Research Report 67 a Practical Guide to Reservoir Management Final

Water Quality Research Australia Membership at December 2008

Industry Members • Australian Water Association Ltd • Degrémont Pty Ltd • Barwon Region Water Corporation “Barwon Water” • Central Highlands Water • City West Water Ltd • Coliban Region Water Corporation • Department of Human Services (Vic) • Goulburn Valley Regional Water Corporation A Practical Guide “Goulburn Valley Water” • Grampians Wimmera Mallee Water Corporation • Hunter Water Corporation to Reservoir Water Quality Research Australia Limited • Melbourne Water Corporation GPO BOX 1751, Adelaide SA 5001 • Power & Water Corporation • South East Water Limited For more information about WQRA visit the website Management • Sydney Catchment Authority www.wqra.com.au • Sydney Water Corporation • United Water International Pty Ltd • Wannon Region Water Corporation • Water Corporation of WA • Yarra Valley Water Ltd Research Report 67 • South Australian Water Corporation • Central Gippsland Regional Water Corporation Research Members • Australian Water Quality Centre • Centre for Appropriate Technology • Curtin University of Technology • Flinders University • Griffith University • Monash University • RMIT University • University of Adelaide • University of NSW • The University of Queensland • University of South Australia • University of Technology, Sydney • University of Wollongong, Faculty of Engineering, The Cooperative Research Centre (CRC) for Water Quality and • Victoria University Treatment operated for 13 years as Australia’s national drinking water research centre. It was established and supported under the General Members Australian Government’s Cooperative Research Centres Program. • Cradle Coast Water • Department of Water (WA) The CRC for Water Quality and Treatment officially ended in October 2008, and has been succeeded by Water Quality • Esk Water Authority Research Australia Limited (WQRA), a company funded by the • Lower Murray Urban and Rural Water Corporation Australian water industry. WQRA will undertake collaborative “LMW” research of national application on drinking water quality, recycled Research Report • NSW Water Solutions, Commerce water and relevant areas of wastewater management. • NSW Department of Health • Orica Australia Pty Ltd The research in this document was conducted during the term of the CRC for Water Quality and Treatment and the final report completed under the auspices of WQRA. 67 A Practical Guide to Reservoir Management

Contributing Authors, Justin Brookes1, Mike Burch2, Matthew Hipsey3, Leon Linden1, Jason Antenucci3, Dennis Steffensen2, Peter Hobson2, Olivia Thorne4, David Lewis1, Stephanie Rinck-Pfeiffer5, Uwe Kaeding5, Paul Ramussen5

1 The University of Adelaide

2 The Australian Water Quality Centre

3 The Centre for Water Research, UWA

4 The University of Cambridge

5 United Water International

Research Report No 67

A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

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.

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

Postal Address: GPO BOX 1751, Adelaide SA 5001

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

A Practical Guide to Reservoir Management

Research Report 67 ISBN 18766 16938

2 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

FOREWORD

Research Report Title: A Practical Guide to Reservoir Management

Contributing authors: Justin Brookes1, Mike Burch2, Matthew Hipsey3, Leon Linden1, Jason Antenucci3, Dennis Steffensen2, Peter Hobson2, Olivia Thorne4, David Lewis1, Stephanie Rinck-Pfeiffer5, Uwe Kaeding5, Paul Ramussen5

Project Leader: Justin Brookes

Research Nodes: 1The University of Adelaide 2The Australian Water Quality Centre 3The Centre for Water Research, UWA 4The University of Cambridge 5United Water International

CRC for Water Quality and Treatment Project No. 2.2.2.3 – Integrated models and guidance manuals for reservoir management

3 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

EXECUTIVE SUMMARY

This report summarises a range of studies and consolidated practical knowledge related to reservoir management in the Australian context, which also have wider international application. The guide provides an overview of processes that impact upon drinking water quality in reservoirs and also a range of selected procedures and tools for reservoir management. This is supported by a series of case study investigations that were largely carried out within the Source Water Program of the CRC for Water Quality & Treatment.

The first section of this guide deals with reservoir monitoring and modeling with discussion about the application of hydrodynamic models in reservoir management. It is followed by sections on catchment- derived contaminants, including description and case studies of the use of a simple model to assess contaminant transport, such as pathogens in reservoir inflows. The guide outlines the importance of natural organic matter dynamics and transformations in reservoirs. Subsequent sections address reservoir-derived contaminants such as cyanobacteria and iron and manganese and provide a case study of the use and importance of variable depth reservoir offtakes to optimise withdrawal of the best quality water. The guide includes discussion of managing the impacts of wildfire on water quality and has an overview of the emerging area of climate change and reservoir management.

Reservoir management should be considered in the context of the Framework for Management of Drinking Water Quality contained in Chapter 2 of the Australian Drinking Water Guidelines (ADWG), (http://www.nhmrc.gov.au/publications/synopses/_files/adwg_11_06_chapter_2.pdf). This Framework offers effective means of assuring drinking water quality and the protection of public health through adoption of a preventive management approach that encompasses all steps in water production from catchment to consumer. When using the Framework there are a number of elements that need to be considered in relation to reservoir management including: assessment of the role and importance of the reservoir in the context of the overall supply system; identification of preventive measures available for water quality management in the reservoir; and the identification of operational and process controls available. It is important to note that reservoirs can form one of the important barriers to contamination in this preventive strategy from ‘catchment to tap’, however they can also be a source of water quality deterioration.

The first step in the development of a reservoir management plan using this key element approach should be to consider the potential hazards from catchment activities and the circumstances under which these will develop as risks to water quality. This risk assessment not only informs the reservoir management plan but may also reveal opportunities for managing contaminants at the source.

The most common contaminants of concern generated from catchments are pathogens, particulates and natural organic matter but in some cases industrial and agricultural chemicals may also be important. High rainfall and associated inflows represent the major risk period to deliver these contaminants.

During storage of water in the reservoir a number of beneficial water quality changes can occur. Reduced water movement increases the rate of sedimentation of particulates. This reduces turbidity and may also result in the sequestering of the contaminants associated with the particles. Many of the pathogens of concern are attenuated by environmental conditions with temperature and UV being the most critical factors. They may also be consumed by grazers. Chemical contaminants may be subject to beneficial transformations that can de-toxify them. Understanding these mechanisms is critical to the understanding of the effectiveness of reservoirs as barriers to contamination. Of particular importance is a good knowledge of the hydrodynamic processes that control water movement in the reservoir as this will determine the length of time the water is retained and the environmental conditions that it will be subjected to. Issues such as the thermal stratification and the short-circuiting of inflows can have significant impacts.

Reservoirs may also be sources of contamination. Contaminants sequestered in the sediments can be resuspended during periods of high flow and bound contaminants may be released if the water chemistry changes. Contaminants that are important in this regard include iron, manganese and nutrients that can be released from sediments under anoxic conditions. Biological growths in the reservoir may also impair water quality and of particular concern is the growth of cyanobacteria which can produce toxins and off-flavours.

4 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Both of these issues are often closely linked with the formation of thermal stratification which promotes stability and dampens mixing, resulting in anoxia in bottom waters.

Reservoirs are subject to complex interactive factors which include hydrology, chemical and biological processes. Most management interventions that will impact on these processes may also result in unintended consequences. For example, artificial mixing to control cyanobacteria may reduce sedimentation rates. The reservoir needs to be managed for all water quality issues.

The successful application of the framework for management and risk assessment depends upon having good knowledge of the system. Historical data from the reservoir and the catchment should be used in the preliminary assessment and to design the on-going monitoring programs. Monitoring should include targeted sampling for particular hazards, on-line sensors for a range of environmental variables and can be supplemented by modeling. This knowledge forms the basis for sound planning and management. The management options available in reservoirs include manipulation of selective off-takes, artificial mixing and the application of chemicals such as algicides or flocculants. The management options for each issue are summarised in Table 1.

Table 1 Causes and Management of Water Quality issues in Reservoirs Problem Impact Cause Management options Cyanobacteria Toxins – Health risks Stratification Destratification Taste, odours, aesthetics Nutrients Nutrient reduction High temperature Algicides Selective off take Application of flocculants Pathogens Health risks Contaminated Selective off take inflows from catchments Iron & Aesthetics Stratification Destratification Manganese Natural Organic Treatment costs Catchment inputs Selective off take Matter (NOM) THM-formation potential – Biological growth health risk in reservoirs Bacterial regrowth- health & aesthetics

MEASURING AND MODELING OF WATER MOVEMENT AND WATER QUALITY

The pattern of water flow and mixing, which is essentially the hydrodynamics, impacts upon all aspects of the chemistry and ecology of the reservoir and represents critical knowledge for management.

Measurement and monitoring

It is often said that you cannot manage what you cannot measure. Fortunately there has been a significant improvement in the methods available for measuring physical, chemical and biological characteristics of reservoirs. On-line sensors combined with telemetry can provide real-time data on a range of reservoir characteristics.

Monitoring of inflows and extractions is the most basic information required for managing reservoirs. Inflow volumes determine the contaminant loads from the catchment. Periods of high inflow represent the greatest risk and it is useful to monitor inflows on-line to have advanced warning of these risk events. Also, weather forecasts should be used to provide advanced warning of potential inflows. This will allow adjustments to be made to the monitoring program and give early warning of possible challenges to the water treatment plant. The balance of the inflow and outflow data determines the overall retention time for the reservoir.

5 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

On-line monitoring of temperature provides valuable information on the mixing regime. The temperature differential of the inflow and the resident water will determine the relative density and therefore influence the way the two waters mix. If the inflow is colder and therefore denser it will tend to undercut the reservoir water while if it is warmer it will flow over the top. A high temperature differential will tend to impede mixing. Temperature profiles provide information on the extent of mixing. Isothermal profiles indicate a high degree of mixing. Stratification of the profile indicates reduced mixing. Stratification can occur when the rate of solar heating of the surface waters exceeds the rate at which the heat can be transferred down through the water column. Density stratification may also occur due to differences between the inflows and the resident water. This can be the result of either temperature or salinity differences. The progress of inflow plumes can be monitored through changes in the temperature profiles.

There are a number of probes that can be used for monitoring chemical and biological characteristics continuously in the reservoir. These include salinity, dissolved oxygen, turbidity and particle counters and algal pigments as surrogates for biomass. It is now possible to have a range of real-time data which provide a detailed picture of the ecology of the reservoir.

Hydrodynamic and water quality models

The application of hydrodynamic and water quality models has significantly increased in recent years. The hydrodynamic models are now well developed and tested and progress is also being made on ecological models, often coupled to the hydrodynamic models. These models may vary in complexity from simple relationships between bioavailable nutrients and the maximum attainable chlorophyll in a reservoir through to sophisticated 3D hydrodynamic models linked to real-time monitoring systems. These models have a key role in reservoir risk assessment and many applications from cyanobacterial control to pathogen fate and transport can be simulated and assessed.

MONITORING PATHOGENS IN INFLOWS

The pathogens of concern from a public health perspective invariably originate from the catchments. Consequently, rain event inflows to reservoirs present one of the greatest challenges to water quality managers. Pathogens can be inactivated by sedimentation, exposure to UV light, grazing and natural mortality. These processes are influenced by the length of time the pathogens spend in the reservoir and on their distribution. This in turn is determined by the processes of dispersion, dilution, horizontal and vertical transport. Of particular importance is the behaviour of the pathogen-containing riverine inflows. Inflows are controlled by their density relative to that of the lake, such that warm inflows will flow over the surface of the lake as a buoyant surface flow and cold, dense inflows will sink beneath the lake water where they will flow along the submerged reservoir basin channel towards the deepest point. In either case the inflow will entrain water from the lake, increasing its volume, changing its density and diluting the concentration of pathogens and other characteristics. In the case of a dense underflow, when its density matches that of the adjacent lake the underflow will become an intrusion. In some cases the underflow is denser than any water in the lake and it will flow all the way to the deepest point. This is of particular interest for in the majority of drinking water reservoirs the deepest point is often where the off-take to drinking water supply is located (i.e. at the dam wall). Given that oocysts will survive longest in cold and dark water, this underflow mechanism can potentially produce the greatest risk.

A comprehensive case study to demonstrate pathogen movement was undertaken at Myponga Reservoir, and is described in detail in this report. This reservoir is an impounded, flooded river valley located 70 km south of Adelaide, South Australia. The reservoir has a capacity of 26,800 ML and maximum depth of 42 m at full supply level. The mean retention time based upon abstraction rates is approximately 3 years and the surface area is 2.8 km2. The reservoir had two meteorological stations installed that detect wind and solar radiation and had thermistor chains that were used to assess the hydrodynamic conditions within the water body. Flow and temperature were also measured in the inflow creek.

In this study the distribution and inactivation of pathogens was predicted using a 3D hydrodynamic model (ELCOM), coupled to a biogeochemical model which incorporated a pathogen settling and inactivation component. The model was validated by intensive monitoring and sampling of Myponga Reservoir during a large inflow event. Rainfall of 42 mm in the catchment in 24 hours generated a

6 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 maximum flow of 9.8 m3 in the single input stream. This riverine water entered the reservoir as an underflow since it was substantially cooler, contained more solutes and therefore had a greater density than the ambient reservoir water. The inflow moved completely through the reservoir as a distinct pulse within 36 hours, which dramatically demonstrated the significance of short–circuiting of contaminated inflows in this reservoir.

The study incorporated sampling for Cryptosporidium and a range of surrogates at four sites through the reservoir for model validation. Cryptosporidium concentrations at the inflow (Myponga Creek) were used as model inputs and then the model was used to predict the concentration at the sampling sites through time. The study demonstrated that: The model provided accurate predictions of the behaviour of the intrusion and of the Cryptosporidium concentrations carried within it The bacterial indicator Clostridium provided the best correlation as a surrogate for the pathogen Cryptosporidium Information on the distribution and the behaviour of contaminated inflows can be used to minimise risks by: o Alerting the treatment operators to high risk periods, o Manipulation of the off-take to avoid contaminated inflowing water.

This guide also describes case studies that used a simple model called ‘INFLOW’ (Antenucci et al., 2005). INFLOW is a simple web-based tool for predicting the dilution, travel times and insertion depths of riverine inflows. The model will also give an indication of the approximate timescale for reduction in viable Cryptosporidium concentrations. The case studies describe where a riverine intrusion was monitored in two reservoirs, and the information from this was used to validate the INFLOW model. The output from the model was used to discuss the quantification of risk reduction for Cryptosporidium transport through the reservoirs. The model was developed to provide a simple tool to assess this risk in reservoirs while balancing an appropriate level of process understanding and still remaining a practical tool for the reservoir operator. The model is freely available on the web: http://www.cwr.uwa.edu.au/services/models.php?mdid=9

NATURAL ORGANIC MATTER

Natural organic matter (NOM) is a complex, heterogeneous residue of decomposing biomass. It is considered a contaminant for potable water supply as it contributes to the cost of treatment by exerting a coagulant demand, acts as a carcinogen precursor during disinfection, can promote bacterial regrowth in the distribution and is an aesthetic (visual and taste) problem to customers at high concentrations.

The nature of the catchment-derived NOM is initially determined by the vegetation and biota but it is subsequently influenced by the catchment soil type and the hydrology. NOM is also generated by biological growth within the reservoir. Both the reservoir- and catchment-derived NOM are influenced by a number of chemical and microbiological processes including exposure to UV. The impact of these processes is influenced by the hydrology and hydrodynamics in a similar way to the fate of the catchment-derived pathogens. The outcome of particular interest is the impact of the eventual NOM characteristics on the subsequent water treatment and disinfection processes.

A case study presented in this report investigated the influence of this range of processes on the character of the NOM during transport through two reservoirs during storm event inflows with the goal being to assess the options for managing reservoirs for optimal treatability of the water.

Two comprehensive field experiments were conducted to track the passage of storm event inflows through Little Para and Myponga Reservoirs in South Australia. During the events samples were collected from the river and reservoir to determine the concentration of dissolved organic carbon at various sites and depths, and analyses were undertaken to determine alum demand, chlorine demand, disinfection by-product formation and the quality of water pre- and post-treatment in the

7 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT associated water treatment plants. The INFLOW model was used to calculate travel time, dilution and insertion depth of the riverine intrusion in each reservoir.

For Little Para Reservoir the inflow formed a distinct underflow. The inflow was rated as a 1 in 10 year event and therefore had a significant impact on the quality of water in the reservoir. The inflow was higher in turbidity, DOC, colour and THM-formation potential. As a consequence of this poor water quality the alum dose required to treat the water increased from 60 to 90 mg/L.

The inflow to the Myponga Reservoir also formed an underflow but this was less distinct than at Little Para Reservoir. This was because this inflow was only a 1 in 2 year event and Myponga Reservoir has a greater retention time than Little Para Reservoir and therefore there was less impact on the overall quality of the water in the reservoir. The inflow was high in turbidity, colour and DOC but the alum dose required to treat the water did not change significantly.

These studies and other information indicate that the NOM introduced during storm water inflows is clearly more difficult to treat. Therefore knowledge of how the intrusion is behaving provides scope and opportunity for manipulating off takes to avoid withdrawing the contaminated water.

CYANOBACTERIA

Many cyanobacteria produce taste and odour compounds making the water objectionable to consumers. Some also produce toxins which have potential for effects upon public health. They are therefore a major water quality risk.

Many of the bloom-forming nuisance cyanobacteria have characteristics that allow them to exploit the environmental conditions found in larger water storage reservoirs. These cyanobacteria generally have gas vacuoles which allow them to adjust their position in the water column to suit their light requirements. The reduced turbulence that results from stable weather and long retention in reservoirs allows for more effective buoyancy control. During summer many of the reservoirs in Australia are subject to thermal stratification where the surface waters are separated from the bottom waters by a density difference across a thermocline. This further dampens the mixing processes. Cyanobacteria tend to be favoured by high temperature which explains their predominance in summer. This advantage is intensified in stratified conditions where the surface waters are significantly warmer and less dense than the bottom waters. A further advantage is the ability of cyanobacteria to utilise low nutrient concentrations. This is important as the nutrient concentrations in the surface waters progressively decline through summer as they are depleted by algal and plant growth. Some species of cyanobacteria also form physically large and almost macroscopic colonies which are not readily grazed by zooplankton.

The most environmentally sound method to control cyanobacterial growth in reservoirs is to manipulate the environment to favour other phytoplankton over the cyanobacteria. The reduced mixing and turbulence in reservoirs is the central factor in promoting cyanobacterial growth in reservoirs, and it is for this reason artificial mixing and destratification has received most attention as a potential management technique to reduce their growth. Artificial mixing not only discourages cyanobacteria but it also can address the release of iron, manganese and nutrients from the sediments, which occurs when reservoirs become stratified.

There are basically two types of artificial destratification systems available; bubble plume aerators and mechanical mixers. Both systems generate turbulence which weakens stratification and allows the influence of the prevailing wind (wind-forcing) to then more readily mix the reservoir. Bubble plume aerators operate by pumping air through a diffuser hose near the bottom of the reservoir. As the small bubbles rise to the surface they entrain water and this rising plume develops unique temperature and density characteristics. This plume will rise to the surface and then plunge back to the level of equivalent density in the reservoir and an intrusion will then propagate horizontally away from the aerator plume at that depth. As the intrusion moves through the reservoir there is return flow above and below the intrusion and these circulation cells facilitate exchange between the surface layer and the deeper water or hypolimnion. Mechanical mixers are usually surface-mounted and pump water downwards through a draft tube. They can also draw water upwards via the tube. Both types of destratifiers have been shown to mix the surface layers very well close to the mixing device but not as effectively outside the immediate influence of the plume and as a consequence there are still often

8 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 stable zones or habitats for buoyant cyanobacteria to exploit. One mixing approach to consider is to use aerators to generate the large basin-wide circulation cells and use mixers to target the surface stratification outside the direct influence of the aerator plume.

Cyanobacterial growth can also be limited by reducing nutrient concentrations in the reservoir with the main target being to reduce phosphorus which is usually the most critical nutrient promoting growth and bloom formation. Managing catchments to reduce the external load is a highly desirable long-term goal but it is complex and costly and often not sufficient by itself to eliminate cyanobacterial blooms. Phosphorus can also be released from the sediments in reservoirs that become stratified particularly if the bottom waters become anoxic as a result of reduced mixing and intense biological activity in the isolated deep water adjacent to the sediments. This is the so called internal nutrient load to the reservoir. Destratification can also be very effective in promoting circulation and oxygen diffusion to deep water to reduce this sediment phosphorus release. The importance and merits of investing in the different management activities to tackle internal and external nutrient loads for controlling nutrient supply to cyanobacteria will depend upon the relative contributions of these external and internal sources to the nutrient budget of the particular reservoir. Good studies to accurately determine the relative sources of nutrients in each particular reservoir are therefore very important to make informed cost-benefit decisions about where to direct management effort.

The report presents a chapter on management of cyanobacteria which describes the conditions which lead to cyanobacterial blooms in reservoirs and from this information identifies opportunities for predicting, detecting and managing cyanobacteria in source waters, particularly by artificial mixing. The report provides a case study at Myponga Reservoir, South Australia that was the site of a CRC for Water Quality and Treatment study on artificial destratification using a bubble plume aerator and two surface-mounted mechanical mixers which were operated for six months each year. More detailed information on the study to evaluate the use of artificial mixing to control the growth of cyanobacteria is contained a separate CRC report (RR 59: Brookes et al, 2008)

The study used numerical models to simulate the reservoir hydrodynamics and cyanobacterial growth. The model was calibrated to simulate algal growth using equations describing nutrient and light-limited growth of Anabaena circinalis and floating velocity. After calibration the model was used to evaluate a series of scenarios based around mixing options with an aerator and combinations of surface-mounted mechanical mixers operated in different configurations and flow rates.

IRON AND MANGANESE

Iron and manganese contribute to taste and dirty water complaints. Iron and manganese are released from sediments under anoxic conditions which typically occur when the reservoir is stratified. The risks can be assessed by monitoring the dissolved oxygen concentration or the redox potential in the bottom waters. As release is related to the presence of stratification the recommended management option is artificial destratification. A case study described at Myponga Reservoir has demonstrated that the aerator was successful for water circulation and for maintaining adequate dissolved oxygen concentrations at depth and as a result the iron and manganese concentrations were maintained below the drinking water guidelines.

WILDFIRES

Wildfires are a feature of the Australian landscape and can have very dramatic impacts upon water quality. The degree of the impact depends on the extent and intensity of the fire and the timing and intensity of the next rainfall event. Burnt catchments are susceptible to erosion and the resulting runoff can have high suspended solids, nutrients and organic carbon. Excessive use of fire retardants can also be an issue. Wildfire prevention is clearly an important catchment management requirement. This is an important consideration when assessing the risks from recreational and other activities in water supply catchments. After the event the main management options relate to reducing sediment transport from the catchment. This can include using barriers such as hay bales for interception and gross filtering of overland flow and sediment or small dams to trap sediments or to reduce discharge rates. The report presents a case describing the design of sediment transport barriers in catchments following wildfire in South Australia. A second case study describes the effects of fire-generated contaminants on reservoir water quality for two drinking water reservoirs in the ACT following a severe fire in the catchment in 2003.

9 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

CLIMATE CHANGE

Climate is a major factor in determining the potential risks to reservoir water quality. It affects the nature of the vegetation in the catchment, the runoff patterns and the ecological processes in the reservoir. Significant climate change will therefore change the risk profile. Of particular concern is the risk of more extreme run off events, changes to vegetation type and increased risk of wildfires.

The report discusses the projected changes in regional climate based upon an ensemble of Global Climate Models (GCMs). These changes may have the potential to impact directly on the catchment hydrology, the thermal structure of the reservoir and the physical, biological, chemical and ecological processes occurring within the reservoir. These will combine to impact on the overall raw water quality and the frequency and magnitude of the potential hazards facing reservoir managers. This section outlines some of the potential impacts of climate change that reservoir managers need to consider when planning future operational strategies. This list is by no means exhaustive and as our knowledge of climate change improves and our understanding of natural and man-made system responses to the changes increases it is possible that additional impacts will be identified.

The guide provides a comprehensive methodology for the design of an aeration system for reservoir destratification in the Appendix.

10 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

TABLE OF CONTENTS Foreword ...... 3 Executive Summary ...... 4 List of Figures ...... 14 List of Tables ...... 17 1 Introduction ...... 18 2 Monitoring and Modeling ...... 21 2.1 Introduction ...... 21 2.2 Monitoring ...... 21 3 Hydrodynamic and biogeochemistry models for risk assessment ...... 24 3.1 Introduction ...... 24 4 Managing water quality and pathogens during rain events ...... 25 4.1 Introduction ...... 25 4.2 Case Study ...... 26 4.2.1 Methods ...... 26 4.2.2 Results ...... 26 5 Predicting the depth of a riverine inflow intrusion ...... 32 5.1 Introduction ...... 32 5.2 Model description ...... 32 5.2.1 Background ...... 32 5.2.2 Development of Inflow model ...... 32 5.2.3 Development of Pathogen Fate model ...... 33 5.2.4 Model inputs and outputs ...... 33 5.3 Case Studies ...... 34 5.3.1 Model application ...... 34 5.3.2 Site description ...... 34 5.3.3 Lake Burragorang ...... 34 5.3.4 Myponga Reservoir ...... 35 5.4 Quantifying risk reduction ...... 36 5.5 Conclusions ...... 37 5.6 References ...... 38 6 Natural Organic Matter Dynamics ...... 39 6.1 Introduction ...... 39 6.2 Methods ...... 40 6.2.1 Site descriptions ...... 40 6.2.2 Sampling ...... 41 6.2.3 Jar Test ...... 41 6.3 Results ...... 42 6.3.1 Little Para Reservoir ...... 42 6.3.2 Myponga Reservoir ...... 45 6.4 Discussion ...... 50 6.5 References ...... 51 7 Cyanobacteria ...... 53 7.1 Introduction ...... 53 7.2 Physical conditions favouring cyanobacteria ...... 54

11 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

7.3 Chemical conditions favouring cyanobacteria ...... 56 7.4 Managing cyanobacteria in source water ...... 56 7.5 Artificial destratification ...... 56 7.6 Case study – Myponga Reservoir ...... 58 7.6.1 The Phytoplankton Community ...... 58 7.6.2 Artificial destratification to control the nutrient load ...... 58 7.6.3 Relating nutrients to algal biomass ...... 59 7.7 Destratification and control of cyanobacterial growth ...... 61 7.8 Simulation of various management strategies ...... 66 7.8.1 No artificial intervention (Strategy 1) ...... 66

7.8.2 Artificial Mixing with no CuSO4 dosing (Strategy 2) ...... 67 7.8.3 Aerator only (Strategy 3) ...... 68 7.8.4 Surface Mixers (Strategy 4) ...... 69 7.8.5 Surface mixers at 5 m3s-1 (Strategy 5) ...... 71 7.8.6 Surface mixers at 8 m3s-1 (Strategy 6) ...... 72 7.8.7 Intermittent operation (Strategy 7) ...... 74 7.8.8 Equivalent aerator energy input using surface mixers (Strategy 8) ...... 76 7.9 References ...... 79 8 Iron and manganese ...... 81 8.1 Introduction ...... 81 8.2 Case study – Myponga Reservoir Fe and Mn reduction ...... 81 8.3 References ...... 83 9 Managing a variable off-take ...... 84 9.1 Introduction ...... 84 9.2 Case Studies ...... 84 9.3 Methods ...... 84 9.4 Results ...... 85 9.4.1 Pathogen transport ...... 85 9.4.2 Natural organic matter transport and treatment ...... 85 9.4.2.1 Vertical Distribution of Cyanobacteria ...... 86 9.5 Discussion ...... 87 9.6 References ...... 88 10 Managing Wildfire Impacts on Water Quality ...... 89 10.1 Introduction ...... 89 10.2 Water quality issues associated with wildfires ...... 89 10.3 Fire retardants ...... 89 10.4 CASE STUDY Managing sediment transport in catchments following wildfire ...... 90 10.5 CASE STUDY – Effects of a severe wildfire and fire-generated contaminants on reservoir water quality ...... 92 10.6 References ...... 97 11 Climate Change and Reservoir Management ...... 98 11.1 Introduction ...... 98 11.2 Understanding future climate change ...... 98 11.2.1 Limitations of GCMs ...... 98 11.3 Climate change projections for Australia ...... 99 11.4 Implications for reservoir management ...... 102 11.4.1 Changes to catchment hydrology ...... 102 11.4.2 Changes to water quality ...... 102

