use of a bayesian network decision tool to mgt environ flow in river

Upload: initiative1972

Post on 03-Jun-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    1/84

    Report No 4

    LWA/MDBC Project UMO43: Delivering Sustainability

    through Risk Management

    Use of a Bayesian Network Decision Tool toManage Environmental Flows in the

    Wimmera River, Victoria

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    2/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    3/84

    Executive Summary

    This report - Use of a Bayesian Network Decision Tool to Manage Environmental Flows in

    the Wimmera River, Victoria- is the forth in a series of five produced by LWA/MDBC

    project UMO43 Developing Risk-based Approaches for Managing Contaminants in

    Catchments.

    Decision making in ecosystem management is a process of balancing multiple objectives,

    constraints, trade-offs and uncertainties against a complex backdrop of socio-economic,cultural and political considerations and limited ecological understanding. A central

    challenge in providing credible, effective and defensible decision support therefore, is to

    provide and apply frameworks and methods that will allow informed choices by providing

    opportunities for genuine, substantive participation in decision making supported by best

    available scientific knowledge that also incorporates uncertainty in an honest, rigorous and

    consistent manner.

    Using Ecological Risk Assessment (ERA) as an organizing framework, stakeholder

    engagement, risk-based modelling and decision-analytic techniques were developed and

    applied to a case study of environmental flow management in a degraded, semi-arid lowland

    river in the Wimmera Catchment in Victoria. The case study focussed on the management of

    the summer environmental flow regime to maintain adequate instream habitat and water

    quality for aquatic biota in the Lower Wimmera River.

    Formal management and expert stakeholder workshops were used to obtain stakeholder

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    4/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    5/84

    Table of Contents

    Executive summary

    1. Background ... 12. Ecological risk assessment .. 33. Wimmera catchment ....3.1.General description3.2.Environmental values

    5

    5

    7

    4. Problem formulation ...4.1.Ecological management issues in the Wimmera River4.2.Managing environmental flows in the Lower Wimmera under uncertainty4.3.

    Planning, problem formulation and stakeholder involvement

    4.4.Stakeholder workshops4.5.Workshop outcomes

    10

    10

    12

    1517

    19

    5. Development of the Bayesian network model for Freshwater Catfish in theLower Wimmera River

    5.1.Bayesian network modelling5 2 O i f th d l t d t ti f th F h t C tfi h BN

    30

    30

    32

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    6/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    7/84

    Figure 8: Predicted probability distributions for variables of management interest. The

    worst case Dry scenario is represented with stippled bars and the SKMflow strategy is shown as solid grey bars 55

    Figure 9: Variation in expected utility for a fixed volume of 240 ML allocated over 1-

    4 freshes for median MDF values of 0, 10 and 30 ML/day 57

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    8/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    9/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 1

    1. Background

    Natural resource managers currently have few quantitative tools to assist them in identifying

    which of their environmental assets are at greatest risk from ecological degradation and then

    to decide upon the best options for managing these risks.

    Land & Water Australia (LWA) and the Murray Darling Basin Commission (MDBC) have

    funded project UMO43Developing Risk-based Approaches for Managing Contaminants in

    Catchments todevelop quantitative risk-based assessment guidelines that will offer a new

    framework and guidance to assist in improving the management of diffuse contaminants in

    catchments. These guidelines have been built around qualitative and quantitative models (the

    backbone of risk assessment) that link catchment contaminant reduction targets (end-of-valley

    targets for nutrients, salinity, SPM and pesticides) with the ecological benefits in receiving

    waterbodies.

    The project was undertaken through a collaborative partnership between Monash University,

    the University of Melbourne and CSIRO.

    The project objectives were:

    To develop guidelines for risk-based approaches that can be used to prioritise ecologicalrisks from multiple contaminants in a catchment context (i.e. multiple-stressors resulting

    in multiple issues),

    To develop quantitative models of ecosystem components and relationships to addressk i ( ) f

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    10/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 2

    These reports are all available at www.sci.monash.edu.au/wsc.

