved (extracellular) to total geosmin for three major blooms of Anabaena circinalis in Myponga Reservoir over the spring and summer of 2006/2007...... 61 Figure 3.14 Photograph of net tow being pulled through an algal bloom ...... 63 Figure 3.15 Photograph showing buoyant Anabaena circinalis cells accumulated at the top of a 1.25 L PET sample bottle...... 64 Figure 3.16 Diagram showing construction and setup for the Flux column inserted in to the sediments and the Loss column placed on top of the sediments...... 66 Figure 3.17 Flux and Loss Columns deployed in the Hope Valley Reservoir...... 66 Figure 3.18 Outline of Hope Valley Reservoir showing orientation and experimental site for Loss and Flux column placement...... 67

13 TASTES AND ODOURS IN RESERVOIRS

Figure 3.19 Change in total geosmin concentration in A) Loss column and B) Flux column over the 24-hour period of Experiment 1 (November, 2006). The experiment commenced on the 6 November, 2006 at 13:30. Graphs include fitted linear regression and associated R2 values. .... 69 Figure 3.20 Change in total geosmin concentration in A) Loss column and B). Flux column over the 48-hour period of Experiment 2 (December, 2006). The experiment commenced on the 11 December, 2006 at 16:00. Graphs include fitted linear regression and associated R2 values. .... 70 Figure 4.1 Schematic representation for the mass balance of a cyanobacterial odourant in a reservoir ...... 74 Figure 4.2 Effect of wind speed on geosmin concentration in Hope Valley Reservoir (HVR)...... 79 Figure 4.3 Effect of biodegradation rate on geosmin concentration in Hope Valley Reservoir (HVR). 79 Figure 5.1 The range of temperatures to which the benthic mat samples were exposed during the experiment...... 82 Figure 5.2 Percentage Relative Humidity to which the benthic mat samples were exposed during the experiment...... 82 Figure 5.3 Chlorophyll a content of the Phormidium mat samples that were allowed to desiccate in natural sunlight over a period of 46 days...... 83 Figure 5.4 Viability of Phormidium sp. in cyanobacterial benthic mat samples that were allowed to desiccate in natural sunlight over a period of 46 days...... 83

14 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

LIST OF TABLES

Table 1.1 Cyanobacteria and actinomycetes identified as geosmin and/or 2-methylisoborneol (MIB) producers...... 20 Table 3.1 Micro-organisms reported to be involved in the biodegradation of MIB and geosmin (Ho et al., 2007) ...... 55 Table 3.2 Degradation rate of geosmin in Myponga Reservoir water at different temperatures (reaction considered to be first-order)...... 57 Table 3.3 Intracellular geosmin content of Anabaena circinalis collected from various reservoirs during cyanobacterial blooms...... 64 Table 3.4 Rates of geosmin release from benthic mats in Flux columns and rates of loss from Loss columns for two separate experiments and comparative geosmin concentrations, both extracellular (dissolved) and intracellular, measured by SA Water at the WTP and cyanobacterial cell numbers, both Anabaena circinalis and Phormidium sp., at Location 9...... 68 Table 4.1: Average monthly temperature and wind speed for Hope Valley (HVR) and Tai-Lake Reservoirs (TLR) - 2005 data ...... 75 Table 4.2 Values for geosmin loss terms in HVR and TLR sourced and calculated using routine monitoring data and referenced papers...... 76 Table 4.3 Summary of geosmin degradation rates measured by different researchers and work presented in the current study...... 76

15 TASTES AND ODOURS IN RESERVOIRS

ABBREVIATIONS

%T percentage transmission Am area covered by cyanobacteria mat As water surface area of lake ASM artificial seawater medium Av surface area of the reservoir BLAST basic local alignment search tool C dissolved odourant concentration CARD catalysed reporter deposition CFU colony forming unit chla chlorophyll a Cin odourant concentration present in the influent cm2 square centimetres cms centimetres Cp cell-bound odourant concentration Cs odourant adsorbed on suspended particles other than cyanobacteria cells CSS SS concentration in water DNA deoxyribonucleic acid DOC dissolved organic carbon e base of the natural logarithm EC50 half maximal effective concentration FISH fluorescent in-situ hybridization Fm flux of the odourant from benthic cyanobacteria fOC organic carbon content of SS Fv flux of the odourant from benthic cyanobacteria g gram GC gas chromatograph HVR Hope Valley Reservoir I0 light intensity at 0 metres depth Iz light intensity at z metres depth KAW mass transfer coefficient between air and water interface KB biodegradation rate Keff rate constant for outflow kJ/m2 kilojoules per square metre km kilometres km/h kilometres per hour KOC partition coefficient for organic compounds between organic carbon and water KOW octanol/water partition coefficient kT light extinction coefficient L litre Li sink terms for odourant Loc location m metres M mass of odourant in the reservoir m/s metres per second m2 square metres m3 cubic metres mg/L milligrams per litre MIB 2-methylisoborneol Min mass of odourant entering the reservoir with water inflow mJ/cm2 millijoules per square centimetre mL millilitres ML megalitres ml/g millilitres per gram ML/year megalitres per year mm millimetres Mout mass of odourant leaving the reservoir with water outflow

16 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

MS mass spectrometer Msed mass of sediments interacting with the odourants MW molecular weight NATA National Association of Testing Authorities NCBI National Centre for Biotechnology Information ng/day nanograms per day ng/h nanograms per hour ng/h/cell nanograms per hour per cell ng/L nanograms per litre ng/L/day nanograms per litre per day ng/m2/hr nanograms per square metre per hour nm nanometres p<0.05 probability less than 5% PAR photosynthetically active radiation PCR polymerase chain reaction PDMS polydimethylsiloxane pg picograms PET polyethylene terephthalate q adsorption capacity of odourant onto the sediment Qin flow rate of influent Qout flow rate of effluent R2 coefficient of determination Rb biodegradation rate of the odourant Rp photodegradation rate of the odourant rRNA ribosomal ribonucleic acid s.d. standard deviation SA Water South Australian Water Corporation Si production rate from different sources of odourant SIM selected ion monitoring SPME solid phase microextraction SS suspended solids t time T&O taste and odour TLR Tai-Lake Reservoir L microlitre USGS United States Geological Survey UV ultraviolet radiation V volume of the reservoir Vb volume of water in which biodegradation may occur VCp volume of water with planktonic cyanobacteria VOC volatile organic carbon Vp volume of water in which photodegradation may occur WTP water treatment plant z water depth

17 TASTES AND ODOURS IN RESERVOIRS

1 INTRODUCTION

A range of organisms that grow naturally in source water are known to cause taste and odour (T&O) problems in drinking water supplies. These include actinomycete bacteria, various types of cyanobacteria and algae, fungi and other aquatic microbiota (Mallevialle and Suffet, 1987). The organisms can be free-living in the water column (i.e. planktonic) or grow as benthic populations in the sediments. An issue of importance for the water industry is the occurrence of these micro-organisms and their relative contribution to imparting tastes and odours thereby affecting water quality. The contribution that these micro-organisms make will vary depending upon the season, and the reservoir and source water supply due to differences in climate, geology and chemistry of the water body. Determining the relative contribution of each of these micro-organisms to the total T&O load in a reservoir will aid in the development of appropriate management plans to reduce their impact upon water quality.

The main compounds of concern in water sources are geosmin and 2-methylisoborneol (MIB) which produce earthy-musty T&Os and which are a significant problem due to their low threshold of detection by humans (approximately 10 ng/L). They are also difficult to remove by conventional treatment processes such as flocculation, sedimentation and filtration. This study concentrates on these two compounds since they are currently the major T&O problem in Australian water supplies.

In early 2005, a survey was conducted of 37 drinking water providers in Australia representing 5 million customers. It was found that 78% of providers had experienced problems with earthy musty odours and half had not positively identified the problem compounds and were not carrying out routine chemical analysis. Interestingly, only 5 respondents reported a definite link between algal numbers and T&Os, indicating a need for further investigation and discussion. In March 2005, approximately 40 people from various Australian water authorities attended a CRC for Water Quality and Treatment Workshop on geosmin and MIB in water supplies, at the National Research Centre for Environmental Toxicology, Queensland. Topics discussed covered source water, treatment and distribution. Delegates discussed their T&O issues and identified their needs and where they felt further research was required. The feedback received from this workshop was used in the formulation of the current project, the results of which are presented in this report.

The problems associated with odours and toxins produced by planktonic cyanobacteria (e.g. Anabaena, Planktothrix - see Table 1) are well known in the Australian water industry. As a consequence, management actions usually focus on the T&O and toxin control of these cyanobacteria. However, benthic cyanobacteria and actinobacteria (actinomycetes) have also been implicated as sources of unexplained T&O incidents in drinking water reservoirs where planktonic cyanobacteria have not been identified. This has occurred in the last few years in reservoirs in South Australia, Victoria, New South Wales, Queensland and Western Australia. There is a distinct lack of knowledge of the relative importance of these two sources to T&O and their seasonal significance. The development of a model covering all inputs and outputs of T&O compounds from a water source, by both planktonic and benthic cyanobacteria would help in the development of predictive and pro- active management plans. There are also no proven management techniques to mitigate or prevent the growth of T&O-producing benthic micro-organisms and further research is required to identify and develop effective control methods.

Benthic cyanobacteria within Australia, which include Aphanizomenon, Oscillatoria, Phormidium and Pseudanabaena, have been linked to T&O problems caused by geosmin and MIB production (Burch and Baker, 2000). A complete list of benthic cyanobacteria associated with geosmin and MIB production is presented in Table 1.1. These cyanobacteria grow in mats or films attached to sediments, rocks and aquatic plants on the banks of lakes and rivers. Light is the major factor determining their distribution: with euphotic depth limiting growth in deeper waters and high irradiance levels suppressing growth in shallow water. Other controlling factors include temperature, nutrient availability and shear stress associated with wave action. Sabater et al. (2003) and Vilaita et al. (2004) investigated the growth of benthic cyanobacteria in the Llobregat River, North-Eastern Spain, and suggested that benthic mats may undergo a process involving 1) growth, 2) the occupation of the river surface area and 3) subsequent detachment and drift. Attached cyanobacteria produced high levels of geosmin but once they developed into the free-floating form geosmin was released and dispersed due to amplified cell lysis caused by increased shear stress and high irradiance.

18 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Vadeboncoeur et al. (2008) modelled the contribution of periphyton (complex mixture of algae, cyanobacteria, heterotrophic microbes and detritus that is attached to submerged surfaces in most aquatic ecosystems) to a lake’s total biomass production. They found that distribution of available benthic habitat for periphyton was influenced by the ratio of mean depth to maximum depth (DR). The model demonstrated that DR and light attenuation strongly determined the maximum possible contribution of benthic algae to lake total biomass production and the benthic proportion of whole-lake primary production declined with increasing nutrients due to higher levels of planktonic biomass. Furthermore, Vadeboncoeur et al. (2008) showed that shallow lakes were insensitive to DR and were dominated by either benthic or pelagic primary productivity depending upon trophic status. Moderately deep oligotrophic lakes had substantial contributions by benthic primary productivity at low depth ratios and when maximum benthic photosynthesis was moderate or high. Extremely large, deep lakes always had low fractional contributions of benthic primary production. This is significant when determining potential coverage of T&O producing benthic algae.

It is therefore important to understand the structural dynamics of reservoirs and their tendencies to promote growth of periphyton which could result in the production of T&Os in drinking water supplies. The identification of organisms and the establishment of a model to assess their impact is the focus of this report.

Although benthic species of cyanobacteria have been generally identified for their T&O problems, toxin production has recently emerged as a possible health issue. Deaths of dogs were linked to the presence of Oscillatoria-like species (Hamill, 2001), Phormidium favosum mats containing anatoxin-a (Gugger et al., 2005) and Phormidium autumnale containing both anatoxin-a and homoanatoxin-a (Wood et al., 2007). Cattle have also died through ingestion of Oscillatoria limosa (Mez et al., 1997; 1998). Baker et al. (2001) investigated Phormidium aff. formosum and Phormidium aff. amoenum from two reservoirs and a recreational lake in South Australia and found them lethal to mice by intraperitoneal injection. Neuro and hepatotoxic effects have also been reported from Calothrix parietina and Phormidium tenue (Mohamed et al., 2006). Izaguirre et al., (2007) found several microcystin-producing Phormidium Oscillatoria and Lyngbya species which were isolated from drinking water reservoirs and Seifert et al. (2007) produced the first evidence of cylindrospermopsin and deoxy-cylindrospermopsin production by Lyngbya wollei.

Both geosmin and MIB were first isolated in actinomycetes: geosmin, an earthy-smelling substance, by Gerber and Lechevalier (1965); and MIB a musty smelling substance, by Medsker et al. (1969) and Gerber (1969). Actinomycetes are a group of gram-positive bacteria which are ubiquitous in aquatic habitats (Goodfellow and Williams, 1983) and share some features with fungi, primarily in the form of mycelium growth, creating a branching network of filaments (Izaguirre and Devall, 1995). They have been found in water samples, aquatic sediments (Boon, 1989) and attached to aquatic plants (Hatano et al., 1994; Wohl and McArthur, 1998). While many researchers have identified a link between the presence of aquatic actinomycetes and T&Os in water supplies (Jensen et al., 1994, Sugiura and Nakano, 2000; Klausen et al., 2004) others suggest that this association is tenuous. While there is a general understanding about the environmental conditions that support their growth on land (i.e. optimum growth 20 – 30°C, pH between 5 and 9, sensitivity to high CO2 and low O2, and association with decomposing organic material) (see Goodfellow and Williams, 1983), little is known about their requirements in aquatic ecosystems. It has been argued that actinomycetes may not actively grow in large bodies of water and that T&Os are produced by terrestrially growing species and transported into water via run-off (Jenson et al., 1994; Zaitlan et al., 2003). These terrestrially growing T&O- producing micro-organisms could also include fungi, bacteria and even amoeba (Hayes et al., 1991). Furthermore, not all actinomycetes produce MIB or geosmin so that their presence in a reservoir with T&O problems doesn’t automatically make them the producers. While they have been recognised as contributors of taste and odours in raw waters for many years (Wood et al., 2001), knowledge of their ecology and the significance of their contribution is limited and requires further research (Jüttner and Watson, 2007, Zaitlin and Watson, 2006, Klausen et al., 2004, Lanciotti et al., 2003).

19 TASTES AND ODOURS IN RESERVOIRS

Table 1.1 Cyanobacteria and actinomycetes identified as geosmin and/or 2-methylisoborneol (MIB) producers.

agnm geminatum Jaaginema splendidum Geitlerinema ebenasubtilis Leibleinia ygy aestuarii Lyngbya siltracurviceps Oscillatoria Microcoleus Hyella granulatum Tychonema bornetii Tychonema mülleri Symplocastrum martensianus Porphyrosiphon prolifica Planktothrix viscosum P. uncinatum P. tenue P. P. Cyanobacteria Benthic .simplissimum P. .tenuis O. .variabilis O. hriimallorgei Phormidium .amoenum P. .chalybeum P. breve P. .cortianum P. P. .formosum P. favosum P. p tanNV 1++ + 51 NIVA strain sp. tanL69+ LM689 strain p + sp. var. p + sp. levis opudPlanktonic Cyanobacteria Compound e M BGoI GeoMIB GeoMIB GeoMIB + + ++ + + + + + + + + ++ + + Odour + + + + + + + + + nbeacircinalis Anabaena .crassa A. .lemmermannii A. .macrospora A. .solitaria A. ypoamuscorum Symploca .viguieri A. paioeo flos-aquae Aphanizomenon siltralimosa Oscillatoria gracile Aphanizomenon lnttrxagardhii Planktothrix .cryptovaginata P. .perornata P. suaaan limnetica Pseudanabaena catenata Pseudanabaena P. perornata var. var. attenuata opudActinomycetes Compound + + + + + + + + + ++ ++ Odour + + + + + Penicillium .expansum P. Aspergillus tetmcsalbidoflavus Streptomyces S. .citreus S. .griseus S. .griseofuscus S. .psammoticus S. .halstedii S. .tendae S. .violaceusniger S. Streptomyces vriii + avermitilis p.++ + spp. spp spp . . Compound + ++ + + + ++ + + ++ ++ Odour

Geo – Geosmin. Table modified from Jüttner and Watson (2007)

20 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

An issue in studying actinomycetes to determine their contribution of T&O compounds in a water body is measuring their activity. Generally, isolation media have been used to determine the presence or absence of actinomycetes in raw water, but this method does not discriminate between actively growing forms and dormant spores and presents them all as colony forming units (CFU’s). It is therefore not possible to determine if the micro-organisms are actively growing or just present in the environment. Furthermore, there are a number of different media available for isolation and the types and numbers of actinomycetes isolated by each of these media varies. A more sensitive method for determining numbers of actinomycetes in raw waters has been developed by Sekar et al. (2003) using fluorescent in-situ hybridisation (FISH) and signal amplification (catalysed reporter deposition, CARD). This method is able to detect larger numbers of actinomycetes by identifying types that are unable to grow on isolation media. However, even though this method has improved identification of actinomycetes, knowledge is still limited regarding the abundance of these micro-organisms in raw waters and growth on isolation agar is still required to determine if a particular actinomycete produces T&Os.

Jüttner and Watson (2007) described a number of studies using labelling and genetic techniques which showed that geosmin and MIB could be produced by several different pathways in both cyanobacteria and actinomycetes (Figure 1.1). However, they identified that there is a lack of understanding as to whether a single pathway is used by a particular micro-organism or multiple pathways are used to produce geosmin or MIB. Furthermore, there is not a complete understanding about the genes and enzymes responsible for the production of these compounds. Generation of this knowledge in the future will help to understand the variation between different species and strains of cyanobacteria and actinomycetes in production of these compounds.

Control of cyanobacterial growth in reservoirs, both planktonic and benthic, is a largely unresolved management problem. The ultimate goal for the control of cyanobacterial growth is to reduce nutrient inputs into the system by implementation of efficient wastewater treatment processes and improvements in watershed management (Izaguirre and Devall, 1995). However, the most common practice for removing planktonic cyanobacteria from source waters has been through the use of algicides particularly copper sulphate. It is generally regarded as effective, economical and safe for operators to use, although increasingly, copper is regarded less favourably as a preferred option due to increased awareness of its adverse environmental impacts on the aquatic ecosystem. The water industry now recognises that ongoing use of copper sulphate is unsustainable and alternatives need to be found. Alternative methods being used and under investigation include algicides such as copper chelates and hydrogen peroxide, modified clays that bind and remove phosphorus (e.g. Phoslock ®), and aerators and mixers which disrupt algal growth by removing stratification and preventing release of nutrients from anoxic sediments.

Measures for the control of benthic cyanobacteria in source waters are similar to those required for other cyanobacteria and algae. However, while a reduction in nutrient input has been seen as important for controlling benthic cyanobacteria, Kahlert and Petterson, (2002) found that benthic algal nutrient status, and to a lesser extent benthic algal biomass, were not readily explained by dissolved nutrient levels in the same way as phytoplankton. Furthermore, they argued that nutrient release from living substrates is an important factor supporting the growth of benthic algae and suggested that control of benthic cyanobacteria is more complex than reducing nutrient inputs alone. As with planktonic cyanobacteria, copper sulphate has been widely used for the control of benthic cyanobacteria. Izaguirre and Devall, (1995) found that for benthic cyanobacteria, copper sulphate was best applied as crystals that are large enough to sink to the bottom of the water body before they can dissolve in the water column. Furthermore, since there will be some loss to the water after the copper sulphate settles on the targeted algal growths, it is generally necessary to use higher doses than would normally be used for the control of planktonic algae. The disadvantage of using copper sulphate, apart from general environmental concerns, is that it lyses the cyanobacterial cells so that the intracellular compounds, including T&O compounds, will be released into the water. This applies equally to both planktonic and benthic cyanobacteria. However, if increases in cyanobacterial numbers can be identified early then the potential quantities of compounds that are released can be reduced. However, research has shown that benthic cyanobacteria can be more resistant to the effects of copper algicides than other algal species. Lembi (2000) investigated the tolerance of various mat-forming green algae and the benthic cyanobacterium Oscillatoria, to Cutrine-Plus® (a chelated copper algicide). They found that Oscillatoria had a high tolerance to copper with a copper concentration 6 times that needed to achieve a similar EC50 in benthic green algae. Lembi (2000)

21 TASTES AND ODOURS IN RESERVOIRS suggested that the reason for this resistance may be a combination of mat density and mucilage covering the filaments which reduced the rate of penetration. In addition, sediment can also partially cover the mats and reduce contact with the algicide. Interestingly, Breemen and Ketelaars (1995) found that artificial mixing was able to control planktonic growth in Biesbosch Reservoirs, in the Netherlands, but suggested that the increased transparency from reduced algal biomass in the water column could stimulate growth of both plants and benthic algae.

In 2001, the Happy Valley Reservoir in South Australia experienced high levels of geosmin due to large floating amounts of biomass of Phormidium which originated from extensive mats growing in shallow bays on sediments to a depth of approximately 2 m. The reservoir was treated in a number of ways with mixed results. A 10mm mesh attached to a boat was able to remove some of the floating material. The reservoir was also drawn down in depth by approximately 1.5 m, to expose Phormidium and allow it to desiccate for 2 weeks. However, Phormidium was still healthy in shallow water and evidence of rejuvenation of partially dried mats was observed after they became re-submerged. Furthermore, the cool wet conditions at the time of the draw down reduced the effect of desiccation by keeping the exposed mats wet. Some of the exposed areas on the shoreline were also sprayed with copper sulphate but while this affected the top layer of algal material it was largely ineffective in penetrating thick mats which were able to re-grow.