12 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

11.4.2.1 Increased water temperatures ...... 102 11.4.2.2 Changes to nutrient loads ...... 103 11.4.2.3 Impacts of increased day-to-day variability ...... 103 11.4.3 Impacts on reservoir hazards ...... 103 11.5 Wildfires ...... 103 11.6 Indirect - climate risks ...... 103 11.7 Adapting the framework for monitoring hazard and risk assessment to incorporate future climate projections ...... 104 11.8 References ...... 105 Appendix 12 Artificial destratification – aerator design and operation ...... 106 12.1 Introduction ...... 106 12.2 Detailed Stage Descriptions ...... 108 12.2.1 Stage 1 Conceptual Design ...... 108 12.2.2 Stage 2 Hydrodynamic modeling of the conceptual design ...... 111 12.2.3 Stage 3 Pneumatic and practical design ...... 113 12.2.4 Summary ...... 114 12.3 References ...... 115

13 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

FIGURES

Figure 1 A Conceptual framework of monitoring and information processes, risk assessment, decision- making, and outcomes for reservoir management and water treatment ...... 20 Figure 2 A combined lake meteorologic station and thermistor chain deployed in Myponga Reservoir, South Australia...... 23 Figure 3 Conceptual model of the major processes affecting pathogen fate and transport through a reservoir ...... 25 Figure 4 Myponga Reservoir showing the inflow from Myponga Creek and two meteorological stations, Met 1 & Met 2, which contain a string of thermistors ...... 27 Figure 5 Myponga Creek temperature (pink) and flow (blue), and reservoir temperature at the surface and bottom measured at Met 2 in the sidearm and Met 1 in the main basin ...... 27 Figure 6 Simulation of Cryptosporidium oocysts entering Myponga Reservoir within a riverine intrusion presenting as an underflow...... 28 Figure 7 Temperatures (°C) measured during the five transects as indicated by the day of the year preceded by year (yyyyddd)...... 29 Figure 8 Predicted and measured Cryptosporidium oocyst concentration in the bottom water of Myponga Reservoir a large rain event in the catchment and subsequent intrusion of this water into Myponga Reservoir from Myponga River (Eastern-most point of map) ...... 30 Figure 9 Predicted Cryptosporidium oocyst concentration at different depths in Myponga Reservoir following a rain event inflow ...... 31 Figure 10 Inflow hydrograph and corresponding inflow temperature for Burragorang event in June/July 1997 ...... 35 Figure 11 Temperature and turbidity profiles collected at Lake Burragorang during a flood event in June/July 1997 ...... 35 Figure 12 Flow rate and inflow temperature in Myponga Creek, with vertical lines showing nominal onset and subsidence of flood waters...... 36 Figure 13 Conceptual representation of the factors affecting carbon input into reservoir, degradation pathways and ultimately the impact on water quality...... 40 Figure 14 Location of Little Para and Myponga Reservoirs, South Australia ...... 41 Figure 15 Turbidity at Little Para Reservoir. The maximum turbidity measurable by the sensors was 50 NTU hence the cut off on the Gould Creek and Little Para turbidity data ...... 43 Figure 16 Apparent molecular weight of dissolved organics in Little Para Reservoir for samples collected before, during and after the inflow event at the Dam Wall ...... 45 Figure 17 Water temperature measured in Myponga Reservoir at seven depths and at the flow monitoring station in Myponga River ...... 47 Figure 18 Molecular weight distribution of Myponga River raw and coagulated before and after a rain event ...... 47 Figure 19 Specific THM formation in Myponga Reservoir before and after a rain event...... 48 Figure 20 Molecular weight distribution of Myponga Reservoir NOM before and after a rain event .... 48 Figure 21 Molecular weight distribution of Myponga Reservoir coagulated waters NOM ...... 49 Figure 22 Myponga Reservoir showing sampling locations, mixer and aerator deployment and sites where meteorological stations are deployed ...... 53 Figure 23 Temperature measurements collected in depth profile with logging thermistors at high frequency in Myponga Reservoir in January 2000 ...... 54 Figure 24 Depth of surface mixed layer, defined as the shallowest depth at which the temperature difference between two adjacent thermistors is 0.05°C or greater...... 55 Figure 25 Flow and circulation fields created by a bubble plume aerator and a surface-mounted mechanical mixer in reservoirs ...... 57 Figure 26 The relative abundance of the different phytoplankton groups in Myponga Reservoir...... 58 Figure 27 Temperature measured weekly at the surface, 10 m, 20 m and 30 m depth adjacent to the off-take point at Myponga Reservoir. The aerator was installed in 1994...... 59

14 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 28 Filterable reactive phosphorus at the surface and 30 m at Location 1 near the dam wall and from Location 4 from October 1998. Aerator installation decreased the internal nutrient load and high concentrations in the hypolimnion were not observed following aerator deployment...... 59 Figure 29 Filterable reactive phosphorus at four depths at Location 4 and chlorophyll concentration integrated over the top 5 m...... 60 Figure 30 Relationship between annual mean Southern Oscillation Index and flow into Myponga Reservoir...... 61 Figure 31 Relationship between maximum total phosphorus and maximum chlorophyll a in the following growing period...... 61 Figure 32 A comparison of observed and simulated total Chl a concentration (μg Chl a L-1), with simulated CuSO4 dosing on 11 and 12 January 2000, surface mixers and aerator operation between 1 October 1999 and 1 April 2000. The R2 and P-value for the comparison were 0.75 and 3E-09, respectively...... 62 Figure 33 Observed and simulated Scenedesmus concentration (μg Chl a L-1), with a R2 and P-value of 0.73 and 4E-09, respectively...... 63 Figure 34 Comparison of the observed and simulated Anabaena circinalis concentration (μg Chl a L-1), with a R2 and P-value of 0.55 and 0.009, respectively...... 64 Figure 35 Observed and simulated Nitzschia sp. concentration (μg Chl a L-1)...... 64 Figure 36 A comparison of observed and simulated total Chl a concentration (μg Chl a L-1), with simulated CuSO4 dosing on 31-January-2000, and surface mixers and aerator operating between the 1-October-2000 and 28-February-2001...... 65 Figure 37 Comparison between the observed and simulated Anabaena circinalis concentration (μg Chl a L-1) from 1-September-2000 to 1-March-2001...... 65 Figure 38 Simulated thermal structure and DO concentration for Myponga Reservoir with no artificial mixing...... 66 Figure 39 Simulated Anabaena circinalis concentration (μg Chl a L-1) with no artificial mixing compared with the observed data under normal operating conditions...... 67 -1 Figure 40 Simulated Scenedesmus concentration (μg Chl a L ) with no CuSO4 dosing compared with the observed data under normal operating conditions, R2 = 0.72...... 67 -1 Figure 41 Simulated Anabaena circinalis concentration (μg Chl a L ) with no CuSO4 dosing compared with the observed data under normal operating conditions...... 68 Figure 42 Temperature and DO profiles for the simulated period using the aerator only (period when aerator operating is marked with a solid black line)...... 68 Figure 43 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the aerator only compared with the observed data under normal operating conditions...... 69 Figure 44 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the aerator only compared with the observed data under normal operating conditions...... 69 Figure 45 Temperature and DO profiles for the simulated period using two surface mixers at their measured flow rate of 3.5 m3 s-1. Note the period when surface mixers were operational is marked with a solid black line...... 70 Figure 46 Simulated green Chl a concentration (μg Chl a L-1) with the use of the surface mixers operating at 3.5 m3 s-1 each compared with the observed data under normal operating conditions...... 70 Figure 47 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the surface mixers operating at 3.5 m3 s-1 compared with the observed data under normal operating conditions...... 71 Figure 48 Temperature and DO profiles for the simulated period using the surface mixers at 5-m3 s-1 (period when surface mixers operating is marked with a solid black line)...... 71 Figure 49 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the surface mixers operating at 5-m3 s-1 compared with the observed data under normal operating conditions...... 72 Figure 50 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the surface mixers operating at 5-m3 s-1 compared with the observed data under normal operating conditions...... 72

15 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Figure 51 Temperature and DO profiles for the simulated period using the surface mixers at 8-m3 s-1. Note: period when surface mixers were operational is marked with a solid black line...... 73 Figure 52 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the surface mixers operating at 8-m3 s-1 compared with the observed data under normal operating conditions...... 73 Figure 53 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the surface mixers operating at 8-m3 s-1 compared with the observed data under normal operating conditions...... 74 Figure 54 Temperature and DO profiles for the simulated period using intermittent mixing using both the surface mixers at 3.5-m3 s-1 and the aerator. The mixing devices operate intermittently (2 days on, 4 days off) throughout the period marked with a solid black line...... 75 Figure 55 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the intermittent artificial mixing compared with the observed data under normal operating conditions...... 75 Figure 56 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the intermittent artificial mixing compared with the observed data under normal operating conditions...... 76 Figure 57 Simulated thermal structure and DO concentration for Myponga Reservoir with 25 surface mixers...... 76 Figure 58 Simulated Scenedesmus concentration (μg Chl a L-1) with 25 surface mixers compared with the observed data under normal operating conditions...... 77 Figure 59 Simulated Anabaena circinalis concentration (μg Chl a L-1) with 100 kW of equivalent surface mixers compared with the observed data under normal operating conditions...... 77 Figure 60 Temperature profiles in Myponga Reservoir...... 82 Figure 61 Soluble manganese concentrations in Myponga Reservoir...... 83 Figure 62 Prediction of Cryptosporidium parvum concentrations at different depths in Myponga Reservoir...... 85 Figure 63 Vertical profiles of a population of the cyanobacterium Anabaena circinalis in a horizontal transect across a South Australian reservoir measured by chlorophyll a fluorescence...... 87 Figure 64 Fire burning in the Little Para Catchment ...... 91 Figure 65 Hay bales contoured on a burnt hill-slope to trap sediments and retain them on the catchment (Photo: Paul Hackney) ...... 91 Figure 66 The burnt hill-slope revegetated after winter rain (Photo Paul Hackney)...... 92 Figure 67 The location of the four main dams of the ACT and their catchments, the Cotter and Queanbeyan Rives...... 93 Figure 68 Total phosphorus concentration at selected depths in Bendora Reservoir ...... 94 Figure 69 Chlorophyll concentration at the surface in Bendora Reservoir...... 94 Figure 70 Total algal abundance in Bendora Reservoir...... 94 Figure 71 Turbidity at selected depths in Bendora Reservoir (note Y axis is logarithmic scale) ...... 95 Figure 72 Total manganese concentration in Bendora Reservoir ...... 96 Figure 73 Total manganese concentration in Bendora post fires in 2003 ...... 96 Figure 74 The propagation of uncertainties through the climate change impact assessment process 99 Figure 75 Projected changes in temperature for the year 2050 (best estimate scenario) ...... 100 Figure 76 Projected changes in annual precipitation for the year 2050 (best estimate scenario) ...... 101 Figure 77 Steps of aerator design showing feedback between stages...... 107 Figure 78 Diagram of a bubble plume, showing a single entrainment and disentrainment cycle...... 108 Figure 79 The relationship between M, C and mechanical efficiency ...... 109 Figure 80 Bubble plume aerators consist of a submerged perforated pipe through which air is pumped...... 115

16 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

TABLES

Table 1 Causes and Management of Water Quality issues in Reservoirs ...... 5 Table 2 Summary information for the two reservoirs...... 34 Table 3 Measured and predicted inflow characteristics for the Burragorang event...... 34 Table 4 Measured and predicted inflow characteristics for the Myponga event ...... 36 Table 5 Percentage reduction in concentration of viable oocysts after 100 days...... 37 Table 6 Jar Test results and water sample analysis for Pre-inflow and Inflow samples collected from Little Para Reservoir at the Dam Wall and at reservoir tributaries (Gould Creek and Little Para River)...... 44 Table 7 Jar Test results and water sample analysis for Pre-inflow and Inflow samples collected from Myponga Reservoir at the Dam Wall and at Myponga River...... 46 Table 8 Anabaena circinalis concentrations (cells mL-1) at five locations at Myponga reservoir. A date with no record (–) signifies A. circinalis not detected in a 1 mL, 10x concentrated sample...... 55 Table 9 Results from existing and simulated water quality management strategies...... 78 Table 10 Jar Test results and water sample analysis for Pre-inflow, Inflow and Post-inflow samples collected at the Dam Wall and at reservoir inlets (Gould Creek and Little Para River)...... 86 Table 11 Total nutrient content of three fire-fighting chemicals (adapted from Couto-Vázquez and González-Prieto, 2006)...... 90 Table 12 Check list of required information for Stage 1 of aerator design process...... 109 Table 13 Check list of required information for Stage 2 of aerator design process...... 112

17 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

1 INTRODUCTION

Reservoir management should be considered in the context of the Framework for Management of Drinking Water Quality contained in Chapter 2 of the Australian Drinking Water Guidelines (ADWG), (http://www.nhmrc.gov.au/publications/synopses/_files/adwg_11_06_chapter_2.pdf) This Framework offers effective means of assuring drinking water quality and the protection of public health through adoption of a preventive management approach that encompasses all steps in water production from catchment to consumer. When using the Framework there are a number of elements that need to be considered in relation to reservoir management including: assessment of the role and importance of the reservoir in the context of the overall supply system; identification of preventive measures available for water quality management in the reservoir; and the identification of operational and process controls available. It is important to note that reservoirs can form one of the important barriers to contamination in this preventive strategy from ‘catchment to tap’, however they can also be a source of water quality deterioration.

It is critical for reservoir managers to understand the processes in the catchment and reservoir that lead to water quality incidents so that they are equipped to manage the risks appropriately and ensure protection of public health. An advanced conceptual framework for risk assessment and management of reservoirs is outlined in Figure 1. The flow chart starts with background System Knowledge which is the prerequisite for management and this is developed from historical data and understanding built up over time to guide monitoring in the catchment and reservoir (Figure 1). The next step is Monitoring to gather information and to detect hazards in particular timeframes. This has usually consisted of the routine collection of samples sent to an analytical laboratory. In addition

18 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 complementary data can now be collected with a range of different on-line sensors which is a powerful addition to system knowledge and routine monitoring to detect hazards.

The detection of hazards requires a Situation Assessment by managers and this can be significantly aided using Hydrodynamic/Ecological Models for the reservoir with different degrees of sophistication for short-term forecasting.

All of these steps inform the Risk Assessment and Management Actions at the appropriate Critical Control Points within this system. In the case of a reservoir there are few critical control points available, other than the drinking water off-take where water of the best quality could be selected or the decision could be made not to extract water if the risk is considered too great. Further options include enhancement of the water treatment process in the treatment plant.

The models referred to in the framework and flowchart are increasingly being used to cope with understanding the complex interactions in reservoirs and these are discussed in detail with examples of applications within this guide. There is a hierarchy that can be used in the development of models. The first order factors are the hydrological and hydrodynamic processes, especially inflows and stratification. Priority should therefore be given to developing the capacity to predict water movement and mixing regimes. Fortunately, there are a number of well-validated hydrodynamic models and acquisition of the data to drive the models has become much more convenient. Weather stations and on-line monitoring of temperature provide the capacity for real-time interactive systems. Once the hydrodynamics is in hand it is possible to develop sub-routines for the ecological processes and the specific contaminants of concern.

The guide provides an overview of processes that impact upon drinking water quality in reservoirs and also a range of selected procedures and tools for reservoir management. This is supported by a series of case study investigations that were largely carried out within the Source Water Program of the CRC for Water Quality & Treatment.

The first section of this guide deals with reservoir monitoring and modeling with discussion about the application of hydrodynamic models in reservoir management. It is followed by sections on catchment- derived contaminants, including description and case studies of the use of a simple model to assess contaminant transport such as pathogens in reservoir inflows. The guide outlines the importance of natural organic matter dynamics and transformations in reservoirs. Subsequent sections address reservoir-derived contaminants such as cyanobacteria and iron and manganese and provide a case study of the use and importance of variable depth reservoir offtakes to optimise withdrawal of the best quality water. The guide includes discussion of managing the impacts of wildfire on water quality and has an overview of the emerging area of climate change and reservoir management.

19 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Figure 1 A Conceptual framework of monitoring and information processes, risk assessment, decision-making, and outcomes for reservoir management and water treatment

20 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

2 MONITORING AND MODELING

2.1 Introduction Lakes and reservoirs across the globe are under pressure from water extraction, modified catchments, eutrophication, fishing pressure and invasive species. These pressures are unlikely to diminish as human populations grow, demand for water resources increases and climate change modifies the drivers of lake ecosystems.

Because many of the modifications to the landscape and climate will be expressed firstly in lake ecosystems, these systems offer a unique opportunity to monitor, analyse and predict future landscape and climate change. By understanding the implications of these changes at a global level the expected ecosystem change can be predicted and planned for. Planning for future lake management and adaptation to meet community needs and expectations will be compromised without knowledge of how lakes respond to both natural and anthropogenic influences.

“Lakes are the canaries in the landscape.” Lake ecosystems are sensitive indicators of catchment modification and climatic conditions. Because lakes integrate across landscape, hydrology and climate, ecosystem change in aquatic systems is often observed more quickly than adjacent terrestrial ecosystems. Therefore changes to catchment or climate may be expressed in lake ecosystems before they are evident elsewhere. Early warning of significant ecosystem change and knowledge of the likely consequences enables communities to respond and adapt to the change. However, to detect changes in lake ecosystems it is necessary to monitor sensitive indicators at appropriate timescales. Events that drive lake ecosystem processes occur at a range of timescales from short-term such as rain event inflows to seasonal changes and longer term phenomena such as El Niño and climate change.

2.2 Monitoring A conventional routine monitoring program usually provides weekly or monthly data on the physical, chemical and biological status of the reservoir. However some of the hazards can develop at timescales that are shorter than these sampling intervals. The keys to developing an enhanced monitoring system for real-time hazard detection are firstly an understanding of the reservoir hydrodynamics, and secondly, integrating the required on-line physical and chemical data and routine data with an understanding of the ecological, chemical and physical processes (i.e. algal growth, metal chemistry, pathogen movement). A process flow chart of these steps is given in Figure 1. Examples of background historical system knowledge would include an understanding of seasonal hydrology, knowledge of sources of pathogens in the catchment, as well as knowledge of the nutrient status and stratification behaviour of the reservoir.

Real-time monitoring with a range of equipment and sensors can be used to supplement a conventional routine program. The establishment of a program with high frequency measurement of local climate, water temperature, dissolved oxygen and phytoplankton biomass (using chlorophyll fluorescence as an indicator) preferably in a depth profile by stations deployed on lakes can capture many of the important ecosystem drivers and responses at timescales necessary to resolve the reservoir processes of interest. These measurements can then be used to inform risk assessment of important water quality hazards such as pathogen fate and transport, cyanobacterial growth and the impact of catchment or lake-derived carbon on lake metabolism and ecosystem health.

To develop the best risk management strategy incorporating on-line monitoring a review of the system outlining the major risks should be developed based on historical data. The monitoring system can then be designed to measure the dominant process in this catchment and reservoir system leading to generation of the hazards. High intensity data collected with on-line sensors are also relevant for input or calibration of hydrodynamic models and is a potential information resource for future data needs in addition to immediate monitoring needs.

Density stratification (temperature stratification) in a lake and riverine inflows are important processes that determine how contaminants travel through a reservoir. To understand and monitor this process the on-line monitoring system should include a string or chain of thermistors spaced over the depth

21 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT profile in the reservoir and a thermistor and flow gauge on the inflow. These measurements alone will enable calculation of:

• the degree and intensity of stratification • the depth of mixing • persistence of the thermocline • the volume and density of inflows so the depth of intrusion can be modeled

Reservoir hydrodynamics is controlled by meteorological processes and measurement of these is important to determine how the reservoir will respond to wind, solar radiation and temperature. Furthermore, when sophisticated hydrodynamic models are to be used, this information is essential for model input and calibration and so for an additional investment in automated monitoring of these on the reservoir itself very valuable data can be collected. The meteorological data sampling should include measurements of:

• wind speed and direction • air temperature • relative humidity • shortwave radiation • incoming and outgoing long-wave radiation

Some research applications require a rapid sampling rate of these variables together with thermistor chain data in order to capture the data necessary to observe processes such as internal waves and turbulence. However, for most management applications a data sampling rate at 10 minute intervals is adequate and recommended.

In summary, on-line monitoring is a valuable tool in day to day management of a reservoir system to supplement the regular sampling for chemical and biological water quality characteristics. The following list of activities should be undertaken when planning and developing a system for on-line monitoring.

• Review the historical reservoir physical chemical and biological data • Identify the hazards to water quality and threats to treatment plants for this system • Introduce an on-line monitoring system that captures the main drivers for the hydrodynamic (temperature, meteorological drivers) and water quality threat processes (inflows, turbidity, etc.) in this system • Regularly review the routine monitoring data of the important variables such as nutrients, cyanobacteria and pathogens • Review the hydrodynamic characteristics of the particular system which lead to hazard generation with the assistance of the on-line data • Develop conceptual model for predicting risk and treatment response, given the prevailing and predicted hydrodynamics • In view of this improved understanding, assess advanced hydrodynamic and water quality modeling that will improve hazard and risk assessment for the system

There are now a considerable number of lakes and reservoirs that have on-line sensor deployments. The Global Lake Ecological Observatory Network (GLEON; www.gleon.org) is a program that represents an array of lake sensors deployed around the globe to address local issues for individual lake ecosystems but also to document changes in lake ecosystems that occur in response to different land use, latitude and climate.

A typical on-line lake monitoring station that measure a range of metrological inputs and temperature profiles is shown in Figure 2.

22 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 2 A combined lake meteorologic station and thermistor chain deployed in Myponga Reservoir, South Australia.

The system has a range of sensors that log at short time intervals and communicate via a radio link or mobile phone network with the option for manual retrieval via a computer modem or automated web- based storage.

23 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

3 HYDRODYNAMIC AND BIOGEOCHEMISTRY MODELS FOR RISK ASSESSMENT

3.1 Introduction Computer models of reservoir hydrodynamics and biogeochemistry can play a key role in management and risk assessment of drinking water supply reservoirs. These models provide prediction of water quality outcomes for various climate and management scenarios and can facilitate informed management and investment in risk mitigation. Models vary in complexity and require investment in model construction, maintenance and training. However, the models also enable utilities to properly evaluate risks from the major hazards; pathogens, cyanobacteria, iron and manganese and chemical pollutants.

The different water quality hazards are generated over different timescales and so it may be necessary to use different models for different situations. These models vary in complexity from simple relationships between bioavailable nutrients and the maximum attainable chlorophyll through to sophisticated 3D hydrodynamic models linked to real-time monitoring systems. A number of models have been developed or expanded within projects run within the Reservoir Program of the CRC for Water Quality and Treatment. While no modeling software is necessarily advocated by the CRC for Water Quality and Treatment there has been a strong collaboration with the Centre for Water Research (CWR) at the University of Western Australia and this group has made considerable advances and investment in the development of a range of models.

Typically water quality hazards that generate over the longer-term (weeks-years) are best represented by one-dimensional models as over these longer timescales the vertical variations in hydrodynamic properties, such as thermal stratification, can adequately explain most of the variability over time. Three dimensional models are the most appropriate tool for prediction when horizontal processes and variability need to be considered. An example of when three dimensional models can be used is given in Chapter 4, which describes the tracking of an inflow that developed a riverine intrusion carrying pathogens. The models can also be used to examine spills and other point source contaminant transport.

Descriptions of the CWR models are available on-line at www.cwr.uwa.edu.au. These include the one- dimensional model DYRESM, a three dimensional model ELCOM, and a biogeochemical model that can be coupled to these is called CAEDYM. The following chapters demonstrate how advanced modeling techniques can be used to aid risk assessment, and to identify and fill knowledge gaps.

24 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

4 MANAGING WATER QUALITY AND PATHOGENS DURING RAIN EVENTS

4.1 Introduction Rain event inflows to reservoirs present one of the greatest challenges to water quality managers. Pathogens, turbidity, and dissolved organic carbon in river water can potentially challenge drinking water treatment plants during rain events and it is important to properly manage this risk.

The processes of dispersion, dilution, horizontal and vertical transport determine the distribution of pathogens in lakes and reservoirs. Settling of pathogen particles operates in conjunction with these hydrodynamic processes. The fate of pathogens is also determined by inactivation in response to UV light and natural mortality. These fate and transport processes are summarised in Figure 3.

Figure 3 Conceptual model of the major processes affecting pathogen fate and transport through a reservoir

The horizontal transport is predominantly driven by inflows and basin-scale circulation patterns including wind-driven currents and internal waves. Although wind-driven currents only influence the surface layer, inflows can occur at any depth in a stratified reservoir and internal waves can generate significant internal currents that can act in different directions at different depths.

Whilst it is acknowledged that recreational activities can contribute to pathogen concentrations in reservoirs (Anderson et al., 1998), the riverine inflow is considered to be the major source of pathogens, and consequently the behaviour of these inflows is of particular importance.

Inflows are controlled by their density relative to that of the lake, such that warm inflows will flow over the surface of the lake as a buoyant surface flow and cold, dense inflows will sink beneath the lake water where they will flow along the submerged reservoir basin channel towards the deepest point. In either case the inflow will entrain water from the lake, increasing its volume, changing its density and diluting the concentration of pathogens and other characteristics. In the case of a dense underflow, when the density matches that of the adjacent lake the underflow will become an intrusion. In some cases the underflow is denser than any water in the lake and it will flow all the way to the deepest point. This is of particular interest for in the majority of drinking water reservoirs the deepest point is often where the off-take to drinking water supply is located (i.e. at the dam wall). Given that oocysts will survive longest in cold and dark water, this underflow mechanism can potentially produce the greatest risk.

25 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

A further complication is introduced where the density difference is derived from particulate matter (turbidity current), in which case the settling of these particles will influence the density and propagation of the inflow.

This chapter describes a comprehensive case study to demonstrate pathogen movement undertaken at Myponga Reservoir, South Australia. In this study the distribution and inactivation of pathogens was predicted using a 3D hydrodynamic model (ELCOM), coupled to a biogeochemical model which incorporated a pathogen settling and inactivation component. The model was validated by intensive monitoring and sampling of Myponga Reservoir during a large inflow event.

The aim of this case study is to demonstrate how pathogens are transported during rain event inflows in reservoirs and how they can be transported rapidly from the river to present a challenge at the drinking water off-take. The sampling was undertaken at Myponga Reservoir during two inflow events; one demonstrated that an underflow occurred and was used to calibrate a three dimensional hydrodynamic model and the other validated the model with extensive sampling.

4.2 Case Study

4.2.1 Methods The study site, Myponga Reservoir, is an impounded, flooded river valley located 70 km south of Adelaide, South Australia (S 35° 24’ 13”, E 138° 25’ 29”). The reservoir has a capacity of 26,800 ML and maximum depth of 42 m at full supply level. The mean retention time based on abstraction rates is approximately 3 years and the surface area is 2.8 km2. The catchment is approximately 124 km2 of mixed land use, including pasture for dairy, beef and hay production, with patchy remnant vegetation. Recent estimates of dominant land uses are 62% grazing and 24% dairy farming.

The reservoir has two meteorological stations that detect wind and solar radiation and collect data at 10 minute intervals. The stations also have thermistor chains that are used to record the hydrodynamic conditions within the water body. One station (Met 1) is located in the main basin and the other (Met 2) in a long narrow sidearm of the reservoir (Figure 4). Flow and temperature were also measured in Myponga Creek at 10 minute intervals.

4.2.2 Results The speed at which an inflow travels through a lake, its entrainment of lake water and resulting dilution of its characteristics and its insertion depth are all of critical importance in determining the hydrodynamic distribution of pathogens in lakes and reservoirs. A small flood with a peak daily total flow of 240 ML was recorded at Myponga Creek (Figure 5) on May 18, 2001 (Brookes et al., 2002). The temperature of the intruding creek water during the elevated flow ranged between 9.5-10.6°C. The intruding flow was detected as a departure from isothermal conditions and cooling at depth. The intrusion was detected at Met 2, in the side arm, (Figure 5) at 17:10 on May 17 and at Met 1, in the main basin 1.3 km away, at 4:00 on May 18. This equates to a velocity of 0.041 ms-1and therefore a potential travel time from inflow to off-take of 30 hours, whereas the retention of the reservoir computed on a whole-of-volume basis is 3.1 years.

26 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 4 Myponga Reservoir showing the inflow from Myponga Creek and two meteorological stations, Met 1 & Met 2, which contain a string of thermistors

Figure 5 Myponga Creek temperature (pink) and flow (blue), and reservoir temperature at the surface and bottom measured at Met 2 in the sidearm and Met 1 in the main basin

The three-dimensional model ELCOM was coupled to a biogeochemical model that incorporated a pathogen settling and inactivation component. The same data that was simulated in Figure 5 was used to input the model and oocysts included in the river inflow to demonstrate pathogen transport (Figure 6). The inflow is dense relative to the reservoir water and the water river enters as an underflow transporting oocysts with it. In this case oocysts travel from the inflow to the dam wall, a distance of 5 km, within 24 hours

27 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Figure 6 Simulation of Cryptosporidium oocysts entering Myponga reservoir within a riverine intrusion presenting as an underflow.

The first column of panels is viable oocysts, the second column is oocysts inactivated by natural loss processes and the third column of panels is the hydrograph of Myponga River and the theoretical concentration of oocyst loading.