    This document is Report 4 of the series and covers the development of a Bayesian Network

    decision tool to assist in the management of environmental flows in the lower Wimmera

    River.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    11/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 3

    2. Ecological Risk Assessment

    Ecological risk assessments evaluate the ecological effects caused by natural events and

    human activities and typically are defined as systematic, formal processes of estimating the

    likelihood of occurrence of adverse events/circumstances and the magnitudes of

    consequences to ecological values that result from these events (Burgman 2005). ERA is

    carried out within a broader framework that includes problem formulation, hazard

    identification and assessment, conceptual modelling, endpoint selection, risk analysis,

    sensitivity analysis, communication of results, decision-making, monitoring, review andupdating (Figure 1). As shown in Figure 1, this is an iterative and cyclic process. Conceived

    in this perspective as a megatool, ERA can facilitate the integration of social, scientific and

    policy dimensions and serve as a comprehensive organizing framework for environmental

    problem-solving in complex settings.

    Learning is an elemental feature of the process and completion of the cycle ensures that (a)

    the appropriate questions are being investigated; (b) a wide range of potential hazards, or

    threats to what stakeholders value and want to protect are canvassed; (c) conceptual modelsof relationships between important interacting variables in the system under study are

    explicated and explicitly communicated; (d) assumptions are probed; (e) risks are quantified

    as rigorously as possible; (f) results are communicated; (g) feedback is collected; (h)

    predictions are validated and (i) knowledge is updated (Burgman, 2005).

    P bl F l tiP bl F l ti

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    12/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 4

    coincide with the publicly-meaningful attributes that are of concern to decision makers

    (Reckhow, 1999a).

    In tailoring the traditional ERA framework for use in complex natural resource management

    problems on large spatial, temporal and ecological scales, the most important innovations are

    an increased emphasis on: communication with stakeholders; stakeholder participation in the

    tasks of problem formulation, hazard identification, endpoint selection and conceptual

    modeling and iteration between the various stages involved in conducting an ERA (see e.g.

    McDaniels et al., 1999; Borsuk et al., 2001; USEPA, 2001, 2002; Serveiss, 2002). On the

    technical side, a great deal of effort is also being invested in the development of modelling

    tools to address the problems of trying to estimate risks from diverse multiple stressors. This

    paradigm recognizes that scientific knowledge is critical for ensuring that the risk assessment

    addresses all important ecological concerns and characterizes uncertainty adequately but at

    the same time acknowledges the validity of socio-cultural perceptions of risks and provides

    procedural mechanisms for incorporating these in the risk assessment process.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    13/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 5

    3. The Wimmera Catchment

    3.1 General description

    The Wimmera region is located in north-west Victoria, Australia and covers almost 30,000

    km2, or about 13% of Victorias total land area. The dominant land use is agriculture, mainly

    dryland cropping of cereals, grain, pulses and oilseeds (Figure 2).

    The climate in the Wimmera region varies from temperate in the Grampians and Pyrenees

    Ranges in the south-east to semi-arid in north (LCC, 1985). Rainfall is highest over the

    ranges in the south and south-east of the catchment, being over 1,000 mm and 700 mm per

    annum in the Grampians and the Pyrenees respectively. In the plains regions of the

    Wimmera, mean annual precipitation steadily decreases from south to north, from 576 mm at

    Stawell, to 449 mm at Horsham, to 384 mm at Jeparit and 368 mm at Rainbow (Bureau of

    Meteorology, 2002). Most of the rainfall occurs over winter and spring. Annual areal

    potential evapo-transpiration in the Wimmera region is high and greatly exceeds annual

    precipitation in all areas except in the south and south-eastern ranges. It varies from about1,100 mm in the south to about 1,200 mm in the north (Wang et al., 2001).

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    14/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    15/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 7

    Flows in the Wimmera River are highly seasonal with most flow occurring in the winter-

    spring period. Low or zero flows typically occur from December to May. Periods of lowflow (consecutive days with less than or equal to 5 ML/day) have lasted for up to 965 days at

    Horsham.