A novel technique that has been successfully employed to resolve geosmin problems resulting from the growth of a benthic cyanobacterium Oscillatoria sp. in the Biesbosch Reservoirs in the Netherlands involved disturbing the sediment surface with chains pulled by a boat (Van Breeman et al., 1991). The area was first identified by divers as the reservoirs involved were rather deep. This process is considered to be very effective and more environmentally friendly than the use of copper sulphate and may have application in Australian reservoirs. A fortnightly treatment was required to prevent recolonisation.

The purpose of this study was to investigate and characterise the occurrence of benthic cyanobacteria and actinomycete T&Os relative to planktonic sources in raw waters and their risk to the water quality. The results presented will allow for better-informed risk characterisation and assessment of their importance in reservoirs. The project developed sampling, isolation and culturing techniques for both benthic cyanobacteria and actinomycetes to achieve these goals. The study also identified and characterised problem species of both benthic cyanobacteria and actinomycetes at a number of Australian reservoirs and assessed the effectiveness of benthic cyanobacterial control methods currently used in Australian reservoirs.

The report will firstly present results from surveys undertaken at five reservoirs in Australia to determine the presence of benthic cyanobacteria and actinomycetes. This is followed by a description of the processes involved in production and loss of tastes and odours in reservoirs and results of laboratory and field based experiments used to determine values for these processes with natural populations of benthic cyanobacteria. A model that incorporates T&O generation and loss processes was developed and applied to two operating reservoirs. Results from experiments investigating the effectiveness of desiccation for control of benthic cyanobacteria in source waters are discussed. Finally, a general overview of management issues that reservoir operators should be aware of in relation to dealing with T&O issues is given.

22 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

2 SURVEY OF RESERVOIRS

2.1 Introduction A hierarchy of factors operating at different spatial and temporal scales determines the distribution of micro-organisms in reservoirs. Climate and geology are the dominant factors, which in turn affect inundation rates, hydrology, geomorphology and biogeography, which collectively will have an effect on growth of actinomycetes and benthic cyanobacteria.

2.2 Aim To determine changes in benthic cyanobacteria and actinomycete numbers over the spring/summer period in Hope Valley and Happy Valley Reservoirs, South Australia, and relate these changes observed to ambient abiotic factors such as sediment type, water temperature, and irradiance. The presence/absence of benthic cyanobacteria and actinomycetes was also determined at three other Australian reservoirs: North Pine Dam in Queensland, Grahamstown Dam in New South Wales and Yan Yean Reservoir in Victoria.

2.3 Materials and Methods Ten transects were located at equal distance around the reservoir, encompassing the range of different substrates, depth gradients and microclimates. Transect locations were chosen using GPS to minimise bias associated with seeking out areas of high benthic algal biomass. Transects started at the water’s edge, gradually progressing deeper to a maximum depth of 2 m. Five quadrats, separated by a distance of approximately 1 m, were assessed by wading at four depths: 0.1, 0.25, 0.5 and 0.75 m, and an additional quadrat was sampled at 2m depth with the assistance of divers. (Figure 2.1). The distance from the water’s edge to a water depth of 0.75 m was measured to provide an estimation of the transect slope using trigonometry. The quadrat consisted of galvanised wire mesh of 16 squares each measuring 0.1×0.1 m (Figure 2.2).

0.1 m 0.25 m 0.5 m 0.75 m 2 m Water’s edge

Figure 2.1 Schematic of the reservoir transects showing sampling depths.

Benthic Mat Coverage, presence or absence of mats was recorded under each of the 25 intersections of the galvanised wire mesh. This minimised potential problems associated with determining if a square was completely covered, half-covered or only partially covered. For each quadrat this gave a count between 0 and 25, depending upon the extent of benthic algal cover. A total of 5 quadrats were measured at each depth and a mean value of benthic mat coverage was calculated.

23 TASTES AND ODOURS IN RESERVOIRS

Intersection

Figure 2.2 0.4m x 0.4m quadrat divided into 16 squares and 25 intersections. Each intersection was used to calculate benthic mat coverage (see text).

To determine the type of actinomycetes and benthic cyanobacteria present, a sediment core was collected from the centre of the third quadrat. Figure 2.3 shows the sediment corer with a diameter of 0.05 m used to collect a core with a surface area of 19.64 cm2. After the sediment core was taken a plunger was inserted (Figure 2.4) and used to push the core out of the end. The top 0.5 cm was scraped off and placed into a sample container. Containers were transported to the laboratory on ice for future analysis.

Benthic Mat Thickness was determined by examining the cross section provided in the sediment core described above and measuring the thickness of the mat. Mat thickness was divided into grades: <0.5 mm, 0.5 – 1.0 mm, 1.0 -2.0 mm, > 2.0 mm, <5.0 mm and >5.0 mm.

Sediment Type at the surface/water interface was also identified using the sediment core. A total of 5 categories containing the majority of a particular particle size were assigned as follows: clay, <1.0 mm, 1.0 -2.0 mm, 2.0 – 5.0 mm, and >5 mm.

24 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

A B Detachable polycarbonate Brass coring column Collar Valve

Figure 2.3 Pictures of Benthic Corer. A. Shows: 1. polycarbonate coring column which is inserted into sediment to required depth 2. threaded brass collar which connects polycarbonate column to valve and allows easy removal of polycarbonate column for insertion of plunger 3. valve which is opened when coring column is inserted into sediment and closed when coring column is removed from sediment – closing the valve prevents sediment core from dropping out of column. B. Shows corer attached to extension column which allows for its use in deeper water from the side of a boat. A rope is attached to the valve which allows remote operation.

Attach rope

25 TASTES AND ODOURS IN RESERVOIRS

A B

Movement of Plunger

Plunger

Figure 2.4 A. Insertion of the plunger into the corer pushes the sediment plug to the top of the column and enables collection of the benthic layer sample. B. Cross section of core showing benthic mat layer at the surface of sediment.

Light Attenuation Coefficient was determined at each location. A Li-cor data logger was connected to two cosine Li-cor light sensors. Incoming irradiance levels, Iz, were then measured simultaneously just below the water surface of the reservoir and at a depth of 0.5 m. The light extinction coefficient was then calculated by using the Lambert-Beer equation

C I S 1 k Dln z T T D I T z E 0 U to determine the extinction coefficient for photosynthetically active radiation (PAR) where I0 and Iz are -1 light intensities at 0 and 0.5 m respectively and kT (m ) is the light extinction coefficient.

2.3.1 Sediment Analysis Core samples collected at each site were analysed for presence/absence of benthic cyanobacteria and actinomycetes. The wet sediment sample was emptied into a tared weigh tray and total weight recorded. Two 1 g samples were taken from this tray and transferred into two separate 10 mL containers and their weight recorded. One sample was used for actinomycete analysis and one sample for benthic cyanobacteria analysis.

Benthic Cyanobacteria

A 5 ml volume of ASM medium and 100 μL of Lugol’s solution was added to each container which was then sealed and mixed. Some samples required further dilution before counting, due to high volume of sediment. Numbers of benthic cyanobacteria were counted under a compound microscope and presented as the number per unit surface area (cm2) of benthic layer.

26 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Actinomycetes Each sample was made up to 5 mL using 0.9% saline solution to achieve a 1/5 dilution. Actinomycete isolation agar plates (Difco™ Actinomycete Isolation Agar), prepared with 5g glycerol per litre of agar, were spread with 100 μL of this dilution. The agar contained the anti-fungal agent ketoconazole at a final concentration of 10 μg/L and the antibacterial agent nalidixic acid at a final concentration of 10 μg/L (Alferova et al., 1989). These compounds were added to suppress bacterial and fungal growth that can smother the agar plate and make identification of actinomycetes difficult. Plates were incubated at 30°C for seven days. Plates were examined for typical actinomycete morphologies and numbers recorded. To help with identification, colonies were gram stained and examined microscopically. Numbers of actinomycetes were presented per unit surface area (cm2) of benthic layer. DNA analysis was also undertaken to further characterise the morphologically confirmed isolates. These isolates were firstly subcultured onto actinomycete agar and incubated as above to obtain a pure culture. Organisms were then harvested and DNA extracted using Soil DNA Extraction Kit (MOBIO Laboratories Inc.).MOBIO SOIL DNA Extraction Kit. 16S rRNA PCR using 11F and 1512R primers was performed as described by Amann et al. (1995). The PCR product was purified using Montage columns (Millipore) and sequenced by the Australian Genome Research Facility using 11F, 1512R, and 1055R primers described by Amann et al. (1995). Resultant data was submitted to GenBank database by a BLAST search at NCBI (National Center for Biotechnology Information) website for similar sequences.

Isolates were also taken from the above pure culture plates added to high purity water (milli-Q®) and analysed for geosmin and MIB. Samples were firstly pre-concentrated by Solid Phase Microextraction (SPME). In this process, MIB, geosmin and other volatile and semi-volatile compounds migrate from the water to the extraction flask headspace with the aid of stirring by a magnetic follower and salting out the solution to saturation point with sodium sulphate. The compounds are adsorbed from the headspace onto the polydimethylsiloxane (PDMS) thin film (100 m) on the SPME fibre. The adsorbed compounds are then thermally desorbed from the SPME fibre directly into the injection port of the gas chromatograph/mass spectrometer (GC/MS) and quantitatively analysed using selected ion monitoring (SIM). This analysis was carried out by a NATA accredited laboratory at the Australian Water Quality Centre, South Australia.

2.3.2 Statistical analysis To determine relationships between different environmental variables and growth of cyanobacteria and actinomycetes a Pearson’s correlation was applied to data collected from each reservoir. Pearson’s correlation determines if a linear relationship exists between two variables. The significance of the correlation was determined with p<0.05 showing a significant relationship. Statistical analysis was calculated using STATISTICA Version 5®.

2.3.3 Variation to Sampling Protocol Where possible, sample methodology as previously described was used. However, for some locations it was not possible to conduct a complete survey at all depths or collect samples as the methodology described due to reservoir terrain and extremely low water levels associated with drought conditions. Differences in sampling protocol will be identified in the Results and Discussion section.

2.4 Results and Discussion 2.4.1 Historical Data from Hope Valley and Happy Valley Reservoirs - South Australia SA Water has measured cyanobacterial cell numbers and geosmin in both Hope and Happy Valley Reservoirs for many years. Analysis of this data provided valuable information about the growth of cyanobacteria and associated geosmin production. Data was obtained from Waterscope® which is the data storage system of SA Water and analysed to identify any trends or correlations between cyanobacterial cell counts and geosmin concentration in the water.

Figure 2.5 presents results for cyanobacterial counts and geosmin concentration at Happy Valley Reservoir since November 1997. Data from Location 1 are presented here because it is the primary location for sampling by SA Water and it is near the reservoir water treatment plant (WTP) intake and

27 TASTES AND ODOURS IN RESERVOIRS therefore has a direct relationship with the geosmin levels measured in the plant. Results show that Anabaena circinalis (Figure 2.5, A) and Phormidium (Figure 2.5, B) numbers are highest during the spring and summer months. Anabaena circinalis is a planktonic cyanobacterium and its presence in water samples is expected. However, Phormidium is a benthic cyanobacterium and finding it in the water column suggests that large amounts would also be growing on sediments around the perimeter of the reservoir. To understand what proportion of the geosmin being measured at the WTP was contributed by Anabaena a theoretical geosmin concentration was calculated (Figure 2.5, D) by combining geosmin cell content measurements made during a large Anabaena bloom in Happy Valley Reservoir in March 2006 (0.041 pg geosmin/A. circinalis cell; see 3.3 Planktonic Cyanobacteria Geosmin Production for details) with A. circinalis cell counts presented in Figure 2.5 A. Two periods: identified as  and  in Figure 2.5 occurred since 1997 where geosmin levels were significantly higher than that expected from A. circinalis cells alone. These higher levels could be attributed to the presence of Phormidium. The period identified by column  shows that intracellular geosmin is significantly higher than extracellular which indicates that the majority of the geosmin entering the WTP is cell bound i.e. from free floating cells of A. circinalis and Phormidium sp.. However, the small amount of extracellular geosmin could be due to lysis of some A. circinalis cells as well as release from benthic Phormidium sp. mats. Intracellular and extracellular measurements have only relatively recently been introduced as a routine category of analysis for South Australian Reservoirs so the record of results for these values are limited.

Figure 2.6 presents historical data for Hope Valley Reservoir. Location 9 is the primary location for sampling by SA Water at Hope Valley and it is near the reservoir WTP intake. Hope Valley Reservoir shows a similar trend to Happy Valley Reservoir with A. circinalis (Figure 2.6, A) and Phormidium sp. (Figure 2.6, B) numbers being higher in the spring and summer months. The period identified by column  shows a similar situation as that recorded at Happy Valley Reservoir i.e. both Phormidium sp. and A. circinalis were detected in the water column but the total geosmin was significantly higher than would be expected from A. circinalis alone as shown in Figure 2.5 D. However, in contrast to Happy Valley Reservoir, columns ,  and  show periods when geosmin was detected at the WTP and only Phormidium, and no A. circinalis, occurred at Location 9. Furthermore, in the period identified by column , extracellular geosmin was significantly higher than intracellular geosmin. The small amount of geosmin identified as intracellular is probably contained within floating filaments of Phormidium. The remaining extracellular geosmin may have entered the reservoir through release from dying and free-floating Phormidium filaments or from benthic mats growing around the reservoir’s perimeter.

In conclusion, this historical data has shown that both Anabaena and Phormidium could contribute to the total geosmin content in different amounts at different times in both Hope Valley and possibly Happy Valley Reservoirs. Furthermore, results from Hope Valley Reservoir show that the total geosmin present in water entering the treatment plant can, on some occasions, come solely from benthic Phormidium sp. In this case the routine historical data has provided important information on the likely sources of and relative contribution of different organisms to odour load in the intake to the water treatment plant by correlation of odour metabolites to the occurrence of cyanobacterial types in the adjacent source water.

While identification of benthic cyanobacteria in routine cell counts can provide a warning of potential problems it provides no indication of the size of the benthic population and their distribution. The extent of the hazard can only be assessed by an appropriately designed benthic sampling program.

28 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

  A 6000 A.circinalis Loc1 4000

2000 cells/mL 0

B 200 Phormidium Loc 1

100 cells/mL

0

C 400 Geosmin WTP

200 ng/L

0

D 400 Geosmin WTP Theoretical from A.circinalis alone

200 ng/L

0

E 400 Extracellular geosmin WTP

200 ng/L

0

F 400 Intracellular Geosmin WTP

200 ng/L

0 1/11/1997 1/11/1998 1/11/1999 1/11/2000 1/11/2001 1/11/2002 1/11/2003 1/11/2004 1/11/2005 1/11/2006 1/11/2007

Figure 2.5 Data collected by SA Water from Happy Valley Reservoir since November 1997 showing A: Anabaena circinalis numbers in water at Location 1, B: Phormidium numbers in water at Location 1 (primary sampling site for reservoir), C: Total geosmin concentration of water entering water treatment plant (WTP) from reservoir, D: Estimated potential geosmin concentration of water entering the WTP (calculated by combining cell counts in A with measured geosmin concentration of 0.041 pg/cell), E: Extracellular geosmin concentration of water entering WTP from reservoir, and F: Intracellular geosmin concentration in water entering WTP from reservoir. Columns  and  represent periods of interest which will be discussed in the report.

29 TASTES AND ODOURS IN RESERVOIRS

    3000 A A.circinalis Loc 9 2000

1000 cells/mL

0

B 100 Phormidium Loc 9

50 cells/mL

0

C 100 Geosmin WTP

50 ng/L

0

100 D Geosmin WTP Theoretical from A.circinalis alone

50 ng/L

0

E 100 Extracellular Geosmin WTP

50 ng/L

0

F 100 Intracellular Geosmin WTP

50 ng/L

0 1/11/1997 1/11/1998 1/11/1999 1/11/2000 1/11/2001 1/11/2002 1/11/2003 1/11/2004 1/11/2005 1/11/2006 1/11/2007

Figure 2.6 Data collected by SA Water from Hope Valley Reservoir since November 1997 showing A: Anabaena circinalis numbers in water at Location 9, B: Phormidium numbers in water at Location 9 (primary sampling site for reservoir), C: Total geosmin concentration of water entering the water treatment plant (WTP) from reservoir, D: Estimated potential geosmin concentration of water entering the WTP (calculated by combining cell counts in A with measured geosmin concentration of 0.041 pg/cell), E: Extracellular geosmin concentration of water entering WTP from reservoir, and F: Intracellular geosmin concentration in water entering WTP from reservoir. Columns, ,  and  represent periods of interest which will be discussed in the report.

30 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

2.4.2 Survey of Hope Valley and Happy Valley Reservoirs – South Australia Both Hope Valley and Happy Valley Reservoirs are important service reservoirs from which water is distributed to the Adelaide metropolitan area after being treated by flocculation, sedimentation, filtration and chlorination processes. Hope Valley has a maximum capacity of 2,840 ML and a surface area of 52 hectares at full supply level of 15 metres. It is supplied by two upstream reservoirs, Kangaroo Creek and Millbrook, which in turn receive water from the River Murray via a pipeline. Happy Valley has a maximum capacity of 11,600 ML; surface area of 178 hectares, and has a maximum depth of 18 metres. It is supplied by the upstream reservoir, Mount Bold which, like Hope Valley Reservoir, receives water from the River Murray.

Hope and Happy Valley Reservoir surveys were undertaken in the spring and summer of 2007, a period when benthic cyanobacteria have historically been observed (Figs 2.5 & 2.6 Historical Data from Hope Valley and Happy Valley Reservoirs - South Australia).

It can be seen that the density of benthic cyanobacteria and Actinomycetes varied from site to site at both Hope Valley and Happy Valley Reservoirs (Figures 2.7 to 2.14). Some sites showed the presence of these organisms at a number of depths whereas at other sites none were found. This indicates that variations in micro-habitat within a reservoir can determine the distribution and growth of these organisms. Furthermore, their growth can vary seasonally i.e. some areas may support growth in spring but not in summer and vice versa. To determine the contribution that these organisms may have on T&Os present in a particular water supply it is important to identify areas where they grow.

Actinomycete numbers in the sediments increased significantly (p<0.05) with increasing water depth in both reservoirs in spring and summer. This is shown in Figure 2.7 for Hope Valley and Figure 2.11 for Happy Valley. There is no explanation for this and further work would be needed to understand this observation.

Interestingly, the same significant (p<0.05) relationship was observed for Oscillatoria sp. i.e. Oscillatoria sp. numbers increased with increasing depth in spring and summer for Happy Valley and summer for Hope Valley. This trend can be seen in Figure 2.8 for Hope Valley and Figure 2.12 for Happy Valley. Higher numbers of Oscillatoria sp. in deeper water may be linked to lower light. Jørgensen et al. (1987) explained that organisms living deeper within the photic zone developed more antenna pigment per reaction centre and became light saturated at lower intensities than those living near the surface and they could also adapt to the prevailing blue-green light at depth by increasing the amount of accessory pigments (e.g. phycoerythrin in cyanobacteria) which can absorb quanta at these wavelengths.

Phormidium sp. has been historically implicated as the main source of geosmin originating from the benthic layer in South Australian reservoirs (2.4.1 Historical Data from Hope Valley and Happy Valley Reservoirs - South Australia). At the time of the spring and summer surveys Phormidium sp. were in low numbers at both Hope Valley (Figure 2.9) and Happy Valley (Figure 2.13) reservoirs. No significant relationship was observed (p<0.05) between Phormidium sp. numbers and all of the environmental variables measured.

Pseudanabaena sp. showed a negative correlation (p<0.05) with depth at Hope Valley Reservoir during spring sampling i.e. higher numbers of Pseudanabaena were found in shallow waters (Figure 2.10). However, the opposite was observed during summer sampling in that a positive correlation was seen indicating that higher numbers were found in deeper waters. This suggests that environmental conditions such as temperature and light levels may affect growth of this cyanobacterial species. Irradiance levels and temperature in the shallow waters during spring was much lower than in the summer months. As temperature and irradiance levels increased Pseudanabaena sp. preferred habitat shifted from the warmer high-light environment found in shallow water to cooler and lower light intensities found in deeper water. No association (p<0.05) was observed between depth and Pseudanabaena sp. numbers in the Happy Valley Reservoir (Figure 2.14).

Benthic mat thickness and coverage increased significantly (p<0.05) with increasing depth at Happy Valley Reservoir in spring and summer as explained by the significant (p<0.05) increase in

31 TASTES AND ODOURS IN RESERVOIRS

Oscillatoria sp. with depth discussed previously. Benthic mat thickness and coverage also increased significantly (p<0.05) with depth at Hope Valley Reservoir in spring but not in summer. This contrasts with observations that numbers of both Oscillatoria sp. and Pseudanabaena sp. did not increase with increasing depth in spring but did in summer. Benthic mats collected from both Hope and Happy Valley Reservoirs were made up of a mixture of benthic cyanobacteria, diatoms and detritus. Therefore, mats collected from deeper water were thicker but were composed of diatoms and detritus.

The effect of other variables that were measured including slope of the bank, light attenuation coefficient, and sediment type, on the growth of actinomycetes and benthic cyanobacteria was found to be not significant (p<0.05).