A field program was developed to specifically validate the pathogen fate and transport model. This involved intensive monitoring and sampling of Myponga Reservoir during a large rain event inflow.

The high rainfall early on Day 178 (42 mm) generated considerable runoff and elevated flow in Myponga Creek. The measured maximum total flow was 9.8 m3s-1 recorded after midday of JD 178 (Figure 7). The riverine water entered the reservoir as an underflow since it was substantially cooler, contained more solutes and therefore had a greater density than the ambient reservoir water (Figure 7). There was a general warming of the river water during the period, which acted to decrease the density of the incoming riverine water. Nevertheless the inflow signal moved through the reservoir as a distinct pulse over 36 hours. The underflow was first detected at Met 2 in the sidearm as a general cooling of the bottom waters at 12:10 on JD 178, approximately nine hours after the Myponga Creek hydrograph began to rise. Cooling at depth occurred subsequently at Met 1 in the main basin, approximately four hours after the intrusion passed Met 2. The inflow pulse was first seen at the dam wall during T4, 36 hours after the hydrograph peak in the inflow stream.

28 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

0 12 5 11.8 10 11.6 15

20 o 11.4 z(m) Temperature ( C) 11.2 25 Transect 1: 2003178.55 30 (27/6/2003 1245 - 1440) 11

35 10.8

40 10.6 0 12 5 11.8 10 11.6 15

20 11.4 z(m) Temperature (oC) 11.2 25 Transect 2: 2003178.8 30 (27/6/2003 1710 - 1930) 11

35 10.8

40 10.6 0 12 5 11.8 10 11.6 15 11.4 20 o z(m) Temperature ( C) 11.2 25 Transect 3: 2003179.5 30 (28/6/2003 1110 - 1310) 11

35 10.8

40 10.6 0 12 5 11.8 10 11.6 15

20 11.4 z(m) Temperature (oC) 11.2 25 Transect 4: 2003179.75 30 (28/6/2003 1545 - 1836) 11

35 10.8

40 10.6 0 12 5 11.8 10 11.6 15

20 11.4 z(m) Temperature (oC) 11.2 25 Transect 5: 2003180.55 30 (29/6/2003 1130 - 1430) 11

35 10.8

40 10.6 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 Distance from inflow (m) Figure 7 Temperatures (°C) measured during the five transects as indicated by the day of the year preceded by year (yyyyddd). Transects were constructed by contouring around the profile locations (indicated by the dot-dash lines) using a rectangular grid with a sigma coordinate transformation.

Sampling for Cryptosporidium and a range of surrogates was undertaken at four sites for model validation. Cryptosporidium concentrations at the inflow (Myponga Creek) were used as model inputs and then the model was used to predict the concentration at the sampling sites through time (Figure 8). As the riverine intrusion travels into the reservoir the oocysts are first detected in the sidearm. The intrusion continues further into the reservoir and oocysts are detected at those sites but at lower concentrations due to dilution with ambient reservoir water (Figure 8).

29 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Figure 8 Predicted and measured Cryptosporidium oocyst concentration in the bottom water of Myponga Reservoir a large rain event in the catchment and subsequent intrusion of this water into Myponga Reservoir from Myponga River (Eastern-most point of map)

This work demonstrates that an understanding of hydrodynamics and river inflows can be used to gauge and mitigate pathogen risk. In the Myponga Reservoir case the riverine inflow occurred as an underflow and so the pathogen risk was predominantly near the bottom of the reservoir whereas the best water quality was found near the surface (Figure 9). Clearly the best water to extract for treatment and potable supply is therefore from the surface. Extraction of water outside of the intrusion maximises water quality and minimises the pathogen challenge to the drinking water plant.

30 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 9 Predicted Cryptosporidium oocyst concentration at different depths in Myponga Reservoir following a rain event inflow

31 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

5 PREDICTING THE DEPTH OF A RIVERINE INFLOW INTRUSION

5.1 Introduction Water-borne outbreaks of pathogenic organisms, such as Cryptosporidium and Giardia, pose a significant human health threat. A critical component in managing pathogen hazards in drinking water is determining the magnitude of the risk. In water supply reservoirs this requires an evaluation of pathogen concentrations which may be of the actual target organism or sometimes of indicator species (surrogates). The high cost of analysis to determine concentrations of Cryptosporidium and other indicator organisms makes it important to invest carefully in an effective sampling strategy to account for the hydrodynamic distribution of pathogens in lakes and reservoirs.

It is well known that water-borne disease outbreaks are often associated with rainfall events when pathogens are washed from the watershed and are transported into reservoirs via rivers. What is not always understood is the fate and transport of pathogens once they enter the reservoir. To determine the pathogen risk it is critical to refine the sampling program to account for pathogen distribution in inflowing waters and in the reservoir.

This chapter describes case studies that used a simple model called ‘INFLOW’ (Inflow Insertion and pathogen Fate Model). The case studies describe where a riverine intrusion was monitored in two reservoirs, and the information from this used to validate the ‘INFLOW’ model. The output from the model was used to discuss the quantification of risk reduction for Cryptosporidium transport through the reservoirs. 5.2 Model description 5.2.1 Background As the greatest risk of transport of pathogens from catchment to the treatment plant is after inflow events, the model of pathogen transport and dilution focuses on describing the dynamics of these events. The model INFLOW was developed to provide a simple tool to assess this risk in reservoirs while balancing an appropriate level of process understanding and still remaining a practical tool for the reservoir operator.

INFLOW is a simple web-based tool for predicting the dilution, travel times and insertion depths of riverine inflows. The model will also give an indication of the approximate timescale for reduction in viable Cryptosporidium concentrations. The model is based on the inflow routine used in the DYRESM model. The model is freely available on the web: http://www.cwr.uwa.edu.au/services/models.php?mdid=9 The model was developed under the American Water Works Association Research Foundation (AwwaRF) project #2752 'Hydrodynamic distribution of pathogens in lakes and reservoirs'. This project was conducted jointly by the Centre for Water Research and the Cooperative Research Centre for Water Quality and Treatment.

5.2.2 Development of Inflow model Inflows into a reservoir are typically colder and denser than the reservoir water. Consequently there are two major stages: downflow and insertion. During the downflow stage, the inflow travels down the water column in the reservoir. Due to shear between the inflow and the ambient reservoir water, entrainment into the inflow occurs, effectively decreasing the density (as well as increasing the volume) of the inflow parcel. When the inflow parcel reaches neutral buoyancy (that is, where the inflow parcel density is the same as the reservoir density), the inflow detaches from the reservoir side, and inserts into the water column. It is possible that an inflow is sufficiently dense not to insert, and travel the full extent of the reservoir to end up at the reservoir base. Consequently knowledge of the water column and river temperature can be used to determine where the riverine intrusions can be found prior to sampling for pathogens.

Whilst this provides useful information for the potential inflow pathways, the ability to quantify where inflows insert, and the associated level of dilution, is also required for risk management. The insertion

32 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 depth and dilution can be determined using DYRESM reservoir model (Imberger and Patterson 1981). In DYRESM, inflows are modeled as being channels of triangular cross-section, with constant bed slope angle ϕ, half-angle (that is, half the angle between the two sides making up the triangular cross- section) α and drag coefficient CD. The entrainment coefficient E and bulk Richardson number Ri for an inflow are calculated and gravity is the driving force for the inflow process. Combining these series of equations it is therefore possible to calculate the insertion height, dilution and travel-time for an inflow event. These are the critical variables for examining pathogen fate and transport in lakes or reservoirs.

5.2.3 Development of Pathogen Fate model In order to estimate risk associated with the inflow dilution, it is useful to consider other processes affecting pathogen concentration and viability. Antenucci et al. (2003) outline a model for Cryptosporidium fate and transport that considers loss of viability due to temperature, ultra-violet light exposure and settling. These processes can be coupled with the above inflow model to give an approximate estimate of risk associated with a particular inflow event.

The model for temperature inactivation follows that of Walker and Stedinger (1999). For UV light, it is also possible to compute the risk reduction due to exposure according to (Antenucci et al 2003)

− 2.1 It = 0eCC , where I is the intensity of UV light [mW cm-2], and t is the duration of exposure [s]. The intensity of UV light can be determined by the integrating the Beer-Lambert law over a relevant time period for the depth and UV light extinction coefficient of interest. The settling model is extremely simple, and requires an estimate of the sedimentation rate of pathogen particles along with knowledge of the insertion depth above the reservoir base.

5.2.4 Model inputs and outputs The sampling strategy to use the model requires the following information:

• Geometric description of river inflow to reservoir (stream bed slope and half-angle). These can be determined from bathymetric maps. • Temperature of inflow waters prior to entering the reservoir. • Flow rate of inflow waters prior to entering the reservoir. • Temperature profile of the reservoir prior to the inflow.

The equations presented above have been incorporated into a web-based application (http://www.cwr.uwa.edu.au/services/models/inflow.html). From this information, the following predictions are made:

• Insertion temperature • Insertion volume • Insertion dilution (relative to initial volume) • Insertion time

This will give the reservoir manager an indication of the potential threat for pathogen contamination.

More sophisticated techniques can be applied, whereby real-time monitoring of thermistor chain data and inflow temperature and flow data can be combined to provide similar information. These data can also be combined with hydrodynamic and water quality models of increasing complexity to provide even more information to the reservoir operator, but at increased cost.

33 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

5.3 CASE STUDIES 5.3.1 Model application The model was applied to two significant inflow events in Australian reservoirs to demonstrate the ability of the method to predict characteristics of the inflows.

5.3.2 Site description Two sites, situated in Australia, were used to test the ability of the model to replicate measured inflow characteristics. The first site, Lake Burragorang, is located in New South Wales and forms the major water supply reservoir for the city of Sydney. Myponga Reservoir, located in South Australia, is an important drinking water supply reservoir for the city of Adelaide (Table 2).

For all locations, inflow temperatures and flow rates were continuously gauged. At Lake Burragorang, in-lake profiles were collected using a profiling instrument at 1-2 day intervals. At Myponga Reservoir, two thermistor chains recorded temperature profiles every 10 minutes. Table 2 Summary information for the two reservoirs. Burragorang Myponga Volume (ML) 2,400,000 27,000 Maximum depth (m) 90 33 Maximum length (km) 80 5 Catchment area (km2) 9050 124 Stream-bed angle (°) 80 80 Stream-bed slope (°) 0.11 0.4 Drag Coefficient 0.015 0.015

5.3.3 Lake Burragorang During June 1997, a large rainfall event led to a significant inflow event into Lake Burragorang (Romero and Imberger 2003). The flow rate of the river when entering the reservoir on June 28 1997 was approximately 510 m3 s-1 (i.e. 44,000 ML d-1), with a temperature of 10.5°C (Figure 10). The temperature and turbidity structure of the reservoir before and after the inflow is presented in Figure 11.

From the field data, the inflow arrived at the dam wall some time after July 3 and prior to July 5, indicating a travel time of 6 to 7 days. The initial temperature of the inflow waters of 10.5°C meant that the inflow was an underflow. The inflow signature is clearly defined in the turbidity profiles at the dam wall, with a large volume of water of high turbidity present in the bottom 42 metres. From the reservoir storage curve (depth versus volume), this corresponds to a volume of approximately 98,000- 117,000 ML. The final temperature of the insertion, determined from the temperature profiles, was approximately 12.5-12.7°C.

Computing the inflow entrainment characteristics based on the above methodology yields an insertion temperature of 12.6°C, an insertion volume of 111,760 ML (that is, an entrained volume of 67,760 ML on top of the original inflow volume of 44,000 ML), and a travel time of 6.3 days. The agreement between the predicted and measured insertion temperature is also excellent, as is the travel time and dilution (Table 3).

Table 3 Measured and predicted inflow characteristics for the Burragorang event. Burragorang 28 June 1997 Measured Predicted Inflow temperature (°C) 10.5 - Inflow volume (ML) 44,000 - Travel time (days) 5-7 6.3 Insertion temperature (°C) 12.5-12.7 12.6 Insertion volume (ML) 98,000-117000 111,760 Inflow dilution 1.9-2.9x 2.5x

34 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 10 Inflow hydrograph and corresponding inflow temperature for Burragorang event in June/July 1997

Figure 11 Temperature and turbidity profiles collected at Lake Burragorang during a flood event in June/July 1997

5.3.4 Myponga Reservoir A small flood with a peak flow of 2.7 ms-1 was recorded at Myponga Creek on May 18, 2001 (Figure 12). The temperature of the inflow showed a diurnal variation between 9.5 and 10.5°C. The volume of the flood waters (defined by the vertical lines in Figure 12a) was 213 ML, with an average temperature of 10.1°C. Two thermistor chain stations were operational during the event, one located 3 km from the dam wall (Met 2), and one located 1 km from the dam wall (Met 1). The arrival time of the flood at the stations was defined by the onset of a departure from isothermal conditions, and is shown by the vertical lines in Figures 12b and 12c. The speed of the inflow, computed between the two in-lake stations and the inflow measurement point, averaged 0.074 ms-1 or approximately 6.4 km d-1.

35 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

From the field data, the arrival time of the flood at the dam wall was approximately 23 hours, with a temperature of 14.7°C and a volume of between 1,700 and 2,500 ML, based on the storage curve of the reservoir. Using the above model, it is possible to compute the insertion temperature to be 14.9°C, the travel time for the downflow to be 26 hours, and the insertion volume to be 1,934 ML, which is approximately 9 times dilution. The calculated insertion temperature matches well with the field data collected at station Met 1, where the temperature at 25 metres was measured as 14.7°C. The measured temperature at station Met 2 is lower than this value, simply as it lies in a transition region between the inflow temperature of 10°C and the final temperature of 14.7°C.

Table 4 Measured and predicted inflow characteristics for the Myponga event Myponga 18 May 2001 Measured Predicted Inflow temperature (°C) 10.1 - Inflow volume (ML) 213 - Travel time (hours) 23 26 Insertion temperature (°C) 14.7 14.9 Insertion volume (ML) 1,700-2,500 1,934 Inflow dilution 8-12 9.1

Figure 12 Flow rate and inflow temperature in Myponga Creek, with vertical lines showing nominal onset and subsidence of flood waters. b) Temperature at station Met 2, with vertical line showing arrival of flood front. c) as for (b), for station Met 1.

5.4 Quantifying risk reduction Using the above model, it is possible to estimate the risk reduction due to inflow dynamics and fate of the pathogens within the receiving reservoirs. Assuming the oocysts are 100% viable upon entering the reservoir, the timescale for temperature inactivation for the insertion temperature in Burragorang of 12.6°C will result in 90% reduction in viability after 205 days, and a 99% reduction in viability after 410

36 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 days. For UV light, assuming a UV light extinction coefficient of 1 m-1, the dosage computed at the insertion into Burragorang is very low. This low dosage results in less than 0.001% reduction in viability in 410 days and so, in the case of Burragorang, UV light exposure is unlikely to be an important process, as temperature alone will render 99% of oocysts nonviable in this time.

Settling is potentially an important loss process, with settling of individual Cryptosporidium particles measured at approximately 0.03 m day-1, whilst aggregated oocysts can settle at up to 2.5 m day-1 (Medema et al. 1998). From the above analysis, the volume of the inflow on reaching the dam wall is known (for example 111,760 ML for the Burragorang case). From a reservoir storage curve or from the temperature and turbidity profiles, the height of the insertion above the reservoir bottom can be determined (40-45 metres for the Burragorang case). It is then a simple calculation to determine the approximate time for particles to settle out of this inflow parcel – 18 days for a settling velocity of 2.5 m day-1, or up to 1,500 days for a settling velocity of 0.03 m day-1.

For each reservoir, the various risk reduction times have been computed and are presented in Table 5. For the ultra-violet light and settling times, the top of the Burragorang insertion was at 45 metres height (40 metres depth), and the top of the Myponga insertion was at 16 metres height (20 metres depth).

Table 5 Percentage reduction in concentration of viable oocysts after 100 days. (For dilution, the timescale is that for the inflow to reach the dam wall, that is, between one and seven days depending upon the reservoir.)

Burragorang Myponga Dilution 60 89 Settling (0.03 m d-1) 6.7 32 Settling (2.5 m d-1) 100 100 Temperature 67 78 UV light 0.0 2

From Table 5 it is clear that for underflows the role of UV light is effectively insignificant and can effectively be ignored. The dilution of the inflow during the underflow process is very important, as it results in a large risk reduction in a short period of time. The settling characteristics of the pathogens entering the reservoir are also relatively important, as they can show great variability in risk reduction depending upon the settling rate.

5.5 Conclusions The application of this model and the determination of the inflow characteristics, provides a tool for improved sampling that will enable a more accurate gauge of pathogen risk. Furthermore, the pathogen risk can be minimised by using the model to select a water harvesting depth from a region not impacted by the inflow. In the two examples presented in this study the riverine intrusions were underflows. Consequently, in these cases, surface grab samples would provide a poor estimate of pathogen risk as the concentration of pathogens and indicators at the surface would be more akin to ambient water quality than the riverine intrusion. The logical off-take depth for water harvesting in reservoirs where dense underflows dominate would therefore be towards the surface.

It should be noted that all models include a number of assumptions. For example, the model presented above assumes that the inflow, once inserted into the water column (after reaching neutral buoyancy), will immediately travel to the dam wall. In reality inflows will take some time to reach the off-takes once they insert after the downflow stage, however the complexity of this process means that the development of a simple predictive tool is not possible. The model presented above therefore represents a conservative estimate of travel time.

The model presented can be used to determine whether inflows act as overflows, insert at mid-depth, or act as underflows. This information allows for the design of a tailored monitoring program to focus on the regions of maximum risk. The model also gives an indication of the relevant timescales for risk reduction, and so gives the reservoir operator a powerful tool in managing to minimise risk.

37 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

5.6 References Antenucci JP, Hipsey MR, Brookes JD and Burch M (2003) A model for Cryptosporidium fate and transport in lakes and reservoirs. Paper oz163, Australia Water Association Ozwater Conference, Perth Western Australia. 7-10 April . Dallimore CJ, Imberger J and Ishikawa T (2001) Entrainment and turbulence in saline underflow in Lake Ogawara. Journal of Hydraulic Engineering, ASCE, 127(11) 937-948. Fischer HB, List EJ, Koh RCY, Imberger J and Brooks (1979) NH For dilution, the timescale is that for the inflow to reach the dam wall, that is, between one and seven days depending upon the reservoir. Mixing in inland and coastal waters. Academic Press, New York, 483 pp. Hebbert R, Imberger J, Loh I and Patterson JC (1979) Collie River underflow into the Wellington Reservoir. Journal of the Hydraulics Division (ASCE), 105(HY5): 533-545. Imberger J and Patterson JC (1981). A dynamic reservoir simulation model - DYRESM:5. In "Transport Models for Inland and Coastal Waters" H.B. Fischer (ed.) Academic Press, New York, pp310-61. Medema GJ, Schets FM, Teunis PFM, and Havelaar AH (1998) Sedimentation of free and attached Cryptosporidium oocysts and Giardia cysts in water. Applied Environmental Microbiology 64(11): 4460-4466. Romero JR and Imberger J (2003) Effect of a flood underflow on reservoir water quality - data and 3D modeling. Archiv für Hydrobiologie, in press. Walker FR and Stedinger JR (1999) Fate and transport model of Cryptosporidium. Journal of Environmental Engineering, 125(4): 325-333.

38 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

6 NATURAL ORGANIC MATTER DYNAMICS

6.1 Introduction Natural organic matter (NOM) is a complex, heterogeneous residue of decomposing biomass. Natural organic matter, while a fundamental component of aquatic ecosystems, is considered a contaminant for potable water supply as it contributes to the cost of treatment by exerting a coagulant demand, acts as a carcinogen precursor during disinfection and is an aesthetic (visual and taste) problem to customers at high concentrations.

Catchment soil type, vegetation and hydrology determine the delivery and chemical character of NOM entering reservoirs. Within the reservoir, the hydrodynamic processes that impact mixing and transport of contaminants are key determinants of the concentration, degradation and risk associated with NOM (Figure 13). NOM derived from the catchment is called allochthonous and that which is generated within the reservoir from algal growth is autochthonous. Photochemical and microbial degradation can reduce the NOM in a reservoir and impact on water quality and treatment costs.

Catchment-derived waterborne contaminants such as NOM and pathogens pose the greatest challenge to the reservoir during rain event inflows when high concentrations can be transported rapidly to the reservoir.

The distribution of inflows entering a reservoir is controlled by inflow water density relative to that of the lake, the volume of the inflow, the slope at entry and the bottom friction. (See Chapter 5) Warm inflows will be buoyant surface intrusions and cold, dense inflows will intrude as interflows or, in the case of very dense inflows, will sink beneath the lake water where they will flow along the lake bed towards the deepest point. In either case the inflow will entrain water from the lake, increasing its volume, changing its density and diluting the concentration of NOM, pathogens and other constituents. The speed at which an inflow travels through a lake, entrainment with ambient lake water, dilution, and insertion depth are all important in determining the distribution of NOM in lakes and reservoirs. Consequently the depth of intrusion relative to the off-take depth may have a significant impact on the quality of the water extracted from the reservoir and the subsequent treatment costs.

The aim of this study was to investigate the relative behaviour of NOM during transport through a reservoir during storm event inflows and to determine the appropriate depth for extraction of water for treatment to minimise chemicals required to achieve a satisfactory quality for potable supply. Two comprehensive field experiments were conducted to track the passage of storm event inflows through Little Para Reservoir and Myponga Reservoir in South Australia. During the storm event inflows samples were collected from the river and reservoir to determine the concentration of dissolved organic carbon at various sites and depths, and analyses were undertaken to determine alum demand, chlorine demand, disinfection by-product formation and the quality of water pre- and post- treatment.

39 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Climate

Stratification Nutrient Load Inflows

Algal Growth

Autochthonous Allochthonous Load Load

Photochemical Reservoir NOM Microbial

Artificial Off-take Catchment Destratification Management Management

Water Quality

Figure 13 Conceptual representation of the factors affecting carbon input into reservoir, degradation pathways and ultimately the impact on water quality.

6.2 Methods 6.2.1 Site descriptions The inflow characteristics and impact on reservoir water quality was examined during significant rain events at two sites in South Australia (Figure 14). Little Para Reservoir is situated approximately 22 km NE of Adelaide (S 34° 44’ 40”, E 138° 43’ 30”) and covers an area of 1.25 km2 with a maximum capacity of 20,800 ML. The reservoir receives both local catchment runoff and operates as a storage reservoir for water pumped from the River Murray, a major water source for the city of Adelaide. The treatment plant at Little Para Reservoir uses a combination of alum coagulation/flocculation, sedimentation and dual media rapid gravity filtration.

The second experiment was undertaken at Myponga Reservoir, an impounded, flooded river valley located 70 km south of Adelaide, South Australia (S 35° 24’ 13”, E 138° 25’ 29”). The reservoir has a capacity of 26,800 ML and maximum depth of 42 m at full supply level. The mean retention time based on abstraction rates is approximately 3 years and the surface area is 2.8 km2. The catchment is approximately 124 km2 of mixed land use, including pasture for dairy, beef and hay production, with patchy remnant vegetation. Recent estimates of dominant land uses are 62% grazing and 24% dairying (Thomas et al., 1999). The Myponga Treatment plant uses alum in combination with a polymer for coagulation, dissolved air flotation/filtration (DAFF) followed by chlorination.

40 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 14 Location of Little Para and Myponga Reservoirs, South Australia

6.2.2 Sampling Sampling was designed to observe reservoir water quality and hydrodynamics prior to and during a major rain event inflow. The characteristics of the riverine intrusion and the ambient reservoir conditions were measured by regular profiling at several sites along the length of the reservoir with a multi-probe water quality analyser (DataSonde® 4 Multiprobe, Hydrolab®). Strings of thermistors for continuous logging of water temperature were deployed at numerous depths in each reservoir and in the inflowing tributaries to determine the depth of the riverine intrusions (StowAway® Tidbit® at Little Para and RBR TR-1050 at Myponga). Greenspan turbidity sensors (Smart Sensor TS300) were also deployed at both tributaries and at depths; 5 m and 20 m in the reservoir adjacent to the dam wall.

The INFLOW model (Antenucci et al., 2005) (Chapter 5) was used to calculate the travel time, dilution and insertion depth of the riverine intrusion. The model is available on-line at www.cwr.uwa.edu.au. Inflow rate at Little Para Reservoir was determined using daily reservoir balance over the period of the inflow event obtained from South Australian Water Corporation. Reservoir balance was calculated by combining total reservoir change, usage and evaporation rates. Inflow rate at Myponga was calculated from flow rating curves at a V-notch weir. Temperature and salinity of inflows and in the main body of both reservoirs was measured using a multi-probe water quality analyser (DataSonde® 4 Multiprobe, Hydrolab®). The stream bed half angle and stream bed slope for Little Para Reservoir were 37.5 degrees and 0.77 respectively. The stream bed half angle and stream bed slope for Myponga Reservoir were 40 degrees and 0.4 respectively.

Water samples were collected from the tributaries and from the reservoir at the surface and 20 m adjacent to the dam wall of Little Para Reservoir and the sampling timed to capture the rain event inflow from 1-5 August 2004. Similarly, samples were collected from Myponga Creek and in the reservoir near the dam wall on 18 July 2006 to capture a riverine intrusion from a rain event that commenced on 15 July 2006. Water samples were analysed for turbidity, colour and DOC and subjected to jar tests to determine alum demand, chlorine demand and disinfection by-product formation.

6.2.3 Jar Test Jar tests were performed on surface and bottom waters collected from adjacent to the dam wall. A variable speed, six paddle gang stirrer with 7.6 cm diameter flat paddle impellers (FMS6V, SEM Pty Ltd. Brisbane Australia) and Gator jars was used for jar testing. After coagulant dosing and 1 minute of flash mixing at 200 rpm (G = 480 s-1), the speed was reduced to 20 rpm (G = 18 s-1) for 14 minutes. The samples were then allowed to settle for 15 minutes and gravity filtered using 11 μm pore filter papers (Whatman, UK). The turbidity, residual aluminium concentration, colour, dissolved organic carbon (DOC) concentration and apparent molecular weight distribution of dissolved organics was determined in the filtered water. Alum doses range from 20 mg L-1 to 120 mg L-1 and jar tests were

41 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT performed with and without pH control. In the Myponga jar tests, a cationic polymer flocculant aid (Magnafloc LT22, Ciba Specialty Chemicals, Australia) was also added to a final concentration of 0.1%. For experiments with pH control, the pH was targeted at 5, 6 and 7 during coagulation. The amount of acid or alkali required to achieve the target pH was determined by prior pH titration. Additional experiments were performed over an expanded pH range of 3 to 7 using a single alum dose.

DOC concentrations were determined using a total organic carbon analyser (Model 820, Sievers Instruments Inc., USA). UV absorbance at 254 nm (UV254Abs) was measured using a 1 cm quartz cell and a Shimadzu UV-1201 UV/Vis spectrophotometer (Shimadzu Oceania Pty Ltd, Australia). Specific UV absorbance (SUVA) is a parameter which has been shown to represent the degree of molecular conjugation and hence aromaticity of organic material and was derived using the equation (UV254Abs × 100/DOC). Colour was determined spectrophotometrically by comparing the absorbance of a sample at 456 nm with a platinum/cobalt standard (50 Hazen Units) (Bennett and Drikas, 1993). Samples for DOC, UV254Abs and colour analysis were pre-filtered through 0.45 μm membranes. Turbidity was measured using a Hach 2100 AN ratio turbidimeter (Hach, U.S.A). Soluble aluminium concentrations of the samples were determined using a Spectroflame inductively-coupled plasma-atomic emission spectrometer (ICP-AES, Analytical Instruments GmbH, Germany). The samples were acidified to ~1% with concentrated nitric acid (Aristar, BDH) prior to analysis.

Chlorine demand was determined by dosing an appropriate volume of saturated chlorine solution into 1 litre of either raw water, collected directly from the reservoir, or water that had already been processed by jar test, and stored in an amber bottle. Doses were selected to achieve a chlorine residual of 0.5 mg L-1 after 72 hours. At predetermined times 100 mL samples were taken for free chlorine analysis. Samples were incubated at 20°C. Chlorine residual was determined using the N,N-diethyl-p-phenylenediamine (DPD) ferrous titration method (APHA et al., 1998).

The trihalomethane formation potential (THMFP) was determined by reacting a buffered water sample (pH 7.4) at 35°C for 4 hours with an excess of chlorine (approximately 20 mg L-1 chlorine). The sample was subsequently quenched with ascorbic acid. THM components were determined using a gas chromatograph with a headspace autosampler, coupled with an electron capture detector.