    The major aquifers of the Wimmera region are the unconfined Parilla Sand aquifer, the

    Murray Group Limestone aquifer and the basal Renmark Group aquifer. Finer grained units

    such as Ettrick Marl, Geera Clay, Winnambol Formation, Bookpurnong Beds and

    Blanchetown Clay behave as aquitards thoughout the region (Macumber, 1990). The Parilla

    Sand and Renmark Group aquifers extend throughout most of the Wimmera, but the Murray

    Group aquifer is restricted to the west of the Wimmera River (Weaver et al., 2002). Within

    the Wimmera region, water quality in the Murray Group Limestone is relatively fresh and

    consistent with a total dissolved solids (TDS) content of approximately 1,500 mg/L. In

    contrast, water quality varies greatly over a small spatial extent within the Parilla Sand (TDS

    650-115,000 mg/L) and Renmark (TDS 600-16,000 mg/L) aquifers (Weaver et al., 2002;

    Swane, 2004).

    3.2 Environmental values

    Since European settlement, about 86% of the total catchment area has been denuded of

    native vegetation (WCCG, 1991). Most of the native vegetative cover that remains is in

    National Parks, State Parks, Flora and Fauna Reserves and Crown land water frontage

    (Fi 3)

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    16/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 8

    3.3 Development of Water Resources: The Wimmera-Mallee System

    European settlement of the Wimmera region began in the 1850s and water regulation works

    commenced soon after, in order to ensure a more reliable supply of water for domestic, stock

    and irrigation purposes, in this region of highly variable streamflow. The resultant

    Wimmera-Mallee system (previously known as the Wimmera-Mallee Stock and Domestic

    System, WMSDS) harvests and distributes water in a complex system of impoundments,

    weirs, diversions, on and off-stream storages and distribution channels (Figure 4). TheWimmera-Mallee system is administered by the rural water authority, Grampians Wimmera

    Mallee Water (GWMW) (formerly Wimmera-Mallee Water, WMW).

    The Wimmera-Mallee system services a total area covering about 30,000 km 2and supplies

    15,760 properties and an estimated 4,500 farming enterprises through a total length of

    approximately 14,000 km of open channels and 32,000 km of rural pipelines. The system

    provides water supply services to approximately 52,000 people in 74 towns throughout the

    region and domestic and stock water supplies for approximately 7,000 people throughout the

    region.

    The water supplied to the Wimmera-Mallee area is provided principally from the Glenelg

    River and Wimmera River and its tributaries. The headworks system consists of 12 storages

    in and around the Grampians, together with their connecting natural streams and man-made

    channels (Figure 4). The system interacts directly with the Wimmera River at Glenorchy and

    H ddl W i O di i f h Wi Ri f H h

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    17/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 9

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    18/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 10

    4. Problem Formulation

    Under the new Victorian Water Allocation Framework, Catchment Management Authorities

    (CMAs) have been invested with the task of managing environmental flows (DSE, 2004).

    This case study of an ERA for environmental flow management in the Lower Wimmera

    River was undertaken in partnership with the Wimmera Catchment Management Authority

    (WCMA), the Environment Protection Authority Victoria (EPA) and a Land and Water

    Australia/Murray Darling Basin Commission (LWA/MDBC) project on Developing Risk-

    based Approaches for Managing Contaminants in Catchments (Contaminants RiskAssessment) (LWA/MDBC Project UMO43, 2005). The project represents a collaborative

    attempt by the scientific, regulatory and natural resource management community to develop

    a system for enhancing the quality of environmental decision making by engaging

    stakeholders and integrating substantive stakeholder concerns into scientific analyses as part

    of a holistic problem solving approach.

    4.1 Ecological management issues in the Wimmera River

    Since European settlement, the Wimmera River and its surrounding catchment has been

    heavily modified by extensive vegetation clearance for agriculture and urbanization and

    intensive river regulation. While the development of land and water resources for

    agricultural, urban and industrial purposes has brought much socio-economic benefit to the

    region and the state, it has exacted a high environmental cost.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    19/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 11

    The primary water quality issue in the Wimmera River is high salinity levels. Salinity has

    historically been high upstream of Glenorchy, decreasing below Glenorchy Weir due to thediluting effect of water entering from the weir pool from the Wimmera-Mallee system and

    then increasing downstream towards the terminal lakes (EPA, 1984). In addition to high

    surface water salinity, the lower reaches are subject to saline groundwater intrusions from

    the Parilla Sand aquifer, leading to localized occurrences of extremely high bottom water

    salinity (eg. 30,000-56,000 EC) which in turn leads to salinity-induced density stratification

    of the water column (Anderson and Morison, 1988a,b).