One of the important variables which can affect growth and distribution of both actinomycetes and benthic cyanobacteria are changes in reservoir depth. Operators at both reservoirs have used level changing as a method for control of benthic cyanobacteria (see Chapter 5). Figure 2.15 shows changes in the water depth in Hope Valley (A) and Happy Valley (B) Reservoirs for the period when surveys were undertaken. It can be seen that for both these reservoirs the level remained constant for approximately two months prior to the spring sampling program. This would have provided time for the communities of both benthic cyanobacteria and actinomycetes to establish. However, in the two months prior to the summer sampling program the level for both reservoirs increased and then decreased until they reached lower levels: 3 m lower for Hope Valley and 1.5 m lower for Happy Valley. This would have influenced the growth of both cyanobacteria and actinomycetes in relation to depth distribution through changes in variables such as water temperature and irradiance at the sediment surface. The dynamic change and range in these variables at the sediment habitat zone however was not quantified over this period during this project.

Actinomycetes were characterised at Hope and Happy Valley Reservoirs by culture isolation onto media and identification by either gram stain or PCR analysis. The results showed that all isolates were from the genus Streptomyces. Difficulties in isolating pure colonies, due to high numbers of overgrowth of other organisms growing the plates, meant that this was the only actinomycete that could be characterised. At both Hope Valley and Happy Valley Streptomyces sp. occurred at all depths across the range from 0.1 to 2m. At Hope Valley Reservoir a total of 55 isolates were characterised from 8 separate locations across depths, and at Happy Valley 37 isolates were obtained from 9 locations and various depths.

The actinomycete isolates from both reservoirs were also characterised for production of geosmin and MIB and all but one isolate from Hope Valley produced geosmin. Approximately 40% of isolates from both reservoirs produced both geosmin and MIB, and the remaining 60% produced geosmin only. From the isolates producing both compounds the relative production varied enormously. For example, an isolate from Hope Valley Reservoir produced 50x more MIB than geosmin (Geosmin/MIB = 0.02) and an isolate from Happy Valley Reservoir produced 1200x more geosmin than MIB (Geosmin/MIB = 1170).

32 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Spring

Actinomycetes (cells/cm2) 0 0 – 100 100 – 1000 1000 – 10,000 >10,000 Not measured Water Depth (m) 2.00 (200 cms) 0.75 (75 cms) 0.50 (50 cms) 0.25 (25 cms) 0.10 (10 cms)

Summer

Figure 2.7 Distribution of Actinomycetes in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites

33 TASTES AND ODOURS IN RESERVOIRS

Spring 10 25 10 50 25 10 0 5 75 5 2 0 200 75 0 5 20 75 0 20

Oscillatoria(cells/cm/cm2 2) 0 0 20 75 0 – 1000 50 25 10 1000 – 100,000 100,000 – 1,000,000 >1,000,000

2 7 0 Not measured 5 5 0 2 0 1 5 0 Water Depth (m)

10 2.000.20 (200 cms) 25 50 75 0.750.75 (75 cms) 2 00 0.500.50 (50 cms) 0.250.25 (25 cms) 20 7 50 5 0 0.10 (10 cms) 2 0.10 1 5 0

10 25 50 75 200

200 75 0 0 50 2 5 7 0 25 5 5 10 2 0 1

10 Summer 25 10 50 5 10 2 0 5 75 5 2 0 200 75 0 5 20 75 0 20

0 20 75 50 25 10

2 7 0 5 5 0 2 0 1 5 0

10 25 50 75 2 00

2 7 0 5 5 0 2 0 1 5 0

10 25 50 75 200

200 75 0 0 50 2 5 7 0 25 5 5 10 2 0 1

Figure 2.8 Distribution of Oscillatoria sp. in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites

34 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Spring 10 0 25 1 0 50 25 1 0 25 75 5 5 0 200 7 0 5 5 20 7 0 20

Phormidium(cells/cm/cm2 2) 0 0 20 75 0 – 1000 50 25 10 1000 – 100,000 100,000 – 1,000,000 >1,000,000

2 Not measured 7 0 5 5 0 2 0 1 5 0 Water Depth (m)

10 2.000.20 (200 cms) 25 50 75 0.750.75 (75 cms) 200 0.500.50 (50 cms) 0.250.25 (25 cms) 2 75 0 5 0 0.100.10 (10 cms) 2 0 10 5

10 25 50 75 200

200 75 0 0 50 2 5 25 7 0 5 5 10 2 0 1

Summer 10 0 25 1 0 50 25 1 0 25 75 5 5 0 200 7 0 5 5 20 7 0 20

0 20 75 50 25 10

2 7 0 5 5 0 2 0 1 5 0

10 25 50 75 2 00

2 7 00 5 5 2 0 1 5 0

10 25 50 75 200

200 75 0 0 50 2 5 25 7 0 5 5 10 2 0 1

Figure 2.9 Distribution of Phormidium sp. in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites

35 TASTES AND ODOURS IN RESERVOIRS

Spring 10 25 10 50 25 10 75 50 25 5 0 200 7 0 5 20 75 0 20

2 PseudanabaenaPseudanabaea/cm/cm(cells/cm22 ) 00 0 20 75 00 – – 1000 1000 50 25 10 10001000 – – 100,000 100,000 100,000100,000 – 1,000,000– 1,000,000 >1,000,000 >1,000,000 2 7 0 Not measured 5 5 0 2 0 1 5 Not measured 0 Water Depth (m) 10 2.000.20 (200 cms) 25 50 7 0.750.75 (75 cms) 2 5 00 0.500.50 (50 cms) 0.250.25 (25 cms) 2 7 0 5 5 0 2 0 0.100.10 (10 cms) 1 5 0

10 25 50 75 200

200 75 0 0 50 2 5 25 7 0 5 5 10 2 0 1

10 Summer 25 10 50 25 75 50 10 200 75 0 25 20 0 5 5 7 0 20

0 20 75 50 25 10

2 7 0 5 5 0 2 0 1 5 0

10 25 50 75 2 00

2 7 00 5 5 2 0 10 5

10 25 50 75 200

200 75 0 0 50 2 5 25 7 0 5 5 10 2 0 1

Figure 2.10 Distribution of Pseudanabaena sp. in the Hope Valley Reservoir during the spring and summer sampling programs at 10 different sites

36 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

5 Actinomycetes (cells/cm2) 6 0 0 – 100 Spring 3 100 – 1000 1000 – 10,000 >10,000 7 4 Not measured Water Depth (m) 2.00 (200 cms) 0.75 (75 cms) 0.50 (50 cms) 2 0.25 (25 cms) 0.10 (10 cms) 8

1 9 10 5 6 3 Summer

7 4

2

8

1 9 10

Figure 2.11 Distribution of Actinomycetes in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites

37 TASTES AND ODOURS IN RESERVOIRS

Oscillatoria 2 5 (cells/cm ) 6 0 0 – 1000 3 1000 – 100,000 Spring 100,000 – 1,000,000 >1,000,000 7 4 Not measured Water Depth (m) 2.00 (200 cms) 0.75 (75 cms) 0.50 (50 cms) 2 0.25 (25 cms) 0.10 (10 cms) 8

1 9 10 5 6 3 Summer

7 4

2

8

1 9 10

Figure 2.12 Distribution of Oscillatoria sp. in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites

38 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

5 Phormidium (cells/cm2) 6 0 3 0 – 1000 Spring 1000 – 100,000 100,000 – 1,000,000 7 >1,000,000 4 Not measured Water Depth (m) 2.00 (200 cms) 0.75 (75 cms) 2 0.50 (50 cms) 0.25 (25 cms) 0.10 (10 cms) 8

1 9 10 5 6 3 Summer

7 4

2

8

1 9 10

Figure 2.13 Distribution of Phormidium sp. in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites

39 TASTES AND ODOURS IN RESERVOIRS

5 Pseudanabaena (cells/cm2) 6 0 3 0 – 1000 Spring 1000 – 100,000 100,000 – 1,000,000 >1,000,000 7 4 Not measured Water Depth (m) 2.00 (200 cms) 0.75 (75 cms) 2 0.50 (50 cms) 0.25 (25 cms) 0.10 (10 cms) 8

1 9 10 5 6 3 Summer

7 4

2

8

1 9 10

Figure 2.14 Distribution of Pseudanabaena sp. in the Happy Valley Reservoir during the spring and summer sampling programs at 10 different sites

40 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

A Hope Valley Reservoir Depth

16 Spring Summer Sampling Sampling 15

14

13 Depth (m) Depth 12

11

10 8/11/2007 1/07/2007 9/09/2007 9/10/2007 8/12/2007 11/07/2007 18/11/2007 28/11/2007 21/07/2007 31/07/2007 10/08/2007 20/08/2007 30/08/2007 19/09/2007 Date29/09/2007 19/10/2007 29/10/2007 18/12/2007 28/12/2007

B Happy Valley Reservoir Depth

12 Spring Summer Sampling Sampling 11

10 Depth (m) Depth 9

8 1/07/2007 9/09/2007 9/10/2007 8/11/2007 8/12/2007 11/07/2007 21/07/2007 31/07/2007 10/08/2007 20/08/2007 30/08/2007 19/09/2007 29/09/2007 19/10/2007 29/10/2007 18/11/2007 28/11/2007 18/12/2007 28/12/2007 Date

Figure 2.15 Changes in water level over a five month period July-Dec 2007: A. Hope Valley Reservoir and B. Happy Valley Reservoir, South Australia. Dates of survey sampling for benthic communities are indicated.

41 TASTES AND ODOURS IN RESERVOIRS

2.4.3 Grahamstown Dam – New South Wales Grahamstown Dam is located approximately 20 km north of Newcastle and is operated by the Hunter Water Corporation primarily as a raw water storage for potable water production. Development of this system is continuing and it is the key water supply scheme for the area providing approximately 40% of the region’s water. Water collected in the dam is sourced from the local catchment and the Williams River in a ratio of approximately 50:50. Water from the dam is pumped about 5 km south to the Grahamstown Water Treatment Plant prior to distribution to Newcastle and the lower Hunter region. This reservoir is approximately 4 km long and 1-2.5 km wide, has a surface area of 2,800 hectares at full supply level of 11.4 m and a capacity of 190,000 ML. Sampling sites are shown in Figure 2.16. Sediment and plant samples were collected from 10 sites around the reservoir. Sampling depths were influenced by reservoir terrain.

Large sections of the reservoir bank were fringed with the emergent macrophyte “torpedo” grass (Panicum repens) which grew out into the reservoir to a water depth of approximately 1 m. Significant amounts of this plant were observed at sites 1, 3, 4, 5, 7, 8, 9 and 10 (Figure 2.17). At some locations there were large amounts of Eleocharis sp. (spike rushes) present on the banks (Figure 2.18). This resulted in large sections of reservoir bank that were shaded up to a depth of 1 m which could retard the growth of benthic cyanobacteria generally found in the shallow sections of reservoirs where light penetration to the sediments is highest. However, benthic cyanobacteria were found attached to torpedo grass at a number of locations which suggests that they may act as a substrate for growth by allowing exposure to higher levels of light.

Benthic Cyanobacteria

Benthic cyanobacteria in sediments were only identified at site 1 with Phormidium sp. at 2.3 x 106 cells/cm2 of sediment surface area. Epiphytic cyanobacteria were also identified at sites 5, 7 and 9 but these were attached to torpedo grass and a quantitative measure was not possible. Cyanobacteria attached to the torpedo grass included Phormidium sp., Pseudanabaena sp. and Anabaena sp.

Actinomycetes

Actinomycetes were detected at all 10 sites. Actinomycetes were only isolated from sediment samples and were not detected on any of the plant material collected. No correlation was found between depth and actinomycete occurrence. However, where isolates were found it was shown that they were all Streptomyces species with Actinomycetale species found at only sites 3 and 8. Isolates of actinomycetes tested for presence of T&O compounds were shown to produce only geosmin.

42 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

1 kilometre

Figure 2.16 Map of Grahamstown Dam showing all 10 sampling locations

43 TASTES AND ODOURS IN RESERVOIRS

Figure 2.17 Location 4 at Grahamstown Dam showing dense growth of Panicum repens (torpedo grass) growing out into the reservoir to a depth of approximately 1 m

Figure 2.18 Location 1 at Grahamstown Dam showing Eleocharis (spike rushes) growing out into the reservoir to a depth of approximately 1 m

44 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

2.4.4 North Pine Dam – Queensland North Pine Dam is located approximately 30 km north of Brisbane and has a capacity of 215,000 ML at a full supply level of 39.6 m (average depth of 9.4 m), resulting in a reservoir with a surface area of 2,181 hectares. It is fed by a local catchment of 348 km2 and supplies approximately 54,750 ML/year to the city of Brisbane and surrounding area.

Sampling sites are shown in Figure 2.19. Sediment samples were collected from 10 sites around the reservoir at depths of 0.1, 0.5 and 0.75 m. Due to the low level of the reservoir at the time of the survey the sediments consisted of fine silt with no identifiable benthic algae or macrophytes (Figures 2.20 and 2.21). This is supported by results of sediment sample analysis which identified Phormidium sp. only at site 1 at a concentration of 1 x 106 cells/cm2 at a depth of 0.5 m. However, this high concentration at one location suggests potential for growth within the reservoir. Actinomycetes were not identified at any of the sites sampled.

1 kilometre

Figure 2.19 Map of North Pine dam showing the sampling locations

45 TASTES AND ODOURS IN RESERVOIRS

Figure 2.20 Location 1 at North Pine Dam near the wall showing the low level of water in the reservoir and fine silt sediment.

Figure 2.21 Location 7 at North Pine Dam showing the low level of water in the reservoir and fine silt sediment (inset).

46 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

2.4.5 Yan Yean Reservoir – Victoria Yan Yean Reservoir is located approximately 30 km north of Melbourne and has a capacity at full supply level of 30,000 ML and a surface area of 560 hectares. The reservoir is approximately 4 km long and 1.5 km wide and has a catchment of 2,250 hectares.

Sampling sites are shown in Figure 2.22. Sediment and plant samples were collected from 10 sites around the reservoir at depths of 0.1, 0.5 and 0.75 m from three sites, northern, centre and southwest, located within the reservoir. Similar to North Pine Dam in Queensland, the very low level of the reservoir at the time of the survey meant that sediments consisted of fine silt, with no identifiable benthic algae or macrophytes seen at any site during the survey (Figures 2.23 and 2.24). Subsequent analysis of sediment samples supported this observation with no cyanobacteria or actinomycetes found in any of the samples.

2.5 Conclusions The historical water quality monitoring data on cyanobacterial occurrence and odour compounds from the two South Australian reservoirs (Hope Valley and Happy Valley) provided a good correlation with the population of odorous cyanobacteria in the reservoir and odour in the water treatment plant inlet water over time. The routine results indicated that both planktonic Anabaena circinalis and benthic Phormidium sp contributed to the total geosmin content in both reservoirs and that water entering the Hope Valley Reservoir treatment plant could contain geosmin derived principally from Phormidium sp. Surveys also mapped the distribution of benthic cyanobacteria and actinomycetes in these two reservoirs; both of which can contribute to T&O problems. Similar surveys at North Pine Dam (Queensland) and Yan Yean Reservoir (Victoria) did not find benthic populations in the reservoir sediments, which may have been due to the low water level and nature of the sediments at the time of the surveys. Benthic Phormidium sp. was found at only one of ten locations sampled in the survey in Grahamstown Dam (New South Wales). However there was a range of epiphytic cyanobacterial communities attached to extensive emergent fringing macrophytes in the reservoir, and these included types that could potentially produce odour metabolites.

47 TASTES AND ODOURS IN RESERVOIRS

500 metres

Figure 2.22 Map of Yan Yean Reservoir showing sampling locations

48 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Figure 2.23 Location 4 at Yan Yean Reservoir showing the low level of water in the reservoir and dried fine silt sediment.

Figure 2.24 Location 8 at Yan Yean Reservoir showing the low level of water in the reservoir and silty sediment.

49 TASTES AND ODOURS IN RESERVOIRS

3 PRODUCTION AND LOSS OF TASTES AND ODOURS FROM RESERVOIRS

There is a lack of knowledge about the mechanisms governing the generation and loss of the odour compounds geosmin and MIB within water storages (Westerhoff et al., 2005). An understanding of these processes would be useful to help predict and manage T&O episodes. As part of this project, a mass balance model has been developed which includes all production and loss processes within a water body (see Section 4 ). Generation processes include growth of benthic and planktonic cyanobacteria, and loss processes include photodegradation, biodegradation, volatilisation and adsorption. This model allows calculation of the concentrations of odour compounds derived from both benthic and planktonic cyanobacteria at the outlet from the reservoir. The model requires rate constants for the various production and loss components. This chapter presents results from a number of experiments undertaken to quantify some of these processes.

3.1 Photodegradation

3.1.1 Introduction Photodegradation through exposure to ultraviolet radiation (UV: wavelengths 280 – 400 nm), has been identified as one of the possible process by which T&O compounds can be lost from natural water (Westerhoff et al., 2005). Studies looking at the effect of UV on T&O compounds have been done with the intention of using it as the water treatment process. Mofidi et al. (2000) exposed “pre- treated natural water” (light transmission of 91 to 95% at 254 nm) spiked with MIB and geosmin at concentrations of 60 ng/L, to UV using a pulsed-UV source (xenon-filled tungsten electrode flashlamp) and a continuously stirred tank reactor. They showed that a UV dose of 10,100 mJ/cm2 reduced MIB and geosmin by 92% and 97%, respectively. A lower dose of 2,600 mJ/cm2 reduced MIB and geosmin by only 9 and 28% respectively. Rosenfeldt et al. (2005) investigated the ability of UV to remove MIB and geosmin from raw source water and treated water. Geosmin and MIB were spiked into water samples to a concentration of 25 and 10 ng/L and exposed to UV using a collimated beam equipped with either a low pressure (set wavelength of 254 nm) or medium pressure (wavelengths of 200 to 400 nm) UV lamp. Results showed that the low pressure and medium pressure lamps removed 10% and 25-50% respectively of both compounds at a dose of 1,000 mJ/cm2. These studies have shown that photodegradation of geosmin and MIB by UV can occur in laboratory based studies. However, there have not been any studies investigating the loss of geosmin and MIB by the UV component of direct sunlight as would occur in a water storage.

3.1.2 Aim The aim of this study was to determine if geosmin and MIB undergo photodegradation after exposure to direct sunlight.

3.1.3 Materials and Methods

Ultra pure water (Milli-Q®) was spiked with MIB and geosmin to achieve a concentration of 3000 ng/L. Mercuric chloride was then added at a concentration of 10 mg/L to prevent any loss of these compounds by biodegradation during the experiment. The experimental design was based on work by King et al. (2008). A total of 4 mL was placed into 50 replicate methylacrylate cuvettes which were capped and then sealed with laboratory film. These cuvettes allow transmission in the wavelength range 285 to 750 nm. The cuvettes were attached to two acrylic sheets (900x70x5 mm), 25 on each sheet, using clear silicon rubber as an adhesive (Figure 3.1). The acrylic sheets were then submerged in a water-filled drum to a depth of 10 cm (Figure 3.2). One sheet of cuvettes was wrapped in aluminium foil to shield them from light for controls to determine any loss by non-radiation processes. A temperature logger was also suspended in the drum to monitor temperature changes over the course of the experiment. The UV exposed and control cuvettes were sampled over a 14-day period and subsamples were stored at 4°C until analysis. A 3 mL volume was taken from each cuvette and made up to 55 mL with milli-Q water to obtain enough volume for analysis of geosmin and MIB by SPME linked to GC/MS and SIM (see 2.3.1

50 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Sediment Analysis). Mercuric chloride was added as a biocide to each bottle after sampling at a concentration of 10 mg/L for sample preservation prior to analysis. Global solar radiation (diffuse plus direct) was measured onsite using a CM3 pyranometer (Kipp & Zonen, Netherlands) connected to a Solrad Integrator data logger (Kipp & Zonen, Netherlands).

Silicon Sealant

Cuvette

Acrylic Sheet

Figure 3.1 Methylacrylate cuvettes used as sample exposure containers were attached to the acrylic sheet using silicon sealant

Temperature Logger

UV exposed Cuvettes UV non-exposed Cuvettes

Figure 3.2 Suspension of sample and control cuvettes immersed in an outdoor tank to a depth of 10 cm to investigate UV degradation of geosmin and MIB in water. A temperature logger was also installed to continuously monitor water temperature.

3.1.4 Results and Discussion Figures 3.3 and 3.4 present results for loss of geosmin and MIB respectively after exposure to natural sunlight for 14 days. Both geosmin and MIB levels in controls reduced at the same rate as UV exposed cuvettes. This suggests that geosmin was lost from the cuvettes but that this loss was not due to breakdown of geosmin or MIB by UV but through absorption onto the plastic walls of the cuvette or through volatilisation through the lid. First-order reaction rates were derived from the data and showed that the loss rate was greater for geosmin than MIB. This indicates that geosmin was more rapidly adsorbed onto the plastic surface and/or loss through the lid was greater.