The apparent molecular weight of natural organics was measured using High Performance Size Exclusion Chromatography (HPSEC) based on the method described by Chin et al. (1994). HPSEC was performed using a Waters Alliance 2690 separations module with a 996 photodiode array detector using a Shodex KW-802.5 packed column (Shoko Co. Ltd., Japan). The carrier solvent consisted of a 0.02 M phosphate buffer solution (pH 6.8) adjusted to an ionic strength of 0.1 M with sodium chloride at a flow rate of 1 mL min-1. Absorbance was measured at 260 nm. Calibration was performed using polystyrene sulphonate (PSS) standards (Polysciences Inc., U.S.A.) of molecular weights 35,000, 18,000, 8,000 and 4,600 Daltons.

6.3 RESULTS

6.3.1 Little Para Reservoir Four consecutive days of rain on the 1, 2, 3, 4 and 5 August 2006 yielded 12, 7.4, 18.4, 18.2 and 12 mm respectively in the Little Para catchment resulting in a 1 in 10 year inflow of water from Gould Creek and Little Para River into the reservoir.

Prior to the inflow the water temperature of the reservoir was approximately 11.2°C. The colder river water (9.3°C at Gould Creek and 10.23°C at Little Para River) entered the reservoir as an underflow decreasing the temperature at the bottom of the reservoir by approximately 1°C, and increasing the temperature difference between the surface and the bottom of the reservoir near the dam wall from 0.2°C to 0.8°C.

42 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Dam Wall 5m Dam Wall 20m 50 Gould Creek Little Para River 45

20 Turbidity (NTU) Turbidity

0 30-Jul-04 01-Aug-04 03-Aug-04 05-Aug-04 07-Aug-04 09-Aug-04 Date

Figure 15 Turbidity at Little Para Reservoir. The maximum turbidity measurable by the sensors was 50 NTU hence the cut off on the Gould Creek and Little Para turbidity data

The elevated flows increased the turbidity in Gould Creek and Little Para River from a baseline of 16 and 24 NTU to 42 and 93 NTU, respectively (Figure 15). The ambient reservoir turbidity in the bottom of the reservoir increased from 13 NTU, prior to the rain, to 30 NTU when the riverine intrusion reached the dam wall. The tributary water carried elevated concentrations of dissolved organic carbon during the rain event, 10 and 13 mg L-1 in Gould Creek and Little Para respectively compared with the dry weather baseline flow conditions of 5.4 and 5 mg L-1 respectively. The elevated organic carbon concentration was also evident in the true colour measurements which ranged from 5-11 Hazen Units (HU) in the tributaries prior to the rain and increased to 73 and 113 HU during the rain event. Although there was some dilution of the dissolved organic carbon as the riverine intrusion mixed with reservoir water, the DOC increased from 5.0 at the surface of the fully mixed reservoir prior to the inflow event to 9.4 mg L-1 in the samples collected from the bottom of the reservoir at the dam wall (Table 6). The INFLOW model was used to calculate insertion type, inflow travel time until insertion, insertion depth and inflow dilution. The model predicted that this inflow event was an underflow that inserted into the overlying water after 15.41 hours. The insertion depth was predicted to be 24 m and the dilution was 1.6 times. Turbidity sensors showed that this inflow reached the dam wall approximately 24 - 36 hours after it had entered the reservoir (Figure 15).

Jar tests were undertaken to determine the amount of alum required to treat water from the reservoir before, during and after the riverine intrusion. Prior to the rain event the required alum dose was 60 mg L-1. The surface water at the dam wall was minimally impacted by the riverine intrusion and the same amount of alum was required both before and after the rain event to achieve a similar water quality. However, water from the bottom of the reservoir, influenced by the intrusion, required 90 mg L-1 of alum to produce a similar treated water quality.

Following the jar tests, water from the bottom of the reservoir had higher residual colour and DOC than the surface waters. This led to a higher chlorine demand; 4.7 at the bottom of the reservoir versus 2.4 mg L-1 Cl at the surface. The higher residual DOC and chlorine demand inevitably contributed to a higher trihalomethane formation potential (90 versus 66 µg L-1).

43 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Table 6 Jar Test results and water sample analysis for Pre-inflow and Inflow samples collected from Little Para Reservoir at the Dam Wall and at reservoir tributaries (Gould Creek and Little Para River). Alum Turbidity UV True Chlorine Dose (NTU) (/cm) Colour DOC Demand THMFP (mg/L) (HU) (mg/L) (mg/L) (μg/L) Gould Creek 0 16 11 5.4 Pre-Inflow Gould Creek 0 42 73 10 During Inflow Little Para River 0 24 5 5 Pre-Inflow Little Para River 0 93 113 13 During Inflow Reservoir Surface 0 13 0.102 11 5. Pre-Inflow 60 0.07 0.042 2 2.9 2.1 57 Reservoir Surface 0 11.3 0.164 31 6.3 During inflow 60 0.08 0.050 2 3.3 2.4 66 Reservoir 20m depth 0 30.2 0.306 67 9.4 During inflow 90 0.11 0.068 5 4.2 4.7 90

The molecular weight distribution of NOM in the reservoir before and during the inflow event at the dam wall was similar but varied mainly in magnitude, reflecting the change in DOC concentration due to the riverine intrusion. In addition, the samples taken during and after the rain event included detectable quantities of ‘colloidal’ organic material. The evidence for this is the peak at the exclusion limit of the column (50,000 Daltons). This multi-component peak has been reported to be composed of NOM-metal complexes (Allpike et al., 2005) and complex amino sugars from bacterial cell walls and other biological sources (Leenheer, 2004 and Makdissy et al., 2004). The organo-metallic complexes are usually indicative of early diagenesis organic matter in the catchment that has not undergone significant natural photo-oxidation or biodegradation, and are not generally detected following conventional treatment or extended storage.

44 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Stormwater Inflow 5m depth 0.016 Stormwater Inflow 20m depth Pre-inflow 0.014 Post-inflow

0.012

0.010

0.008

0.006

UV Abs at 260nm at Abs UV 0.004

0.002

0.000

100 1000 10000 100000 Apparent Molecular Weight (Da)

Figure 16 Apparent molecular weight of dissolved organics in Little Para Reservoir for samples collected before, during and after the inflow event at the Dam Wall

6.3.2 Myponga Reservoir Myponga experienced a similar rain event inflow to Little Para following an extended dry spell but there were significant contrasts in the inflow character and resulting water quality. Three consecutive days of rain from 15, 16, 17 July 2006 yielded 10, 13.2 and 17.6 mm, respectively. This rainfall resulted in peak flows in Myponga River of 218, 363 and 392 ML day-1, respectively. Due to the temperature difference between the river (9.2°C) and the reservoir water (10.5°C) an underflow developed, observed as a cooling of the bottom waters in the Myponga Reservoir side arm (Figure 17). The reservoir water was slightly cooler than it would normally be at this time of year due to climatic conditions. The mean minimum temperature for June 2006 was 5.6°C, which is 2.5°C below the average June mean minimum of 8.1°C (Australian Bureau of Meteorology). The temperature of Myponga River increased over the period (Figure 17) and the temperature difference between the River and the reservoir was not as great as has occurred historically. Despite this, an intrusion was evident as a cooling of the bottom water in the sidearm of the reservoir (Figure 17). The INFLOW model was used to calculate insertion type, inflow travel time until insertion, insertion depth and inflow dilution. The model predicted that this inflow event was an underflow that would travel along the reservoir thalweg to the dam wall in 47.7 hours diluting 2.25 times.

Under pre-inflow conditions the turbidity in Myponga River was 1.87, colour was 44 HU and DOC was 7.0 mg L-1. During the inflow event when samples were collected on 18 July 2006, turbidity increased to 15.8 NTU, colour was 218 HU and DOC was 20.2 mg L-1 (Table 6). This is reflected in the molecular weight distribution difference between pre-inflow and during inflow which indicated significant increases in the absorbance response across the molecular weight range, but especially in the high MW and colloidal ranges (Figure 17). Results show that this inflow did have an effect on both turbidity and DOC in the main body of the reservoir. Turbidity increased from 0.46 NTU for pre-inflow reservoir

45 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT water to 0.54, 0.85 and 0.65 NTU during the inflow event at the surface and at depths of 15 m and 30 m respectively. DOC also increased from 11.9 mgL-1 in pre-inflow reservoir water to 12.3, 12.3 and 12.7 mgL-1 during the inflow event at the surface and at depths of 15 m and 30 m respectively. However, colour remained stable (62 ± 1 HU) for both pre-inflow and during inflow and UV254Abs was marginally lower at all depths during the inflow compared to the pre-inflow value. The reduction in these water quality parameters is further supported by the HPSEC data which showed a marginal reduction in absorbance response across the molecular range measured (Figure 19).

Jar tests were performed to determine how the Myponga Creek water influenced the relative treatability of the reservoir water and contributed to chemical demand. Table 7 shows the water quality parameters of raw and treated Myponga River water, both before and during the rain event. The regular alum dose used in Myponga treatment plant at the time of the investigation was 91 mg L-1. It was decided to use 90 mg L-1 as the dose rate for comparison purposes. For a 90 mg L-1 dose prior to the rain event, the DOC concentration in the reservoir water was reduced from 11.9 mg L-1 in the raw water to 4.6 mg L-1 after treatment (61% removed), turbidity was reduced from 0.46 to 0.14 and colour from 62 to 4 (Table 7). THMFP was reduced from 374 to 156 µg L-1 (Table 7). Following the rain event, residual DOC after addition of 90 mg L-1 of alum dropped from 12.3 to 7.4 mg L-1 at the surface and from 12.7 to 8.7 mgL-1 at a depth of 30 m which equated to 40% and 31% removal respectively (Table 6). Compared with the initial ‘baseline’ samples, chlorine demand was reduced by between 26 and 36% for untreated samples following the rain event, and between 43 and 64% for samples treated with 90 mg L-1 alum. In both cases, chlorine demand reduction was less apparent in water from the bottom (30 m depth) than water at the surface (Table 7). Despite this, THM formation in untreated samples increased. However, if evaluated in terms of specific THM formation (μg THMs per mg DOC), no significant change is observed. This differs from the reactivity of the coagulated waters. Compared with the initial ‘baseline’ samples, specific THM formation following treatment increased from 13 μg per mg C to over 20 μg per mg C at all sampling depths (Figure 19).

Table 7 Jar Test results and water sample analysis for Pre-inflow and Inflow samples collected from Myponga Reservoir at the Dam Wall and at Myponga River.

Alum Filtered UV True Chlorine Dose Turbidity (/cm) Colour DOC Demand THMFP (mg/L) (NTU) (HU) (mg/L) (mg/L) (μg/L) River 0 1.87 0.258 44 7.0 2.9 259 Pre-Inflow 90 0.13 0.058 2 3.2 0.4 96 River 0 15.8 0.888 218 20.2 5.8 861 During Inflow 90 0.20 0.147 11 7.4 2.2 245 Reservoir Surface 0 0.46 0.428 62 11.9 5.0 374 Pre-Inflow 90 0.14 0.088 4 4.6 1.4 156 Reservoir Surface 0 0.54 0.409 61 12.3 3.2 440 During inflow 90 0.19 0.093 10 7.4 0.5 157 Reservoir 15m depth 0 0.85 0.415 62 12.3 3.6 427 During inflow 90 0.35 0.089 6 6.3 0.5 155 Reservoir 30m depth 0 0.65 0.420 63 12.7 3.7 419 During inflow 90 0.10 0.099 7 8.7 0.8 174

Molecular weight distributions of the treated waters from different depths, shows that the small differences apparent in the untreated waters (Figure 20) translated to treatment results with little difference in absorbance response across the MW range at an alum dose of 90 mg L-1 (Figure 21) i.e. all the high MW colloidal materials were effectively removed.

46 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

9 Temperature (°C)

8 0.1 m 1 m 2 m 7 4 m 6 m 10 m

15 m 14-Jul-06 15-Jul-06 16-Jul-06 17-Jul-06 18-Jul-06 19-Jul-06 Myponga Creek Time

Figure 17 Water temperature measured in Myponga Reservoir at seven depths and at the flow monitoring station in Myponga River

0.045

0.040 Myponga creek post rain event 0.035 Myponga Creek pre rain event

0.030 90 mg/L alum post rain event 91 mg/L alum pre rain event 0.025

0.020

UV Abs @ 260 nm 0.015

0.010

0.005

0.000 100 1000 10000 100000 Apparent Molecular Weight

Figure 18 Molecular weight distribution of Myponga River raw and coagulated before and after a rain event

47 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

50 ) Untreated 90mg/L alum 40

specific THM formation (ug/mg C 0 Before rain event Post event Post event Post event (Surface) (15m depth) (30m depth)

Figure 19 Specific THM formation in Myponga Reservoir before and after a rain event.

0.035

0.030 Myponga Res. surface before rain

0.025 Myponga Res. 30m after rain Myponga Res. 15m after rain 0.020 Myponga Res. surface after rain

0.015 UV Abs @ 260 nm 0.010

0.005

0.000 100 1000 10000 100000 Apparent Molecular Weight

Figure 20 Molecular weight distribution of Myponga Reservoir NOM before and after a rain event

48 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

(a) 0.035 (b) 0.030 Myponga Myponga surface 0.030 0.025 surface before after rain 0.025 rain 0.020 50 mg/L 0.020 alum 0.015 0.015 70 mg/L 90mg/L alum alum 0.010 UV Abs UV nm 260 @ UV AbsUV nm 260 @ 0.010 90 mg/L 0.005 0.005 alum

0.000 0.000 100 1000 10000 100000 100 1000 10000 100000 Apparent Molecular Weight Apparent Molecular Weight

(d) (c) 0.030 Myponga 0.030 Myponga 15m after 30m after 0.025 rain 0.025 rain 50 mg/L 0.020 0.020 50 mg/L alum alum 0.015 70 mg/L 0.015 70 mg/L alum alum 0.010 0.010

UV AbsUV @ 260 nm 90 mg/L AbsUV @ 260 nm 90 mg/L alum 0.005 0.005 alum

0.000 0.000 100 1000 10000 100000 100 1000 10000 100000 Apparent Molecular Weight Apparent Molecular Weight

Figure 21 Molecular weight distribution of Myponga Reservoir coagulated waters NOM at (a) surface before rain event, (b) surface after rain event, (c) 15 m depth after rain event and (d) 30 m depth after rain event.

49 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

6.4 DISCUSSION Reservoirs have often been viewed as barriers to catchment derived contaminants, such as pathogens and NOM. This is because the nominal retention times are often long compared with abstraction volumes. However, riverine intrusions can rapidly travel through reservoirs, transport contaminants and challenge treatment plants within days of the rain event (Romero and Imberger, 2003; Hipsey et al., 2005; Antenucci et al., 2005, Brookes et al., 2005). Two such riverine intrusions which showed slightly different hydrodynamics were presented here. The Little Para inflow was extreme for this climatic region and was rated a one in ten year flow. The Myponga rain event resulted in a flow magnitude that occurs about twice every year. Both inflow events had an impact on water quality but the Myponga inflow did not cause as significant a change, either because the weak stratification caused by the inflow was readily overcome by wind mixing or the already high DOC in Myponga Reservoir could adsorb the diluted DOC concentration in the inflow with minimal effect. The significant change in the concentration of solutes on the rising and falling limbs of the hydrograph are often observed in Mediterranean catchments (Butturini et al. 2006). This feature has been observed in the dissolved organic carbon concentration in Myponga River (Linden 2007). In combination with the differing character of organic matter that is contributed from the smaller subcatchments draining into Myponga Reservoir, the organic carbon concentration dynamics are difficult to predict without an untenable amount of sampling and analysis (Linden 2007).

In Little Para Reservoir, the depth-dependent variability in water quality caused by the inflows had a dramatic effect on the chemicals required to treat the water to a potable standard and an impact on the product water. Large amounts of particulate and dissolved matter entered the reservoir after the large flow events and travelled rapidly as underflows to the dam wall at the bottom off-take level where water was being taken for supply. The high levels of turbidity, colour and DOC in Little Para Reservoir inflow water meant that higher doses of alum and chlorine were required to treat the water to an acceptable level. The higher level of dissolved organic carbon and chlorine in turn produced a higher trihalomethane formation potential.

At the time of the inflow events, water from both reservoirs was being extracted from off-takes sited near the bottom of the reservoir. Little Para would have required a change in the treatment plant to a higher alum dose to adequately remove DOC and to reduce chlorine demand and disinfection by-product formation. The Myponga plant dosing at 91 mg L-1 alum did not need to alter treatment as dramatically. Based on the 2.25 times dilution predicted by the INFLOW model, a dose of 106 mg L-1 alum was predicted for optimum removal of DOC (mEnCo, CRCWQT, Australia). The MW distribution and absorbance response of the NOM determined by HPSEC does not support this prediction, showing instead a marginal reduction in UV absorbance response following the intrusion (Figure 20). This can also be observed in the UV254 Abs and colour which did not vary, despite the increases in DOC measured with the inflow. Chow et al. (2006) have observed that the high molecular weight compounds are readily removed during the coagulation and the lower molecular weights compounds are more recalcitrant to treatment. Ninety (90) mgL-1 was the optimal dose for removal of NOM of all sizes from water from the surface in Myponga Reservoir (Figure 20) whereas an alum dose of 105 mgL-1 removed additional NOM in the range 900-1000 Da in water from 15 and 30 m compared with a 90 mgL-1 alum dose.

Interestingly, the >50,000 Da colloidal peak was evident in all raw water samples flowing from the catchments. These high molecular weight compounds were not observed in the reservoir prior to the inflow in Little Para (Figure 16) but were clearly evident in all Myponga sampling sites (Figure 18 and Figure 19). It has been suggested by Chow et al. (2006) that these fractions are probably derived from vegetation decay in the catchment. Generally, this MW fraction has been observed to increase in surface water storages following rain events and then reduces with extended residence time. In Little Para Reservoir, colloidal NOM was not detected prior to the rain event indicating that the organic material in the reservoir had been subject to degradation, through both biological and physical processes.

In the water sources studied, a change in the off-take level from the bottom of the reservoir to the surface would have avoided the dense, contaminated water and reduced the amount of alum and chlorine required to treat the water. In response to this sampling, the off-take depths are now actively managed in these reservoirs during rain event inflows to avoid extracting water from future riverine intrusions. This work has shown the potential advantages of monitoring and modeling of water quality changes that occur with inflow events and the potential to improve management of water treatment

50 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 plant off-take choices. Within the HACCP approach to drinking water supplies (Deere and Davidson, 1998) the reservoir off-take remains the only critical control point which can be actively monitored and managed to control water quality. Where multiple off-takes are available, water should be extracted from outside of influence from the riverine intrusions to reduce costs and reduce the risk to human health.

6.5 REFERENCES Allpike BP, Heitz A, Joll CA, Kagi RI, Abbt-Braun G, Frimmel FH, Brinkmann T, Her N & Amy G (2005) Size Exclusion Chromatography to Characterize DOC Removal in Drinking Water Treatment. Environmental Science & Technology 39(7), 2334-2342. Antenucci J, Brookes JD & Hipsey M (2005) A simple model for quantifying Cryptosporidium transport, dilution, and potential risk in reservoirs. Journal of the American Water Works Association 97(1), 86-93 APHA. 1998 Standard Methods for the examination of Water and Wastewater. Prepared and published jointly by American Public Health Association, American Water Works Association and Water Environment Federation, Washington DC, USA. Bennett LE & Drikas M (1993) The evaluation of colour in natural waters. Water Research 27(7), 1209-1218. Brookes JD, Davies C, Antenucci J & Hipsey M (2006) Association of Cryptosporidium with bovine faecal particles and implications for risk reduction by settling within water supply reservoirs. Water and Health 4, 87-98 Collins MR, Amy GL & Steelink C (1986) Molecular weight distribution, carboxylic acidity and humic substances content of aquatic organic matter: implications for removal during water treatment. Environmental Science & Technology 20, 1028-1032. Butturini A, Gallart F, Latron J, Vazquez E & Sabater F (2006) Cross-site comparison of variability of DOC and nitrate c-q hysteresis during the autumn-winter period in three Mediterranean headwater streams: A synthetic approach. Biogeochemistry 77(3), 327-349. Chin YP, Aiken GR & O'Loughlin E (1994) Molecular weight, polydispersity, and spectroscopic properties of aquatic humic substances. Environmental Science & Technology 28, 1852–1858. Chow C, Fabris R, Wilkinson K, Fitzgerald F & Drikas M (2006) Characterising NOM to assess treatability. Water, March 36 – 42. E&WS. 1984 Little Para Water Filtration Plant. Designed and produced by Public Relations Branch, Printed by DJ Woolman, Government Printer, South Australia. Deere D & Davison A (1998) Safe drinking water: Are food guidelines the answer? Water, November, 21-24 Fisher I, Kastl G, Sathasivon A, Chen P, van Leeuwen J, Daly R & Holmes M (2004) Tuning the enhanced coagulation process to obtain best chlorine and THM profiles in the distribution system. Water Science & Technology: Water Supply, 4(4), 235-243. Hipsey MR, Brookes JD, Regel R, Antenucci JP, & Burch MD (2005) In situ evidence for the association of Total Coliforms and Escherichia coli with suspended inorganic particles in an Australian Reservoir. Water Air Soil Pollution 170(1), 191-209. Leenheer JA (2004) Comprehensive assessment of precursors, diagenesis, and reactivity to water treatment of dissolved and colloidal organic matter. Water Science & Technology: Water Supply 4(4), 1-9. Linden LG (2007) Impact of destratification on the treat-ability of natural organic matter in drinking water reservoirs. PhD Thesis, School of Earth and Environmental Sciences, University of Adelaide, March, 2007. Linden LG, Lewis DM, Burch MD and Brookes JD (2004) Nutrient load is determined by high flow episodes in the Mediterranean Myponga Reservoir. International Journal of River Basin Management 2(3), 1-11.

51 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Makdissy G, Croué J-P, Amy G & Buisson H (2004) Fouling of a polyethersulfone ultrafiltration membrane by natural organic matter. Water Science & Technology: Water Supply 4(4), 205- 212. Romero JR & Imberger J (2003) Effect of a flood underflow on reservoir water quality - data and three- dimensional modeling. Archiv für Hydrobiologie 157(1), 1-25. Thomas D, Kotz S & Rixon S (1999) Watercourse survey and management recommendations for the Myponga River catchment. Environment Protection Agency, Adelaide, July, 1999.

52 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

7 CYANOBACTERIA

7.1 Introduction

Cyanobacteria are notorious as indicators of degraded aquatic systems. The toxins produced by the cyanobacteria have been responsible for livestock (Saker et al., 1999) and human deaths, albeit under unusual circumstances (Joachimsen et al., 1998). Cyanobacteria are an extremely well adapted group of photo-autotrophic organisms which dominate the freshwater phytoplankton community during stratified conditions. Whilst the cyanobacteria have existed on earth for more than 2.5 billion years (Lau et al., 1980), there is a general opinion that “cyanobacterial blooms” are increasing in frequency due to anthropogenic eutrophication (Hallegraff, 1993) and the regulation of waterways with weirs and dams. The features that cyanobacteria evolved in a resource-poor environment 2.5 billion years ago enable them to exploit a resource-rich environment in lakes with higher nutrient concentrations and more suitable habitat with impounded waterways.

However, recent research has shown that with appropriate treatment most toxins can be effectively removed from drinking water (Drikas et al., 2001), significantly reducing the risk to consumers. Cyanobacteria also produce taste and odour compounds which are not easily removed during conventional water treatment and compromise the quality of potable water. Even low numbers of cyanobacteria can produce concentrations of these compounds which exceed the taste threshold and attract customer complaints. The water industry has been active in seeking methods to control cyanobacteria, however, controlling to a level which ensures no taste and odour problems is difficult.

The aim of this chapter is to describe the conditions which lead to cyanobacterial blooms and from this information identify means of predicting, detecting and managing cyanobacteria in source waters. The site used as a case study was Myponga Reservoir in South Australia (S 35° 24’ 13” E 138° 25’ 13”). The reservoir has a maximum depth of 36 m, a volume of 26,000 ML and a catchment area of 124 km2. Land-use in the catchment is predominantly dairy and beef grazing. The reservoir was the site of a CRC for Water Quality and Treatment study on artificial destratification using a bubble plume aerator and two surface-mounted mechanical mixers which were operated for six months each year. As part of this study two meteorological stations with thermistor chains have been permanently deployed in the reservoir to record physical data at 10 minute intervals (Figure 22).

Figure 22 Myponga Reservoir showing sampling locations, mixer and aerator deployment and sites where meteorological stations are deployed

53 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

7.2 Physical conditions favouring cyanobacteria The success of the cyanobacteria is, in part, attributable to the gas vesicles which provide buoyancy (Walsby, 1994; Oliver, 1994; Walsby et al., 1987). The ability to float into the illuminated surface water in stratified water-bodies provides a distinct advantage over other phytoplankton that rely entirely upon turbulence to remain entrained. By floating upwards the cyanobacteria can significantly increase light capture and consequently increase productivity (Walsby et al., 1997; Walsby, 1997, Mitrovic et al., 2001), nitrogen fixation (Stal-Lucas and Walsby, 1998) and growth (Sherman and Webster, 1994).

The scum-forming cyanobacteria do particularly well when there is a shallow surfaced mixed layer, that is, when stratification persists close to the water surface. An example of rapid Anabaena growth in a shallow surface layer occurred in Myponga Reservoir in January 2000. Myponga Reservoir is generally a well-mixed site and cyanobacterial concentrations are low. However, during summer the physical conditions can become suitable for Anabaena circinalis to grow rapidly. Myponga Reservoir was well-mixed in early January 2000, however, there was significant heating of the surface water between January 7 and January 21 (Figure 23). The result of this was that the diurnal surface layer remained shallow (Figure 24) and cyanobacteria were not entrained deep into the water column.

25 T100mm T3_0m 24 T5_0m T10_0m 23 T20_0m T30_0m 22

20 Temperature (Degrees C)

18 31/12/99 07/01/00 14/01/00 21/01/00 28/01/00 04/02/00 Date

Figure 23 Temperature measurements collected in depth profile with logging thermistors at high frequency in Myponga Reservoir in January 2000

54 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

1/01/00 8/01/00 15/01/00 22/01/00 29/01/00 5/02/00 0.0

-5.0

-10.0

-15.0 Depth (m) Depth -20.0

-25.0

-30.0

Figure 24 Depth of surface mixed layer, defined as the shallowest depth at which the temperature difference between two adjacent thermistors is 0.05°C or greater.

The euphotic depth (depth at which the light intensity is 1% of the surface light intensity) in January was 3.6 m (light attenuation coefficient: kd=1.22) and nutrient concentrations were sufficiently high to support rapid growth rate and high yield: Ammonia – 0.027 mg L-1, Filterable reactive phosphorus – 0.038 mg L-1, Total phosphorus – 0.051 mg L-1, Total Kjeldahl Nitrogen – 0.98 mg L-1, Nitrate and nitrite – 0.131 mgL-1. Anabaena circinalis was not detected on 14 December 1999, and was first recorded later in December, albeit at low numbers (Table 8). As the water column stratified, A. circinalis growth accelerated and by 10 January 2000 the highest recorded concentration was 3,891 cells mL-1. The mean growth rate between 4 January - 10 January was 0.36 day-1 and concentrations were great enough to present a geosmin (taste and odour) threat to the treatment plant.

In this circumstance the A. circinalis population was controlled with a chemical algicide on 11 January 2000, however, in reservoirs where algicides are not used, early warning of cyanobacterial risks can enable treatment, such as activated carbon dosing, to be anticipated.

Table 8 Anabaena circinalis concentrations (cells mL-1) at five locations at Myponga Reservoir. A date with no record (–) signifies A. circinalis not detected in a 1 mL, 10x concentrated sample. Location 21/12/99 29/12/99 4/01/00 10/1/00 18/1/00 25/1/00 1 4 9 43 3891 45 2 4 - - - 2186 2 - 5 29 - - 146 12 - 6 - - 163 1470 30 - 7 - - 459 448 8 -

Although there are two different destratifying systems in Myponga Reservoir, there is still strong persistent stratification in the surface layer as high nocturnal temperatures and low wind speeds inhibit cooling. However, modeling studies have shown that the destratifiers have significantly reduced the period when Anabaena can grow. The phytoplankton biomass at Myponga Reservoir is dominated by green algae and diatoms, which rely on turbulence to remain entrained, and the conditions when cyanobacteria grow is narrowed to a brief period in summer each year (Figure 26).

55 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

7.3 Chemical conditions favouring cyanobacteria Nutrients can be limiting to cyanobacteria in two ways; they can limit the rate of growth and they can limit the maximum biomass or magnitude of the bloom. There is a common perception that cyanobacterial blooms are the direct result of eutrophication, however, the reality is that cyanobacteria can exceed problematic levels at concentrations barely above the detection limit. Concentrations of phosphorus less than 0.01 mgL-1 as filterable reactive phosphorus (FRP) are considered to be growth limiting (Vollenweider and Kerekes, 1980) and 0.1 mgL-1 soluble inorganic nitrogen is considered the minimum concentration to maintain growth during the growing season (Reynolds, 1992). Higher concentrations support rapid growth and higher biomass.