    In the Lower Wimmera River, the major in-stream problem from an environmental point of

    view is not salinity per se, but rather salinity-induced density stratification of the water

    column, which results in restricted vertical mixing in the water column during low flow

    periods and subsequent severe and persistent deoxygenation (with levels of dissolved oxygen

    (DO) frequently

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    20/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 12

    4.2 Managing environmental flows in the Lower Wimmera under uncertainty

    As caretakers of river health, the Wimmera CMAs primary responsibilities in terms of

    operational management of the Environmental Water Reserve (EWR) to address the problem

    of altered flow regimes and water quality issues in the Wimmera River are as follows

    (WCMA, 2003a):

    Implement Stressed Rivers Project: Environmental Flows Study recommendations bySinclair Knight Merz (SKM, 2002a),

    Continue to investigate the environmental flow requirements of tributaries anddistributaries, in light of new understanding and environmental water requirement

    priorities,

    Monitor and review the delivery and effectiveness of environmental flows.Studies of environmental flow requirements for the Wimmera River as part of the BulkEntitlements Conversion process have resulted in detailed environmental flow

    recommendations for the Lower Wimmera River (SKM, 2002a,b). Essentially, a modular

    approach was adopted for characterizing the complete flow regime, thus allowing different

    flow types/events to be represented as separate flow components. The full suite of

    recommendations, couched in terms of flow components, and the underlying rationale and

    specific management objectives each recommendation addresses are summarized in Table 1

    ( f SKM 2002 )

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    21/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 13

    past, but no direct evidence was gathered to enable assessment of causality (Anderson and

    Morison, 1988b; M. Burns, Fisheries Victoria, pers. comm., 2004).

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    22/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 14

    Table 1: Environmental flow recommendations for the Lower Wimmera River (between McKenzie River and Lake Hindmarsh)

    (adapted from SKM, 2002a)

    Season Flow Component Magnitude Frequency Duration Rationale & Target Management Objective

    Summer Cease to Flow

    (CTF)

    0 ML/day annual 5-24 days Disturbance mechanism to maintain benthic

    community diversity

    Summer

    Autumn

    Minimum Median

    Flow

    5 ML/day annual Dec-May except

    during periods of

    CTF and Fresh

    Minimum flow to maintain habitat for Murray

    Cod, Macquarie Perch and Freshwater Catfish.

    Summer Summer Fresh* >20 ML/day 4 per summer 7-15 days Maintain or improve water quality in pools via

    mixing and/or destratification. Unlikely to mix

    saline pools in Lower Wimmera River, but may

    prevent significant increases in salinity that

    occur at flows of less than 10-20 ML/day.

    Spring Minimum Median

    Flow

    34 ML/day annual July-Nov except

    during periods of

    Fresh

    Minimum flow to maintain habitat for Murray

    Cod, Macquarie Perch and Freshwater Catfish.

    Spring Spring Fresh* >334 ML/day 5 per annum Minimum 14 days May stimulate migration and/or spawning of

    native fish species that inhabit the region.

    Makes available instream habitats such as bars,

    benches and undercuts. May mobilize and

    redistribute fine sediment which might

    otherwise smother habitats.

    Promote recruitment of Murray Cod and

    Macquarie Perch.

    Annual Bankfull Flow > 3,000 ML/day annual Minimum 2 days Maintain or improve water quality in pools via

    mixing and/or destratification.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    23/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 15

    Annual Overbank Flow 6,000 ML/day annual 3-5 days Provide lateral connectivity for exchange of

    energy and nutrients between in channel and

    floodplain habitats. Disturbance mechanism

    and fluvial power to shape and maintain

    diversity in channel geomorphological

    characteristics.

    Maintain and promote recruitment of Yarra

    Gum communities.

    *A fresh is a small peak flow event that exceeds the median flow for a given period and lasts several days.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    24/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 16

    Environmental flows released after a prolonged dry spell in the river may result in a sudden

    influx of nutrients and organic material with high biochemical oxygen demand. The organic

    matter may cause rapid depletion of dissolved oxygen causing hypoxia or anoxia and

    nutrient influxes may fuel harmful algal blooms or encourage proliferation of aquatic

    vegetation. Environmental flows released from storages with high levels of dissolved salts or

    suspended sediments could increase salinity or turbidity and lead to adverse impacts on in-

    stream biota. Large overbank flows recommended for the maintenance of lateral connectivity

    and channel geomorphological processes (Table 1) also carry with them the risk of causing

    accelerated streambank erosion in highly degraded river sections.