51 TASTES AND ODOURS IN RESERVOIRS

4000

3500 UV non-exposed UV exposed

3000

2500

2000 y = 2085e-0.0826x 1500 R2 = 0.6423

1000 y = 2198.3e-0.0895x

Geosmin Concentration (ng/L) Concentration Geosmin 500 R2 = 0.7246 0 0 2 4 6 8 10 12 14 Time (days)

Figure 3.3 Changes in geosmin concentration in high purity water after exposure to natural sunlight for 14 days

4000

3500 UV non-exposed UV exposed

3000 -0.0349x y = 3052.1e 2 2500 R = 0.8874 y = 3208.9e-0.0356x 2000 R2 = 0.843 1500

1000 MIB Concentration (ng/L) Concentration MIB 500

0 02468101214 Time (days)

Figure 3.4 Changes in MIB concentration in high purity water after exposure to natural sunlight for 14 days

52 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

25

20

15

10 Temperature (oC) 5

0 0 2 4 6 8 10 12 14 Time (days)

Figure 3.5 Temperature of water in the water-filled drum used in the experiment

Ambient temperature experienced by the experimental samples ranged from 10 to 20°C for the period of the experiment (Figure 3.5). Samples were exposed to a total irradiance (cumulative dose) of 216,777 kJ/m2 of global solar radiation over a 14-day period. Given that approximately 5% of global solar radiation is comprised of UV wavelengths (Foyo-Moreno et al., 1998) the total UV dose received by the samples can be estimated as 10,839 kJ/m2 or 1.08x106 mJ/cm2.

This contrasts to findings by both Mofidi et al. (2000) and Rosenfeldt et al. (2005) who reported high geosmin loss at UV-dose approximately 500X lower than that estimated here. Albeit the UV light in their studies were generated by an artificial source which may be important, as it is not known if the artificial light sources produce higher levels of energy at the specific wavelengths that are involved in the destruction of the geosmin and MIB.

Researchers have found that the reduced penetration of UV light (UVA 400 nm – 320 nm; UVB 320 nm - 290 nm; UVC 290 nm - 100 nm) with depth in water bodies is predominantly determined by dissolved organic carbon (DOC) content. Scully and Lean (1994), Morris et al. (1995) and Huovinen et al. (2003) all showed that as DOC increased the depth of UV penetration decreased. Huovinen et al. (2003) investigated the attenuation of solar UV radiation in five Finnish lakes and found UV-B radiation was attenuated to 1% of subsurface irradiance within 0.5 m and UV-A in the upper 1 m for the clearest lake (DOC = 4.9 mg/L). In the lake with the highest DOC (13.9 mg/L) UV-B radiation was attenuated to 1% of the subsurface irradiance within 0.10 m and UV-A within 0.25 m. DOC in the Hope Valley and Happy Valley Reservoirs for 2006 to 2007 was 6.6 and 6.4 mg/L respectively indicating that transmission of UV radiation would be similarly affected by DOC. The transmission of light through samples of Hope Valley and Happy Valley Reservoir water was measured from 200 – 800 nm using a Cary 1 UV-Visible Spectrophotometer (Varian®) and compared with ultra purified (Milli-Q®) water using a quartz cuvette with 5 cm path length (Figure 3.6). It can be seen from these figures that the transmission of UVA and UVB is significantly reduced, hence its effect on dissolved geosmin and MIB would be diminished with increasing water depth in a reservoir. Only geosmin and MIB at the very surface of the reservoir is likely to be exposed to significant UV radiation.

Therefore, the small loss of geosmin and MIB in exposure to UV radiation in these trials combined with low penetration into a water column, particularly in reservoirs with high DOC, suggests that loss by photodegradation in natural water is likely to be minimal.

53 TASTES AND ODOURS IN RESERVOIRS

Ultra-pure Water

Hope Valley Reservoir

Happy Valley Reservoir

Figure 3.6 Transmission spectra of Ultra-pure water and water from Hope Valley and Happy Valley Reservoirs for the range 200-800nm through a 5 cm path length.

54 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

3.2 Biodegradation 3.2.1 Introduction A number of researchers have investigated the removal of T&Os in water treatment plants by sand and activated carbon filters (Hoefel et al 2006; Ho et al., 2006), or by strains of bacteria previously isolated from treatment plants and reservoirs (Table 3.1). However, relatively little is known about biodegradation in the natural environment of the reservoirs. Recent studies by Westeroff et al. (2005) in three Arizona reservoirs identified biodegradation as the most important loss process compared with volatilisation, photodegradation and adsorption. Furthermore, they observed that laboratory biodegradation rates using reservoir water were equivalent to the total net loss rates recorded in the field.

Table 3.1 Micro-organisms reported to be involved in the biodegradation of MIB and geosmin (Ho et al., 2007)

MIB Geosmin

Microorganism Source Microorganism Source

Pseudomonas sp. Izaguirre et al. (1988), Bacillus cereus Silvey et al. (1970), Egashira et al. (1992), Narayan and Nunez (1974) Tanaka et al. (1996)

Pseudomonas Egashira et al. (1992) Bacillus subtilis Narayan and Nunez (1974), aeruginosa Yagi et al. (1988)

Pseudomonas Oikawa et al. (1995) Arthrobacter Saadoun and El-Migdadi putida atrocyaneus (1998)

Enterobacter sp. Tanaka et al. (1996) Arthrobacter Saadoun and El-Migdadi globiformis (1998)

Candida sp. Sumitomo (1988) Rhodococcus Saadoun and El-Migdadi moris (1998)

Flavobacterium Egashira et al. (1992) Chlorophenolicus Saadoun and El-Migdadi multivorum strain N-1053 (1998)

Flavobacterium sp. Egashira et al. (1992)

Bacillus sp. Ishida and Miyaji (1992), Lauderdale et al. (2004)

Bacillus subtilis Yagi et al. (1988)

3.2.2 Aim To determine biodegradation rates of geosmin and MIB in natural source water.

3.2.3 Materials and Methods Water was collected from the surface of two drinking water supply reservoirs: Myponga Reservoir (10th May 2007), and the Blue Lake, (29th May 2007). Myponga Reservoir has a history of blooms of the geosmin producing cyanobacterium Anabaena circinalis and had a series of major blooms 2 which finished months prior to sample collection, whereas the Blue Lake is a pristine water source with no history of cyanobacterial growth.

The water from both reservoirs was filtered through a 5 μm cellulose acetate membrane to remove any contaminating algae or cyanobacteria (e.g. Anabaena circinalis) that may have been present. The water was then spiked with geosmin and MIB for an initial concentration of 100 ng/L for both compounds. This water was then transferred to 110 mL brown glass bottles, which were filled with

55 TASTES AND ODOURS IN RESERVOIRS no air gap, sealed and then maintained in the dark at 10, 15, 20 and 25°C in constant temperature environment cabinets. Controls consisted of either Myponga Reservoir or Blue Lake water with the preservative mercuric chloride added and stored in the dark at 20°C. Triplicate samples were destructively analysed for geosmin and MIB concentration at 0, 6, 12, 24, 48 hours for both Reservoirs and then at 168 and 336 hours for Blue Lake. Mercuric chloride (10 mg/L) was added as a preservative to prevent further biodegradation before analysis. Geosmin and MIB were analysed by SPME linked to GC/MS and SIM (see 2.3.1 Sediment Analysis).

3.2.4 Results and Discussion Biodegradation of geosmin was greater at higher temperatures for Myponga Reservoir (Figure 3.7). While a small loss in geosmin occurred at 10°C the greatest loss was found at 20 and 25°C. This is not surprising and indicates that the geosmin-degrading bacteria had a higher metabolic activity at higher temperatures. In contrast, geosmin loss was much lower in the Blue Lake water (Figure 3.8). In Myponga Reservoir water there was 70% reduction in geosmin after 48 hours at 20°C compared to no loss in Blue Lake water for the same period. However, after 14 days (336 hours) geosmin concentration in Blue Lake water had reduced to 45% of the control indicating that there were micro-organisms in the Blue Lake water capable of degrading geosmin, but that numbers were low compared to the Myponga.

Biodegradation rates for MIB were also measured for both Myponga (Figure 3.9) and Blue Lake (Figure 3.10) water. Interestingly, biodegradation of MIB did occur in Myponga Reservoir water after 48 hours at all temperatures although it was significantly lower than for geosmin. This suggests that micro-organisms with capacity to degrade MIB were present in Myponga Reservoir. In contrast, there was no loss in MIB from Blue Lake water even after 14 days (336 hours) at 20°C implying that while there were geosmin-degrading bacteria present there was no evidence of micro- organisms that could remove MIB.

The occurrence of geosmin- or MIB-degrading bacteria is likely to be influenced by a prior history of these compounds being available as a substrate in a water body. Certainly a history of large and persistent blooms of geosmin-producing Anabaena circinalis in Myponga Reservoir appears to have allowed the development of a microbial community capable of rapid biodegradation of geosmin. Biodegradation also occurred in Blue Lake over time, but at a much lower rate, indicating it required a lag time for a suitable population to develop.

If we consider the geosmin-degrading bacteria in Myponga Reservoir were in excess then a first- order reaction can be fitted to the data which can be used to predict loss rates for geosmin by biodegradation (Table 3.2). For Myponga Reservoir at the time that water samples were collected, it would take approximately 24 hours for geosmin to reach 50% of its original concentration through biodegradation alone, at temperatures of 20 and 25°C.

In the seven month period prior to the water samples being collected from Myponga Reservoir there were three large blooms and one minor bloom of Anabaena circinalis. Figure 3.11 shows surface cell numbers near the dam wall adjacent to the water supply offtake and mean cell number for the sampling throughout the reservoir. The figure shows cell numbers reached a maximum concentration of 280,000 cells/mL at the dam wall. It can also be seen that total geosmin (including both dissolved and cell bound), in the water leaving the reservoir was generally correlated with cell numbers of Anabaena circinalis (Figure 3.12) indicating that geosmin was predominantly intracellular. This is also indicated in Figure 3.13 showing cell numbers at Location 1 (key sampling location near the dam wall and off-take), and corresponding intracellular and extracellular geosmin concentrations. Extracellular geosmin was significantly lower than intracellular geosmin. The ratio of dissolved/total geosmin was in the range 20-30% during the early stages of the blooms, which reflects a period of active growth. This later increased progressively to up to 80% at the later stages and after collapse of the bloom. This is best illustrated in the bloom from October- November, 2006 (Figure 3.13). Each bloom generally ended with a sudden collapse and rapid loss of cell numbers in the reservoir (see Figure 3.11). This would have been expected to have been accompanied by mass cell lysis and release of geosmin into the dissolved phase. However this was not the case as reflected in the geosmin monitoring data measured in the water at the reservoir outlet (BC tap) which then entered the water filtration plant.

56 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

The expected or likely concentration of geosmin associated with the bloom at its peak can be calculated from cell population number and the geosmin cell quotas (0.025 pg/cell of Anabaena circinalis) that were measured for this Myponga population

The potential increase in dissolved geosmin was calculated for each bloom using the intracellular geosmin concentration measured for this population (See Section 5) together with the total cell number at the peak of the bloom. Cell numbers measured at the dam wall location were used for the calculation as this was closest to the reservoir outlet and where the loss of cells would have been expected to have had an immediate effect on the water leaving the reservoir. Figure 3.12 shows that the dissolved geosmin concentration would be expected to increase significantly immediately after the collapse of each of the four blooms. The blooms were generally concentrated in the top 5 m of the reservoir which equates to approximately 15% of the total volume of the reservoir. Using the first bloom in October/November 2006 as an example, the potential total dissolved geosmin concentration in the top 5 m water layer would have been 7,150 ng/L as a result of lysis of 286,000 cells/mL over 4 days. Allowing for dilution throughout the entire water column a concentration of greater than 1,000 ng/L would be expected. However, the actual geosmin concentration remained extremely low, less than 70 ng/L after 4 days.

This is an important observation from a water supply operations perspective as the T&O hazards from these Anabaena circinalis blooms were not realised based upon population size and odour production rates. The explanation for the rapid disappearance of geosmin most likely relates to losses by biodegradation and/or volatilisation.

Reservoir water temperature at that time was approximately 19°C and laboratory experiments presented here showed that biodegradation rates in Myponga Reservoir water at this temperature were rapid i.e. geosmin had a half life of approximately one day. Therefore, it is suggested that biodegradation could partially explain the dramatic loss of geosmin from the water following the collapse of the bloom. Alternatively, it is possible that the Anabaena cells within Myponga Reservoir sedimented and were lost rapidly from the water column. Further work is needed to understand these processes.

Table 3.2 Degradation rate of geosmin in Myponga Reservoir water at different temperatures (reaction considered to be first-order).

2 Temperature First-Order Reaction R Half-life KB (ºC) equation (days) (day-1) (Geosmin concentration at time t) 10 = e-0.198.t 0.903 3.5 0.20 20 = e-0.681.t 0.947 1.0 0.68 25 = e-0.701.t 0.971 0.99 0.70

57 TASTES AND ODOURS IN RESERVOIRS

140 10oC 20oC

25oC Control 120

100

80

60 Geosmin Concentration (ng/L) Concentration Geosmin 40

20

0 0 6 12 24 48

Time (hours)

Figure 3.7 Loss of geosmin in water from Myponga Reservoir at different temperatures (10-25°C) over 48 hours. Samples were spiked with geosmin to reach 100 ng/L and sampled at various times up to 48 hours. Error bars are ± standard deviation.

120

20oC

100 Control

80

60

Geosmin Concentraion (ng/L) 40

20

0

0 6 12 24 48 168 336

Time (hours)

Figure 3.8 Loss of geosmin over 14 days in water from Blue Lake at 20°C. Samples were spiked with geosmin to reach 100 ng/L and sub-sampled at various times up to 14 days. Error bars are ± standard deviation.

58 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

140 10oC 20oC

25oC Control 120

100 …..

80

60

40 MIB Concentration (ng/L) Concentration MIB

20

0 0 6 12 24 48 Time (hours)

Figure 3.9 Loss of MIB in water from Myponga Reservoir at different temperatures (10-25°C) over 48 hours. Samples were spiked with MIB to reach 100 ng/L and sampled at various times up to 48 hours. Error bars are ± standard deviation.

120 20oC Control

100

80

60

40 MIB Concentraion (ng/L) Concentraion MIB

20

0 0 6 12 24 48 168 336 Time (hours)

Figure 3.10 Loss of MIB over 14 days in water from Blue Lake at 20°C. Samples were spiked with MIB to reach 100 ng/L and sub-sampled at various times up to 14 days. Error bars are ± standard deviation.

59 TASTES AND ODOURS IN RESERVOIRS

350000 1200 AnabaenaAnabaenaAnabaena circinaliscircinalis Circinalis atat LocationLocation at 1 1 (cells/mL) (cells/mL) Dam Wall

300000 AnabaenaAnabaenaAnabaena circinaliscircinalis circinalis reservoirreservoir reservoir meanmean mean (cells/mL)(cells/mL) 1000

GeosminGeosminGeosmin concentrationconcentration concentration atat BC BCin reservoir taptap (ng/L)(ng/L) outlet 250000 800

(cells/mL) 200000 600 150000

400 Geosmin (ng/L) 100000 Anabaena circinalis 200 50000

0 0 8-Apr-07 7-Jun-07 7-Feb-07 9-Dec-06 10-Oct-06 30-Oct-06 28-Apr-07 18-Jan-07 19-Mar-07 27-Feb-07 20-Sep-06 19-Nov-06 29-Dec-06 18-May-07 Date

Figure 3.11 Cell numbers of Anabaena circinalis at a location the near water supply offtake (Location 1), mean cell number in the reservoir and total geosmin concentration in water at the reservoir outlet (BC tap) for three major blooms of Anabaena circinalis in Myponga Reservoir which occurred over the spring and summer of 2006/2007.

350000 1200 AnabaenaAnabaena circinalis Circinalis at Locationat Dam Wall 1 (cells/mL) 300000 ActualActual geosmin Geosmin concentration Concentration at BCat Reservoir tap (ng/L) Outlet (ng/L) 1000

Expected geosmin concentration calculated 250000 fromSeries2 cell numbers (ng/L) 800 (cells/mL) 200000 600 150000

400 Geosmin (ng/L) 100000 Anabaena circinalis 200 50000

0 0 8-Apr-07 7-Jun-07 7-Feb-07 9-Dec-06 10-Oct-06 30-Oct-06 28-Apr-07 18-Jan-07 27-Feb-07 19-Mar-07 20-Sep-06 19-Nov-06 29-Dec-06 Date 18-May-07

Figure 3.12 Cell numbers of Anabaena circinalis at a location near water supply offtake (Location 1); the actual measured total geosmin concentration at the water supply offtake (BC tap); and expected geosmin concentration calculated assuming complete lysis of all A. circinalis cells and release of geosmin into water at the reservoir outlet for three major blooms of Anabaena circinalis in Myponga Reservoir over the spring and summer of 2006/2007.

60 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

300000 Anabaena circinalis at Location 1 250000 200000 150000

cells/mL 100000 50000 0

1200 Intracellular Geosmin Concentration at WTP 1000 800 600 ng/L 400 200 0

1200 Extracellular Geosmin Concentration at WTP 1000

800

600 ng/L 400

200

0

1.0 Extracellular/Total Geosmin Ratio at WTP 0.8

0.6

0.4

0.2

0.0 7/02/2007 7/03/2007 4/04/2007 2/05/2007 20/09/2006 18/10/2006 15/11/2006 13/12/2006 10/01/2007 30/05/2007

Figure 3.13 Cell numbers of Anabaena circinalis near dam wall (Location 1), intracellular and extracellular (dissolved) concentrations of geosmin measured at the reservoir water treatment plant (WTP) and the ratio of dissolved (extracellular) to total geosmin for three major blooms of Anabaena circinalis in Myponga Reservoir over the spring and summer of 2006/2007.

61 TASTES AND ODOURS IN RESERVOIRS

3.3 Volatilisation

Odor compounds can be lost from a water body through the process of volatilisation. Volatilisation is the movement of volatile chemicals from the bulk water phase of a water body across the air–water interface into the air. It is usually expressed as a first-order process, where its rate is directly proportional to the concentration of the chemical undergoing the process. A number of models have been developed (Liss and Slater, 1974; Upstill-Goddard, 2006) using the concept of a diffusive sub- layer or two films to describe the transfer of these volatile compounds at the air–water interface.

Mackay and Leinonen, 1975 showed that the half-lives for volatile compounds such as pesticides, polychlorinated biphenyls, or hydrocarbons at a water depth of 1 m show rapid evaporation from solution. In situations where the water body is turbulent with frequent exchange between the surface water layer and the bulk, for example, in a fast-flowing shallow river, or during whitecapping on a lake or ocean, the liquid phase mass transfer coefficient may be considerably increased and the evaporation rate increased correspondingly. For depths greater than 1 m the half-life is correspondingly increased. Interestingly, Mackay and Leinonen, 1975 explain that mass transfer coefficients are relatively temperature insensitive.

Prabhakar et al., 2007 investigated the influence of physical factors on trichloroethylene evaporation from surface water using open containers that were exposed to varying wind speeds and agitation. They showed that the evaporation mechanism was more complex than a simple first order process and found that factors such as wind and turbulence in water were highly critical in determining volatilisation rates.

Leyris et al., 2000 identified a need for the development of reliable methods for the measurement of odour emissions to obtain an accepted standard. They developed a wind tunnel that could measure odour emissions from areal sources and undertook a laboratory study measuring concentrations of the odorous pollutant diethyl sulphide above a liquid surface under different wind conditions. A dynamic flux chamber was also developed to measure emission rates.

In the current study the flux of odour compound lost from a reservoir through volatilisation, Fv, was determined using two-film theory and the following equation:

 FV KAw (Cwater - Cair ) KAwCwater where

Fv is the flux of the odorant due to volatilisation, Cwater is the dissolved odourant concentration Cair is the odourant concentration in air KAW is the mass transfer coefficient.

The amount of loss per unit time through the whole reservoir with an area of As can be determined using the equation shown below.

FVAs KAwAsCwater

The gaseous/aqueous phase mass transfer coefficient, KAW, was calculated from the empirical correlations developed by Wanninkhof and Bliven. The significance of odour loss from water storages through volatilisation will be described at two reservoirs, Hope Valley Reservoir in South Australia and Tai-Lake Reservoir in Taiwan, in Chapter 4. KAW for geosmin and MIB was determined using the average wind speed and Henry’s law constants for the compounds at 20°C. While this work has provided basic knowledge about volatilisation in these reservoirs, further work, is needed to fully understand the rate of volatilisation and its significance in relation to other loss processes. This work should involve the use of wind tunnels and dynamic flux chambers.

62 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

3.4 Geosmin Production by Planktonic Cyanobacteria

3.4.1 Introduction The accurate determination of cell content and production of T&O compounds by natural cyanobacterial populations can used in conjunction with other data such raw water cell counts to estimate the potential odour load and corresponding hazard to the water treatment process and eventually to water supply. This information is also an important input for the T&O Model that has been developed as part of this project and which is presented in Section 4.

3.4.2 Aim The purpose of this study was to determine intracellular concentrations of geosmin (i.e. cell quotas) in natural populations of Anabaena circinalis from different Australian reservoirs.