The major nutrient sources are the catchment, the internal load derived from sediment release and atmospheric deposition. In some instances catchments are naturally high in phosphorus and consequently attempts to reduce phosphorus to limiting levels would be unsuccessful. In these cases alternative strategies to control algae should be sought.

The nutrient source that reservoir managers often do have the ability to control is the internal nutrient load or sediment-derived nutrient load. The internal nutrient load is most often controlled by oxygenation of the hypolimnion either by artificial destratification or by direct oxygen injection. In systems where the internal nutrient load contributes significantly to the total nutrient load, a reduction in nutrient release from sediment can significantly decrease the sustainable algal biomass (Sherman et al., 2000).

In a typical phosphorus cycle, phosphorus is remobilised from sediment or decaying organic matter and entrained into the water column, where it is taken up by algae. From there the phosphorus is either passed on to higher levels of the food web or lost to the bottom as the algae sediment. In deep lakes the resolubilisation of phosphorus at the sediment is vertically separated from the algae and so phosphorus can only be accessed with entrainment from the hypolimnion to the epilimnion. In strongly stratified deep lakes this may happen only once or twice a year during significant ‘over-turn’ events. In contrast shallow lakes have the zone of P resolubilisation much closer to the zone of greatest productivity and a single molecule may be recycled a number of times during the growing season (Reynolds, 1997) and thereby sustain a high algal biomass for longer. This is why shallow lakes prove so much more difficult to restore than deep ones (Reynolds, 1997; Sas, 1989)

7.4 Managing cyanobacteria in source water Cyanobacteria are difficult to manage once they have become established in a water body. Often copper based algicides are used, but there are environmental issues concerning the addition of heavy metals to aquatic systems. The most environmentally sound methods to control cyanobacterial growth in reservoir are to manipulate the environment to favour other phytoplankton over the cyanobacteria. Historical management practices have included manipulating the physical, chemical and biological components of the reservoir. Manipulation of catchment nutrient sources takes a number of years before any impact is observed. Manipulation of the biological environment to encourage grazing requires rigorous monitoring and is not applicable in some systems (Boon et al., 1994). On the other hand manipulation of the physical environment is a practical means of changing the phytoplankton habitat and also can manipulate the internal nutrient load. The reduced mixing and turbulence in reservoirs is the central factor in promoting cyanobacterial growth in reservoirs. Consequently artificial mixing is an important option for reducing cyanobacterial growth.

7.5 Artificial destratification During stratification the hypolimnion (the water below the thermocline) is effectively separated from the atmosphere and becomes depleted of oxygen. This leads to reducing conditions at depth and contaminants such as iron, manganese and nutrients are released from sediment.

56 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 equivalent density in the reservoir and an intrusion will then propagate horizontally away from the aerator plume at that depth. As the intrusion moves through the reservoir there is return flow above and below the intrusion and these circulation cells facilitate exchange between the surface layer and the deeper water or hypolimnion (Figure 25). The role of bubble plume aerators is to weaken stratification and work synergistically with wind to mix the reservoir and to oxygenate the hypolimnion. To control contaminant resolubilisation the hypolimnion must receive sufficient oxygen to satisfy the sediment oxygen demand. Artificial destratification has been relatively successful at controlling the release of contaminants from sediments (Brookes et al., 2000) but has been less successful in controlling cyanobacteria (McAuliffe and Rosich, 1990). The inability of destratifiers to mix the stratified surface layers, outside the immediate influence of the plume or mixer, has meant there is still a habitat for buoyant cyanobacteria to exploit (Visser et al., 1996).

Mechanical mixers are usually surface-mounted and pump water downwards through a draft tube. They can also draw water upwards via the tube. Both types of destratifiers have been shown to mix the surface layers very well close to the mixing device but not as effectively outside the immediate influence of the plume and as a consequence there are still often stable zones or habitats for buoyant cyanobacteria to exploit. One mixing approach to consider is to use aerators to generate the large basin-wide circulation cells and use mixers to target the surface stratification outside the direct influence of the aerator plume.

surface

Upper circulation

intrusion

Lower

bottom

raft impeller surface

Flow in Draft tube

intrusion

Buoyant plume

bottom

Figure 25 Flow and circulation fields created by a bubble plume aerator and a surface-mounted mechanical mixer in reservoirs

57 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

7.6 Case study – Myponga Reservoir

7.6.1 The Phytoplankton Community Current management at Myponga Reservoir includes both artificial destratification and chemical algicides to control cyanobacteria. Although there are two different destratifying systems in Myponga Reservoir, there is still strong persistent stratification in the surface layer at particular times when high nocturnal temperatures and low wind speed inhibit cooling (Figure 27). However, modeling studies have shown that the destratifiers have significantly reduced the period over which Anabaena can grow. The phytoplankton community in Myponga Reservoir is dominated by green algae and diatoms, which rely on turbulence to remain entrained, and the conditions when cyanobacteria growth is narrowed to short periods each year (Figure 26).

Figure 26 The relative abundance of the different phytoplankton groups in Myponga Reservoir.

7.6.2 Artificial destratification to control the nutrient load Seasonal temperature stratification was evident at Myponga Reservoir during summer from 1984 until 1994. Since installation of the aerator in 1994, close to isothermal conditions have been maintained at the sampling site (Figure 27). However, surface layer heating is evident at other sites in the reservoir outside of the immediate bubble plume which is consistent with other reservoirs where bubble plume aerators are operating (Visser et al., 1996; Sherman et al., 2000). Dissolved oxygen concentrations were below 4 mg L-1 for extended periods during 1992/93 and 1993/94, which provided conditions suitable for contaminant resolubilisation. Since aerator operation in 1994 the dissolved oxygen concentration at 30 m has been maintained above 4 mg L-1.

Prior to 1994 the concentration of filterable reactive phosphorus (FRP) at 30 m depth was consistently higher than the surface concentrations during summer and autumn (Figure 28). This coincides with the periods of extreme temperature stratification and low dissolved oxygen in the hypolimnion. Filterable reactive phosphorus at 30 m depth reached a maximum concentration of 0.259 mg L-1 in April 1986. The vertical gradient in FRP concentration has decreased since deployment of the bubble plume aerator and the large flux events have been eliminated.

58 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 27 Temperature measured weekly at the surface, 10 m, 20 m and 30 m depth adjacent to the off-take point at Myponga Reservoir. The aerator was installed in 1994.

Surface 0.3 30m 0.25 Surface loc 4 cont 30m loc 4 cont 0.2

0.15

FRP (mg/L) FRP 0.1

0.05

0 01-Jun-84 01-Jun-85 01-Jun-86 01-Jun-87 31-May-88 31-May-89 31-May-90 31-May-91 30-May-92 30-May-93 30-May-94 30-May-95 29-May-96 29-May-97 29-May-98 29-May-99 28-May-00 28-May-01

Date

7.6.3 Relating nutrients to algal biomass In Myponga Reservoir the nutrient loading from the catchment occurs predominantly during winter and early spring. The nutrient pool is not utilised immediately as phytoplankton growth is limited by cool water temperatures and grazing pressure. As water temperature increases the phytoplankton grow rapidly and chlorophyll a concentration increases with an associated decrease in FRP (Figure 29). FRP concentrations decrease to below the detection limit (0.005 mgL-1) and the chlorophyll decreases some time later and the seasonal cycle is repeated.

59 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

0.08 0.03

0.07 ) 0.025 -1

) 0.06 -1 0.02 Surface 0.05 10m 0.04 0.015 20m 0.03 0.01 FRP (mg L 30m 0.02 0.005 Chla 0.01 Chlorophyll a (mg L 0 0 24-Jul-98 19-Jul-99 13-Jul-00 08-Jul-01 20-Jan-99 15-Jan-00 09-Jan-01 21-Mar-99 15-Mar-00 10-Mar-01 22-Sep-98 17-Sep-99 11-Sep-00 21-Nov-98 16-Nov-99 10-Nov-00 20-May-99 14-May-00 09-May-01 Time

Figure 29 Filterable reactive phosphorus at four depths at Location 4 and chlorophyll concentration integrated over the top 5 m.

With the internal nutrient load largely controlled in Myponga Reservoir by the aeration system, the catchment is the dominant source of nutrients. In Myponga Reservoir two tributaries contribute the majority of the nutrient load, but loading is both seasonally and inter-annually variable.

Rainfall in the Australian context is highly variable in space and time with significant seasonal and inter-annual variability. Inter-annual variability in rainfall patterns over eastern Australia is strongly influenced by “El Niño” events, which are characterised by sustained warming over a large part of central and eastern Pacific Ocean, and low rainfall on land. El Niño events are opposed by “La Niña” events that show essentially the opposite patterns of sea surface temperature and bring higher than average rainfall (Allan et al., 1996). Nicholls & Kariko (1993) concluded that the El Niño-Southern Oscillation (ENSO) mostly affects the frequency and intensity of rainfall in Australia and exerts less influence upon the length of events.

Years of high mean annual Southern Oscillation Index (SOI) coincide with years of high flow volume in the Myponga catchment (Figure 30). Deviations from the trend are explained by variation within years. For example, in 1999 the mean annual SOI was influenced by strong positive values in January and December but were otherwise low, resulting in low inflows. In 1992 the monthly mean SOI in May, August and September was slightly positive and all three months received higher than average rainfall. The low inflows found in 1988 were possibly due to the strong negative SOI values in the first half of the year remnant from the strong negative SOI values in 1987. The years 1984 and 1985 also had a number of winter months with negative SOI values.

We have seen that high flow results in high TP loads and reservoir concentrations. The result of a high maximum TP concentration in Myponga Reservoir results in a high chlorophyll a concentration (Figure 31). This figure shows the relationship between the maximum annual TP concentration and the maximum chlorophyll a found in the following growth period in the years between 1985 and 2000. Two outlier years, 1988 and 1993, are excluded from the regression in Figure 31. 1988 was unusual in that it had early rains and consequently there was 6 months between the TP and Chl a maximum. In 1993, hypolimnetic anoxia caused by thermal stratification, released higher than usual FRP concentrations from the sediments, sustaining high algal biomass and resulted in a high maximum chlorophyll a concentration. The operation of the bubble plume aeration system since 1994 has most likely prevented this situation from recurring (Brookes et al., 2000).

60 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

30000 15

Total inflow 25000 10 SOI

20000 5

15000 0 SOI

Total Inflow (ML) 10000 -5

5000 -10

0 -15 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Year

Figure 30 Relationship between annual mean Southern Oscillation Index and flow into Myponga Reservoir.

50 )

-1 outlier 1993 y = 195.57x + 1.7132 g L 2 μ 40 R = 0.7818

30 concentration (

a 20

outlier 1988 10 Maximum Chl

0 0 0.05 0.1 0.15 0.2 0.25 Maximum TP concentration (mg L-1)

Figure 31 Relationship between maximum total phosphorus and maximum chlorophyll a in the following growing period.

7.7 Destratification and control of cyanobacterial growth Because weather and limnological conditions are never constant it is difficult to determine whether destratification has an impact on cyanobacterial growth without very extensive historical data sets. An alternative approach is the use of numerical models to simulate the hydrodynamics and cyanobacterial growth. DYRESM-CAEDYM is a coupled hydrodynamic, water quality and algal growth model available as free-ware from the Centre for Water Research, The University of Western Australia (http://www.cwr.uwa.edu.au/). The modeling approach has been used in these studies to evaluate destratification in Myponga Reservoir.

61 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Meteorological variables measured at the stations on the reservoir were used for model inputs. Algal growth was simulated using equations describing nutrient and light-limited growth of Anabaena circinalis and floating velocity. The model was first calibrated and then used to evaluate the following scenarios: • No artificial intervention

• Aerator and surface mixers with no CuSO4 dosing • Aerator only • Surface Mixers at measured flow rate (3.5 m3 s-1) • Surface mixers at design flow rate (5 m3 s-1) • Surface mixers at increased flow rate (8 m3 s-1) • Intermittent operation • Equivalent aerator energy input using surface mixers

The observed phytoplankton data were obtained from the monitoring program (integrated samples from the top 5 m of the water column). Subsequently the simulated data presented in the same integrated format, i.e. simulated daily concentrations are averaged over the top 5 m. The comparison of measured and simulated total Chl a concentration for the simulated period, with a R2 and P-value of 0.75 and 3E-09 respectively, is shown in Figure 32. The correlation was the best achieved using the calibration method described above. The R2 and P-value obtained indicated a reasonably strong correlation between observed and simulated data, which was also supported by visual inspection.

Maintaining the calibration parameters within a reasonable range of their published values and accurately reproducing the total Chl a concentration for the modeled period proved to be difficult. All of the parameters were found to be sensitive and slight changes in their values produced dramatic changes in the simulation.

30 Aerator and surface mixers 25 )

-1 20 gL μ

( 15 a

Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00 CuSO dosing Date 4 - - - - - - - - - Observed Simulated Figure 32 A comparison of observed and simulated total Chl a concentration (μg Chl a L-1), with simulated CuSO4 dosing on 11 and 12 January 2000, surface mixers and aerator operation between 1 October 1999 and 1 April 2000. The R2 and P-value for the comparison were 0.75 and 3E-09, respectively.

The model over-predicted the total Chl a concentration for the simulated period, which could not be rectified without completely inhibiting the growth of the cyanobacteria and diatoms due to the dominance of the greens. The artificial mixed conditions at Myponga Reservoir and the dominance of the green algae, resulted in the minimal growth of cyanobacteria, with the maximum concentration of Anabaena circinalis peaking at ~1.2 μg Chl a L-1 (~1,600 cells mL-1). Phytoplankton concentrations at this low level are inherently very difficult to simulate, especially when other simulated groups dominate and grow to concentrations that are orders of magnitude greater.

62 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Subsequently the model output was accepted as a reasonable representation of the phytoplankton assemblage observed in Myponga Reservoir under artificially mixed conditions, which included CuSO4 dosing.

The individual phytoplankton groups were analysed using the measured and simulated total chlorophyll data. The total Chl a concentration for the three simulated phytoplankton groups represented the entire phytoplankton assemblage in Myponga Reservoir, i.e. the sum of the individual Chl a concentration for the three groups was equal to the total Chl a concentration of the phytoplankton assemblage. Using individual and total Chl a concentrations the percentage contribution for each group was determined.

The observed phytoplankton community at Myponga Reservoir was dominated by green algae (96.3% of the total biomass measured as Chl a) for the entire monitoring period and was accurately reproduced in the model simulation with greens dominating 96.6% of the total biomass (Figure 33). The relationship between observed and simulated data was a R2 and P-value of 0.73 and 4E-09, respectively.

The comparison between observed and simulated Anabaena circinalis chlorophyll concentration is shown in Figure 34. The observed peak 1.2 μg Chl a L-1 (~1,600 cells mL-1) which occurred on the 10 January 2000, was not reproduced in the simulation as a result of simulated copper sulphate dosing and due to a low cell concentration. Although, the observed concentration represented 0.5% of the total biomass, the simulated concentration represented 0.8% of the total biomass. The correlation between observed and simulated data was a R2 and P-value of 0.55 and 0.009, respectively.

30 Aerator and surface mixers 25 )

-1 20 gL μ

( 15 a

Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00 CuSO dosing Date 4 - - - - - - - - - Observed Simulated

Figure 33 Observed and simulated Scenedesmus concentration (μg Chl a L-1), with a R2 and P-value of 0.73 and 4E-09, respectively.

63 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

1.4 Aerator and surface mixers 1.2

1 ) -1

gL 0.8 μ (

a 0.6

Chl 0.4

0.2

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00 CuSO dosing Date 4 - - - - - - - - - Observed Simulated

Figure 34 Comparison of the observed and simulated Anabaena circinalis concentration (μg Chl a L-1), with a R2 and P-value of 0.55 and 0.009, respectively.

The observed growth of Nitzschia sp. was not accurately reproduced in the simulation (Figure 35). In particular, the model did not capture the occurrence of a single peak in late March to early April. However, the observed diatom concentration represented 3.2% of the phytoplankton biomass, while the model output simulated 2.7% of the total biomass (measured as Chl a). Modeling also showed that Nitzschia sp. growth would persist at a low concentration.

2.0 Aerator and surface mixers

1.5 ) -1 gL μ

( 1.0 a Chl 0.5

0.0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00 Date CuSO4 dosing - - - - - - - - - Observed Simulated

Figure 35 Observed and simulated Nitzschia sp. concentration (μg Chl a L-1).

To further validate the DYRESM-CAEDYM simulation of the phytoplankton community, the modeled period was extended from September 2000 to March 2001. The observed and simulated total Chl a concentrations for the extended period are shown in Figure 36. The simulated biomass captures the timing of the summer peak that was observed in the field data, but did not simulate the unseasonal peak that occurred in December 2000. This peak was attributed to the excessive growth of Chroomonas sp., a species which was not included in the model. The simulated growth of Anabaena circinalis from September-2000 to March-2001 produced a reasonable match with the observed field

64 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 data (Figure 37), although the simulated growth started earlier in the season than observed in the field. 30 Aerator and surface mixers 25

) 20 -1 gL μ

( 15 a

Chl Chl 10

0 01-Sep-00 01-Nov-00 01-Jan-01 01-Mar-01 Date CuSO4 dosing

- - - - - - - - - Observed Simulated

Figure 36 A comparison of observed and simulated total Chl a concentration (μg Chl a L-1), with simulated CuSO4 dosing on 31-January-2000, and surface mixers and aerator operating between the 1-October-2000 and 28-February-2001.

0.5 Aerator and surface mixers 0.4 ) -1 0.3 gL μ ( a 0.2 Chl Chl

0.1

0.0 01-Sep-00 01-Nov-00 01-Jan-01 01-Mar-01

Date CuSO4 dosing

- - - - - - - - - Observed Simulated

Figure 37 Comparison between the observed and simulated Anabaena circinalis concentration (μg Chl a L-1) from 1-September-2000 to 1-March-2001.

The simulation of the 3 types of phytoplankton that were representative of the assemblage in Myponga Reservoir from September 1999 to March2001 produced reasonable results considering the limitations of the model. The observed phytoplankton community consisted of more than the three species simulated in this model. Other species will dominate with changes in nutrients, light and temperature as highlighted by the excessive growth of Chroomonas. An improvement to the CAEDYM model would be to increase the number of species simulated, although this would require intensive calibration and a trial and error approach as used in this study would be insufficient. An alternative method that will be explored in the future will be to use non-linear parameter-fitting software such as NLFIT (Kuczera and Parent, 1998).

65 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

7.8 Simulation of various management strategies The CAEDYM model output compared with observed field data gave a reasonable representation of phytoplankton biomass (as total Chl a) for three species in Myponga Reservoir. The comparison between observed and simulated for Scenedesmus showed a strong correlation whereas a moderate correlation was observed with Anabaena circinalis. The next step involved using the model to determine the individual and combined impact of the surface mixers and the aerator for destratification and control of cyanobacteria. The following strategies were investigated for their ability to maintain DO greater than 4 mg L-1 and to limit Anabaena circinalis below 2,000 cells mL-1. • No artificial intervention

7.8.1 No artificial intervention (Strategy 1)

The simulation of Myponga Reservoir with no artificial mixing for the period of 1 September 1999 to 1 September 2000 indicated that stratification would persistent for several months (Figure 38). The stratification would cause DO levels in the hypolimnion to decrease below 4 mg L-1 for approximately six months from mid-spring to mid-autumn.

Figure 38 Simulated thermal structure and DO concentration for Myponga Reservoir with no artificial mixing.

Simulated Anabaena circinalis growth showed that the number of cells would reach 4,400 cells mL-1 in mid-April and have a mean concentration of 1,000 cells mL-1 (Figure 39). These simulation results highlight the potential degradation of water quality that could occur without the use of artificial mixing. However, it is worth noting that if green algae were to dominate earlier in the summer season, nutrient concentrations may decrease, potentially leading to nutrient-limited growth of other species including cyanobacteria later in the season.

66 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

3.5

2.5 ) -1

gL 2 μ (

a 1.5

Chl 1

0.5

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 39 Simulated Anabaena circinalis concentration (μg Chl a L-1) with no artificial mixing compared with the observed data under normal operating conditions.

7.8.2 Artificial Mixing with no CuSO4 dosing (Strategy 2)

The use of artificial mixing with no CuSO4 dosing in the simulation resulted in a slightly increased Scenedesmus biomass and a decreased correlation with the observed data (Figure 40). The simulated single summer bloom occurred at the correct time compared with observed data. The influence of artificial mixing with no CuSO4 on Anabaena circinalis is shown in Figure 41. The model simulation indicates that cell numbers would be approximately 300 cells mL-1 for about 6 months and the maximum concentration would reach 1,100 cells mL-1 or 0.78 mg Chl a L-1.

35 Aerator and surface mixers 30

25 ) -1

gL 20 μ ( a 15 Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00 Date - - - - - - - - - Observed Simulated

-1 Figure 40 Simulated Scenedesmus concentration (μg Chl a L ) with no CuSO4 dosing compared with the observed data under normal operating conditions, R2 = 0.72.

67 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

1.4 Aerator and surface mixers 1.2

1 ) -1

gL 0.8 μ (

a 0.6 Chl 0.4

0.2

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00 Date - - - - - - - - - ObservedOb d Si l d Simulated

-1 Figure 41 Simulated Anabaena circinalis concentration (μg Chl a L ) with no CuSO4 dosing compared with the observed data under normal operating conditions.

7.8.3 Aerator only (Strategy 3) The DYRESM-CAEDYM model was run with only the aerator algorithm, which was operational between 1 October 1999 and 1 April 2000. The temperature profile and DO profiles indicate that mixed conditions would be maintained for the majority of the simulated period (Figure 42), with stratification limited to a couple of weeks during late February. The DO concentrations would remain greater than 4 mg L-1; although some DO depletion is evident towards late April at the sediment surface. The simulated growth of Scenedesmus would produce a lower total biomass (92.9%) without the combined use of the surface mixers, and the timing of growth was unchanged compared with growth under normal operating conditions (Figure 43). The growth of Anabaena circinalis would increase substantially and accounted for 4.0% of the total biomass (Figure 44). A maximum peak of 1.0 μg Chl a L-1 or 1,400 cells mL-1 occurred in mid-April. As with the previous strategies, sustained growth of Anabaena was maintained, but with a higher mean concentration of ~450 cells mL-1.

Figure 42 Temperature and DO profiles for the simulated period using the aerator only (period when aerator operating is marked with a solid black line).

68 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

35 Aerator 30

) 25 -1

gL 20 μ (

a 15

Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00 Date - - - - - - - - - Observed Simulated

Figure 43 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the aerator only compared with the observed data under normal operating conditions.

1.4

1.2

1 ) -1

gL 0.8 (

a 0.6 Chl 0.4

0.2

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 44 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the aerator only compared with the observed data under normal operating conditions.

7.8.4 Surface Mixers (Strategy 4) The operation of the two surface mixers alone at their actual measured flow rate of 3.5 m3 s-1 was not capable of maintaining fully mixed conditions (Figure 45). Significant stratification occurred during the summer months, which was reflected in the DO profile with concentrations falling below 3.0 mg L-1 at the sediment surface. DO at this level could cause poor water quality conditions.

69 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Figure 45 Temperature and DO profiles for the simulated period using two surface mixers at their measured flow rate of 3.5 m3 s-1. Note the period when surface mixers were operational is marked with a solid black line.

The timing and magnitude of the simulated growth of Scenedesmus was significantly different to the observed data (Figure 46). Scenedesmus grew earlier in the summer period, peaking at the beginning of December 1999 which was similar to the simulated growth with no mixing (strategy 1). Anabaena circinalis growth also started earlier than that in the previous scenarios, with persistent and significant growth occurring early autumn (Figure 47). The maximum concentration of Anabaena circinalis was ~2,400 cells mL-1, with a mean concentration of ~680 cells mL-1. These results indicate that the sole use of the surface mixers operating at 3.5 m3 s-1 will not limit the growth of Anabaena circinalis and associated geosmin production could become a problem in the water supply.

30 Surface mixers 25 )

-1 20 gL μ

( 15 a

Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 46 Simulated green Chl a concentration (μg Chl a L-1) with the use of the surface mixers operating at 3.5 m3 s-1 each compared with the observed data under normal operating conditions.

70 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

1.8 Surface mixers 1.6 1.4 )

-1 1.2

gL 1 μ (

a 0.8

Chl 0.6 0.4 0.2 0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 47 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the surface mixers operating at 3.5 m3 s-1 compared with the observed data under normal operating conditions.

7.8.5 Surface mixers at 5 m3s-1 (Strategy 5) Increasing the flow rate of the surface mixers to 5 m3s-1 improved their destratification ability and decreased the temperature gradient through the water column (Figure 48). DO was maintained at levels greater than 4 mg L-1 throughout the simulated period. As with the use of the surface mixers alone operating at 3.5 m3 s-1, the growth of Scenedesmus occurred earlier in the season with similar magnitude (Figure 49). Scenedesmus contributed 95.3% of the total biomass (as Chl a). When the two surface mixers operated at 5 m3s-1, Anabaena circinalis peaked at a cell concentration of ~1500 cells mL-1 or 1.12 μg Chl a L-1 in mid-April 2000, and persisted with a mean concentration of ~480 cells mL-1 (Figure 50). The sustained growth of cyanobacteria would trigger operator intervention with algicides to maintain water quality.

Figure 48 Temperature and DO profiles for the simulated period using the surface mixers at 5-m3s-1 (period when surface mixers operating is marked with a solid black line).

71 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

30 Surface mixers 25 )

-1 20 gL μ

( 15 a

Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 49 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the surface mixers operating at 5-m3s-1 compared with the observed data under normal operating conditions.

1.4 Surface mixers 1.2

1 ) -1

gL 0.8 μ (

a 0.6

Chl 0.4

0.2

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 50 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the surface mixers operating at 5-m3 s-1 compared with the observed data under normal operating conditions.

7.8.6 Surface mixers at 8 m3s-1 (Strategy 6) A further increase in surface mixer flow rate 8 m3 s-1 resulted in a decrease in stratification and a increase in DO levels above 4 mg L-1 throughout the water column (Figure 51). However, the timing and magnitude of Scenedesmus growth was similar to when the surface mixers were operated at the lower flow rates (Strategies 4 and 5), with growth occurring in late spring (Figure 52). Scenedesmus also maintained their dominance contributing to 96.4% of the total biomass. The timing of Anabaena circinalis growth was also similar to strategies 4 and 5, however the mean concentration was significantly reduced to ~330 cells mL-1, with a maximum peak of ~1,000 cells mL-1 or 0.73 μg Chl a L-1 occurring mid-April (Figure 53). With the sole use of the surface mixers running at 8 m3 s-1, the growth of Anabaena circinalis was maintained at manageable levels and additional intervention involving CuSO4 could be avoided.

72 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 51 Temperature and DO profiles for the simulated period using the surface mixers at 8-m3 s-1. Note: period when surface mixers were operational is marked with a solid black line.

30 Surface mixers 25 )

-1 20 gL μ

( 15 a

Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 52 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the surface mixers operating at 8-m3 s-1 compared with the observed data under normal operating conditions.

73 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

1.4 Surface mixers 1.2

) 1 -1

gL 0.8 μ (

a 0.6

Chl 0.4

0.2

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 53 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the surface mixers operating at 8-m3 s-1 compared with the observed data under normal operating conditions.

7.8.7 Intermittent operation (Strategy 7) Intermittent artificial mixing has been shown to be effective in managing phytoplankton by preventing any one group of species becoming dominant (Reynolds et al., 1994). To test a similar scenario for Myponga Reservoir the aerator and surface mixers (at 3.5 m3 s-1) were operated intermittently in the DYRESM-CAEDYM model for the period 1 December 1999 to 19 April 2000. This period of operation was initially based on the corresponding dates of anoxic periods, but was further refined to reduce the number of days that artificial mixing would be required. The surface mixers and aerators were switched on for an arbitrary period of 2 days every 4 days throughout the operational period.

The temperature profile in Figure 54 shows a similar trend as when the surface mixers and aerators were run continuously and the DO concentration was adequately maintained throughout the modeled period. The growth of Scenedesmus occurred early in December, which was similar to the results that were produced when the surface mixers used alone at various flow rates was simulated (Figure 55). The growth of Anabaena circinalis also occurred at a similar time to when the surface mixers were operated alone (Figure 56). However, with intermittent mixing, the magnitude of Anabaena circinalis growth was significantly reduced with a maximum peak of ~670 cells mL-1 or 0.48 μg Chl a L-1occurring in mid April and a mean concentration of 235 cells mL-1.

Simulating intermittent operation of the surface mixers and aerator shows potential as a management strategy aimed at reducing costs, as isothermal conditions were maintained and Anabaena circinalis concentrations were kept below 2,000 cells mL-1.