    The quantities of water available for environmental flows are typically highly variable and it

    may not be possible for the complete suite of flow recommendations (making up the full

    flow regime) to be implemented. What then are the benefits (if any) of piecemeal

    implementation of the environmental flow recommendations? How should river managers

    allocate the fixed available volumes of water for environmental flows among the competing

    targets? For instance, should environmental flows be allocated to freshes in summer rather

    than freshes in autumn, or for a single high/bankfull flow event in spring rather than a fewfresh events in spring. Will implementation of certain flow components produce greater

    relative benefit than others? Or to reframe the perspective, can the implementation of certain

    flow components reduce the relative risks to particular values?

    The application of an ERA approach to addressing these management issues is described in

    the remainder of this report. Scientifically credible management of environmental flows

    within an ERA framework rests on operationalizing sensitive endpoints that encapsulate

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    25/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    26/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 18

    prioritize key hazards and values requiring more detailed risk analysis and develop a setof candidate assessment endpoints for the risk analysis phase.

    4.4 Stakeholder workshops

    There were a total of 23 attendees at the first workshop held on the 20 July 2004 in

    Horsham, the major service and commercial centre for the Wimmera region. This included

    the 14 participants mentioned in the previous section and the nine members of the project

    team, seven of whom doubled up as expert stakeholders for the workshop. The workshop

    was mainly facilitated by Yung En Chee, with assistance from Anne-Maree Westbury, Jan

    Carey and Mark Burgman. The participants included seven natural resource managers from

    the local and regional CMAs, two officers from Fisheries Victoria, one regional officer from

    the EPA, and 11 scientists with a range of expertise in hydrology, freshwater ecology, water

    resource and quality management, ecotoxicology, population ecology, conservation biology,

    forestry, risk assessment and decision analysis. Details of the list of participants at this

    workshop are given in Appendix 1.

    At the commencement of the workshop, an overview was given of the project context, aims

    and objectives. Participants were invited to briefly introduce themselves, their credentials,

    professional affiliations and responsibilities within their organizations. This was followed by

    a presentation on the ERA process by Jan Carey to familiarize participants with the

    underlying concepts, rationale and motivation for adopting the ERA approach. The

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    27/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 19

    stakeholders. Hazard identification and assessment in the workshop proceeded through

    brain-storming, group discussion, hazard matrices and conceptual modelling.

    The identification of hazards and their assessment depends on an understanding of how a

    system is believed to work. The diverse pool of participants encompassed a range of

    different experiences and perspectives on the system and different mental conceptions of

    cause-effect relationships within the system. Group discussion and the development of

    conceptual models helped to draw out and visually represent these mental models so that

    they could be mapped, shared, documented and scrutinized.

    Hazard matrices were used to depict the interactions between values and hazards arising

    from natural events or human activities and helped in identifying hazards that could have

    multiple effects. They also helped to reduce the probability of overlooking interactions. In

    this way, a more comprehensive list of hazards could be generated. The hazard matrices

    constituted a visual summary of interactions between values and activities and provided a

    rudimentary but useful foundation for the hazard assessment phase.

    Detailed notes and outcomes of the first workshop were organized and summarized. They

    were then disseminated to all workshop participants for corrections, additions and

    clarification.

    Further deliberation took place with a second workshop in Horsham held on the 12 August

    2004. There were a total of 11 attendees at the workshop. They included four participants

    from WCMA, two participants from Fisheries Victoria, four participants from the EPA and

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    28/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 20

    in the second workshop and their relative merits deliberated. In selecting assessment

    endpoints, social relevance was evaluated by the local natural resource managers whilescientists provided technical input on ecological relevance, susceptibility to hazards/stressors

    and operational feasibility of measurement.