3.4.3 Materials and Methods Highly concentrated cell samples were collected by plankton net tows during blooms of the geosmin producing cyanobacterium Anabaena circinalis (Figure 3.14) at several of the key study sites including Myponga, Happy Valley and Mount Bold Reservoirs in South Australia and Yan Yean Reservoir in Victoria. These samples were transferred into 1.25 L PET bottles and transported back to the laboratory on ice. Samples were then left overnight on the bench at ambient temperature (22°C). Due to the buoyancy of cyanobacterial cells, a thick scum formed on the surface of the sample (Figure 3.15) which was transferred by pipette to a sterilized measuring cylinder. These concentrated cells were diluted with ASM culture medium (Gorham et al., 1964) to produce a uniform suspension at a cell density of approximately 2x106 cells/mL. This mixture was then sub-sampled and analysed for total geosmin concentration. A further sample was also taken and preserved with Lugol’s solution for cell counts. Due to the consistency of the cell suspension it was assumed that the geosmin analysed was likely to be entirely from the cell material and the minor contribution from reservoir water was ignored. Furthermore, the scum material was comprised entirely of Anabaena circinalis as other algae and detritus sedimented to the bottom of the sample bottle while standing overnight. Geosmin was analysed by SPME linked to GC/MS and SIM (see Section 2.3.1). Cell numbers were determined using a compound microscope at 200X magnification and a Sedgewick-Rafter chamber. The number of cells in the first 30 trichomes was counted and the mean number of cells per trichome determined. The number of cells per trichome was multiplied by number of trichomes counted, at least 100 in total, and then converted to cell count using the appropriate dilution factor. This gave a counting error of ± 20% (Hötzel and Croome, 1999).

Figure 3.14 Photograph of net tow being pulled through an algal bloom

63 TASTES AND ODOURS IN RESERVOIRS

Figure 3.15 Photograph showing buoyant Anabaena circinalis cells accumulated at the top of a 1.25 L PET sample bottle.

3.4.4 Results and Discussion Geosmin content of cells varied significantly between reservoirs (Table 3.3). The lowest value, 0.021 pg/cell, was recorded for a population at Myponga Reservoir and the highest was 0.051 pg/cell, at .

Table 3.3 Intracellular geosmin content of Anabaena circinalis collected from various reservoirs during cyanobacterial blooms. Reservoir Date Geosmin (pg/cell)

mean s.d.

Myponga 24/10/2006 0.029 0.006 9/11/2006 0.021 0.004 Happy 30/3/2006 0.041 0.009 Valley Mount Bold 9/11/2006 0.051 0.011 Yan Yean 20/4/2006 0.046 0.007

64 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

3.5 Geosmin Production by Benthic Cyanobacteria 3.5.1 Introduction Benthic cyanobacteria in natural water bodies produce significant emissions of odour metabolites and these can often be readily detected by the human nose , even if the material is not easily visible. It follows that the production rates and emission rates in natural water bodies must be significant. It is important to measure the rates of production and emission of odours by natural populations in situ to understand the relative significance of the hazard these benthic organisms pose to water quality. The production or ‘flux’ are also required for developing the mass balance model of T&O production and loss for a reservoir described in Section 4.

Flux chambers are often employed in measuring emissions of oxygen and carbon dioxide (Boucher et al., 1994), trace metals (Point et al., 2007) and phosphorus (Gächter et al., 1988; Clavero et al., 2000) from the sediment/water interface of lakes and reservoirs. They have also been used to measure the movement of volatile organic compounds from contaminated soil to the air (Woodbury et al., 2006; Persson et al., 2005; Gao and Yates 1998). In this component of the project flux chambers (columns) were used to measure the rate of geosmin released from benthic mats attached to reservoir sediments in the Hope Valley Reservoir, South Australia.

3.5.2 Aim To determine the geosmin release rate in situ from a natural benthic mat attached to reservoir sediment, and also the loss or emission of geosmin from the water column to the atmosphere in a natural water body (Hope Valley Reservoir, South Australia).

3.5.3 Materials and Methods Two separate experiments were conducted in the Hope Valley Reservoir: Experiment 1 (November, 2006) and Experiment 2, one month later (December, 2006). This is historically the seasonal period when significant growth of Phormidium sp. occurs in the reservoir (see Section 2.4.1). Experiment 1 ran for 24 hours and commenced at 13:30 on 6 November and Experiment 2 ran for 48 hours and commenced at 16:00 on 11 December, 2006.

The experimental design involved the installation of separate “Flux” and “Loss” columns or chambers in shallow water (approximately 500 mm depth) to determine geosmin production by the benthic cyanobacterial community and loss processes from the overlying water in the reservoir.

The chamber or columns were constructed from clear acrylic plastic tube (length: 500 mm; internal diameter: 190 mm; wall thickness: 5 mm). The tubes also had a flange welded at one end that could be used to fix a base plate to seal the column. The “Loss Column” was sealed at one end by an acrylic plate fastened to a flange with a rubber gasket, silicon sealant and eight 20 mm bolts (Base plate dimensions: 470 mm diameter; thickness: 10 mm. The “Flux Column” was left open at both ends and was used to isolate a section of benthic (Phormidium) mat (283.5 cm2) (Figure 3.16).

The Loss column was filled with 10 L of reservoir water and spiked with geosmin standard. This was then positioned on top of the reservoir sediments at a depth where the level of water in the column matched the reservoir depth. The columns were immersed at a suitable depth that would avoid water exchange, but at the same time would also avoid as much as possible any heating of the column structure which has the potential to affect water temperature and volatilisation from the enclosed water. Therefore, loss of geosmin from this column was directly due to photodegradation, biodegradation and volatilisation.

The Flux column was pushed into the sediment, isolating a section of benthic mat, at the same depth as the Loss column to achieve a volume of 10 L of water above the mat. The depth of water in the columns was measured at each time a sample was taken to compensate for any changes in depth due to evaporation. The columns were supported in the water by securing them to steel posts (Figure 3.17). Sample volumes of 50 mL were collected in the morning and afternoon from both columns and analysed for geosmin concentration as described in Chapter 3. The position of columns at Hope Valley Reservoir is shown in Figure 3.18. At the end of each experiment the majority of the water was removed from the Flux column. The remaining water and benthic mat

65 TASTES AND ODOURS IN RESERVOIRS was collected and placed into a plastic container. This was then transported back to the laboratory on ice. The combined benthic mat/reservoir water sample was gently mixed to obtain a homogeneous sample. This was then weighed to give a total wet weight for this sample. A sub- sample was taken, its wet weight determined, and ASM media added to achieve a dilution suitable for determining Phormidium sp. cell number by microscopy. Lugol’s solution was added to preserve the sample. Flux Loss

open top open top

Water level

Geosmin movement

Benthic mat sealed bottom open bottom

Figure 3.16 Diagram showing construction and setup for the Flux column inserted in to the sediments and the Loss column placed on top of the sediments.

Figure 3.17 Flux and Loss Columns deployed in the Hope Valley Reservoir.

66 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Experimental site

Figure 3.18 Outline of Hope Valley Reservoir showing orientation and experimental site for Loss and Flux column placement.

3.5.4 Results and Discussion The change in geosmin in both the Flux and Loss columns for the two experiments is presented in Figures 3.19 and 3.20 respectively. The concentration change over time was analysed by simple linear regression and regression lines and associated R2 values are shown on the Figures. The slope of the linear relationship represents a rate of increase in the Flux column, or decrease in the Loss column respectively of geosmin concentration. Geosmin was lost at a constant rate over the time periods used in both experiments. However, the overall flux rate was different between the separate experiments (Figures 3.19 and 3.20).

Production rates were greater in November (0.0938 ng/cm2/h), than in December (0.0067 ng/cm2/h), which related to a greater cell density of Phormidium sp. on the surface of the sediment contained within the flux column at the time of the two experiments (November: 32.38 x 105; December: 3.38 x 105 cells/cm2)

Release of geosmin from benthic mats is a result of cell lysis (Vilaita et al. 2004; Wu and Jüttner 1998; Utkilen and Froshaug, 1992) and is caused by exposure to high levels of irradiance, high temperatures, shear stresses from water movement and nutrient limitation. Therefore, the rate of geosmin release could vary considerably throughout the reservoir both spatially and over time. Furthermore, studies by Sabater et al. (2003) and Vilaita et al. (2004) suggested that benthic mats don’t always stay attached to the sediment. They explained that once mats reach a critical mass they detach from the sediment due to oxygen build up from increased rates of photosynthesis. These mats then float free on the surface of the water where exposure to high temperatures and high levels of irradiance result in increased rates of cell lysis and release of geosmin. Further work is required to determine the variation in geosmin release and the factors that determine this variation to obtain a more accurate estimate of geosmin flux from benthic mats.

The loss rate was greater in November (3.74 ng/L/h) than in December (1.06 ng/L/h) (Table 3.4). A number of variables can be responsible for this difference; higher wind speeds across the air-water interface at the surface of the column can increase the rate of volatilisation, higher water temperature will increase volatilisation rates as well as biodegradation rates i.e. greater numbers and more active biodegrading bacteria. The average water temperature at the time of the experiment was slightly lower in November (18°C), than in December (20°C), so it may be expected that loss of geosmin by volatilisation would be greater in December, however this was not the case. Average wind speed was 14 km/h from the South-East for the experiment in November and 17 km/h from the North-West to South-West for the experiment in December. It can be seen

67 TASTES AND ODOURS IN RESERVOIRS from the reservoir map (Figure 3.18) that winds from the west would have more effect on the columns than a southerly direction due to the experimental site being more open to the west and protected from the south. Again, this seems not to have influenced geosmin loss, as lower losses were recorded in December when the columns were more exposed to the influence of wind. It would seem from these observations that a higher biodegradation rate is the most probable explanation for higher geosmin losses seen in November.

The experiments show that loss rate for geosmin was higher than the release rate from the benthic mat. This would suggest that there should have been no geosmin in the water entering the water treatment plant. However, routine monitoring results undertaken by the reservoir manager, SA Water, showed that geosmin was present in the reservoir water entering the WTP at the time of the experiments. Interestingly, a higher extracellular (dissolved) geosmin concentration compared to intracellular geosmin concentration was recorded during both the November and December experiments. The planktonic cell counts at Location 9, situated next to the WTP, showed that A. circinalis was not found, but that the benthic cyanobacterium Phormidium sp. was present (Table 3.4). The detection of benthic cyanobacteria in a planktonic count (i.e. in the water column), suggests that they are actively growing on sediments around the perimeter of the reservoir and become dislodged and contribute to the planktonic population. Greater numbers of benthic cyanobacteria around the area of the WTP could explain the presence of geosmin in the water entering the plant. This supports the importance of a detailed survey of a reservoir to identify areas where cyanobacterial mats are growing.

Table 3.4 Rates of geosmin release from benthic mats in Flux columns and rates of loss from Loss columns for two separate experiments and comparative geosmin concentrations, both extracellular (dissolved) and intracellular, measured by SA Water at the WTP and cyanobacterial cell numbers, both Anabaena circinalis and Phormidium sp., at Location 9. Experiment Experiment 2

1 (6/11/2006) (11/12/2006) Flux Column: increase in geosmin above 0.0938 0.0067 benthic mat (ng/cm2/h) Loss Column: 3.74 1.06 loss rate of geosmin from water (ng/L/h) Total number of Phormidium sp. cells on 32.38 x 105 3.38 x 105 sediment surface in Flux Column (cells/cm2) Extracellular (dissolved) geosmin concentration 17 9 at Water Treatment Plant (ng/L) Intracellular geosmin concentration at Water 4 Not detected Treatment Plant (ng/L) A. circinalis cell numbers at Location 9 Not detected Not detected (cells/mL) Phormidium sp. cell number at Location 9 7 2 (cells/mL)

68 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

A) 3000

2500

2000 y = -37.437x + 2594.2 1500 R2 = 0.9986

1000

500 Total Geosmin in Tube Loss (ng) Total Geosmin 0 0 102030 Time (hours)

B) 1600 1400 y = 26.57x + 611.17 1200 R2 = 0.7634 1000

800 600 400

200 Total Geosmin inTotal Geosmin Flux Tube (ng) 0 0102030 Time (hours) Figure 3.19 Change in total geosmin concentration in A) Loss column and B) Flux column over the 24-hour period of Experiment 1 (November, 2006). The experiment commenced on the 6 November, 2006 at 13:30. Graphs include fitted linear regression and associated R2 values.

69 TASTES AND ODOURS IN RESERVOIRS

A) 1000 900 800 700 600 500 y = -10.59x + 841.46 400 R2 = 0.9315 300 200 100 Total Geosmin in Tube Loss (ng) Total Geosmin 0 0 1020304050 Time (hours)

B) 250

200

150 y = 1.907x + 112.22 R² = 0.3674 100

50

Total Geosmin in Flux Tube (ng) Tube in Flux Geosmin Total 0 0 1020304050

Time (hours)

Figure 3.20 Change in total geosmin concentration in A) Loss column and B). Flux column over the 48-hour period of Experiment 2 (December, 2006). The experiment commenced on the 11 December, 2006 at 16:00. Graphs include fitted linear regression and associated R2 values.

70 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

3.6 Conclusions This study set out to measure the production of odour metabolites by natural populations of both planktonic and benthic cyanobacteria. It also aimed to determine the rates and relative significance of loss processes including biodegradation, volatilisation and photodegradation in reservoirs.

Potential production of geosmin by planktonic cyanobacteria was estimated by the determination of the cell content or ‘cell quota’ of the compound in a range of natural populations of Anabaena circinalis. Samples of natural blooms of A. circinalis in Myponga, Happy Valley and Mount Bold Reservoirs in South Australia and Yan Yean Reservoir in Victoria had a geosmin content ranging from approximately 0.02 to 0.05 pg/cell. These cell quotas can be combined with cyanobacterial cell counts to estimate the potential range of odour yield for natural populations and provide an estimate of the likely water treatment challenge.

The process by which T&Os are released from benthic cyanobacterial mats is poorly understood. A novel set of experiments using specially designed chambers measured the release rate or “flux” of geosmin from benthic mats in the reservoir. These flux experiments were done in Hope Valley Reservoir in November and December 2006. This is historically a period of high growth of Phormidium sp. in this reservoir. Production rates ranged from 0.0067 to 0.0938 ng/cm2/h in December and November respectively. The difference was directly related to the cell density of Phormidium sp. in the mats at the time of the experiments (November: 32.38 x 105; December: 3.38 x 105 cells/cm2).

Production rates were greater in November (0.0938 ng/cm2/h), than in December (0.0067 ng/cm2/h), which related to a greater cell density of Phormidium sp. on the surface of the sediment contained within the flux column at the time of the two experiments (November: 32.38 x 105; December: 3.38 x 105 cells/cm2)

The loss rate of geosmin from the experimental columns was also correspondingly higher in November (3.74 ng/L/h) than in December (1.06 ng/L/h). This could be due to a range of variables including higher wind speeds across the air-water interface at the surface of the column, which can increase the rate of volatilisation, and also higher water temperature which will increase both volatilisation and biodegradation.

The study demonstrated that the rate of odour metabolite release can be measured in situ and is dependent upon the density of cyanobacterial cells within the mat. The density of benthic cyanobacteria per unit area can also be estimated at individual locations by cell counts, and this can be used to derive a reservoir odour production rate or yield. The largest source of error in deriving the total reservoir odour yield is likely to be the estimation of total benthic coverage given that surveys indicated that the distribution and density of benthic mat is highly variable throughout a reservoir.

This study found that loss processes such as biodegradation and volatilisation have a major effect on the concentration of T&Os in a reservoir. Biodegradation was found to be temperature dependent with optimum rates occurring between 20° and 25°C in Myponga Reservoir, South Australia. Biodegradation followed a first-order reaction rate and, based upon this, the half-life of geosmin ranged from 1 to 3.5 days at 20 and 10°C respectively in Myponga Reservoir at the time of these experiments. This means that it would take approximately 24 hours for geosmin to reach 50% of its original concentration through biodegradation alone, at 20-25°C. Based upon these findings it is proposed that biodegradation could partially explain the observation of very rapid loss of extracellular T&O compounds following the sudden collapse of Anabaena circinalis blooms in Myponga Reservoir in the summer of 2006/2007.

Volatilisation rates of odourants from the reservoir surface were calculated by determining the overall mass transfer coefficient, which was estimated from the two-film theory, and the gaseous and aqueous phase mass transfer coefficients were calculated from the empirical correlations previously published for lakes (See Section 4). Volatilisation can be shown to be significantly affected by wind speed i.e. high winds increased loss of both geosmin and MIB. A concentration gradient exists between the water/air interface and high winds can rapidly remove volatile compounds from the surface of the water allowing the movement of fresh volatiles to the interface. High winds can also create waves on the surface which act to increase the area of water/air interface resulting in increased rates of volatilisation.

71 TASTES AND ODOURS IN RESERVOIRS

Photodegradation has been identified as a possible loss mechanism for T&Os in the environment. While photodegradation has been observed by other researchers this project showed that both geosmin and MIB were not significantly affected by exposure to sunlight over 14days in an outdoor tank experiment at ambient temperature. In addition, measurements of the transmission of UV radiation through reservoir water demonstrated relatively high absorbance, which is most likely due to high levels of dissolved natural organic matter. It follows that significant exposure to UV would occur only near the water surface in the South Australian reservoirs in this study.

72 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

4. TASTE AND ODOUR MODEL

4.1 Introduction While the problems associated with geosmin and MIB in drinking water have been extensively studied, the mechanisms governing the generation and loss of these compounds in reservoirs are not well understood (Westerhoff et al., 2005). An understanding of these processes would be advantageous to source water managers by helping to predict T&O episodes and to develop suitable management plans for their elimination.

4.2 Aim To develop a model that is able to estimate the generation and loss processes involved in the production of cyanobacterial T&O compounds in reservoirs and to apply it to two drinking water supply reservoirs with a history of T&O problems.

4.3 Model Development Figure 4.1 illustrates the schematic representation of the model. In developing the model, the reservoir was assumed to be completely mixed. The mass balance of target odourant, geosmin or MIB, in the reservoir can be expressed as equation (1).

 BMi i BM -BBM S -BL (1)  in,i out,i i i t i iii

Where: M is the mass of the odourant in the reservoir, Min is the odourant entering the reservoir with water inflow, Mout is the odourant leaving the reservoir with water outflow, S represents the production rate from different sources of odourant, and L is the combination of sink terms for the odourants, which may include volatilisation, biodegradation, and photodegradation.

Equation (1) can be expanded to account for all component terms to produce Equation (2).

 C C q ( P Cs ) M Q C  Q C C V F A  R V  R V F A V sed in in ( P ) R m m p p b b - v s (2) t t out Cp Cp where C is the dissolved odourant concentration, Cp is the cell-bound odourant concentration, Cs is the odourant adsorbed on suspended particles other than cyanobacteria cells, Msed is the mass of sediments interacting with the odourants, q is the adsorption capacity of odourant onto the sediment, Qin and Qout are the flow rate of influent and effluent of the reservoir, respectively, Cin is the odourant concentration present in the influent, RCp is the production rate of the odourant by planktonic cyanobacteria, Rp is the photodegradation rate of the odourant, Rb is the biodegradation rate of the odourant, Fm is the flux of the odourant from benthic cyanobacteria, Am is the area covered cyanobacterial mat, Fv is the flux of the odourant due to volatilisation, As is the surface area of the reservoir, V is the volume of the reservoir, VCp is the volume of water with planktonic cyanobacteria, Vp is the volume of water in which photodegradation may occur, and Vb is the volume of water in which biodegradation may occur.

73 TASTES AND ODOURS IN RESERVOIRS

Figure 4.1 Schematic representation for the mass balance of a cyanobacterial odourant in a reservoir

In equation (2), the first term represents the mass change of the odourants, which may include the changes in dissolved phase, cell-bound phase, and adsorbed phase on suspended non- cyanobacterial particles. The second term represents the mass change of odourant adsorbed onto the sediment. On the right-hand side, the terms include the mass entering and leaving with water flow, production of the odourant by planktonic cyanobacteria and benthic cyanobacteria, degradation by photo-oxidation and biodegradation, and loss caused by volatilisation.

4.4 Site Description

4.4.1 Hope Valley Reservoir The model was validated by applying it to Hope Valley Reservoir, South Australia. A description of the reservoir is given in 2.4.2. The average residence time for HVR is ~ 40 days. Geosmin is the problem odourant in this reservoir and it is known to be produced by the planktonic cyanobacterium Anabaena circinalis and the benthic cyanobacterium Phormidium sp. Meteorological data for the Reservoir was obtained from the Australian Bureau of Meteorology Weather Station at Kent Town, Adelaide, which is approximately 9 km from HVR. This data is summarised in Table 4.1.

4.4.2 Tai-Lake Reservoir Tai-Lake Reservoir (TLR) is located in the eastern part of an offshore island, Kinmen Island, Taiwan. The storage capacity and surface area of the reservoir are 1.65106 m3 and 3.6105 m2 respectively, and the average depth is approximately 4.5 m. The average residence time for TLR is ~ 200 days. The reservoir was monitored for major water quality parameters seasonally by Taiwan EPA (TWEPA), from 2002 to 2008. The average monthly wind speed and temperature on the Kinmen Weather Station, which is about 5 km from TLR, are summarised in Table 4.1. Tai-Lake has a history of Microcystis spp. blooms producing the T&O compound ß-cyclocitral and Oscillatoria sp. which produces MIB.