74 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 54 Temperature and DO profiles for the simulated period using intermittent mixing using both the surface mixers at 3.5-m3 s-1 and the aerator. The mixing devices operate intermittently (2 days on, 4 days off) throughout the period marked with a solid black line.

30 Aerator and surface mixers

) 20 -1 gL η

( 15 a

Chl 10

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 55 Simulated Scenedesmus concentration (μg Chl a L-1) with the use of the intermittent artificial mixing compared with the observed data under normal operating conditions.

75 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

1.4 Aerator and surface mixers 1.2

1 ) -1

gL 0.8 (

a 0.6 Chl 0.4

0.2

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 56 Simulated Anabaena circinalis concentration (μg Chl a L-1) with the use of the intermittent artificial mixing compared with the observed data under normal operating conditions.

7.8.8 Equivalent aerator energy input using surface mixers (Strategy 8) The final strategy that was investigated was related to the energy requirements of the bubble plume aerator (100 KW) versus the surface mixers (4 KW). Based upon energy consumption, 25 surface mixers (3.5 m3 s-1) equate to one bubble plume aerator. The DYRESM-CAEDYM was run with 25 surface mixers which resulted in fully mixed conditions and DO concentrations above 4 mg L-1 (Figure 57). Using this strategy, the growth of Scenedesmus started early in November and dominated the biomass with a 98.3% contribution (Figure 58). The growth of Anabaena circinalis was almost insignificant, but persisted all year with a mean concentration of ~80 cells mL-1 or 0.06 μg Chl a L-1 and a maximum concentration of ~150 cells mL-1 or 0.11 μg Chl a L-1 (Figure 59).

Figure 57 Simulated thermal structure and DO concentration for Myponga Reservoir with 25 surface mixers.

76 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

25 Surface mixers

20 ) -1 15 gL μ ( a 10 Chl

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 58 Simulated Scenedesmus concentration (μg Chl a L-1) with 25 surface mixers compared with the observed data under normal operating conditions.

1.4 Surface mixers 1.2

1 ) -1

gL 0.8 μ (

a 0.6

Chl 0.4

0.2

0 01-Sep-99 01-Dec-99 01-Mar-00 01-Jun-00 01-Sep-00

Date - - - - - - - - - Observed Simulated

Figure 59 Simulated Anabaena circinalis concentration (μg Chl a L-1) with 100 kW of equivalent surface mixers compared with the observed data under normal operating conditions.

The effectiveness of the various operational strategies used to limit the growth of Anabaena circinalis and maintain DO concentration in the water column is summarised in Table 9. The simulation employed for validation, including the surface mixer, bubble plume aerator and CuSO4 dosing algorithms, produced similar results to the observed field data. If no artificial mixing or CuSO4 dosing were employed, excessive growth of Anabaena circinalis would occur and permanent stratification would lead to the presence of anoxic conditions. The use of the aerator without CuSO4 dosing adequately maintained well-mixed conditions and DO throughout the water column. However, the growth of Anabaena circinalis could exceed 1,000 cells mL-1 (for a total of 16 days) but would not reach the threshold of 2,000 cells mL-1.

When the aerator is coupled with the surface mixers (at 3.5 m3 s-1), the growth of Anabaena circinalis was further reduced with the peak concentration falling from ~1,400 cells mL-1 to ~1,000 cells mL-1.

77 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

The current operation of the surface mixers (3.5 m3 s-1) being used alone would not be able to destratify the water column and maintain DO at acceptable levels, and importantly the growth of Anabaena circinalis would exceed 2,000 cells mL-1. Increasing the flow rates of the surface mixers improves their destratification ability and reduces the growth of Anabaena circinalis. With a surface mixer flow-rate of 8 m3s-1, optimal results were achieved maintaining DO above 4 mg L-1 and limiting the maximum concentration of Anabaena circinalis to ~1,000 cells mL-1.

Using intermittent mixing, the growth of cyanobacteria was restricted to a maximum concentration of -1 ~700 cells mL and well-mixed conditions were maintained. The use of CuSO4 dosing would not be required under this strategy and operational costs would be lower due to the reduced use of the aerator and surface mixers. The use of 25 surface mixers, using the same energy as the existing aerator, adequately destratified Myponga Reservoir and almost completely inhibited the growth of Anabaena circinalis.

Table 9 Results from existing and simulated water quality management strategies.

Maximum Artificial mixing cyanophyte Days above Minimum DO Simulated phytoplankton assembly composition operation concentration 1000 cells.mL-1 (mgL-1) -1 (cells.mL ) Chlorophytes Cyanophytes Diatoms Existing - Field 1625 1 ~5.00 96.30% 0.50% 3.20% Existing - Sim 278 0 4.70 96.60% 0.70% 2.70% Strategy 1 4444 196 1.00 91.30% 6.80% 1.90% Strategy 2 1069 3 4.70 94.10% 2.90% 3.00% Strategy 3 1389 16 4.70 92.90% 4.00% 3.10% Strategy 4 2361 133 1.20 93.90% 4.70% 1.40% Strategy 5 1556 21 4.70 95.30% 3.40% 1.30% Strategy 6 1014 1 4.70 96.40% 2.40% 1.20% Strategy 7 667 0 4.70 97.10% 1.70% 1.20% Strategy 8 153 196 4.70 98.30% 0.60% 1.10%

The addition of the surface mixer and CuSO4 dosing algorithms to DYRESM-CAEDYM enabled the phytoplankton succession and DO concentration to be adequately simulated and validated against observed field data for the period 1 September 1999 to 1 September 2000. This enabled various management strategies to be investigated. Modeling showed that the potential for growth of Anabaena circinalis would occur during periods of thermal stratification and with the presence of a shallow surface mixed layer. This coincided with oxygen depletion in the hypolimnion and adequate levels of -1 -1 nutrients (FRP> 0.01 mg L and NOx > 0.1 mg L ).

The actual mixing program with an aerator at Myponga Reservoir adequately maintains DO throughout the water column, and coupled with CuSO4 dosing, limits the growth of Anabaena circinalis to a maximum concentration of ~1,600 cells mL-1 or 1.17 μg Chl a L-1 (0.5% of the total biomass as Chl a). The simulation of the existing aerator, surface mixers and CuSO4 dosing produced similar results, affirming the need for intervention to maintain manageable levels of cyanobacteria and DO concentrations. The simulation showed that when the surface mixers and aerator are used without CuSO4 dosing (strategy 2) the Anabaena circinalis would not exceed concentrations that would be of concern for water supply. The sole use of the surface mixers was found to be adequate at maintaining water quality if the flow rate could be increased to 8 m3s-1. However, at their current flow rate (3.5 m3 s-1) they are unable to fully destratify Myponga Reservoir and limit the growth of Anabaena circinalis to below 2,000 cells mL-1.

The use of intermittent artificial mixing would reduce operational costs as the aerator and surface mixers would run at 50% less than the current operational schedule. Using this technique, destratified conditions are maintained, DO concentrations are kept high and the growth of Anabaena circinalis is minimal and importantly, the use of CuSO4 dosing is not necessary. Under the current operating conditions, the simulation demonstrated that the use of CuSO4 dosing is not necessary, as Anabaena circinalis concentrations did not exceed 2,000 cells mL-1. As demonstrated with DYRESM-CAEDYM, the current nutrient concentrations, light climate, meteorological forcing and artificial mixing operations

78 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 at Myponga Reservoir do not favour the excessive growth of Anabaena circinalis. However, even at these concentrations taste and odours can be problematic and require additional treatment.

The successful simulation of Myponga Reservoir with the DYRESM-CAEDYM model demonstrates that modeling is an effective tool to assess and optimise water quality management strategies. Upon successful simulation of the limnological processes in a water body, any myriad of strategies can be investigated. The modeler can alter meteorological conditions, nutrient loadings, timing and use of artificial mixing devices to gain insight into the management of the water body. The addition of the surface mixer and CuSO4 dosing algorithms has successfully been incorporated into the DYRESM- CAEDYM model and has broadened its use.

More detailed information on the study to evaluate the use of artificial mixing to control the growth of cyanobacteria is contained a separate CRC report (RR 59: Brookes et al, 2008)

7.9 References Allan RJ, Lindesay J & Parker D (1996) 'El Niño Southern Oscillation & Climatic Variability' (CSIRO publishing: Collingwood, Victoria, Australia) Boon PI, Bunn SE, Green JD and Shiel RJ (1994) Consumption of cyanobacteria by freshwater zooplankton: implications for the success of “Top-down” control of cyanobacterial blooms in Australia. Australian Journal of Marine and Freshwater Research. 45: 875-887. Brookes JD, Burch MD and Tarrant P. (2000) Artificial destratification: Evidence for improved water quality. Water: Official Journal of the Australian Water and Wastewater Association. 27 (3): 18- 22. Brookes JD, Burch MD, Lewis DM, Regel RH, Linden L, Sherman B. (2008) Artificial mixing for destratification and control of cyanobacterial growth in reservoirs. Research Report 59. CRC for Water Quality and Treatment Drikas M, Newcombe G and Nicholson B (2001) Water treatment options for cyanobacteria and their toxins. Blue green algae: Their significance and management within water supplies. CRC for Water Quality and Treatment, Occasional paper 4. Hallegraeff, G.M. (1993). A review of harmful algal blooms and their apparent global increase. Phycologia 32: 79-99. Jochimsen E M, Carmichael WW, An JS,, D.M Cardo DM, Cookson ST, Holmes CE, Antunes MB, de Melo Filho DA, Lyra TM, Barreto VS, Azevedo SM, and Jarvis WR (1998) Liver failure and death after exposure to microcystins at a hemodialysis center in Brazil. New England Journal of Medicine 338: 873-8. Lau RH, Sapienza C and Doolittle WF (1980) Cyanobacterial plasmids: Their widespread occurrence, and the existence of regions of homology between plasmids in the same and different species. Molecular and General Genetics 178: 203-211. Mc Auliffe TF and Rosich RF (1990) The triumphs and tribulations of artificial mixing in Australian waterbodies. Water August, pp 22-23 Mitrovic SM, Bowling LC and Buckney (2001) Vertical disentrainment of Anabaena circinalis in the turbid, freshwater Darling River, Australia: Quantifying potential benefits from buoyancy. Journal of Plankton Research 23: 47-55. Nicholls N. & Kariko A. (1993) East Australian rainfall events: Interannual variations, trends, and relationships with the Southern Oscillation. Journal of Climate 6: 1141-1152. Oliver RL (1994) Floating and sinking in gas-vacuolate cyanobacteria. Journal of Phycology 30: 161- 173. Reynolds, C.S. (1992). Eutrophication and the management of planktonic algae: what Vollenweider couldn’t tell us. In J.G. Jones and D.W. Sutcliffe (Eds). Eutrophication: Research and Application to Water Supply. Freshwater Biological Association, Ambleside, U.K. Reynolds CS (1997) Vegetation Processes in the Pelagic: A model for Ecosystem Theory. Ecology Institute Oldendorf/Luhe Germany.

79 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Saker ML, Thomas AD, and Norton JH (1999) Cattle mortality attributed to the toxic cyanobacterium Cylindrospermopsis raciborskii in an outback region of North Queensland. Environmental Toxicology 14: 179-182. Sas H (1989) Lake restoration by reduction of nutrient loading: Expectations, Experiences, Extrapolations. Sankt Augustin: Academia Verlag Richarz. Sherman BS and Webster IT (1994) A model for the light-limited growth of buoyant phytoplankton in a shallow, turbid waterbody. Australian Journal of Marine and Freshwater Research 45: 117-132. Sherman BS, Whittington J and Oliver RL (2000) The impact of destratification on water quality in Chaffey Dam. Archives Hydrobiologica Special Issues on Advanced Limnology 55: 15-29. Stal-Lucas J and Walsby A (1998). The daily integral of nitrogen fixation by planktonic cyanobacteria in the Baltic Sea. New Phytologist 139: 665-671. Visser PM, Ibelings BW, van der Veer B, Koedoods J and Mur LR (1996) Artificial mixing prevents nuisance blooms of the cyanobacterium Microcystis in Lake Nieuwe Meer, The Netherlands. Freshwater Biology 36: 435-450 Vollenveider RA and Kerekes J (1980) The loading concept as basis for controlling eutrophication philosophy and preliminary results of the OECD program on eutrophication. Progress in Water Technology 12: 5-38. Walsby AE (1994) Gas Vesicles. Microbiological Reviews 58: 94-144. Walsby AE (1997) Numerical integration of phytoplankton photosynthesis through time and depth in a water column. New Phytologist 136: 189-209. Walsby AE, Hayes PK, Boje R and Stal LJ (1997) The selective advantage of buoyancy provided by gas vesicles for planktonic cyanobacteria in the Baltic Sea. New Phytologist 136: 407-417. Walsby AE, Reynolds CS, Oliver RL, Kromkamp J and Gibbs MM (1987) The role of buoyancy in the distribution of Anabaena sp. In Lake Rotongaio. New Zealand Journal of Marine and Freshwater Research 21: 525-526.

80 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

8 IRON AND MANGANESE

8.1 Introduction Excessive iron and manganese in source water can lead to discoloured water at customer taps if these metals are not oxidised and flocculated in the treatment process. Moderate concentrations of soluble iron and manganese entering the distribution system can also be problematic as they may be accumulated by bio-films which dislodge during periods of high flow and compromise water quality.

The development of iron and manganese problems in reservoirs is usually a seasonal phenomenon related to the evolution of thermal stratification and associated anoxia in sediments and overlying water. This can take some time to evolve over the summer period and depends upon the sediment oxygen demand, and the degree of oxygenation and circulation in the overlying water. With the establishment of stratification the sediment oxygen demand can rapidly drawdown the oxygen in the overlying water column and permit soluble iron and manganese to stay in solution. It may take some time for concentrations to reach levels that can cause water quality problems.

It is possible to monitor a range of indicators to gain advanced warning of likely iron and manganese issues in the reservoir. Monitoring of either dissolved oxygen or redox potential in the water adjacent to the sediments, either with manual probes or ideally with on-line sensors can be an indicator of the potential for iron and manganese to be released from sediments. This approach can be integrated into an on-line monitoring program (described in Chapter 2 of this guide).

To utilise measurements of redox potential in the reservoir for prediction of soluble metal release requires some understanding the chemistry of the process in the sediments. Ferrous ions (Fe2+) are readily released from sediments when the redox potential declines to about 200 mV while resolubilisation of Manganese (Mn2+) from the sediment occurs at redox potentials in the order of 400 mV at neutral pH. In more alkaline environments, such as lake sediments, iron and manganese require a lower redox potential to remain soluble. Release of manganese therefore precedes iron, and manganese will remain soluble if the oxygen saturation is less than about 50% (~4 mg L-1; Wetzel, 2001). For on-line monitoring in reservoirs experience has shown that dissolved oxygen probes are likely to be more reliable than redox probes, even though redox is a more direct measure of oxygen reduction potential and therefore redox-sensitive metal release. In particular the relatively new generation of optical dissolved oxygen sensors have been developed to provide accurate dissolved oxygen measurements over long periods of time without the need for re-calibration compared to the traditional galvanic and polargraphic dissolved oxygen sensors, and are better suited to on-line deployment.

Iron has a taste threshold of 0.3 mg L-1 in water and becomes objectionable above 3 mg L-1 (NHMRC/ARMCANZ, 1996). Soluble manganese concentrations greater than 0.l mg L-1 are considered problematic although concentrations as low as 0.05 mg L-1 can cause ‘dirty water’ problems and require oxidation in the treatment plant. Many operators aim to keep soluble Mn as low as 0.01-0.02 mg L-1 to avoid subsequent problems in the distribution system.

8.2 Case study – Myponga Reservoir Fe and Mn reduction Myponga Reservoir in South Australia had a history of recurring seasonal issues with iron and manganese release; however, these have been significantly reduced with the operation of a bubble plume aerator for destratification.

Strong seasonal temperature stratification associated with low dissolved oxygen in the hypolimnion occurred at Myponga Reservoir during summer-autumn seasons from 1984-1994 (Figure 60). This was associated with the concentrations of iron and manganese at depth that were consistently higher than at the surface. Soluble iron at 30 m depth reached a maximum concentration of 2.84 mg L-1 in March 1990. The vertical gradient in iron concentration decreased after the installation of the bubble plume aerator in 1994. The mean concentration of iron at 30 m decreased from 0.71 mg L-1 in 1986 to 0.345 mg L-1 in 1996, and the large flux events have been eliminated. Manganese concentrations at depth have responded to destratification in a similar manner to iron (Figure 61). The mean concentration of manganese at 30 m was 0.41 mg L-1 in 1986 and was reduced to 0.052 mg L-1 in 1996 due to destratification.

81 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

The maximum concentration measured in 1986 was approximately 1.8 mg L-1. These large flux events observed prior to 1994 have been reduced by operation of the aerator.

The concentration of iron and manganese in the hypolimnion depends upon the rate of flux from sediment and the duration of stratification. Once the hypolimnion becomes anoxic the conditions suitable to maintain iron and manganese in the soluble form will persist whilst there is little mixing and the hypolimnion remains separated from the atmosphere. To allow a comparison between years the duration of stratification was quantified as number of days per year (July–June) where the temperature difference between 10 m and 30 m was greater than 1°C. Measurements were generally taken in the morning, however, the temperature difference between 10 m and 30 m was selected as the best indicator of persistent stratification as this avoids confounding effects due to diurnal surface heating, and time of sampling.

There was a temperature difference of greater than 1°C for 224 days in the year 1985/86 which gave rise to maximum iron and manganese concentrations of 2.55 mg L-1, 1.8 mg L-1. In 1995/96 there were only 28 days where this temperature difference occurred and consequently the maximum iron and manganese concentrations were considerably low, 0.386 mg L-1 and 0087 mg L-1, respectively. The yearly maximum concentration of iron and manganese at 30 m was significantly correlated with the duration of stratification.

The bubble plume aerator operated in Myponga Reservoir has been successful at reducing the concentration of iron and manganese to achieve compliance with the Australian Drinking Water Guidelines (1996). The guidelines recommend that the concentration of iron in drinking water should not exceed 0.3 mg L-1. This guideline is based upon the concentration at which iron precipitates from solution and the taste threshold. Prior to the installation of aerator at Myponga Reservoir the soluble iron concentration frequently exceeded the guideline. However, since aerator operation the iron concentration is maintained closer to the guideline concentration.

Based upon aesthetic considerations the concentration of manganese in drinking water should not exceed 0.1 mg L-1 (Australian Drinking Water Guidelines, 1996). The bubble plume aerator at Myponga has been successful at maintaining manganese concentrations below this level. Manganese would not be considered a health threat unless the concentration exceeded 0.5 mg L-1, a level which has not occurred since aerator operation.

30 Surface 10 m 25 20 m 30 m

5 Temperature ( degrees C) degrees ( Temperature 0 01-Jun-84 01-Jun-85 01-Jun-86 01-Jun-87 31-May-88 31-May-89 31-May-90 31-May-91 30-May-92 30-May-93 30-May-94 30-May-95 29-May-96 29-May-97 29-May-98 29-May-99 Date

Figure 60 Temperature profiles in Myponga Reservoir.

82 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

) Surface -1 1.6 30 m

1.2

0.8 Aerator insatlled

0.4 Soluble manganese (mg L (mg manganese Soluble

0 01-Jun-84 01-Jun-85 01-Jun-86 01-Jun-87 31-May-88 31-May-89 31-May-90 31-May-91 30-May-92 30-May-93 30-May-94 30-May-95 29-May-96 29-May-97 Date

Figure 61 Soluble manganese concentrations in Myponga Reservoir.

8.3 References Australian Drinking Water Guidelines (1992) National Health and Medical Research Council. Raman RK and Arbuckle BR (1989) Long-term Destratification in an Illinois Lake. Journal of the American Water Works Association. 81: 66-71 Stumm W and Morgan JJ (1996) Aquatic Chemistry: chemical equilibria and rates in natural waters. Third edition. John Wiley and Sons, New York. Wetzel (1983) Limnology. Second Edition. CBS College Publishing, Philadelphia.

83 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

9 MANAGING A VARIABLE OFF-TAKE

9.1 Introduction There are few management options available to control water quality within the reservoir itself and ideally upstream catchment management practices should be implemented for long-term sustainability. The sustainable reservoir management options are limited to artificial destratification and active management of the depth from which water is harvested. The fact that most of the water quality problems are not uniform throughout the water column means that this vertical variability can be exploited to select an off-take depth with the lowest contaminant concentrations.

The aim of this chapter is to demonstrate how the water off-take can be actively managed to select and harvest water that has the lowest contaminant concentrations for treatment, and therefore presents a reduced challenge to the treatment plant that results in reduced risk, lower chemical use and better plant performance.

Pathogens and natural organic matter (NOM) are carried into reservoirs during rain event inflows and so present the greatest challenge during storms. The distribution of inflows are controlled by their density relative to that of water in the reservoir, such that warm inflows will travel over the surface of the reservoir as a buoyant surface flow and cold, dense inflows will sink beneath the reservoir water and flow along the lake bed towards the deepest point (see Chapters 4, 5 & 6). In either case the inflow will entrain water from the lake, increasing its volume, changing its density and diluting the concentration of pathogens and other contaminants. The speed at which an inflow travels through a lake, the degree of entrainment of ambient lake water, the dilution, and the insertion depth are all important in determining the distribution of pathogens or NOM in lakes and reservoirs. Consequently it is important to know the depth of the riverine intrusion so water can be harvested outside of this intrusion.

9.2 Case Studies We present three case studies that demonstrate how managing the off-take can reduce the risk from pathogens and reduce treatment costs during riverine intrusions associated with rain events.

9.3 Methods Two rain event inflows were monitored to assess contaminant transport through a reservoir.

The first study was conducted at Myponga Reservoir. Myponga Reservoir is a flooded river valley impounded by a concrete arch-dam that was completed in 1962. Myponga Reservoir is located 70 km south of Adelaide, South Australia and has a capacity of 26,800 ML at a full supply level, with a maximum depth of 41 m. The mean retention time based upon abstraction is approximately 3 years. The catchment area is 124 km2 and the land use is dominated by cattle grazing and dairy farming.

A sampling program was structured to track a riverine intrusion, including its physical properties and the passage of pathogens and indicators, through the reservoir during a large runoff event following intense rainfall. The experiment was conducted from June 26–29, 2003 during which time the total rainfall was 58.4 mm.

Sampling for microbiological analysis was undertaken at the surface and at depth (2.5 m above the bottom) at three sites in the reservoir and in the inflows. Sampling was timed to capture the stream inflow, within the reservoir ahead of the intrusion, and subsequently within the riverine intrusion.

The hydrodynamic model ELCOM was coupled to the ecological model CAEDYM to describe pathogen transport, settling and inactivation within the reservoir (Hipsey et al., 2004) (Also see Chapter 4).

The second study was undertaken at Little Para Reservoir, North-East of Adelaide, in August 2004. Dissolved organic carbon and sediment transport was measured at the surface and depth within and out of a riverine intrusion that occurred following heavy rainfall and inflow. Jar tests were performed to determine the alum dose required to treat the water in the inflow as it varied in water quality over the

84 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67 time course of the event. Chlorine demand, THM formation potential and product water turbidity, colour and DOC concentration were measured following jar tests.

The third example to illustrate management of reservoir water off-take is the management of a cyanobacterial bloom in a South Australian reservoir. The vertical distribution of cyanobacteria was examined with cell counts and regular profiling with a multi-parameter water quality probe fitted with a sensor for chlorophyll fluorescence. This profiling data was used for selection of the best depth for water supply off-take.

9.4 Results 9.4.1 Pathogen transport A rain event inflow was monitored at Myponga Reservoir, South Australia, in June 2003. Four locations were monitored over the course of several days. Cryptosporidium parvum was transported in the underflow riverine intrusion and arrived at the dam wall approximately 24 hours after the river flow began to increase. The concentration of Cryptosporidium at the dam wall was significantly lower than concentrations in the creek inflow but still posed a threat to the water treatment plant.

The riverine intrusion travelled as an underflow so the greatest pathogen risk occurred near the bottom of the reservoir. This was also shown by modeling simulations of Cryptosporidium concentrations in Myponga Reservoir for an inflow event (Figure 62). Following the inflow the greatest concentrations are near the bottom so withdrawing water from higher in the water column would significantly reduce this risk.

Figure 62 Prediction of Cryptosporidium parvum concentrations at different depths in Myponga Reservoir.

9.4.2 Natural organic matter transport and treatment In-flow samples were collected at Little Para Reservoir in August 2004 to monitor the movement of sediment and natural organic matter through the reservoir following a large rainfall and runoff event. Jar tests were performed to determine the alum required to adequately coagulate water from within

85 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT and outside of the riverine intrusion. Water from within the intrusion (bottom water) had higher DOC concentration, higher turbidity, higher alum demand and higher chlorine demand (Table 10). Harvesting water for treatment outside of the influence of the riverine intrusion will significantly reduce the chemical use in the treatment plant and result in better quality product water.

Table 10 Jar Test results and water sample analysis for Pre-inflow, Inflow and Post-inflow samples collected at the Dam Wall and at reservoir inlets (Gould Creek and Little Para River).

Alum Turbidity True DOC1. Chlorine THMFP2. Dose (NTU) Colour (mg/L) Demand (µg/L) (mg/L) (HU) (mg/L) Pre-Inflow Dam Wall 60 13 11 4.99 (2.10) (57) (“baseline”) (0.073) (2) (2.93) Gould Creek 16 11 5.4 Little Para 24 5 5 River During Dam Wall 60 11.3 31 6.29 (2.37) (66) stormwater Top (0.080) (2) (3.31) inflow Dam Wall 90 30.2 67 9.41 (4.66) (90) Bottom (0.106) (5) (4.20) Gould Creek 42 73 10 Little Para 93 113 13 River Post inflow Dam Wall 60 5.0 51 7.80 (2.83) (123) Top (0.066) (5) (4.15) Dam Wall 90 4.23 51 7.80 (2.79) (115) Bottom (0.066) (4) (3.95) Gould Creek 5 47 8.4 Little Para 4.8 49 8 River 1. DOC; Dissolved Organic Carbon 2.THMFP: Tri-Halo Methane Formation Potential Note: ● values in brackets ( ) are results after alum treatment using identified dose

9.4.2.1 Vertical Distribution of Cyanobacteria The phytoplankton population was monitored in a special investigation in a reservoir in the spring of 2006, and at the time of monitoring was dominated by the cyanobacterium Anabaena circinalis. Vertical distribution of A. circinalis was determined with profiles of chlorophyll fluorescence. The buoyant nature of A. circinalis meant that it migrated towards the surface and accumulated at the greatest concentrations above 5 m (Figure 63). Cyanobacterial distribution showed significant horizontal and vertical heterogeneity throughout the reservoir. The off-take to the water treatment plant was at 25 m.

86 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 63 Vertical profiles of a population of the cyanobacterium Anabaena circinalis in a horizontal transect across a South Australian reservoir measured by chlorophyll a fluorescence. Scale is µgL-1 Chl a.

9.5 Discussion Different contaminants become problematic under different hydrodynamic and meteorological conditions. To maximise water quality entering a treatment plant it is necessary to have an understanding of the conditions that give rise to the various hazards and the vertical distribution of the contaminants. This vertical variability in reservoir water quality can be exploited to select an off-take depth with the lowest contaminant concentrations.

Nuisance cyanobacteria in the Australian context tend to be dominated by Microcystis, Anabaena and Cylindrospermopsis. These genera tend to be distributed near the surface and the selection of an appropriate off-take depth, where this is available, can be used to avoid extreme concentrations in the vertical profile. However, this risk from cyanobacteria in stratified water bodies is a relatively dynamic phenomenon and requires careful monitoring. Also, toxins and other metabolites may be released from the cyanobacterial cells and these dissolved components will be distributed through the water column. These hazards must also be balanced against other hazards associated with stratification including iron and manganese.

The greatest concentrations of pathogens and DOC are transported from the catchment and are strongly associated with storm event inflows. Off-take selection to avoid these high concentrations must identify where the riverine inflow is and harvest water outside the inflow. The potential improvements in water quality that could be achieved by actively detecting and avoiding the storm inflow was clearly demonstrated in the example from Little Para Reservoir following the inflow event in winter, 2004 (Table 8). The alum dose required to coagulate turbidity and DOC was approximately 50% greater for the inflow water travelling at depth than water at the reservoir surface in this case, and the surface water also had a much lower Chlorine Demand and THMFP at the time of the inflow.

87 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

A simple model called ‘INFLOW’, developed by Antenucci et al. (2005), enables prediction of the depth at which the riverine inflow will occur and the anticipated dilution as it travels through the reservoir.

The equations have been incorporated into a web-based application (http://www.cwr.uwa.edu.au/services/models.php?mdid=9). From this information, the following predictions are made:

• Insertion temperature • Insertion volume • Insertion dilution (relative to initial volume) • Insertion time

The development and use of this model is described in more detail in Chapter 5 of this guide.