    4.5 Workshop outcomes

    Workshop participants tended to emphasize broad-scale, aggregate values such as a natural

    flow regime, biodiversity, native endemic plant and animal species and plant communities,

    diverse biofilm, phytoplankton, benthic and macrophyte communities. Values nominated by

    individual participants tended to reflect their area of expertise and/or professional

    responsibility. In particular, WCMA participants strongly identified with their role as

    managers for broad big picture goals such as river health and biodiversity and were

    reluctant to parse these concepts at a finer level of detail. Even as other workshop

    participants contributed values of a more specific nature, such as particular fish and

    vertebrate species, concerns were expressed throughout over reductionism and excessive

    narrowing of objectives.

    A broad range of hazards arising from natural events such as prolonged drought, fire and

    algal blooms and human activities such as river regulation, environmental flow management,

    agricultural practices, urbanization and industrial development were identified in the

    workshop. Hazards were considered mainly in terms of their potential effects on broad

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    29/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 21

    as reduced habitat availability, low DO and salt toxicity. The latter two were also deemed to

    be the key hazards to water quality for sustaining biota.

    Physicochemical parameters of water quality are the most commonly measured metrics in

    river management and conventionally, it has been assumed that if target levels of water

    quality parameters are attained, the aquatic environment could be considered to be healthy

    (Norris and Thoms, 1999). However, this premise is problematic for the following reasons:

    appropriate parameters might not be measured (eg. measurement only of surface waterDO when the problem may lie with bottom water DO),

    the approach does not take into account possible antagonism or synergism betweenstressors which may affect aquatic biota (eg. high temperatures and low DO increase the

    toxicity of ammonia to fish (Alabaster and Lloyd, 1980),

    spot measurements of parameters which frequently exhibit high natural variability areunlikely to reflect water quality over timescales relevant to aquatic biota and

    furthermore, may miss intermittent inputs and flood events (Norris and Thoms, 1999;ANZECC/ARMCANZ, 2000), and

    merely monitoring physicochemical water quality parameters provides no informationon other factors that may affect biotic distribution (e.g. quantity of hydraulic habitat,

    flow velocity, availability of food resources and competitive interactions between biota

    and/or exotic species).

    Water quality is essential and if socially valued as an end in itself is appropriate as an

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    30/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 22

    Table 2: Examples from the workshop of some of the hazards identified and the assessment of their potential effects

    Event/Activity Hazard/Threat/Stressor Potential primary effects Potential secondary effects

    Natural Event

    Fire in riparian zone Increased input of sediment

    and particulate matter

    Loss of riparian habitat

    Increased turbidity levels and reduced

    light penetration in water column

    Transport into waterway of contaminants

    bound to particulate matter

    Smothering of benthic biota

    Reduced feeding efficiency of filter-

    feeding and visual feeding biota

    Burial of structural habitat features such

    as gravel and coarse debris

    Reduced riparian habitat

    Reduced connectivity between aquatic

    and terrestrial ecosystems

    Reduced primary productivity in deeper

    waters

    Acute toxic effects/indeterminate

    sublethal effects on aquatic biota

    Reduced abundance and/or diversity of

    benthic community

    Loss of benthic feeding, refuge and

    spawning habitat

    Prolonged drought Extended period of low/zero

    flow

    Loss of longitudinal

    hydrologic connectivity

    Lack of overbank flows ->

    loss of lateral hydrologicconnectivity, loss of channel-

    floodplain exchange

    mechanism

    Reduced quantity and quality of physical

    instream hydraulic habitat

    Reduced quantity and quality of

    floodplain wetland habitat

    Reduced productivity of riverine-

    floodplain ecosystem

    Reduced abundance and/or diversity of

    aquatic biota

    Reduced abundance and/or diversity ofriverine and floodplain biota

    Saline groundwater Increased salinity Acute toxic effects/indeterminate

    Reduced abundance and/or diversity of

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    31/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 23

    intrusion

    Salinity induced density

    stratification

    sublethal on aquatic biota

    Lack of vertical mixing of water column

    leading to development of hypoxia/anoxia

    in water column and bottom sediments

    Lack of vertical mixing resulting in

    increased residence time of algae in

    euphotic zone

    aquatic biota

    Reduced quantity and quality of physical

    instream hydraulic habitatBuild up of toxic levels of ammonia and

    hydrogen sulphide due to activity of

    anaerobic bacteria

    Increased rates of nitrogen and

    phosphorus mineralization which may

    fuel algal blooms

    Algal blooms

    High temperatures insummer Temperature induced densitystratification Lack of vertical mixing of water columnleading to development of hypoxia/anoxia