74 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Table 4.1 Average monthly temperature and wind speed for Hope Valley (HVR) and Tai-Lake Reservoirs (TLR) - 2005 data

Temperature (°C) Wind Speed (m/s) Temperature (°C) Wind Speed (m/s) Hope Valley Reservoir Tai-Lake Reservoir Jan 22.4 3.2 12.2 3.6 Feb 20.5 3.2 12.0 3.7 Mar 19.7 2.6 12.8 3.4 April 19.6 2.7 19.2 2.8 May 15.0 1.9 23.4 3.0 June 12.9 3.1 26.0 3.5 July 11.3 2.4 28.1 2.9 Aug 12.8 3.3 27.5 3.2 Sep 14.1 3.1 27.2 3.6 Oct 16.6 3.7 23.9 4.2 Nov 19.0 3.4 20.8 3.7 Dec 21.4 3.9 14.1 4.2

4.5 Estimation of Model Parameters

4.5.1 Emission of Geosmin from Benthic Mats If the benthic mat coverage can be determined for the whole reservoir then an estimate of the potential geosmin contribution from the benthic mat to the total reservoir geosmin load can be calculated. Section 3.4 of the current report presents results from experiments carried out at HVR to determine total quantities of geosmin released into the reservoir water per day.

4.5.2 Planktonic Generation of Geosmin Section 3.3 presents intracellular geosmin concentration for Anabaena circinalis from several Australian reservoirs. Using this information, in combination with routine determination of planktonic counts, it was possible to determine the potential geosmin contribution to a particular reservoir from A. circinalis. The results have shown that the intracellular geosmin concentration of A. circinalis populations can vary between reservoirs even those in relatively close proximity: 0.21 pg/cell at Myponga Reservoir and 0.51 pg/cell at Mount Bold Reservoir which are located approximately 50 km apart. Therefore, to achieve the best estimation of T&O compound contribution from a known number of cyanobacterial cells, the intracellular concentration of the compound must be determined for the cyanobacteria and reservoir under investigation.

4.5.3 Mass in the Influent and Effluent of the Reservoir The mass of geosmin entering the reservoir with inflow and the mass leaving of the reservoir via the treatment plant off-take were determined based on the monitoring data obtain from SA Water and Taiwan Water Authority for the calendar year 2005 (Table 4.2).

75 TASTES AND ODOURS IN RESERVOIRS

Table 4.2 Values for geosmin loss terms in HVR and TLR sourced and calculated using routine monitoring data and referenced papers.

Flow Out Volatilisation Biodegradation -1 -1 -1 Parameter Keff = Qout/V (day ) KAW (As/V) (day ) KB (day )

Geosmin 2-MIB Geosmin 2-MIB Hope Valley 0.025** 0.041 0.045 0.68# Reservoir Tai-Lake Reservoir 0.005** 0.057 0.062 0.03 – 0.07##

Reference/Method Data for 2005 Two-film Theory # Biodegradation rate

Calendar year from Calculation in this report SA Water and ## Westerhoff et al. Taiwan Water (2005) Authority EPA**

4.5.4 Rate of Biodegradation Degradation of geosmin in the reservoir water and water treatment plants has been conducted by several research groups (Westerhoff et al., 2005; Ho et al., 2007; McDowall et al., 2009). Either a first-order or zero-order decay model was employed to fit the experimental data. The extracted rate constants, together with those obtained in this study, are listed in Table 4.3.

Table 4.3 Summary of geosmin degradation rates measured by different researchers and work presented in the current study.

Temperature, KB Rate Water Matrix Reference °C Law 20 1.1-1.4 First- Reservoir water after McDowall day-1 order sedimentation processes, South (2008) Australia 20 0.12- First- Finished water with biofilms Ho et al. 0.58 order spiked, South Australia (2007) day-1 22 0.80 Zero- Reservoir Water, Arizona, US Westerhoff et day-1 order al. (2005) 0.20 First- Reservoir Water, South Australia This study 10 day-1 order 0.68 First- 20 day-1 order 0.70 First- 25 day-1 order

4.5.5 Rate of Photodegradation The limited loss of geosmin and MIB through exposure to UV radiation, reported in experiments presented as part of this report in Section 3.1. Photodegradation, combined with low penetration of UV into a water column, particularly in reservoirs with high DOC, suggests that loss of these compounds by photodegradation is minimal.

4.5.6 Rate of Volatilisation

For calculating the volatilisation loss, the overall mass transfer coefficient, KAW, was estimated from the two-film theory, and the gaseous and aqueous phase mass transfer coefficients were calculated from the empirical correlations developed by Wanninkhof and Bliven 1991 for lakes. To calculate KAW , the average wind speed for the two reservoirs, and Henry’s law constants for geosmin and MIB at 20°C were used.

76 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

4.5.7 Rate of Adsorption

The adsorption of odourants onto Suspended Solids (SS) (CS) in reservoir water may be estimated from equation (4):

CS K OC f OC CCSS (4) where KOC is the partition coefficient for organic compounds between organic carbon and water, fOC is the organic carbon content of SS, and CSS is the SS concentration in water.

Based on equation (4), it is clear that the mass of geosmin and MIB adsorbed onto suspended solid (SS) may be important if the reservoir has high concentrations of SS, or if the SS has high organic carbon content. According to Kile et al. (1995), KOC may be estimated from the octanol/water partition coefficient, KOW (KOC ~ 1/2.6-1/5 KOW). For the two compounds, KOW was 1,350 for MIB and 5,010 for geosmin. Therefore, assuming KOC ~ 1/4KOW, and fOC ~ 0.01 (a high value for suspended solid in water), CS is ~ (12.5 mL/g)(C CSS) for geosmin and ~ (3.4 mL/g)(C CSS) for MIB. For a typical reservoir with either geosmin or MIB concentration of 20 ng/L, Cs is still < 1 ng/L even if the SS concentration (CSS) is as great as 2,000 mg/L. Therefore, in most reservoirs, the partitioning of odourants into free and particle-bound (other than those bound with cells) indicates that particle adsorption may be negligible. However, adsorption onto sediment may need to be considered separately, as the sediment surface involved in the reservoir system may represent a more significant sorption pathway than the suspended phase.

4.6 Assessing and Modelling the Importance of Loss Terms for Odourants Due to the relatively low rate of geosmin loss by photodegradation and adsorption onto suspended solids these mechanisms were not included in the current analysis. Therefore, the three mechanisms of outflow, volatilisation and biodegradation were assessed. It was assumed that volatilisation in the reservoirs can be modeled using two-film theory, and biodegradation follows a first-order reaction. To simplify the analysis, the reaction volume for biodegradation was assumed to be the same as total volume of the reservoirs, as both TLR and HVR are shallow reservoirs. Then, the three loss terms may be expressed in equation (3) as:

Q A  out (C C ) S Loss P - KBC - K C (3) V AW V where KB is the first-order reaction rate constant for biodegradation, and KAW is the overall mass transfer coefficient between air and water interface.

Table 4.3 shows the magnitude for each of the three pathways for odourant loss in the two reservoirs. In the analysis, a first-order rate constant, Keff, is used to represent outflow, which is equal to Qout/V. This is estimated by the inversion of mean residence time. Note that for TLR, (Cp+C) is ~ C, as most of the odourant (2-MIB in this reservoir) is in the dissolved phase. However, for HVR, (Cp+C) is ~ 4C, since about 80% of the geosmin in the reservoir is cell-bound, and only 20% is in the dissolved phase. For calculating the volatilisation loss, the overall mass transfer coefficient, KAW, was estimated from the two-film theory, and the gaseous and aqueous phase mass transfer coefficients were calculated from the empirical correlations developed by Wanninkhof and Bliven (1991) for lakes. In calculating KAW, the average wind speed for the two reservoirs, and Henry’s law constants for geosmin and 2-MIB at 20°C were used. The KB value for Happy Valley Reservoir was that measured for laboratory experiments using Myponga Reservoir water at 20°C presented in the current report (Section 4.).

The losses for the three mechanisms of outflow, volatilisation and biodegradation are of the same order of magnitude for HVR, while that for outflow is smaller for TLR (Table 4.3). The difference may be due to the fact that the water residence time of TLR is much longer than that for HVR. In addition, the loss caused by outflow in HVR needs to account for cell-bound odourants, which make this mechanism more important. It is also apparent that volatilisation was a very important loss mechanism for both reservoirs. This is very different to that reported by Westerhoff et al. (2005) for three reservoirs

77 TASTES AND ODOURS IN RESERVOIRS in Arizona, U.S. where they concluded that volatilisation was not important. The three reservoirs studied by Westerhoff et al. (2005) were all much deeper (all > 35 m) than TLR and HVR. In addition, the average wind speed was much lower, being only 0.5 m/s compared to 3.1 and 3.5 m/s for HVR and TLR, respectively. Therefore, the mass losses caused by volatilisation in the three U.S. reservoirs are not likely to be significant.

Table 4.3 also shows that KAW (AS/V) is approximately 10% greater for 2-MIB than for geosmin. The two odourants are different in molecular weight (MW) and Henry’s Law constants. The lower MW and higher Henry’s Law constant for 2-MIB leads to slightly higher KAW and tendency to be volatile.

Estimated KB values based on laboratory data sourced from the literature and presented in the current report provide an indication of the importance of this loss mechanism. Westerhoff et al. (2005) showed that the average degradation rate in laboratory batch experiments for both geosmin and 2-MIB, presented as a zero-order decay, was approximately 1.0 ng/L/day. Considering that the typical concentration of geosmin in HVR is ~ 20 ng/L, and that of 2-MIB in TLR is ~ 50 ng/L, 10-25 days would be needed for degrading half the odourant concentration in the two reservoirs. This would be -1 equivalent to a KB ~ 0.028-0.069 day , if a first-order reaction is assumed.

4.7 Effect of Environmental Parameters on Loss Terms A sensitivity analysis of the major environmental parameters involved in the loss terms, including biodegradation and volatilisation, was conducted for Hope Valley Reservoir. A deterministic model, called LAKEVOC, was employed for the simulation. The LAKEVOC model was developed by USGS (Bender et al., 2003) for the simulation of the VOC budget in lakes, and is very similar to a simplified version of the model proposed in this study. Only input, output, biodegradation, and volatilisation were considered in the LAKEVOC model. To apply this model, the temperature and wind speed for the reservoir (Table 4.1) were used. In addition, an annual average inflow and outflow of 2005 was used for hydrodynamic data.

An initial concentration of 20 ng/L of geosmin was assumed for the reservoir. In addition, the geosmin production was assumed to be equivalent to 2 ng/L/day for the whole reservoir during the warmer season (October to March), and was 0 during the cooler season (April to September).

The effect of wind speed on the geosmin concentration in Hope Valley Reservoir is illustrated in Figure -1 4.2, in which biodegradation rate constant (KB) is assumed to be 0.1 day . It is apparent that the wind speed has very strong impact on the geosmin concentration in the reservoir. For example if monthly wind speed were reduced by half, geosmin concentration in the reservoir may build up by a factor of 2, suggesting the relative importance of volatilisation in the reservoir. Figure 4.2 also shows the effect of relative humidity and atmospheric pressure on the geosmin concentration. Changes of up to 30% in relative humidity and 10% in atmospheric pressure, had negligible effect upon the simulated results, suggesting that these variables do not have a strong impact on the geosmin concentration in the reservoir.

The effect of biodegradation on the geosmin concentration in Hope Valley Reservoir is shown in Figure 4.3. For the analysis, the average wind speed shown in Table 4.2 was used. The figure indicates that biodegradation does have an impact on the geosmin concentration in the water. However, it is not as significant as wind speed. For example, when KB is increased by a factor of 2 (from 0.05 to 0.1 day-1), the geosmin concentration only changes by 10%. This is smaller than that observed for wind speed, in which geosmin concentration may decrease by a factor of 2 with a doubling of wind speed.

In both Figures 4.2 and 4.3, the geosmin concentration increases from day 0 to day 70, and decreases from day 290 to day 350, although the monthly input of geosmin was constant during the entire period. This may be partly due to the effect of temperature changes between months. The temperature gradually decreased from January to March (day 0 to day 70), and then increased from October to December (day 290 to day 350), causing the volatilisation rate to become progressively lower and then higher in the two time periods, respectively.

78 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

1.5 0.1 Wind Speed 1.25

1

0.75 0.5 Wind Speed 0.5 Model Tolerance 0.25

Geosmin 1000 ng/LConcentration, Avg. Wind Speed 0 0 50 100 150 200 250 300 350 400 Time, Day

Figure 4.2 Effect of wind speed on geosmin concentration in Hope Valley Reservoir (HVR).

0.35

0.3 K = 0.1 day-1 0.25

0.2

0.15 K = 0.5 day-1 0.1 K = 1.0 day-1 0.05 Geosmin Concentration, 1000 ng/LGeosmin Concentration, 0 0 50 100 150 200 250 300 350 400 Time, Day

Figure 4.3 Effect of biodegradation rate on geosmin concentration in Hope Valley Reservoir (HVR).

4.8 Conclusion A model was developed to determine the concentration of T&O compounds based upon generation and loss processes in reservoirs. The model incorporated the major variables of inflow and outflow of water to the reservoir, production of odour metabolites by planktonic and benthic cyanobacteria, their degradation by photo-oxidation and biodegradation, and loss by volatilisation presented in the previous chapter. As a result, this model was able to critically assess the relative importance of these processes for determining the concentration of geosmin and 2-methylisoborneol (MIB) in both a Taiwanese and an Australian reservoir.

79 TASTES AND ODOURS IN RESERVOIRS

Further development of this model could allow for the incorporation of the rate constants for the processes into a deterministic reservoir process model to then do comprehensive forecasting and to use the information for management. An example of such a model is the hydrodynamic and water quality model ELCOM-CAEDYM developed by the Centre for Water Research at the University of Western Australia (http://www.cwr.uwa.edu.au/services/models.php). ELCOM-CAEDYM couples the three-dimensional hydrodynamic model ELCOM (Estuary, Lake and Coastal Ocean Model) with the aquatic ecological model CAEDYM (Computational Aquatic Ecosystem Dynamics Model). This allows for investigations into the spatial and temporal relationships between physical, biological and chemical variables in water bodies, over single events or seasonal to annual timescales. The application of such a model to a reservoir could enable operators to predict taste and odour episodes caused by planktonic and benthic cyanobacteria.

80 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

5 CONTROL OF BENTHIC CYANOBACTERIA BY DESICCATION

5.1 Introduction Desiccation or ‘drying out’ has been identified as a possible method for the control of benthic cyanobacterial mats in water supply reservoirs. In 2001, Happy Valley Reservoir, South Australia experienced high levels of geosmin due to large floating mats of Phormidium which originated from extensive mats growing in shallow bays on sediments to a depth of approx 2 m. The reservoir was treated in a number of ways with mixed results. A 10 mm mesh was towed from a boat and was able to remove some of the floating material for disposal. The reservoir water level was also drawn down, approximately 1.5 m, to expose Phormidium and allow it to desiccate or air-dry for 2 weeks. However, Phormidium was still healthy in shallow water and evidence of rejuvenation of partially dried mats was observed after they became re-submerged. Furthermore, the cool wet conditions at the time of the draw-down reduced the effect of desiccation by keeping the exposed mats wet. Some of these exposed areas on the shoreline were also sprayed with copper sulphate but this affected only the top layer of algal material and was largely ineffective in penetrating thick mats which were able to re-grow. Hawes et al. (1992) studied Antarctic benthic cyanobacteria and observed that Phormidium recovered slowly following desiccation of 26 days with these mats also providing viable cells for subsequent regrowth. Benthic algal regrowth in intermittent streams in the Grampians National Park, situated in South-Eastern Australia, showed regrowth capability of algal biofilms following seven weeks of desiccation (Robson, 2000). Results showed that the dry residual biofilm was an important drought refuge for cyanobacteria and played an important role in recolonisation once flows had resumed.

Davis (1972) provided a comprehensive overview of desiccation studies which included over 130 species of green algae, diatoms and cyanobacteria. Davis found that species from the green and blue-green algae showed the greatest resistance to desiccation and temperature extremes, whereas species of red and brown algae showed the least. The filamentous green alga Klebsormidium rivulare was found to increase cell wall thickness, increase cellular concentrations of low-molecular- weight solutes resulting in changes to osmotic potential and shift from a protein-synthesis metabolism to one of carbohydrate and lipid accumulation upon desiccation (Morison and Sheath, 1985). Peterson (1996) citing the former, stated that benthic microalgae employ a variety of mechanisms to withstand long-term desiccation and the onset of drying can induce physiological changes in algal cells that can, in fact, enhance resistance.

5.2 Aim To determine the effectiveness of desiccation or drying out using natural air drying as a viable method for controlling benthic cyanobacteria in reservoirs.

5.3 Materials and Method Large sections of Phormidium sp. mat were freshly harvested from the sediments at a depth of 50 cm from Hope Valley Reservoir, South Australia. The mats were covered with reservoir water and transported on ice in an insulated container to the laboratory where they were stored at 4°C overnight. The mat was laid out flat and circular disks, with a diameter of 28 mm, were cut from the mat using a 50 mL syringe with the end removed. These were placed onto small pre-weighed plastic trays (70 mm square by 25 mm deep). The wet weight of each disc was measured. A total of 55 trays were produced. The trays were then transferred to large shallow open containers to enable the trays to be easily moved outside during the day for 8 hours into the sun between 09:00-17:00 and returned inside the laboratory at night. Fine plastic bird net (1 mm thickness; 20 mm square mesh) was placed over the containers to prevent the small trays from being blown away in strong winds and to protect from falling leaves. Replicate sample trays (5 replicates) were sampled destructively approximately every 7-10 days over 46 days. Replicate trays were reweighed to determine percentage of water lost from the mat. A small amount of mat, less than 0.5%, was removed to determine viability by autofluorescence microscopy, and the remainder was analysed for chlorophyll a content using a spectrophotometric method. Viability of benthic mat material was determined using chlorophyll autofluorescence produced under a reflected light fluorescence microscope at a magnification of x100 or x200. The principal of the assay was that when healthy cells containing active

81 TASTES AND ODOURS IN RESERVOIRS chlorophyll were exposed to green excitation light, they emitted longer wavelength red fluorescence. The number of fluorescent cells was expressed as proportion of total cells counted.

5.4 Results and Discussion Temperature and humidity conditions during the exposure of the benthic mat samples are presented in Figure 5.1 and 5.2. Humidity values were a combination of outside values measured at the Adelaide office of the Bureau of Meteorology and the humidity level provided by the air-conditioned laboratories (40%). Temperature values were a combination of the constant temperature provided in the air-conditioned laboratory (22°C), and ambient outside air temperatures recorded at 9am, 2pm and 5pm at the Adelaide office of the Bureau of Meteorology.

40 35 C)

o 30 25 20 15 10 Temperature ( Temperature 5 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 Time (days)

Figure 5.1 The range of temperatures to which the benthic mat samples were exposed during the experiment.

100 y 80

60

40

20 % Relative Humidit 0 0246810121416182022242628303234363840424446 Time (days)

Figure 5.2 Percentage Relative Humidity to which the benthic mat samples were exposed during the experiment.

The figures showed the relative humidity and temperature varied throughout the experiment due to the samples being stored in an air-conditioned laboratory overnight and then taken outside between 09:00-17:00. The air-conditioned building provided a constant temperature of 22°C and relative humidity of 40%.

82 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

40

30

20

10 ug chl a/g dry weight 0 0 2 4 6 8 10121416182022242628303234363840424446 Time (Days)

Figure 5.3 Chlorophyll a content of the Phormidium mat samples that were allowed to desiccate in natural sunlight over a period of 46 days.

t 100 90 80

cells tha cells 70 60 50 40

were viablewere 30 20 phormidium 10 0 % of 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 Time (Days)

Figure 5.4 Viability of Phormidium sp. in cyanobacterial benthic mat samples that were allowed to desiccate in natural sunlight over a period of 46 days.

The reduction in chlorophyll a content occurred at a similar rate to loss in viability with a drop from 27.6 to 10.9 g chl a /g dry weight over the 46 day period (Figure 5.3). A total of 93.8 % of the Phormidium cells in the sample mats were identified as viable at the commencement of the experiment (Time 0) (Figure 5.4). This reduced gradually over 46 days to reach a viability of 23.7%. If the loss in viability is considered to be first-order reaction then the reduction rate can be expressed by:

% viable cells = 95.16e -0.0319.Time (days)

From this relationship it can be estimated that 5% of cells in the mat would still be viable after 90 days of desiccation under the conditions used in this experiment. It is possible that if cells remained whole they could still contain geosmin which could be released into the water when the water level is raised again.

5.5 Conclusions Desiccation or ‘drying out’ was investigated as a possible method for the control of benthic cyanobacterial mats collected from Happy Valley Reservoir, South Australia. This involved an air- drying trial and assessment of cell viability over a period of approximately 7 weeks. The results showed that some cells within a benthic mat could remain viable following long periods of desiccation. The trial indicated that around 20% of the population of Phormidium was still viable after 46 days of drying out in the air and it is expected that around 5% would still be viable after 90 days i.e. 3 months. It is very possible that this proportion of healthy cells could provide a seed population for re-growth of benthic mats when they are re-wet.

83 TASTES AND ODOURS IN RESERVOIRS

6 MANAGEMENT OF BENTHIC CYANOBACTERIA

While there is considerable experience with the management of planktonic cyanobacteria there is very little knowledge on the management of benthic cyanobacterial problems. A discussion on the options for management of benthic cyanobacteria is provided here, however given that knowledge on the ecology of these organisms is relatively limited it is difficult to develop a comprehensive management plan at this stage.

Assessing the potential risk of T&Os from Benthic Cyanobacteria

There are some general reservoir features that may be used to assess the potential risk of benthic cyanobacteria developing in a reservoir and the generation of T&O problems.