The case studies presented here show that a range of options or tools are available to assist in selection the optimum depth from which to take the best quality water. These include the simple INFLOW model to predict the depth of inflow intrusion; fluorescence profiles can be used to determine the distribution of cyanobacteria; and more sophisticated hydrodynamic models can also assist in modeling the distribution of a range of water quality characteristics both vertically and horizontally. allow the optimal off-take depth to be selected. These techniques can potentially be used to reduce the risk of the breakthrough of contaminants and lead to improved water quality and result in reduced cost associated with water treatment chemical use.

9.6 References Antenucci J, Brookes JD and Hipsey M (2005) A simple model for quantifying Cryptosporidium transport, dilution, and potential risk in reservoirs. Journal of the American Water Works Association 97 (1) 86-93. Brookes JD, Davies C, Antenucci J and Hipsey M (2006) Association of Cryptosporidium with bovine faecal particles and implications for risk reduction by settling within water supply reservoirs. Water and Health 4: 87-98. Brookes JD, Hipsey MR, Linden L, Regel RH, Burch MD and Antenucci JP (2005) Investigation of the relative value of surrogate indicators for detecting pathogens in lakes and reservoirs. Environmental Science and Technology 39(22): 8614-8621. Hipsey M, Antenucci J, Brookes JD, Burch M, Regel R and Linden L (2004). A three-dimensional model for Cryptosporidium dynamics in lakes and reservoirs. International Journal of River Basin Management 2 (3): 181-197. Hipsey MR, Brookes JD, Regel R, Antenucci JP, and Burch MD (2005) In situ evidence for the association of Total Coliforms and Escherichia coli with suspended inorganic particles in an Australian reservoir. Water, Air and Soil Pollution. 170 (1): 191-209.

88 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

10 MANAGING WILDFIRE IMPACTS ON WATER QUALITY

10.1 Introduction Wildfire is a characteristic feature of the Australian landscape and has shaped the ecology of the continent. However, wildfires can cause catastrophic damage and challenge water quality in burnt catchments. Climate change is expected to increase the frequency and possibly the intensity of wildfires. Fried et al. (2004) coupled a climate model to wildfire records to predict how a doubling in CO2 concentrations would affect wildfires in northern California, a Mediterranean climate with summer drought similar to much of Southern Australia. The warmer, windier conditions expected under climate change were predicted to produce more fires that burned more intensely and spread faster. Fried et al. (2004) predicted that the fire return interval for grass and brush vegetation would be halved under these climate change scenarios. Frequent fires would impact vegetation and increase the frequency with which combustion products are exported from the catchment into waterways. The significant risks associated with wildfires in drinking supply catchments warrants contingency planning in water utilities for the eventuality of fire.

10.2 Water quality issues associated with wildfires A large suite of water quality issues can arise from wildfires in drinking water supply catchments. Most do not eventuate until there is rain in the catchment to transport the contaminants. The erosion of burnt catchment leads to the transport of sediment, nutrients, dissolved organic carbon (DOC) and combustion products into water ways. Severe intense wildfire can have an effect on the hydrology similar to clear felling (Yao, 2003). A further issue for water utilities arising from wildfires is the reduction in water yield in revegetating forest catchments where plant water demand is greater than in established intact forests (Langford, 1976; Kuczera, 1987). The issue of modified hydrology and reduced water yields will not be covered in this summary of water quality but there is ongoing research in Australia to quantify and document impact of bushfire and regrowth on water yield.

The intensity of the fire, the revegetation and the intensity of subsequent storms will interact to determine the impact on receiving streams or lakes. Prosser and Williams (1998) documented increased local runoff and erosion through enhanced soil hydrophobicity and reduced ground cover. However, they found the impacts were confined to the hill slope catchment scale and large rainfall events would be required to mobilise sediments. This may support the use of controlled burns at appropriate times to reduce wildfire intensity and moderate erosion. Shakesby et al. (2006) suggest that a number of natural features also act to reduce post-fire erosion under light-moderate intensity rainfall. These include mats of fine roots, litter dam-microterrace complexes and ant or small mammal activity which provide sinks for overland flow and trap mobilised sediment.

10.3 Fire retardants Advanced fire fighting techniques involve the use of foams and fire retardants which are often administered from fire-fighting aircraft. Fire fighting chemicals are effective and efficient means for fighting fire by retarding combustion and preventing re-ignition (Rawet et al., 1996). Foams tend to have short term fire suppressant action and are principally surfactants whereas the retardants are longer acting, forming a combustion barrier between the fire and fuel (Adams and Simmons, 1999). Retardants have a chemical composition similar to fertilizer and are essentially ammonium and phosphate salts. The chemical content of three fire-fighting agents was measured by Couto-Vázquez and González-Prieto (2006) including a foaming agent (Auxquimica RFC-88), Firesorb and ammonium phosphate, a retardant (Table 11). As its name suggests the fire retardant, ammonium phosphate is very high in nitrogen and phosphorus which could fertilise waterways in catchments where this is applied.

When ammonium polyphosphate was applied to burnt soil the ammonium level was initially 200 times higher than unburnt soil treatments (Couto-Vázquez and González-Prieto, 2006). This was gradually converted to nitrate over 90 days and reduced to pre-application concentrations after one year. Available phosphorus was also very high initially in the top 2 cm of burnt soil with ammonium polyphosphate added, which reduced by 75% over the first 30 days and declined gradually over the year. Rainfall soon after fire may not just mobilise eroded sediment if fire retardants are used, but considerable additional nutrients may be transported to waterways. Adams and Simmons (1999) concluded that unnecessary use of retardants and foams for prescribed burning operations should be

89 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT avoided but that foams were preferable over traditional retardants which potentially had a greater ecological impact.

Table 11 Total nutrient content of three fire-fighting chemicals (adapted from Couto-Vázquez and González-Prieto, 2006). Chemical Foaming agent Firesorb Ammonium polyphosphate (Auxquimica RFC-88) (FR Cross) Na (gL-1) 19.7 8.7 10.2 K (mg L-1) 22.8 0 1588 Ca (mg L-1) 10.8 38.5 1210 Mg (mg L-1) 0 0 3124 P (gL-1) 0 0.2 933 N (gL-1) 3.7 12.6 683

The dissolved and particulate contaminants that wash-off following fire can present considerable water quality challenges. The relevant mechanisms of transport and risks that may occur have been covered in previous chapters but the scale at which they present may increase tremendously.

The immediate risk to reservoir water quality follows rainfall onto burnt catchments and the contaminants transported in the inflowing water. High turbidity, high dissolved organic carbon and potentially high metal concentrations are likely to be found in the inflow. In the longer term the increased nutrient load can lead to higher algal biomass. Iron and Manganese problems can also arise following fires in catchments, which can be of immediate concern and persist in the water column for some time.

Risk mitigation can include reducing transport from the catchment and managing the polluted water in the reservoir. Managing water quality in the reservoir can draw on techniques described in other chapters e.g. off-take selection, iron and manganese control.

10.4 CASE STUDY Managing sediment transport in catchments following wildfire A fire in the catchment of Little Para Reservoir (Figure 64) on 8 March 2004 swept through the area surrounding the reservoir, burning right up to the water’s edge in many areas. A total area of 600 hectares was burnt, which represents about 7% of the reservoir’s natural catchment of 8,200 hectares. Little Para Reservoir is located near Adelaide, South Australia, and is one of the main reservoirs operated by SA Water, having a capacity of 20,800 ML, The reservoir supplies Little Para Water Treatment Plant, which is a filtration plant with a capacity of up to 160 ML/day. SA Water undertook an investigation and catchment remedial works to mitigate sediment transport from the catchment. Hay bales were positioned along contours in the catchment (Figure 65) to trap sediment in an attempt to reduce sediment loading to the reservoir following rain. The net benefit of this action was not quantified but it is one of the few practical works available to reduce sediment transport until revegetation occurs (Figure 66).

90 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 64 Fire burning in the Little Para Catchment

Figure 65 Hay bales contoured on a burnt hill-slope to trap sediments and retain them on the catchment (Photo: Paul Hackney)

91 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Figure 66 The burnt hill-slope revegetated after winter rain (Photo Paul Hackney).

10.5 CASE STUDY – Effects of a severe wildfire and fire-generated contaminants on reservoir water quality The Bendora and Corin Reservoirs are sited on the Cotter River and provide water supply storage to the city of Canberra (Figure 67). Devastating wildfires were ignited by lightning in the bushland surrounding Canberra and burnt in remote areas for about 10 days before weather conditions changed and the fires increased to a greater level of intensity. On 18 January 2003 the fires swept rapidly through urban bushland, crossed the bush-urban interface destroying 501 homes and causing four deaths. The drinking water catchments were burnt by these fires with Bendora and Corin Reservoir catchments being heavily burnt.

Soon after the fires over 60 mm of precipitation fell on the steep burnt catchment from an intense storm cell. This intense rainfall eroded the catchment transporting not just sediment but also boulders down the denuded hill-slope. Wasson et al. (2004) estimated that 19,300 tonnes of inorganic sediment and 1,900 tonnes of organic matter were deposited in Bendora Dam following the fire. The water quality impacts of this fire are summarised below.

92 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 67 The location of the four main dams of the ACT and their catchments, the Cotter and Queanbeyan Rives.

The fires in the Cotter Catchment had a noticeable impact on water quality. Following rain the nitrogen and phosphorus concentrations increased considerably over the long-term historical concentrations (Figure 68). Nitrate concentrations increased by an order of magnitude to a maximum of 0.25 mgL-1. The total phosphorus concentration also increased by an order of magnitude. The net result of the increased nutrient concentration was an increase in the maximum phytoplankton biomass as represented by the chlorophyll concentration (Figure 69) and total algae counts (Figure 70). There is very little that can be done to avoid algal issues once the nutrients have entered the water body. Relatively small increases in nutrient export are evident as a change in water quality. Lamontagne et al. (2000) concluded that the increased rates of N and P export represented negligible loss of nutrients from the fire but were important supplementary nutrients for the receiving lakes.

Fortunately in the case of the Canberra reservoirs the increase in nutrients following the fire did not initiate a cyanobacterial bloom. The background nutrient concentrations are typically low in the soils in the catchment and in the reservoir. Following prescribed fires burning in the Lake Tahoe catchment (California) stream phosphorus concentrations were found not to increase which may be due to the fire releasing calcium and raising soil pH which could have acted to incorporate phosphorus into insoluble forms (Stephens et al., 2004). While an increase in phytoplankton occurred in the Canberra reservoirs this did not present a major ongoing water quality problem. This may not be the case for all reservoirs and catchment protection from intense wild fires, controlled burning and water treatment options should all be considered as part of a plan for water security in bushfire prone regions.

93 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Bendora Reservoir Total Phosphorus

1.2 1 0.3m 0.8 3m 0.6 27m 0.4 TP (mg/L) 30m 0.2 0 16-Jan-93 16-Jan-94 16-Jan-95 16-Jan-96 15-Jan-97 15-Jan-98 15-Jan-99 15-Jan-00 14-Jan-01 14-Jan-02 14-Jan-03 14-Jan-04 Date

Figure 68 Total phosphorus concentration at selected depths in Bendora Reservoir

Bendora Reservoir Chlorophyll

9 8

g/L) 7 μ 6 5 0.3m 4 3 2

Chlorophyll ( 1 0 16-Jan-93 16-Jan-94 16-Jan-95 16-Jan-96 15-Jan-97 15-Jan-98 15-Jan-99 15-Jan-00 14-Jan-01 14-Jan-02 14-Jan-03 14-Jan-04 Date

Figure 69 Chlorophyll concentration at the surface in Bendora Reservoir.

Bendora Reservoir Total Algae

20000 16000

12000 0.3m 8000

4000

Total Algae (cells/mL) Algae Total 0 16-Jan-93 16-Jan-94 16-Jan-95 16-Jan-96 15-Jan-97 15-Jan-98 15-Jan-99 15-Jan-00 14-Jan-01 14-Jan-02 14-Jan-03 14-Jan-04 Date

Figure 70 Total algal abundance in Bendora Reservoir.

94 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Although the elevated nutrient concentrations following the Canberra fire did not cause a significant water quality incident, the very high manganese concentrations did cause concern. White et al. (2006) reported that the 2003 fires increased turbidity, iron and manganese by up to 30 times concentrations in previous inflow events. Turbidity was typically below 10 NTU but post-fire turbidity exceeding 1,000 NTU was recorded (Figure 71). The concentration of total manganese in Bendora Reservoir in the decade before the fires was typically below 0.5 mgL-1 with a few excursions to 0.8 mgL-1 at the deepest sites (Figure 72). Following the fires and subsequent rainfall, significant amounts of manganese were washed into Bendora and Corin Reservoirs. High turbidity and manganese concentrations meant that these reservoirs were unsuitable for potable water supply and Googong Dam water was used preferentially. However, the treatment plant was unable to meet demand and water restrictions were imposed (White et al., 2006).

Interestingly, high concentrations of soluble manganese persisted even in oxygenated waters where typically they would oxidise. This ‘soluble’ fraction was classified by size, passing through a 0.45 µm filter. This may still have consisted of fine colloidal particles that sink very slowly hence soluble manganese persistence. The persistence of manganese was a concern as it caused dirty water problems and rendered the water unusable with the level of treatment available. Because the highest concentrations were at depth (Figure 73) a decision was made to flush the poorer quality water from the hypolimnion during a natural inflow event. This effectively removed poor quality water from the system and restored manganese concentrations to pre-fire concentrations. The limited water availability in Australia precludes this as an option for many reservoirs but was effective in the Cotter catchment reservoirs.

Bendora Reservoir Turbidity

10000 1000 0.3m 100 12m 10 18m 1 24m 27m Turbidity (NTU) 0.1 0.01 16-Jan-93 16-Jan-94 16-Jan-95 16-Jan-96 15-Jan-97 15-Jan-98 15-Jan-99 15-Jan-00 14-Jan-01 14-Jan-02 14-Jan-03 14-Jan-04 Date

Figure 71 Turbidity at selected depths in Bendora Reservoir (note Y axis is logarithmic scale)

95 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Bendora Reservoir Total Manganese

3.5 0.3m 3 3m 2.5 6m 2 9m 1.5 12m 1 15m Total Mn (mg/L) 0.5 18m 0 21m 24m

16-Jan-93 16-Jan-94 16-Jan-95 16-Jan-96 15-Jan-97 15-Jan-98 15-Jan-99 15-Jan-00 14-Jan-01 14-Jan-02 14-Jan-03 14-Jan-04 27m Date 30m

Figure 72 Total manganese concentration in Bendora Reservoir

Bendora Reservoir Total Manganese

3.5 0.3m 3 3m 2.5 6m 2 9m 1.5 12m 1 15m Total Mn (mg/L) 0.5 18m 0 21m 24m 28-Jul-03 29-Apr-03 26-Oct-03 29-Jan-03 30-Mar-03 28-Jun-03 28-Feb-03 29-May-03 27-Aug-03 25-Nov-03 26-Sep-03 25-Dec-03 27m Date 30m

Figure 73 Total manganese concentration in Bendora post fires in 2003

96 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

10.6 References Adams R and Simmons D (1999) Ecological effects of fire fighting foams and retardants. Australian Forest 62:307-314. Couto-Vázquez A and González-Prieto SJ (2006) Short and medium-term effects of fire fighting chemicals on the properties of a burnt soil. Science of the Total Environment 371: 353-361. Fried JS, Torn MS and Mills E (2004) The impact of climate change on wildfire severity: A regional forecast for Northern California. Climatic Change 64:169-191. Kuczera GA (1987) Prediction of water yield reductions following a bushfire in ash-mixed species Eucalyptus forest. Journal of Hydrology. 94: 215-236. Lamontagne S, Carignan R, D’Arcy P, Praire YT and Paré D (2000) Element runoff from eastern Canadian Boreal Shield drainage basins following forest harvesting and wildfires. Canadian Journal of Fisheries and Aquatic Science. 57(suppl. 2): 118-128. Langford KJ (1976) Changes in water yield following a bushfire in a forest of Eucalyptus regnans. Journal of Hydrology. 29: 87-114. Prosser IP and Williams L (1998) The effect of wildfire on runoff and erosion in native Eucalyptus forests. Hydrological Processes 12: 251-265. Rawet D, Smith R and Kravainis G (1996) A comparison of water additives for mopping-up after forest fires. International Journal of Wildland Fire 6: 37-43. Shakesby RA, Wallbrink PJ, Doerr SH, English PM, Chafer CJ, Humphreys GS, Blake WH and Tomkins KM (2007) Distinctiveness of wildfire effects on soil erosion in south-east Australian eucalypt forests assessed in a global context. Forest Ecology and Management. 238: 347-364. Stephens SL, Meixner T, Poth M, McGurk B and Payne D (2004) Prescribed fire, soil and stream water chemistry in a watershed in the Lake Tahoe Basin, California. International Journal of Wildland Fire 13: 27-35. Wasson RJ, Worthy M, Olley J, Wade A and Mueller N (2004) Sources of Turbidity in Bendora reservoir. Report to ActewAGL March 2004. Centre for Resource and Environmental Studies, Australian National University. White I, Wade A, Worthy M, Mueller N, Daniell T and Wasson R (2006) The vulnerability of water supply catchments to bushfires: Impacts of the January 2003 wildfire on the Australian Capital Territory. Australian Journal of Water Resources 10: 115. Yao S (2003) Effects of fire disturbance on forest hydrology. Journal of Forestry Research. 14:331-334.

97 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

11 CLIMATE CHANGE AND RESERVOIR MANAGEMENT

11.1 Introduction The Australian climate has always been highly variable, characterised by a series of major droughts interspersed with extreme wet events. Reservoir managers are faced with precipitation patterns that are not only seasonal but also exhibit significant inter-annual variability. These variations are strongly influenced by the El Niño - Southern Oscillation – cyclical oscillations of air pressure and ocean temperatures over the Pacific Ocean. Climate change is likely to further increase the variability of the Australian climate. Increasing temperatures and changes in the spatial and temporal patterns of precipitation are expected. ‘Heatwaves and fires are virtually certain to increase in intensity and frequency’ and ‘floods, landslides, droughts and storm surges are very likely to become more frequent and intense’ (Hennessy et al., 2007).

While it is likely that the nature of the major hazards facing reservoir managers will be the same, the projected changes in regional climate variability combined with non-climatic changes including population growth, changes in water use patterns, land use changes and the use of alternative water sources will present reservoir managers with a changed risk profile.

11.2 Understanding future climate change To assess the risks associated with climate change, an understanding of the potential changes in the climate are required. Global climate models (GCMs) provide the best source of information regarding possible future climate scenarios. These computer models simulate the processes that govern the earth’s climate, solving complex mathematical functions that represent the highly non-linear physical processes underlying the Earth’s climate system. Calibrated using historic measured data, the models attempt to replicate large scale climatic behaviour over a range of time scales from sub-daily to inter- annual. GCMs have been developed independently by a range of scientific and research institutions around the world and the models are continually being refined in response to improved climatic understanding and increased computational capabilities. In the Intergovernmental Panel for Climate Change’s (IPCC) Fourth Assessment Report (AR4), a total of 23 GCMs from 16 organisations were included in the analysis. These included output from the CSIRO’s Mark 3.0 and Mark 3.5 GCMs. Output from all of the AR4 models can be accessed through the Earth Systems Grid CMIP3 database (ESG).

11.2.1 Limitations of GCMs While GCMs are useful tools for generating broad-scale projections of future climate, they are not perfect and care must be taken when interpreting the model output, particularly for the purpose of regional or catchment scale impact assessments. Climate change science is an inexact science which is characterised by deeply embedded uncertainties that are propagated through the climate change impact assessment as illustrated in Figure 74. These uncertainties result in significant inter-model variability in climate projections. For example, a review of output from 10 GCMs for the Adelaide region found average annual precipitation projections for 2080s ranging from an 8% increase to a 35% decrease. No single greenhouse gas emission scenario or GCM can be considered ‘best’ and hence when considering future climate change projections for risk analysis it is recommended that an ensemble of GCMs and emission scenarios be analysed in an effort to quantify the magnitude of these uncertainties.

98 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

EmissionEmission scenariosscenarios

Global climate modeling

Temporal and spatial INCREASING UNCERTAINTY downscaling

Engineering modeling used for impact assessment

Figure 74 The propagation of uncertainties through the climate change impact assessment process

11.3 Climate change projections for Australia The Australian Greenhouse Office in cooperation with CSIRO and the Bureau of Meteorology (BoM) has developed regional climate change projections for Australia, to assist industry in understanding the likely magnitudes of climate change and assess the associated impacts. Based on the ensemble of GCMs used in AR4, regional projections for precipitation, temperature, wind, solar radiation and potential evaporation and relative humidity are provided. The extent of inter-model variability is accounted for with the inclusion of 10th and 90th percentile projections (reflecting the lowest 10% and highest 10% of the model results). An example of the output available is shown in Figures 75 and 76, with seasonal temperature and precipitation projections for precipitation and temperature for the 2050 ‘best guess’ (50th percentile) scenario. Seasonal projections and projections for other time periods, parameters and percentiles are available at www.climatechangeinaustralia.gov.au.

99 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Figure 75 Projected changes in temperature for the year 2050 (best estimate scenario)

100 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 76 Projected changes in annual precipitation for the year 2050 (best estimate scenario)

The above projections are indicative of the trends expected. While the changes will be region-specific, the Climate Change in Australia report (CSIRO, 2007), makes the following predictions: • Increasing temperatures, with the magnitude of increase highly dependent on the emission scenario. • A decrease in average annual rainfall across most of Australia, with winter and spring rainfall decreases greater than those for summer and autumn rainfall. • Minimal changes in solar radiation across most of Australia. • Small decreases in relative humidity. • Increases in annual potential evaporation. • An increase in drought occurrence, particularly over south-western Australia. • A substantial increase in fire weather at most sites in south-eastern Australia.

The evidence relating to the changes in the amplitude and frequency of the ENSO phenomenon is inconclusive. However the IPCC (2007) states that despite the uncertainty relating to ENSO changes,

101 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT in south-east Australia El Niño events are projected to become drier and La Niña events will tend to become wetter. Projected wind speed changes were not well defined due to high variability in model projections.

11.4 Implications for reservoir management The projected changes in regional climate have the potential to impact directly on the catchment hydrology, the thermal structure of the reservoir and the physical, biological, chemical and ecological processes occurring within the reservoir. These will combine to impact on the overall raw water quality and the frequency and magnitude of the potential hazards facing reservoir managers. This section details some of the potential impacts of climate change that reservoir managers need to consider when planning future operational strategies. This list is by no means exhaustive and as our knowledge of climate change improves and our understanding of natural and man-made system responses to the changes increases it is possible that additional impacts will be identified.

11.4.1 Changes to catchment hydrology One of the primary challenges that climate change will pose to water resource managers is hydrological changes. Decreases in precipitation combined with increases in temperature and potential evaporation lead to reduced stream flows, resulting in an increased risk of supply shortages. Hydrological modeling of the Myponga River catchment in South Australia has found that the projected reductions in rainfall are amplified in the stream flow projections by an average factor of 3.9. Similar results have been found noted in other Australian catchments, with Chiew (2007) finding that the ‘rainfall elasticity of runoff is between 2 and 3.5 with even greater reductions in drier catchments’.

Despite an overall decrease in annual average precipitation across the southern parts of Australia, GCMs indicate a possible increase in high rainfall events in some regions. The models point towards an increase in the day-to-day variability of rainfall with an increase in daily rainfall intensity (defined as rain per rainy day) combined with an increase in the number of dry days. The Climate Change in Australia Report states that ‘extreme daily rainfall tends to increase in many areas but not in the south in winter and spring when there is a strong decrease in mean rainfall’. Detailed modeling of the Myponga catchment showed no significant changes in the overall number of high rainfall (>25 mm/day) days, although for the 2050’s and 2080’s there was a slight projected increase in the number of high rainfall days over the summer months. Flood frequency analysis showed an increase in the magnitude of extreme daily flow events (average recurrence intervals greater than 10 years) for the 2020’s, however for the 2050’s and 2080’s the magnitudes of these events compared to the baseline scenario were reduced. For all future climate scenarios considered, the hydrological modeling showed an overall increase in the number of low flow (<25 ML/day) days.

An increase in the magnitude of extreme flow events may be associated with flash flooding and infrastructure damage. However it is more likely that the major impact facing reservoir managers will be deterioration in water quality attributable to the projected changes in day-to-day rainfall variability.

11.4.2 Changes to water quality Climate change will directly impact on reservoir water quality via two primary pathways - it will alter the water quality characteristics of the inflows as well as impacting on the physical, chemical and biological processes occurring within the reservoir. The potential changes to water quality and the possible implications for reservoir management are discussed further below.

11.4.2.1 Increased water temperatures The Intergovernmental Panel for Climate Change’s (IPCC) Fourth Assessment Report (AR4) projects increases in average air temperatures of 0.2°C per decade for the next two decades and 1°C per decade after that. However, under a high emissions scenario it is possible that temperatures could rise by up to 4°C above the 1961-1990 averages. Increases in air temperature will be reflected by increases in water temperature, with the following potential consequences: • Increases in the duration and intensity of thermal stratification. • Changes in the timing of the onset of stratification. • Changes in the depth of the riverine inflow intrusion.

102 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

• Altered biota composition both within the stream and reservoir. For example, the length of the potential growing season of nuisance algal blooms may be extended. • Changed chemical composition of the inflow and reservoir water including decreased dissolved oxygen concentrations. • Increased chemical reaction rates which will impact on the treatment process, in particular chlorine demand and disinfection by-product formation within the distribution system.

11.4.2.2 Changes to nutrient loads The impacts of climate change on reservoir nutrient loads will be site-specific and are strongly related to the catchment characteristics. For predominantly non-urban catchments, with no point pollution sources, reduced stream flow and a corresponding reduction in the number of ‘high flow’ days may result in reduced organic matter and nutrient loads being transported into the reservoir (e.g. Thorne and Fenner, 2008). In urbanised catchments with point pollution sources such as wastewater outlets, reduced stream flows provide less dilution capacity and hence higher concentrations of nutrients in the inflow stream (e.g. Mimikou et al., 2000).

11.4.2.3 Impacts of increased day-to-day variability The changes in the day-to-day variability of rainfall, characterised by an increase in the number of dry days along with an increase in the daily rainfall intensity will result in spikes in the nutrient and sediment concentrations in the reservoir inflows due to first flush effects. Such events may adversely impact on reservoir water quality through increased turbidity, decreased light availability and/or algal blooms.

High rainfall events are associated with an increased pathogen risk and hence a change in the frequency or seasonal distribution of such events may increase the risks posed by pathogen contamination at the reservoir off-take.

11.4.3 Impacts on reservoir hazards The aforementioned impacts of climate change on catchment hydrology and reservoir water quality do not themselves constitute hazards and it is important to understand how these changes will impact on the hazards facing reservoir managers. It is unlikely that climate change will change the nature of the hazards; rather they will impact on the frequency and severity of the events, thus creating a changed risk profile. A conceptual model of factors contributing to three hazards (cyanobacterial, metal and pathogen) is presented in Chapter 1 of this management guide. Review of the hazard model reveals that the overarching factors influencing all of the hazards are sediment and nutrient concentration, inflow volume and temperature, inflow rate and pathogen concentration. This chapter has demonstrated that climate change will influence the timing and magnitude of changes in all of these characteristics thus impacting on the occurrence and severity of the existing hazards facing reservoir managers.

11.5 Wildfires Wildfires are a further example of an existing hazard facing reservoir managers that may become more prevalent with climate change. The Climate Change in Australia report investigated the incidence of wildfires across south-eastern Australia under future climate scenarios and found ‘a likely increase in fire weather risk at most sites’. As discussed in Chapter 10 wildfires can initiate a series of adverse impacts on reservoir operations in the space of a short period of time (days) with consequences that may last for several decades. Typical impacts include an initial increase in runoff following an intense fire, a decrease in runoff quality with high levels of ash and charcoal increasing turbidity and then, as the catchment vegetation begins to re-establish, there will be a marked decrease in runoff volumes which may last for many decades (e.g. Howe et al., 2005).

11.6 Indirect - climate risks In addition to the altered risk profiles resulting from changes in regional climate, there are a number of associated risks resulting from changes in human behaviour and policy that may compound the projected changes and must be taken into account when developing future management strategies. These changes include (but are not limited to):

103 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

• Land use changes – not only will land use types within the catchment continue to change, but changes in regional climates may result in changed agricultural practises (such as increased fertiliser and pesticide use). Such changes may influence the quantity and quality of reservoir inflows. • Changes in demand – with a projected decrease in precipitation and increasing temperatures, there will be an increase in water demand particularly from the agricultural sector. Such increases may be counteracted by decreases in the domestic and industrial sectors due to mandatory water restrictions. The end result will be changed patterns of water demand which may impact on reservoir operations. • Alternative water supplies – in a push towards sustainability, alternative water sources are being developed, especially in the urban environment. Water saving schemes such as rainwater collection, grey water reuse and aquifer storage and recovery will further change the patterns of potable water demand. In urban catchments they may also impact on the quantity and quality of runoff.