Reservoir bathymetry

The distribution of benthic cyanobacteria is restricted by the depth of light penetration. Shallow reservoirs, especially those with high water transparency, will have greater area available for benthic cyanobacterial growth than deep reservoirs. Shallow reservoirs with gently sloping margins as opposed to deep, steep sides will have higher sediment surface to water volume ratios, resulting in less potential for dilution of the odours. The proximity of the benthic mats to the water off-take is also an important consideration. As a general guide, benthic cyanobacteria need about 1% of the surface irradiance to grow, however this may be lower depending upon the species or type. The area of the reservoir potentially available for benthic cyanobacterial growth can be calculated from the extinction co-efficient of the water and the bathymetry of the reservoir.

Integrity of the benthic mats

Sections of the benthic mats can break off and the fragments dislodge from the sediment and float into the water column. These fragments may release more odourants than stable benthic mats. This dislodgement is often observed to be the result of high photosynthetic activity within the mat leading to the formation of oxygen bubbles which can be quite visible and which make the mat likely to lift from the sediment. Wind and wave action will also be a factor to facilitate this. The integrity of the mats may also be influenced by the substrate e.g. hard substrates may provide better attachment than soft sediments.

Retention time of the Water

The length of time that the water is retained in the reservoir may also influence the impact of benthic sources of T&Os but it is very difficult to assess this factor. Longer retention times may allow more build up of the T&O compounds in the water column. Conversely long retention times may also allow processes such as volatilisation and biodegradation to operate and reduce the concentration of these compounds.

History of T&O Problems

The occurrence of T&Os in a reservoir in the absence of known planktonic producers is the most direct indicator of a potential benthic source. Within Australia spring appears the most likely time for benthic sources to present water quality problems.

A review of historical data from the reservoir can provide significant information about taste and odour issues associated with benthic cyanobacteria. Benthic cyanobacteria can become dislodged from the sediment surface and fragments can float into the water column and can then be detected in routine water samples. The relative concentration of total and dissolved odour compounds can help to identify the source. High levels of dissolved or extracellular odours can suggest a source other than planktonic cyanobacteria.

Survey and Monitoring

Reservoirs at high risk of benthic cyanobacterial growth and any reservoir which has a history of unexplained T&O problems should be surveyed to determine the abundance and species composition

84 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

of benthic cyanobacteria and compared with known T&O producers (see Table 1.1). Samples should be analysed using appropriate analytical methods for T&O compounds.

Visual observations of the sediment surface can provide very useful information on the distribution of benthic cyanobacteria. More detailed surveys can be conducted using underwater cameras or divers. This requires access to relatively sophisticated expertise and resources. More basic survey monitoring involves collection of sediment samples using a benthic sampler such as an ‘Eckman’ grab or a rigid pole mounted corer (e.g. PVC or polycarbonate pipe) (See Section 2).

A quantitative method for sampling and analysis of cyanobacterial populations in sediments is given in Section 2. This uses a corer and describes a lab counting method based upon reconstitution of a defined surface area of sediment sample to provide estimates of numbers related to sediment surface area (e.g. cells/cm2). The survey of reservoirs in South Australia indicated that natural populations of benthic cyanobacteria develop with numbers of individual species in the range of 103 - >106 cells/cm2 (see Section 2).

To obtain a representative assessment of numbers and the distribution of benthic cyanobacteria, samples should be collected from a number of transects throughout or around the perimeter of a reservoir. Particular attention should be paid to shallow protected bays and any areas where benthic mats have been observed. Samples at varying depths may need to be collected down to approximately 5 metres, although this will depend upon light attenuation in the water body. Shallow depths can be sampled with a corer, while greater depths will require a closing ‘grab’ sampler lowered with a rope or cable.

An example of a grab sample is the ‘Ekman’ Bottom Grab sampler which is designed for sampling in soft-bottomed lakes and rivers with sediments composed of sand, silt or mud. As the sampler is lowered, two hinged upper lids swing open to let water pass through and close upon retrieval preventing sample washout. When the sampler reaches the bottom, a messenger is sent down the line tripping the overlapping spring-loaded scoops. These samplers are designed to collect an accurate representative sample of the sediment bottom. It is not recommended for rocky bottoms or those with moderate macrophyte growth because small pebbles or macrophyte stems prevent proper jaw closure. The bite of the sampler should be deep enough so all depths are sampled equally. The closing mechanism is required to completely close and hold the sample as well as prevent wash-out during retrieval. Likewise, during descent the sampler should be designed to minimise disturbance of the topmost sediment by the pressure wave as it is lowered to the bottom. These samplers are readily available commercially.

It may be possible to assess the distribution and density of benthic algal communities using a specialised chlorophyll ‘fluoroprobe’ technique to estimate biomass from algal or cyanobacterial pigment fluorescence. This is a relatively new area of commercial technological development and warrants investigation for quantitative surveys of the distribution of benthic algae. Although the authors of this study have no experience with the technique for sediments, the principle of in vivo fluorometers for measuring cyanobacterial and algal distribution and abundance in water is now being widely used for cyanobacterial monitoring in drinking water reservoirs. The same issues that apply to the use of fluoroprobes in surface water would apply to benthic measurements, which include that the instruments require careful calibration against the local population by accurate cell counts and/or pigment measurement using an extractive technique.

Benthic cyanobacteria may also be attached to structures within the reservoir e.g. dam walls, concrete surfaces and offtake pipes, and samples can be removed from these by scraping. Some cyanobacteria that produce T&Os may also grow attached to the leaves of emergent or submerged macrophytes as a biofilm and this can be referred to as having an epiphytic growth habit. Surveys and scraping of macrophytes leaves or stems is important to include in any investigation for sources of T&Os in reservoirs.

An alternative method for quantitative assessment of benthic cyanobacterial or algal growth within or between sites is the use of removable artificial substrates. Artificial substrates are often made of hardwood plates of known surface area which can be installed in appropriate locations and depths within the water body for a period of time to allow benthic cyanobacteria to colonise the surface of the substrate. This is usually for a period of greater than 28 days depending upon the season. Once the

85 TASTES AND ODOURS IN RESERVOIRS substrates are retrieved the benthic cyanobacteria can be scraped off and measured quantitatively to determine the population size per unit surface area. These substrates can be cleaned and reused.

Control Measures

Experience with the control of benthic cyanobacteria across the water industry is limited and the few options that have been used have varying success. The following is a brief summary of the methods that have been used.

Water Level Changes

Changing reservoir water levels to expose benthic cyanobacteria and dry them out does lead to a reduction in the contribution of odours to the overlying water. This technique has been used in small reservoirs in South Australia which are operated as balancing storages. It only applies in circumstances where there is the operational control and flexibility to manipulate levels. However the limitations are that the level must remain down for an extended period as the benthic mats are quite resistant to desiccation as has been shown in this study. The surviving cyanobacteria in the dry mat will also be rejuvenated and regrow when they are again inundated with water. Another potential issue is that lowering the level may allow cyanobacteria to colonise a new band of sediments that were previously deeper and may become a zone with a suitable light climate for benthic growth.

Nutrient Control

Reduction of nutrients from the catchment may decrease the cyanobacterial growth potential but the relationship between nutrient loads and growth of benthic cyanobacteria or of T&O producers is not as well understood as it is with nuisance planktonic cyanobacteria. Many benthic cyanobacteria are also known to grow in relatively nutrient poor water, so it is likely that other factors are equally important or more significant drivers in their distribution and abundance.

Physical Control

Physical removal by raking or towing drag lines across the sediment surface can be effective to scour the surface and disrupt and dislodge the benthic mat communities. This technique has not been widely used in Australia and is reported from North America and Europe. The technique is only applicable for reservoirs of a certain type, size and bathymetry and may require repeated application over the growing season. This practice is also likely to generate turbidity and possibly localised spikes in odour compounds in the water above the sediment which may be an issue depending upon the proximity of the supply offtake.

Chemical Control

The application of algicides such as copper sulphate has been shown to be effective but this may not be acceptable for environmental reasons in local circumstances. An extensive discussion on the use of algicides, including recommendations on application techniques, monitoring and withholding periods is provided by Newcombe et al (2009). In relation to the specific treatment of benthic cyanobacteria, copper sulphate has been applied as large crystals directly to the areas of benthic mat growth with the aim that the crystals sink to the sediment and dissolve adjacent to the target benthic organisms, to enhance the toxic effect of the copper. This application can be targeted to the area where the mats are found and is therefore dependent upon having good survey information. Dose rates can also be adjusted to treat only the affected areas and not the entire reservoir. It is important to ensure that any chemical that is applied as an algicide is a registered (i.e. approved) product for that purpose. This applies to crystalline copper sulphate, which is a different form to the granular chemical - the form registered specifically as an algicide. It should be noted that the T&O producing genus Phormidium is recognised as being more tolerant to copper sulphate than common nuisance planktonic species of Microcystis and Anabaena.

The limited amount of work on management and control that has been reported here clearly indicates that more effective methods for controlling benthic cyanobacteria are required.

86 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

7 SUMMARY AND CONCLUSIONS

This study set out to investigate and characterise the occurrence of T&Os produced by benthic cyanobacteria and actinomycetes and the potential risk they pose to water quality in a number of Australian reservoirs. This study developed a model that incorporates processes for production and loss of odour metabolites in a reservoir. Investigations and experiments were undertaken to calibrate the magnitude and significance of the components of the model.

The options to control benthic cyanobacteria are limited and this project evaluated the use of desiccation or drying out by changing water level as a possible control technique. While much has been provided in the literature regarding management of planktonic cyanobacteria little is available for control of benthic cyanobacteria due to limited knowledge about their ecology.

Survey of Australian Reservoirs

A survey was carried out of major drinking water supply reservoirs in four states: Hope Valley and Happy Valley Reservoirs (South Australia), North Pine Dam (Queensland), Grahamstown Dam (New South Wales) and Yan Yean Reservoir (Victoria). The aim was to characterise the benthic community (cyanobacteria and actinomycetes) and determine their relative contribution to odour production in these reservoirs. The initial focus of investigation was on benthic cyanobacteria, however, actinomycetes have recently been identified in the literature as possible sources of T&Os in water supplies and were also included.

In addition, historical data collected from both Hope Valley and Happy Valley Reservoirs by SA Water since 1997 were analysed to determine relationships between cyanobacteria and odour episodes. Both planktonic Anabaena and the benthic cyanobacterium Phormidium were shown to contribute significantly to the total geosmin content in Hope Valley and possibly Happy Valley Reservoir. Total geosmin present in water entering Hope Valley Reservoir treatment plant on some occasions came entirely from Phormidium. These results showed the important role that historical data can provide in identifying sources of T&Os in reservoirs.

A detailed spatial survey was undertaken at these two reservoirs in spring and summer to gain a better understanding about the ecology of both benthic cyanobacteria and actinomycetes. The survey consisted of ten transects equally spaced around the perimeter where samples were collected at various depths down to 2 m. The aim was to assess the variation both within and between locations. The density of benthic cyanobacteria, Oscillatoria sp., Phormidium sp. and Pseudanabaena sp., varied significantly from site to site at both Hope Valley and Happy Valley Reservoirs. These organisms were found at all depths at some sites and were completely absent from other sites. This suggested variations in micro-environments and habitats within the reservoirs at relatively closely spaced locations. Growth was also found to vary seasonally with growth at some sites in spring but not in summer and vice versa. There was a significant increase (p<0.05) in the number of Oscillatoria sp. with increasing depth in both spring and summer for Happy Valley Reservoir and summer for Hope Valley Reservoir. This suggested that Oscillatoria sp. preferred lower irradiance and possibly lower temperatures which are both found in deeper waters. Such observations of differences in growth preferences are important for developing a monitoring program for particular reservoirs.

Actinomycetes producing both geosmin and MIB were isolated from Hope Valley and Happy Valley Reservoirs. These were all members of the genus Streptomyces. All but one isolate from a total of 92 produced geosmin and approximately 40% of isolates from both reservoirs produced both geosmin and MIB in varying ratios. Correspondingly the remaining 60% of isolates produced geosmin only. The presence of T&O producing actinomycetes in these reservoirs confirms that they could be contributing to the reservoir odour load but further work is required to determine their significance. An issue with the technique of culture isolation used in this study is that it is not possible to determine whether the isolates that grew from reservoir samples were from actively growing cells or from dormant spores and it is therefore not possible to determine whether they were producing geosmin. The relationship between actively growing actinomycetes and geosmin production requires further study.

The benthic cyanobacterium, Phormidium sp., was identified in the sediments of Grahamstown Reservoir at a single depth at one site only. Large areas of the reservoir supported a dense coverage of torpedo grass (Panicum repens) which grew down to a depth of approximately 1 m and which could

87 TASTES AND ODOURS IN RESERVOIRS retard the growth of benthic cyanobacteria by shading and light limitation. However, an epiphytic community consisting of Pseudanabaena sp. and Anabaena sp. was found attached to the macrophytes indicating that they provide a substrate to support cyanobacterial growth. Actinomycetes were detected in the sediments at all 10 sites surveyed, but not at all depths, and included both Streptomyces sp. and Actinomycetale sp. Both species were found to produce only geosmin.

The water levels at both North Pine Dam and Yan Yean Reservoir were historically low at the time of the study, which was a result of drought conditions. As a consequence sediments around the margins consisted of fine silt representing material which would normally be found in deeper sections of the reservoirs. This sediment was devoid of benthic cyanobacteria or macrophytes apart from a single finding of Phormidium sp. at one site in North Pine Dam. The numbers of Phormidium sp. at this site were high however, and under appropriate conditions its distribution could be expected to be more widespread. Actinomycetes were not found at either North Pine Dam or Yan Yean Reservoir.

Production and Loss of Tastes and Odours in Reservoirs

This study set out to measure the production of odour metabolites by natural populations of both planktonic and benthic cyanobacteria. It also aimed to determine the rates and relative significance of loss processes including biodegradation, volatilisation and photodegradation in reservoirs.

Potential production of geosmin by planktonic cyanobacteria was estimated by the determination of the cell content or ‘cell quota’ of the compound in a range of natural populations of Anabaena circinalis. Samples of natural blooms of A. circinalis in Myponga, Happy Valley and Mount Bold Reservoirs in South Australia and Yan Yean Reservoir in Victoria had a geosmin content ranging from approximately 0.02 to 0.05 pg/cell. These cell quotas can be combined with cyanobacterial cell counts to estimate the potential range of odour yield for natural populations and provide an estimate of the likely water treatment challenge.

The process by which T&Os are released from benthic cyanobacterial mats is poorly understood. A novel set of experiments, using specially designed chambers, measured the release rate or “flux” of geosmin from benthic mats in the reservoir. These flux experiments were done in Hope Valley Reservoir in November and December 2006. This is historically a period of high growth of Phormidium sp. in this reservoir. Production rates ranged from 0.0067 to 0.0938 ng/cm2/h in December and November respectively. The difference was directly related to the cell density of Phormidium sp. in the mats at the time of the experiments. (November: 32.38 x 105; December: 3.38 x 105 cells/cm2)

The loss rate of geosmin from the experimental columns was also correspondingly higher in November (3.74 ng/L/h) than in December (1.06 ng/L/h). This could be due to a range of variables including higher wind speeds across the air-water interface at the surface of the column which can increase the rate of volatilisation, and also higher water temperature which will increase both volatilisation and biodegradation.

The study demonstrated that the rate of odour metabolite release can be measured in situ and is dependent upon the density of cyanobacterial cells within the mat. The density of benthic cyanobacteria per unit area can also be estimated at individual locations by cell counts, and this can be used to derive a reservoir odour production rate or yield. The largest source of error in deriving the total reservoir odour yield is likely to be the estimation of total benthic coverage given that surveys indicated that the distribution and density of benthic mat is highly variable throughout a reservoir.

This study found that loss processes such as biodegradation and volatilisation have a major effect on the concentration of T&Os in a reservoir. Biodegradation was found to be temperature dependent with optimum rates occurring between 20 and 25°C in Myponga Reservoir, South Australia. Biodegradation followed a first-order reaction rate and based upon this the half-life of geosmin ranged from 1 to 3.5 days at 20 and 10°C respectively in Myponga Reservoir at the time of these experiments. This means that it would take approximately 24 hours for geosmin to reach 50% of its original concentration through biodegradation alone, at 20-25°C. Based upon these findings it is proposed that biodegradation could partially explain the observation of very rapid loss of extracellular T&O compounds following the sudden collapse of A. circinalis blooms in Myponga Reservoir in the summer of 2006/2007.

88 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Volatilisation rates of odourants from the reservoir surface were calculated by determining the overall mass transfer coefficient, which was estimated from the two-film theory, and the gaseous and aqueous phase mass transfer coefficients were calculated from the empirical correlations previously published for lakes (See Section 4). Volatilisation can be shown to be significantly affected by wind speed i.e. high winds increased loss of both geosmin and MIB. A concentration gradient exists between the water/air interface and high winds can rapidly remove volatile compounds from the surface of the water allowing the movement of fresh volatiles to the interface. High winds can also create waves on the surface which act to increase the area of water/air interface resulting in increased rates of volatilisation.

Photodegradation has been identified as a possible loss mechanism for T&Os in the environment. While photodegradation has been observed by other researchers this project showed that both geosmin and MIB were not significantly affected by exposure to sunlight over 14-days in an outdoor tank experiment at ambient temperature. In addition measurements of the transmission of UV radiation through reservoir water demonstrated relatively high absorbance, which is most likely due to high levels of dissolved natural organic matter. It follows that significant exposure to UV would occur only near the water surface in the South Australian reservoirs investigated in this study.

Taste and Odour Model

An important outcome of this study was the development and validation of a model characterising the mass balance of T&O compounds in a reservoir. The model incorporated major variables of odourants in the inflow and outflow for the reservoir, production of T&Os by planktonic and benthic cyanobacteria, their degradation by photo-oxidation and biodegradation, and loss by volatilisation. The model critically assessed the relative importance of geosmin and 2-methylisoborneol (MIB) in both a Taiwanese and an Australian reservoir. Rate constants were developed for the production of geosmin and MIB by planktonic and benthic cyanobacteria and loss by photo-oxidation, biodegradation, and adsorption onto suspended solids for use in the model. Volatilisation and biodegradation were found to be the most important mechanisms governing loss of geosmin and MIB from the two reservoirs. Wind speed was shown to be an important variable influencing volatilisation i.e. high winds lead to increased loss of geosmin and MIB by volatilisation. Adsorption of geosmin and MIB onto suspended solids was estimated to be relatively low, however, further work is needed in this area to assess adsorption onto reservoir sediments.

Further development of this model could allow for the incorporation of the process rate constants into a deterministic reservoir process-based model and to then do comprehensive forecasting and to use the information for management. These models, which are often built around 3D-hydrodynamic models linked to water quality process models, and allow for investigations into the spatial and temporal relationships between physical, biological and chemical variables in water bodies, over single events or seasonal to longer timescales. The application of such a model to a reservoir could enable operators to predict taste and odour episodes caused by planktonic and benthic cyanobacteria.

Control of Benthic Cyanobacteria by Desiccation

Desiccation or ‘drying out’ was investigated as a possible method for the control of benthic cyanobacteria using mats collected from Hope Valley Reservoir, South Australia. This involved an air- drying trial to assess cell viability over a period of approximately 7 weeks. The results showed that some cells within a benthic mat could remain viable following long periods of desiccation. Around 20% of the population of Phormidium was still viable after 46 days exposure and it is expected that around 5% would still be viable after 90 days i.e. 3 months. It is very possible that this proportion of healthy cells could provide a seed population for re-growth of benthic mats when they are re-wet.

89 TASTES AND ODOURS IN RESERVOIRS

Management of Benthic Cyanobacteria

While there is considerable experience with the management of planktonic cyanobacteria there is very little knowledge on the management of benthic cyanobacterial problems. A brief discussion on options for management of benthic cyanobacteria is provided in this report, however given that the knowledge about their ecology is relatively limited it is difficult to develop a comprehensive management plan at this stage.

The occurrence of T&Os in a reservoir in the absence of known planktonic producers is the most direct indicator of a potential benthic source. There are some general reservoir features that are likely indicators of the risk of benthic cyanobacteria growing in a reservoir.

The distribution of benthic cyanobacteria is restricted by the depth of light penetration. Shallow reservoirs, especially those with high water transparency, will have greater area available for benthic cyanobacterial growth than deep reservoirs. Shallow reservoirs with gently sloping margins as opposed to deep steep-sides will have higher sediment surface to water volume ratios, resulting in less potential for dilution of the odours. The proximity of the benthic mats to the water off-take is also an important consideration. As a general guide, benthic cyanobacteria need about 1% of the surface irradiance to grow, however this may be lower depending upon the species or type. The area of the reservoir potentially available for benthic cyanobacterial growth can be calculated from the extinction co-efficient of the water and the bathymetry of the reservoir.

A review of historical data from the reservoir can provide significant information about taste and odour issues associated with benthic cyanobacteria. Benthic cyanobacteria can become dislodged from the sediment surface and fragments can float into the water column and can then be detected in routine water samples. The relative concentration of total and dissolved odour compounds can help to identify the source. High levels of dissolved or extracellular odour compounds can suggest a source other than planktonic cyanobacteria.

Reservoirs at high risk of benthic cyanobacterial growth and any reservoir which has a history of unexplained T&O problems should be surveyed to determine the abundance and species composition of benthic cyanobacteria and these should be compared with known T&O producers. Samples should be analysed using appropriate analytical methods for T&O compounds. Visual observations of the sediment surface can provide very useful information on the distribution of benthic cyanobacteria. More detailed surveys can be conducted using underwater cameras or divers. This requires access to relatively sophisticated expertise and resources. More basic survey monitoring involves collection of sediment samples using a benthic sampler such as an ‘Eckman’ grab or a rigid pole-mounted corer (See Section 2).

The aim of a survey is to obtain a representative assessment of numbers and distribution of benthic cyanobacteria. Samples should be collected from a number of transects around the perimeter of a reservoir. Particular attention should be paid to shallow protected bays and any areas where benthic mats have been observed. Samples at varying depths may need to be collected down to approximately 5 metres, although this will depend upon light attenuation in the water body.

Methods for control of benthic cyanobacteria are limited and have varying success. This study evaluated the control of benthic cyanobacteria by desiccation or drying out through lowering of reservoir levels and showed that the organisms are quite hardy and resistant to desiccation. The limitation to doing this is having operational flexibility so that the level can remain down for an extended period.

Reduction of catchment derived nutrients may decrease cyanobacterial growth potential but it is likely to be less effective on benthic forms.

Physical removal by raking or towing drag lines across the sediment surface can be effective to scour the surface and disrupt and dislodge the benthic mat communities. This technique has not been widely used in Australia and is reported from North America and Europe. This technique is only applicable for reservoirs of a certain type, size and bathymetry, and may require repeated application over the growing season. This practice is also likely to generate turbidity and possibly localised spikes in odour

90 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

compounds in the water above the sediment which may be an issue depending upon the proximity of the supply offtake.

The application of algicides such as copper sulphate has been shown to be effective but this may not be acceptable for environmental reasons in local circumstances. In relation to the specific treatment of benthic cyanobacteria, copper sulphate has been applied as large crystals directly to the areas of benthic mat with the aim that the crystals sink to the sediment and dissolve adjacent to the target organisms to enhance the toxic effect of the copper. The application should be targeted to the area where the mats are found and this is therefore dependent upon having good survey information. Dose rates can be adjusted to treat only the affected areas and not the entire reservoir. It is important to ensure that any chemical that is applied as an algicide is a registered (i.e. approved) product for that purpose. It should be noted that the T&O producing genus Phormidium is recognised as being more tolerant to copper sulphate than common nuisance planktonic species of Microcystis and Anabaena.

A number of alternative methods for the control of cyanobacteria, both chemical and non-chemical, are available in the market place but most have not had rigorous and scientifically valid testing with evaluations published in the peer-reviewed scientific literature. It is crucial that such testing be undertaken so an appropriate alternative to copper sulphate can be identified that will be effective not only for planktonic but also for benthic cyanobacteria.

Actinomycetes have emerged in a number of places around the world as a possible contributor to T&Os in water supplies. This study has identified them as potential sources of both geosmin and MIB in a range of Australian reservoirs but the significance of their contribution to the total T&O load is unclear at this time. There is insufficient quantitative information to incorporate actinomycetes into the model developed as part of this project, however the study confirmed their presence in several drinking water reservoirs in Australia and their production of geosmin and/or MIB. This is an important first step in determining their significance as a T&O contributor and further research is required in this area.

91 TASTES AND ODOURS IN RESERVOIRS

8 REFERENCES

Alferova IV, Terekhova LP, and Prauzer KH (1989) Antibiotiki i khimioterapiia = Antibiotics and chemoterapy [sic] / Ministerstvo meditsinskoi i mikrobiologicheskoi promyshlennosti SSSR 34(5):344-348.

Amann RI, Ludwig W and Schleifer KH (1995) Phylogenetic identification and in situ detection of individual microbial cells without cultivation. Microbiology Review. 59: 143-169.

Baker PD, Steffensen DA, Humpage AR, Nicholson BC, Falconer IR, Lanthois B, Fergusson KM, Saint CP (2001). Preliminary evidence of toxicity associated with the benthic cyanobacterium Phormidium in South Australia. Environmental Toxicology 16(6): 506-511.

Bender DA, Asher WE, Zogorski JS (2003) LakeVOC—A Computer Model to Estimate the Concentration of Volatile Organic Compounds in Lakes and Reservoirs. U.S. Geological Survey, Fact Sheet 104-03.

Boon PI (1989) Relationships between actinomycete populations and organic matter degradation in , Southeastern Australia. Regulated Rivers: Research Management 4: 409-418.

Boucher G, Clavier J, Garrigue C (1994). Oxygen and carbon dioxide fluxes at the sediment water interfaceof a tropical lagoon. Marine Ecology, Progressive Series 107: 185-193.

Burch M, Baker P (2000). Common Algal Species in South Australian Waters Potential Toxicity and Odours. Australian Water Quality Centre Report.

Clavero V, Izquierdo JJ, Fernández JA, Niell FX (2000) Seasonal fluxes of phosphate and ammonium across the sediment–water interface in a shallow small estuary (Palmones River, southern Spain). Marine Ecology, Progressive Series 198: 51-60

Davis JS (1972). Survival records in the algae, and the survival role of certain algal pigments, fat and mucilaginous substances. Biologist 54: 52-93.

Egashira K, Ito K., Yoshiy Y (1992). Removal of musty odour compound in drinking water by biological filter. Water Science & Technology 25: 307-314.

Foyo-Moreno I, Vida J, Alados-Arboledas L (1998) Ground-based ultraviolet (290–385 nm) and broadband solar radiation measurements in southeastern Spain. International Journal of Climatology, 18: 1389-1400.

Gao F and SR Yates (1998). Laboratory study of closed and dynamic flux chambers: experimental results and implications for field application. Journal of Geophysical Research 103(D20): 26115-26125.

Gerber N (1969). A volatile metabolite of actinomycetes, 2-methylisoborneol. Journal of Antibiotics 22: 508.

Gerber N, Lechevalier HA (1965). Geosmin, an earthy smelling substance isolated from actinomycetes. Applied Microbiology 13(6): 935-938.

Goodfellow M and Williams ST (1983). Ecology of Actinomycetes. Annual Review Microbiology 37: 189–216.

Gorham PR, McLachlan RW, and Hammer UT (1964). Isolation and culture of toxic strains of Anabaena flos-aquae (Lyngb.) de Bréb. Mitteilungen der Internationalen Vereinigung für Limnologie 19: 796–804.

Gugger M, Lenoira S, Berger C, Ledreux A, Druart JC, Humbert JF, Guette C and Bernard C (2005) First report in a river in France of the benthic cyanobacterium Phormidium favosum producing anatoxin-a associated with dog neurotoxicosis Toxicon 45(7): 919-928.

92 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Hamill KD (2001). Toxicity in benthic freshwater cyanobacteria (blue-green algae): first observations in New Zealand. New Zealand Journal of Marine & Freshwater Research 35(5): 1057-1059.

Hatano K, Frederick DJ, Moore JA (1994) Microbial ecology of constructed used for treating pulp mill wastewater. Water Science & Technology 29(4): 233–239.

Hawes I, Howard-Williams C and Vincent WF (1992) Desiccation and recovery of Antarctic cyanobacterial mats. Polar Biology, 12: 587-594.

Hayes, SJ, Hayes KP and Robinson BS (1991) Geosmin as an odorous metabolite in cultures of a free-living amoeba, Vannella species (Gymnamoebia,Vannellidae). Journal of Protozoology 38: 44-47.

Ho L, Hoefel D, Bock F, Saint CP, Newcombe G (2007) Biodegradation rates of 2-methylisoborneol (MIB) and geosmin through sand filters and in bioreactors. Chemosphere 66(11): 2210-2218

Ho L, Meyn T, Keegan A, Hoefel D, Brookes J, Saint CP, Newcombe G (2006) Bacterial degradation of microcystin toxins within a biologically active sand filter. Water Research 40(4): 768-774.

Hoefel D, Ho L, Aunkofer W, Monis PT, Keegan A, Newcombe G Saint CP (2006) Cooperative biodegradation of geosmin by a consortium comprising three gram-negative bacteria isolated from the biofilm of a sand filter column. Letters in Applied Microbiology 43(4): 417-423 2006.

Hötzel G and Croome R (1999) A Phytoplankton Methods Manual for Australian Freshwaters. Occasional Paper 22/99, Land & Water Resources Research Development Corporation, Canberra.

Huovinen PS, Penttilae H, Soimasuo MR (2003) Spectral attenuation of solar ultraviolet radiation in humic lakes in Central Finland. Chemosphere 51(33), 205-214

Ishida H and Miyaji Y (1992) Biodegradation of 2-methylisoborneol by oligotrophic bacterium isolated from eutrophied lake. Water Science & Technology 25: 269–276.

Izaguirre G and Devall J (1995) Resource Control for Management of Taste-and Odour Problems. In Advances in Taste-and Odour Treatment and Control, Seffet IH, Mallevialle and Kawczynski (eds), American Water Works Association Research Foundation, Denver, USA: 23-74.

Izaguirre G, Wolfe RL, Means EG (1988). Degradation of 2- methylisoborneol by aquatic bacteria. Applied and Environmental Microbiology 54: 2424–2431.

Izaguirre G, Jungblut AD, Neilan BA (2007) Benthic cyanobacteria (Oscillatoriaceae) that produce microcystin-LR, isolated from four reservoirs in southern California. Water Research 41(22): 492-498.

Jensen SE, Anders CL, Goatcher LJ, Perley T, Kenefick S, Hrudey SE (1994) Actinomycetes as a factor in odour problems affecting drinking water from the North Saskatchewan River. Water Research 28(6): 1393–1401.

Jütttner F and Watson SB (2007) Biochemical and Ecological Control of Geosmin and 2- Methylisoborneol in Source Waters. Applied and Environmental Microbiology 4395–4406.

Kahlert M, Pettersson K (2002) The impact of substrate and lake trophy on the biomass and nutrient status of benthic algae. M Kahlert, K Pettersson. Hydrobiologia 489:11, 161-169.

Kile DE, Chiou CT, Zhou H, Li H, and Xu O, (1995) Partition of nonpolar organic pollutants from water to soil and sediment organic matters, Environmental Science and Technology, 29: 1401- 1406.

93 TASTES AND ODOURS IN RESERVOIRS

King BJ, Hoefel D, Daminato DP, Fanok S, Monis PT (2008) Solar UV reduces Cryptosporidium parvum oocyst infectivity in environmental waters Journal of Applied Microbiology 104(5) 1311-1323.

Klausen C, Jørgensen NOG, Burford M, O’Donohue M (2004) Actinomycetes may also produce taste and odour. Water, August: 58-62.

Lanciotti E, Santini C, Lupi E, Burrini D (2003) Actinomycetes, cyanobacteria and algae causing tastes and odours in water of the River Arno used for the water supply of Florence. Journal of Water Supply: Research and Technology – AQUA 52(7): 489-500.

Lembi CA (2000). Relative tolerance of mat-forming algae to copper. Journal of Aquatic Plant Management 38: 68-70.

Mallevialle J, Suffet IH (1987) Identification and treatment of tastes and odours in drinking water. American Water Works Association Research, Denver, USA.

McDowall B, Hoefel D, Newcombe G, Saint CP, Ho L (2009) Enhancing the biofiltration of geosmin by seeding sand filter columns with a consortium of geosmin-degrading bacteria. Water Research 43(2):433-40.

Medsker LL, Jenkins D, Thomas JF, Koch C (1969) Odorous compounds in natural waters. 2-exo- hydroxy-2-methylbornate, the major odorous compound produced by actinomycetes. Environmental Science and Technology 3: 476.

Mez K, Beattie KA, Codd GA, Hanselmann K, Hauser B, Naegeli H, Preisig HR (1997) Identification of a microcystin in benthic cyanobacteria linked to cattle deaths on alpine pastures in switzerland. European Journal of Phycology 32(2): 111-117.

Mez K, Hanselmann K, Preisig HR (1998) Environmental conditions in high mountain lakes containing toxic benthic cyanobacteria. Hydrobiologia 368: 1-15.

Mofidi AA, Coffey BM, Chou CI, Liang S, Green JF (2000) Using UV light to achieve multiple water quality objectives. AWWA Water Quality Tech. Conf., Salt Lake City, Utah (Nov. 5-9).

Mohamed ZA, El-Sharouny HM, Ali WSM (2006) Microcystin production in benthic mats of cyanobacteria in the Nile River and irrigation canals, Egypt. Toxicon 47(55) 584-590.

Morison MO and Sheath RG (1985) Responses to desiccation stress by Klebsormidium rivulare (Ulotrichales, Chlorophyta) from a Rhode Island stream. Phycologia 24: 129-145.

Morris DP, Zagarese H, Williamson CE, Balseiro EG, Hargreaves BR, Modenutti B, Moeller R, Queimalinos C (1995) The attenuation of solar UV radiation in lakes and the role of dissolved organic carbon. Limnology and Oceanography 40:1381-1391.

Narayan LV and Nunez WJ (1974). Biological control: isolation and bacterial oxidation of the taste and odour compound geosmin. Journal of the American Water Works Association 66: 532- 536.

Newcombe G, House J, Ho L, Baker P, Burch M (2009) Management strategies for cyanobacteria (blue-green algae) and their toxins: a guide for water utilities. Research Report No 74. Water Quality Research Australia, Adelaide. Australia.

Oikawa E, Shimizu A, Ishibashi Y (1995) 2-Methylisoborneol degradation by the CAM operon from Pseudomonas putida PpG1. Water Science & Technology 31: 79 86.

Persson NJ, Pettersen H, Ishaq R, Axelman J, Bandh C, Broman D, Zebuhr Y, Hammar T. (2005) Polychlorinated biphenyls in polysulfide sealants–occurrence and emission from a landfill station. Environmental Pollution, 138:18-27.

94 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 73

Peterson CG (1996) Response of benthic algal communities to natural physical disturbance. In ‘Algal Ecology: Freshwater Benthic Ecosystems’. (Eds R. J. Stevenson, M. L. Bothwell and R. L. Lowe.) pp. 375-402. (Academic Press: San Diego.)

Point D, Monperrus M, Tessier E, Amouroux D, Chauvaud L, Thouzeau G, Jean F, Amice E, Grall J, Leynaert A, Clavier J, Donard OFX (2007) Biological control of trace metal and organometal benthic fluxes in a eutrophic lagoon (Thau Lagoon, Mediterranean Sea, France) Estuarine, Coastal and Shelf Science 72(3): 457-471

Robson B (2000) Role of residual biofilm in the recolonization of rocky intermittent streams by benthic algae, Marine and Freshwater Research, 51: 725-732.

Rosenfeldt EJ, Linden KG, Melcher B (2005) UV and UV/H2O2 treatment of methylisoborneol (MIB) and geosmin in water. Journal of Water Supply Research and Technology - Aqua 54: 423-434

Saadoun I, El-Migdadi F, (1998) Degradation of geosmin-like compounds by selected species of gram-positive bacteria. Letters in Applied. Microbiology 26: 98-100.

Sabater S, Vilalta E, Gaudes A, Guasch H, Munoz I, Romani A (2003) Ecological implications of mass growth of benthic cyanobacteria in rivers. Aquatic Microbial Ecology 32(2): 175-184.

Scully NM, and Lean DRS (1994) The attenuation of UV radiation in temperate lakes. Archiv für Hydrobiology 43: 135-144.

Seifert M, McGregor G, Eaglesham G, Wickramasinghe W, Shaw G (2007) First evidence for the production of cylindrospermopsin and deoxycylindrospermopsin by the freshwater benthic cyanobacterium, Lyngbya wollei (Farlow ex Gomont) Speziale and Dyck Harmful Algae. 6(1): 73-80.

Sekar R, Pernthaler A, Pernthaler J, Warnecke F, Posch T, Amann R (2003) An improved protocol for quantification of freshwater Actinobacteria by fluorescence In-Situ Hybridization. Applied and Environmental Microbiology 69(5): 2928-2935.

Silvey JKG, Henley AW, Nunez WJ, Cohen RC (1970) Biological control: control of naturally occurring taste and odors by micro-organisms. In: Proceedings of the National Biological Congress, Detroit, USA.

Sugiura N and Nakano K (2000) Causative micro-organisms for musty odor occurrence in the eutrophic Lake Kasumigaura. Hydrobiologia 434(11): 145-150.

Sumitomo H (1988) Odor decomposition by the yeast Candida. Water Science & Technology 20: 157-162.

Tanaka A, Oritani T, Uehara F, Saito A, Kishita H, Niizeki Y, Yokota H, Fuchigami K (1996) Biodegradation of a musty odour component, 2 methylisoborneol. Water Research. 30: 759- 761.

Utkilen H and Froshaug M (1992) Geosmin production and excretion in a planktonic and benthic Oscillatoria. Water Science and Technology 25(2): 199-206.

Vadeboncoeur Y, Peterson G, Vander Zanden MJ, and Kalff J (2008) Benthic algal production across lake size gradients: interactions amoung morphometry, nutrients, and light. Ecology 89(9):2542-2552.

Van Breemen LWCA, Ketelaars HAM, Visser P and Ebbeng JH (1991) A new method to control growth of geosmin-producing benthic cyanobacteria. Verhandlungen der Internationalen. Vereinigung für Limnologie. 24: 2168-2173

95 TASTES AND ODOURS IN RESERVOIRS

Van Breemen L, Ketelaars H (1995) Influence of artificial mixing and other factors on algal biomass in the Biesbosch reservoirs. Journal of Water Supply Research and Technology-Aqua 44(1):65-71.

Vilaita E, Guasch H, Munoz I, Romani A, Valero F, Rodriguez JJ, Alcaraz R, Sabater S (2004) Nuisance odours produced by benthic cyanobacteria in a Mediterranean river. Water Science and Technology. 49(9): 25-22

Wanninkhof RH, Bliven LF (1991) Relationship Between Gas-Exchange, Wind-Speed, and Radar Backscatter in a Large Wind-Wave Tank. Journal of Geophysical Research-Oceans, 96 C2: 2785-2796

Westerhoff P, Rodriguez-Hernandez M, Baker L, Sommerfeld M (2005) Seasonal occurrence and degradation of 2-methylisoborneol in water supply reservoirs. Water Research 39:2020, 4899- 4912, 2005.

Wohl DL and McArthur JV (1998) Actinomycete-flora associated with submersed freshwater macrophytes. FEMS Microbial Ecology 26(2): 135-140.

Wood S, Williams ST, White WR (2001) Microbes as a source of earthy flavours in potable water - a review. International Biodeterioration and Biodegradation 48: 26-40.

Wood SA, AI Selwood, A Rueckert, PT Holland, JR (2007) First report of homoanatoxin-a and associated dog neurotoxicosis in New Zealand. SA Wood, AI Selwood, A Rueckert, PT Holland, JR Milne, KF Smith, B Smits, LF Watts, CS Cary Toxicon 50(22): 292-301.

Woodbury BL, Miller DN, Eigenberg RA, Nienaber JA (2006) An inexpensive laboratory and field chamber for manure volatile gas analysis. Transactions ASAE 49:767-772.

Wu J and Juttner F (1998) Effect of environmental factors on geosmin production by Fischerella muscicola. Water Science & Technology 20(8/9): 143-148.

Yagi M, Nakashima S, Muramoto S (1988). Biological degradation of musty odour compounds, 2- methylisoborneol and geosmin, in a bio-activated carbon filter. Water Science & Technology 20: 255-260.

Vadeboncoeur Y, Peterson G, Jake Vander Zanden M, Kalff J (2008) Benthic algal production across lake size gradients: interactions among morphometry, nutrients and light. Ecology: 89(9) 2542-2552.

Zaitlin B and Watson SB (2006) Actinomycetes in relation to taste and odour in drinking water: myths, tenets and truths. Water Research 40:1741-1753.

96 Water Quality Research Australia Membership at September 2009

Industry Members Australian Water Association Ltd Barwon Region Water Corporation Ben Lomond Water Central Gippsland Regional Water Corporation Central Highlands Water Ltd Coliban Region Water Corporation Degrémont Pty Ltd Dept of Human Services (Vic) Grampians Wimmera Mallee Water Corporation Goulburn Valley Regional Water Corporation Hunter Water Corporation Water Quality Research Australia Limited Corporation GPO BOX 1751, Adelaide SA 5001 Power & Water Corporation South Australian Water Corporation For more information about WQRA visit the website South East Water Limited www.wqra.com.au Sydney Catchment Authority Sydney Water Corporation United Water International Pty Ltd Wannon Region Water Corporation Water Corporation of WA Ltd Research Members Australian Water Quality Centre Centre for Appropriate Technology Curtin University of Technology Flinders University Griffith University Monash University Murdoch University RMIT University University of Adelaide University of NSW University of Queensland University of South Australia The Cooperative Research Centre (CRC) for Water Quality and University of Technology, Sydney Treatment operated for 13 years as Australia’s national drinking University of Wollongong water research centre. It was established and supported under the Victoria University Australian Government’s Cooperative Research Centres Program. General Members The CRC for Water Quality and Treatment officially ended Department of Water (WA) in October 2008, and has been succeeded by Water Quality Research Australia Limited (WQRA), a company funded by the Lower Murray Urban and Rural Water Corporation Australian water industry. WQRA will undertake collaborative NSW Department of Health research of national application on drinking water quality, recycled NSW Water Solutions, Commerce water and relevant areas of wastewater management. Orica Australia Pty Ltd Syme and Nancarrow Water The research in this document was conducted during the term of Water Futures the CRC for Water Quality and Treatment and the final report completed under the auspices of WQRA.