11.7 Adapting the framework for monitoring hazard and risk assessment to incorporate future climate projections This chapter has presented a qualitative analysis of the potential impacts of climate change on reservoir operations. It has demonstrated that it is unlikely that climate change will present any new hazards; however it will alter the risk profile associated with existing hazards, presenting new challenges for reservoir management. To ensure that such risks are adequately addressed in the development of management plans, it is critical that information regarding future climate change is incorporated into the hazard identification and risk assessment framework presented in Chapter 1 of this management guide. To assist reservoir managers with assessing the potential future ‘situations’, synthetic climate time series that are representative of future climates should be fed into the hydrological, hydrodynamic and ecological models that are currently used to support management decisions.

Due to the inherently uncertain nature of climate change science, it is not possible to accurately define changes in future climate. Instead a scenario-based approach should be used to develop future climate projections based on the output from an ensemble of GCMs. An ensemble analysis helps to capture the magnitudes of uncertainty in the climate projections and will allow probabilities of occurrence to be attached to certain scenarios. Further difficulties in scenario development arise due to the mismatch in temporal and spatial scales of the GCM and decision support models. Regional climate change information at a seasonal timescale can be extracted from the Climate Change in Australia Report. However if the models require daily input and/or it is considered that changes in the day-to-day variability of certain climatic parameters are important in the impact analysis then downscaling of the GCM data is required.

Use of the synthetic climate time series that are representative of climate scenarios with associated probabilities of occurrence with the existing models will provide quantitative information about the potential changes to the risks facing reservoir operations, enabling managers to develop informed, risk-based planning strategies.

104 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

11.8 References Chiew F (2007) Climate Impacts on Water Availability. Presented at Hydrological consequences of climate change, CSIRO Cutting Edge Science Symposium, 15-16 Nov. CSIRO (2007) Climate Change in Australia. Pearce K, Holper P, Hopkins M, Bouma W, Whetton P, Hennessy K and Power S (eds), 137 pp. Hennessy K, Fitzharris B, Bates BC, Harvey N, Howden SM, Hughes L, Salinger J and Warrick R, (2007): Australia and New Zealand. Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, ML Parry, OF Canziani, JP Palutikof, PJ van der Linden and CE Hanson, (eds.), Cambridge University Press, Cambridge, UK, 507-540. Howe C, Jones RN, Maheepala S and Rhodes B (2005). Implications of Potential Climate Change for Melbourne's Water Resources. Melbourne Water and CSIRO Urban Water and Climate Impact Groups: 26pp. Mimikou MA, Baltas, E, Varanou E and Pantazis K (2000). Regional impacts of climate change on water resources quantity and quality indicators. Journal of Hydrology 234: 95-109. Thorne OM and Fenner RA (2008) Risk based climate change impact assessment for the water industry: a South Australian case study. Paper and poster presented at the IWA Young Professionals Conference, Brisbane, Feb. Earth System Grid (ESG) WCRP CMIP3 multi model GCM database. Available on-line at: https://esg.llnl.gov:8442/index.jsp.

105 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

APPENDIX

12 ARTIFICIAL DESTRATIFICATION – AERATOR DESIGN AND OPERATION

12.1 Introduction Artificial destratification is one of the main interventional management methods available to reservoir managers to control water quality in drinking water reservoirs.

A significant interest in artificial destratification developed in the 1970s in Australia with many researchers reporting improvements in water quality after its introduction (e.g. AWRC 1979, Bowles et al. 1979). It has been implemented in order to manage a number of water quality problems associated with thermal stratification, particularly release of soluble iron and manganese, midges and their larvae, nuisance cyanobacteria (blue-green algae) and restoration or maintenance of oxygenated habitat for ecological health. Artificial destratification has achieved good results in reducing iron and manganese problems for water treatment plants (Brookes et al. 2000, Burns 1998, Ismail et al. 2002). The results in relation to control of nuisance algae have been more variable. This is most likely due to the complex interaction of the effects of destratification on the availability of nutrients and light required for the growth of photosynthetic organisms such as algae and cyanobacteria.

Destratification systems operating in deep reservoirs (mean depth>15m) have generally been more successful in changing the composition of the phytoplankton community (e.g. Visser et al. 1996, Heo and Kim 2004), while studies in shallower water bodies have had less impact (Barbiero et al. 1996, Sherman et al. 2000). It is likely, in situations where artificial destratification has failed to reduce cyanobacterial growth, that neither nutrient nor light resources were limited sufficiently to impact on growth. Either there was a sufficient external load to provide adequate nutrients, and therefore limiting the internal load was inconsequential, or the mixing was inadequate to light-limit the cyanobacteria.

The access of light resources in cyanobacteria is enhanced by their ability to regulate their buoyancy. Many cyanobacteria contain structures called gas vesicles, which allow them to reduce the overall density of their cells. In combination with their ability to store carbohydrate fixed by photosynthesis, which then is used as a ballast during the day, they are able to further control their buoyancy. If they do not have access to light then they metabolise their carbohydrate ballast and become positively buoyant, while during the daylight hours they synthesise ballast and lose their positive buoyancy. Through this process they are able to regulate their vertical position within a stable water column, and therefore access the light near the surface (in the epilimnion) and the nutrients at depth in the hypolimnion. Furthermore there is an interaction between light and nutrients in determining the buoyancy of cyanobacteria where nutrient replete cells regain their buoyancy more rapidly than nutrient depleted cells (Brookes et al. 1999).

The successful reduction of cyanobacterial biomass by artificial destratification is dependent upon the depth to which the water column is mixed (Zmix) and the depth of the penetration of photosynthetically active radiation (PAR, 400-700nm). This is often given as the euphotic depth (Zeu) which is the depth to which 1% of the subsurface irradiance penetrates (Reynolds 1984). The ratio between these depths is often used to evaluate the potential for light to limit the growth of phytoplankton entrained in the surface mixed layer. For example a Zmix:Zeu ratio of 2.5 (Heo and Kim 2004) or 3 (Sherman et al. 2000) are regarded as ratios that have not supported cyanobacterial growth. Therefore the clarity of the water column also determines the Zmix:Zeu ratio, and if a water body is inherently turbid or coloured it is theoretically more suitable for the use of mixing as a control technique than a clear water column, because the euphotic depth is shallower.

Construction of artificial destratification infrastructure is a considerable capital investment, and as such, confidence in the effectiveness of the final implemented design is required. Bubble plume aerators are a popular choice due to their simplicity, lack of complex moving parts and the location of their components which require the most maintenance (the compressor) situated on land.

106 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Aerator Design An effective bubble plume aerator for destratification requires a suitable number of bubble plumes to impart sufficient energy to destratify the water column. The design process that arrives at a suitable aerator configuration is divided into a number of stages that ensure confidence in the effectiveness of the commissioned infrastructure (Figure 77). The design process consists of a number of stages, which are described here briefly and in more detail below.

Stage 1 Conceptual Design Prediction of the total aerator flow rate and number of bubble plumes using models of bubble plumes and the characteristics of the stratification found in the reservoir.

Stage 2 Hydrodynamic modeling of the conceptual design Modeling the effectiveness of the conceptual design using real meteorological and thermal structure data. Further implementation of ecological models to assess the impact on dissolved oxygen, dissolved metals, phytoplankton dynamics and cyanobacterial biomass is possible where biological and chemical information is available.

Stage 3 Pneumatic and practical design Any installation must consider the availability of power, the practicality of construction site access and the aerator diffuser design. Aerator diffuser line design includes calculations relating to line operating pressure, pressure loss, line diameter, port configuration and separation.

Stage 4 Construction Unforseen issues may arise during construction that may require the reconsideration of certain minor aspects of the aerator design and configuration. This stage will not be considered in this chapter.

Stage 1. Conceptual Design Process

Stage 2. Hydrodynamic Modeling of the Conceptual Design

Stage 3. Pneumatic and Practical Design

Stage 4. Construction

Figure 77 Steps of aerator design showing feedback between stages.

107 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

12.2 Detailed Stage Descriptions 12.2.1 Stage 1 Conceptual Design The conceptual design process aims to predict the flow rate and number of plumes of the aerator required to destratify the reservoir and therefore determine the capacity of the compressor required. There are a number of ways of estimating the capacity of an aerator system required for a given water body. Lake morphology, such as surface area (Lorenzen and Fast 1977) or volume has been used. Other design methodologies have assumed a fixed efficiency of mixing of the water column by the bubble plume (Davis 1980). However, as the mechanical efficiency of a plume changes with the degree of thermal stratification of the water column and interactions with other plumes, these design methodologies make untenable assumptions. This is particularly relevant for shallow water columns, where bubble plumes tend to have low mechanical efficiencies, often less than the 5% assumed by Davis (1980). The changing efficiency of a bubble plume in stratified fluid can be described by the single core model of Schladow (1992) or the double core model of Asaeda and Imberger (1993) and are applied in the design methods of Lemckert et al. (1993), Schladow (1993) and as modified in Sahoo and Luketina (2003).

In these models, the bubble plume is conceptualised as a rising mass of buoyant air which entrains water (Figure 78). When the upward forces of buoyancy are equal to the downward forces of gravity the rising water becomes disentrained and forms an outward radiating plume of neutral buoyancy. From the point of disentrainment, the bubble plume continues to rise and entrain water until another disentrainment event occurs or the bubble plume reaches the surface. At this point the excess kinetic energy contained in the vertical movement is dissipated as the disruption of the water surface in the form of waves and sound, for example. Therefore, the maximum efficiency of mixing (volume water entrained per unit energy supplied) is achieved when the rising bubble plume disentrains immediately below the surface and the minimum amount of energy is ‘wasted’ on the disruption of the water surface. Therefore, the greatest efficiency will be found when the number of entrainment- disentrainment events is close to an integer value.

Figure 78 Diagram of a bubble plume, showing a single entrainment and disentrainment cycle.

The characteristics of the bubble plume are commonly described in terms of two non-dimensional parameters; C and M represent the degree of stratification and the source strength of the bubble plume, respectively. If the mechanical efficiency of the bubble plume is described in terms of these parameters then these peaks in efficiency can be seen in Figure 79. The volume of water entrained by a plume of certain characteristics can then be calculated and the time taken to circulate the volume of the reservoir is estimated. As a general rule, bubble plumes are more efficient in deeper water columns. In shallow water columns (< 5.0 m depth) the individual air flow rates of the plumes must be very small to maintain efficiency.

108 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 79 The relationship between M, C and mechanical efficiency

Stage 1. Preparation The first step of each stage in the aerator design process is to gather the relevant information required to perform the calculations. If some of the more vital information is unavailable it may be necessary to make appropriate measurements in order to complete the process satisfactorily. A checklist of the required information is given in Table 12.

Table 12 Check list of required information for Stage 1 of aerator design process. x Data required Map of reservoir hypsography (reservoir depth contours, depth vs volume curves) Information about suitable site placement for compressor. Consider: • vehicle access • power • proximity to deepest part of the reservoir Maximum reservoir volume (not service volume) Maximum depth of reservoir (true depth not just gauge height) Seasonal water temperature data from deepest location possible • maximum summer surface temperature • minimum winter bottom temperature An initial estimate of the maximum length of the diffuser line. This may be determined by: • reservoir morphology • capital costs

Stage 1. Calculations

The following is essentially a reproduction of the method described in Lemckert et al. (1993). 1. Choose a design stratification to which the aerator will be designed. If the reservoir is currently not artificially mixed then this should be the most severe stratification experienced. If a good temperature profile of these conditions is not available then a synthetic data set can be created by joining the minimum winter bottom temperature with the maximum summer surface temperature with a sigmoid curve. The equivalent linear stratification is then determined using an equation to calculate the linear stratification that has the equivalent potential energy as the sigmoid stratification. 2. Determine the most efficient and therefore appropriate individual plume air flow rates using equations 1-3. Depending on the depth of the water column there will be multiple peak

109 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

efficiencies. Note that the flow volume is given at the diffuser depth pressure NOT the free air flow rate at atmospheric pressure. 3. Determine the total number of plumes and therefore total flow rate required to achieve circulation of the entire volume of the reservoir in a given period of time (usually 3 weeks) using equations 4 to 6. 4. Calculate the mechanical efficiencies and compare the two methods of calculating destratification time (Eq. 5 & 6). This step is to test if the stratification encountered is suitable for this design methodology. We plan to make a numerical simulation of the destratification system in the next stage of the process even if there isn’t a large discrepancy in the two predicted values. 5. Determine the distance between plumes required to prevent interactions (Eq. 1). If the maximum length of the diffuser line is less than number of plumes multiplied by the required separation distance then plume interaction must be considered. 6. If so, then a compromise between the preferred per plume flow rate and the number of plumes to be applied must be found. This can be achieved using the typical reduction in mixing efficiency found during plume-plume interactions (0.75; Robertson, et al. 1991). However, plume interactions are not considered in DYRESM (stage 2) and are therefore preferentially avoided from the perspective of efficiency and our confidence in the modeling of the device.

Equations: 23 4πα hvs QM = g Eq.1 Nh34 QP = g Eq. 2 Q Q = P R Q M Eq. 3

α = entrainment coefficient (0.083) h = head of water above diffuser (m) -1 vs = slip velocity of bubbles relative to the rising bubble plume (0.3 ms ) g = acceleration due to gravity N = buoyancy frequency (Brunt Vaisaila frequency)

1 *34 *0⎛⎞()QgB .11 QMfH= 0.56⎜⎟ N 5 ⎝⎠E Eq. 4 * Qf = effective bubbler flow through rate * QB = volumetric air flow rate at diffuser pressure (m3s-1) NE = buoyancy frequency of equivalent linear stratification MH = represents source strength compared to pressure head

V Eq. 5 TV = ** mQf

TV = time taken to destratify the reservoir V = reservoir volume m* = number of diffuser ports

ΔPE TC = Eq. 6 hPC

110 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

TC = time taken to destratify the reservoir ΔPE = change in potential energy during destratification PC = the work per unit time done by the compressor

1 * 4 * ⎛⎞Qg0 ls = 3.88⎜⎟3 Eq. 7 ⎝⎠NE

* ls = separation required to avoid plume interaction * Q0 = volumetric air flow per diffuser

12.2.2 Stage 2 Hydrodynamic modeling of the conceptual design

To extend the conceptual design applied in stage 1 and provide further confidence in the predicted aerator configuration and capacity, a hydrodynamic modeling approach is applied. The use of the 1-dimensional (1D) hydrodynamic model DYRESM (www.cwr.uwa.edu.au) is described here. However if the system under consideration is strongly influenced by lateral factors and therefore violates the assumptions of applying a 1D model then a more sophisticated 3D modeling approach may be required. This is most likely to occur if the lake or reservoir is especially shallow or its heat budget is dominated by large volumes of in- and out- flowing water, relative to the storage volume.

Hydrodynamic modeling has significantly greater data demands than the initial conceptual design. It is strongly recommended to perform the hydrodynamic modeling, even if entirely site-specific data is unavailable. Data from nearby locations can be adapted for the purposes of the modeling if necessary. The 1D hydrodynamic model developed at the Centre for Water Research (DYRESM see http://www.cwr.uwa.edu.au/) is suitable for use at this stage. It may be necessary to consider the license aspects of using of this software in a commercial environment. Of course, other hydrodynamic models may also be used; however this design process will assume the use of the CWR package. Given the high capital investment of a new destratification system it is worth investing in some data collection prior to and during the design phase to ensure the input data is of good quality.

DYRESM is used to predict the variation of water temperature and salinity with depth and time. The hydrodynamic component is a based model (as opposed to being empirical) and therefore does not require calibration. The success of any simulation is determined by the quality of the data and the validity of the assumptions. DYRESM is based on a Lagrangian layer scheme, where the lake is modeled by a series of horizontal layers of uniform property but variable thickness. The layer positions change as inflow, outflow, evaporation and rainfall affect the stored volume, and layer thicknesses change as the layers are moved vertically to accommodate volume changes. The additional advantage of this layer scheme is that it lends itself to the vertical structure of the lake. For example, if the properties of part of the water column are mostly constant over depth, this part of the water column can be modeled as a single layer. On the other hand, in parts of the water column where significant gradients occur the layers can be narrower to provide better description of the gradient of, for example, temperature, oxygen or salinity.

The stability of layers is determined by comparing adjacent layer densities starting at the surface layer and working downwards; if the density of the upper layer is greater than the one under it, then the layers are amalgamated. When layers merge, temperature, salt, energy and momentum are conserved. The surface dynamics are predominantly driven by the exchange of heat, mass and momentum between the surface layer and the atmosphere. These surface exchanges are responsible for the input of the majority of the energy for heating, mixing and stratifying the lake. The surface exchanges include heating due to short wave radiation penetration into the lake and the fluxes at the surface due to evaporation, sensible heat (i.e. convection of heat from the water surface to the atmosphere), long wave radiation and wind stress. The equations that describe these surface fluxes are beyond this chapter and can be found in the model science manual (available at http://www.cwr.uwa.edu.au/)

Stage 2. Preparation Again, preparation for stage 2 is a process of preparing the data required for the modeling exercise in the formats required for the input files of the model. Some conversion of certain data may be required,

111 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT generally using established methods as described in the model documentation. It is important to perform rigorous data quality control before attempting to run the model as any modeling performed will need to be repeated if erroneous data is subsequently discovered. The data required for the stage 2 calculations is shown in Table 13.

The period chosen to ‘test’ (by simulation) the aerator design configurations should preferably include periods of stability and surface heating and periods of unstable, windy weather, where the reservoir is expected to be well mixed. The performance of the model can then be examined under a range of conditions to ensure that the input data selected is suitable.

Table 13 Check list of required information for Stage 2 of aerator design process. x Data required (in addition to that used in stage 1) Reservoir hydraulic budget • inflows • outflows • change in reservoir level • estimated or measured evaporation Meteorological data and information about recording site and instrumentation • solar radiation • wind speed • wind direction (especially if long narrow reservoir) • air temperature Details of the current destratification device (if applicable) • operation logs • free air flow rate • depth distribution, separation and number of diffusers Water temperature data for model performance validation: • a number of vertical temperature profiles through the period to be modeled • on-line monitored temperature data at a number of depths (preferred)

Stage 2 Calculations The first goal of this stage of the design is to establish a hydrodynamic model that satisfactorily captures the surface layer dynamics of the reservoir before the implementation of the new aerator, whether that be without destratification or under the influence of a previous destratification device. To do this, the model is set up to run with a set of data and the output is compared with (validated against) a time series of data collected from the reservoir.

The comparison of the model output to the field data can be performed through simple visual means or using more time consuming numerical methods, such as generating fit statistics, such as modeling efficiency (see Mayer and Butler 1993).

In the event that significant discrepancies are found between the modeled values and measured data, then the reasons for this should be found and corrected. Check: 1. The time of day the data were collected compared to the model output 2. The prediction of surface and depth temperatures during stable (calm) and unstable (windy) periods 3. The prediction of surface mixed layer depth during stable, unstable and intermediate periods 4. Proximity of weather station data to reservoir and potential influence of topography on differences in weather actually experienced by the reservoir. Compare with other meteorology data from other adjacent sites. 5. Check the quality of your meteorology data. How often was the instrument calibrated or checked? 6. Did you account for the altitude of the wind anemometer? 7. Is your reservoir particularly influenced by topography, sheltered by trees or elongated?

112 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Once a satisfactory simulation of the before design device stratification is achieved, then the design configurations may be implemented. This is facilitated by altering the destratification file (*.mix?), implementing the design as a total flow rate and a number of plumes. Again, these plumes are simulated assuming that the plumes do not interact, so if they do, due to diffuser length and plume separation, an appropriate reduction in flow rate should be implemented.

Design configurations The design configurations that are simulated to test whether they are suitable for the reservoir again depend on the water quality issues being managed and what the goals of the system are. The results should be assessed on the basis of the probable effect of the water column structure on the water quality issue of concern. This can be further assessed by the application of coupled chemical and ecological models (for example CAEDYM), however this chapter will only consider more simple indices of mixing and its effects on water quality hazards. Iron (Fe) and manganese (Mn) are generally controlled when dissolved oxygen is maintained above 50% saturation (Brookes et al. 2000, Wetzel 2001). Consequently the degree of mixing that is effective to significantly reduce the flux of Fe and Mn from the sediment is far less than that for phosphorus, for example, which is predominantly mobilised under anoxic conditions.

Determining the potential effect on phytoplankton is more difficult and should include calculations of the changes in euphotic depth to mixed depth ratio. Further calculations of the daily light dose experienced in the surface layer will give better indications of the potential to reduce the growth of a certain species, usually a taste, odour or toxin producing blue-green algae. The growth of a number of common blue-green algae under certain daily light doses can be found in the literature (Reynolds 1984, 1997).

12.2.3 Stage 3 Pneumatic and practical design The final ‘design’ stage of the design process consists of performing the relevant calculations to determine the physical characteristics of the system. The characteristics to be determined include: the flow and operating pressure demands for the compressor, the size and configuration of the diffuser holes. In general, the diffuser line should be operated in the deepest part of the reservoir. This provides the greatest distance of water column to entrain water and facilitate mixing.

The following calculations are suitable for determining the port configuration and were modified from Patterson and Schladow (1992).

Diffuser Holes: Diameter and Layout The type of diffuser ports used should be kept simple, single holes or clusters of holes between 1.0 and 1.5 mm are commonly used. The appropriate size and number of holes can be determined by calculating the flow through a single hole using:

0.07ψ (CPIN A ) QH = 1 T 2

QH = volume of air flow through single diffuser hole C = coefficient of discharge A = hole area (mm2) T = internal absolute air temperature (°K) ψ = flow coefficient, which varies between 0.0 and 0.48 according to:

1.43 1.71 0.5 ⎡⎤⎛⎞PP ⎛⎞ ψ =−1.867 ⎢⎥⎜⎟22 ⎜⎟ ⎢⎥⎝⎠PPIN ⎝⎠IN ⎣⎦

P2 = external hydrostatic pressure PIN = internal pipe pressure

113 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

If [P2/PIN] < 0.5 then the flow through the whole is super-critical and independent of hydrostatic pressure (ψ = 0.48). When [P2/PIN] > 0.5 then the flow is subcritical and ψ is dependent on [P2/ PIN], and fluctuates between 0 and 0.48. Most aerator systems are designed to operate at sub-critical flows as super-critical flow is inefficient and requires higher capacity compressors. In sub-critical conditions flow per hole fluctuates according to [P2/ PIN] and therefore with depth and losses from the delivery system. Some variation can be expected from the flow distribution, especially if the installation follows the sloped benthos of the reservoir.

The choice of the diameter and the number of holes per port is essentially a matter of trial and error, with the simplest configuration being a single hole per port. The reality is likely to be that a number of holes are required, to balance the total hole area and the technical difficulty of drilling a sufficiently small hole (less than 1.0 to 1.2 mm). In multi-hole port configurations it is suggested that one or more holes be drilled on the underside of the pipe to ensure proper drainage of any water that enters the pipe during installation. All drilled holes should be carefully deburred. Selection of the pipe pressure capacity should include significant safety margins and scoring of the pipe should be avoided; preservation of diffuser pipe strength is far more important than exact hole placement.

The port flow equations are resolved for the series of ports on the basis of their respective depths dependent on their distance from the first port, which should be located at the deepest location in the reservoir. As there is pressure head loss due to both the pipe internal surface roughness (a simple exponential decay) and loss of air through the ports, the loss of pressure over the length of the line is balanced with the decrease in hydrostatic head as the diffuser line runs into shallower water. Therefore a series of calculations can be constructed to determine the supply line pressure and flow rate required to operate the aerator. From these values a suitable compressor can be chosen, after accounting for the head loss associated with lubrication oil filtration devices and other fittings.

12.2.4 Summary • Reservoirs can be artificially destratified to disrupt thermal stratification, control cyanobacteria and reduce sediment release of contaminants. • The most common and effective artificial destratification devices are bubble plume aerators, consisting of a submerged perforated pipe through which air is pumped from a land-based compressor. • The rising air bubbles expand and entrain water from throughout the water column into the bubble plume. • When the plume reaches the surface, the bubbles are released to the atmosphere and the entrained water plunges to the depth of equivalent density (temperature) and moves through the reservoir. • This flow displaces water and creates a return flow, generating circulation and weakening the prevailing stratification. This enables deeper mixing generated by wind and these deeper mixing events can induce light limitation in cyanobacteria (see Figure 80).

114 CRC FOR WATER QUALITY AND TREATMENT – RESEARCH REPORT 67

Figure 80 Bubble plume aerators consist of a submerged perforated pipe through which air is pumped.

Mixing is generated and stratification is weakened. Surface heating and stratification can still occur away from the immediate impact of the bubble plume.

12.3 References Asaeda T and Imberger (1993) Structure of bubble plumes in linearly stratified environments. Journal of Fluid Mechanics 249: 35-57. AWRC (1979) Various Destratification of Lakes and Reservoirs to improve water quality, Melbourne, Australia, Australian Government Publishing Service. Barbiero RP, Speziale BJ and Ashby SL (1996) Phytoplankton community succession in a lake subjected to artificial circulation. Hydrobiologia 331(1-3): 109-120. Bowles BA, Powling IJ and Burns FL (1979) Effects on water quality of artificial aeration and destratification of Tarago Reservoir. Canberra, Department of National Development, Australian Water Resources Council: 239. Brookes JD, Burch M and Tarrant P (2000) Artificial destratification: Evidence for improved water quality. Water 27(4): 18-22. Brookes JD, Ganf GG, Green D and Whittington J (1999) The influence of light and nutrients on buoyancy, filament aggregation and flotation of Anabaena circinalis. Journal of Plankton Research 21(2): 327-341. Burns FL (1998) Case study: Automatic reservoir aeration to control manganese in raw water Maryborough town water supply Queensland, Australia. Water Science and Technology 37(2): 301-308. Davis JM (1980) Destratification of reservoirs - a design approach for perforated-pipe compressed-air systems. Water Services 84: 497-504. Heo WM and Kim B (2004) The effect of artificial destratification on phytoplankton in a reservoir. Hydrobiologia 524(1): 229-239. Ismail R, Kassim MA, Inman M, Baharim NH and Azman S (2002) Removal of iron and manganese by artificial destratification in a tropical climate (Upper Layang Reservoir, Malaysia). Water Science and Technology 46(9): 179-183. Lemckert CJ, Schladow SG and Imberger J (1993) Destratification of reservoirs. Some rational design rules. 15th Federal Convention of the Australian Water and Wastewater Association, Gold Coast, Australia, Australian Water and Wastewater Association. Lorenzen MW and Fast AW (1977) A guide to aeration/circulation techniques for lake management. Ecological Research Series EPA-600/3-77-004. Mayer DG and Butler DG (1993) Statistical validation. Ecological Modeling 68:21-32.

115 A PRACTICAL GUIDE TO RESERVOIR MANAGEMENT

Patterson JC and Schadlow SG (1992) Destratification of Lower Hollywood Reservoir. Report to Los Angeles Department of Water and Power. Centre for Water Research, University of Western Australia, Nedlands, Western Australia. Reference: WP 754 JP. Reynolds CS (1984) The Ecology of Freshwater Phytoplankton. Cambridge, Cambridge University Press. Reynolds CS (1997) Growth of pelagic plants. Vegetation processes in the pelagic: A model for ecosystem theory. O. Klinne. Oldendorf/Luhe, Ecology Institute: 101-128. Robertson DM, Schladow SG and Patterson JC (1991) Interacting bubble plumes: The effect on aerator design. In Environmental Hydraulics, J.H.W. Lee and Y.K. Cheung (eds), Proceedings of the Symposium of Environmental Hydraulics, Hong Kong, December 1991, A.A. Balkema, Rotterdam, Netherlands, pp 167-172. Sahoo GB and Luketina D (2003) Bubbler design for reservoir destratification. Marine and Freshwater Research 54(3): 271-285. Schladow SG (1992) Bubble plume dynamics in a stratified medium and the implications for water- quality amelioration in lakes. Water Resources Research 28(2): 313-321. Schladow SG (1993) Lake destratification by bubble-plume systems: Design methodology. Journal of Hydraulic Engineering ASCE 119(3): 350-368. Sherman B, Whittington J and Oliver R (2000) The impact of artificial destratification on water quality in Chaffey Reservoir. Archiv für Hydrobiologie Special Issues Advances in Limnology 55: 15-29. Visser PM, Ibelings BW, vanderVeer B, Koedood J and Mur LR (1996) Artificial mixing prevents nuisance blooms of the cyanobacterium Microcystis in Lake Nieuwe Meer, the Netherlands. Freshwater Biology 36(2): 435-450. Wetzel RG (2001) Limnology. R. G. Wetzel. New York, Academic Press

116 Water Quality Research Australia Membership at December 2008