    in water column and bottom sediments

    Lack of vertical mixing resulting in

    increased residence time of algae in

    euphotic zone

    Reduced quantity and quality of physicalinstream hydraulic habitat

    Build up of toxic levels of ammonia and

    hydrogen sulphide due to activity of

    anaerobic bacteria

    Increased rates of nitrogen and

    phosphorus mineralization which may

    fuel algal blooms

    Algal blooms

    Occurrence of algal

    bloom

    Reduced light penetration

    Algal bloom is harmful

    Reduced primary productivity in deeper

    waters

    Illness/death of biota that graze on algae

    that produce toxins

    Bioaccumulation of toxins in

    taxa/organisms at higher trophic levels

    Die-off of al al bloom ma roduce DO

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    32/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 24

    Injury/death of fish via gill and tissue

    damage caused by algal species that have

    physical structures such as spines (eg.

    some diatoms, dinoflagellates and

    raphidophytes.)

    sag.

    River Regulation

    Activities

    Abstraction, diversion

    and distribution

    Reduced flow magnitude in

    all seasons

    Reduced frequency and

    duration of cease-to-flow

    events -> loss of disturbance

    mechanism

    Excessive baseflow stability -

    > loss of flow variability

    Reduced quantity and quality of physical

    instream hydraulic habitat

    Reduced quantity and quality of

    seasonally inundated low-lying backwater

    and anabranch habitat

    Reduced diversity of benthic community

    Proliferation of monospecifc stands of

    aquatic vegetation (eg.Azollaspp.,

    Phragmites australiaand Typha spp.)

    Impacts on life-history strategies of native

    aquatic biota (eg. loss of spawning cues)

    Reduced abundance and/or diversity of

    aquatic biota

    Reduced diversity of macrophyte

    community

    Reduced diversity of instream structural

    habitat

    Respiratory requirements of large

    standing crop can create hypoxic

    conditions before sunrise and during days

    of overcast weather.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    33/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 25

    May favour establishment and

    proliferation of exotic species

    Die-off of large standing crop of aquatic

    vegetation may produce DO sag.

    Impaired survival, competitive ability and

    reproduction of affected biota.

    Reduced abundance and/or diversity of

    native biota

    Leakage from

    Wimmera-Mallee

    distribution system

    Raised water table -> reduced

    frequency and duration of

    cease-to-flow events -> loss

    of disturbance mechanism

    Excessive baseflow stability -

    > loss of flow variability

    Reduced diversity of benthic community

    Proliferation of monospecifc stands of

    aquatic vegetation (eg.Azollaspp., P.australiaand Typha spp.)

    Impacts on life-history strategies of native

    aquatic biota (eg. loss of spawning cues)

    May favour establishment and

    proliferation of exotic species

    Reduced diversity of macrophyte

    community

    Reduced diversity of instream structural

    habitat

    Respiratory requirements of large

    standing crop can create hypoxic

    conditions before sunrise and during days

    of overcast weather.

    Die-off of large standing crop of aquatic

    vegetation may produce DO sag.

    Impaired survival, competitive ability and

    reproduction of affected biota.

    Reduced abundance and/or diversity of

    native biota

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    34/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 26

    Environmental Flow

    Management

    Fresh magnitude too

    large

    Sudden purging of saline

    pools containing highlysaline, bottom waters

    containing low DO and/or

    low pH and/or high levels of

    ammonia/hydrogen sulphide

    Acute toxic effects/indeterminate

    sublethal effects on aquatic biota

    Flow release allocated

    from storage with high

    levels of salinity

    Input of water with high

    levels of salinity

    At best, no improvement in water quality,

    at worst, acute toxic effects/indeterminate

    sublethal effects on aquatic biota

    Flow release allocated

    from distant storage

    within the Wimmera-

    Mallee system

    Flow magnitude diminished

    through seepage and

    evaporative losses ->

    extended period of low/zero

    flow

    Loss of longitudinal

    hydrologic connectivity

    Reduced quantity and quality of physical

    instream hydraulic habitat

    Reduced abundance and/or diversity of

    aquatic biota

    Lack of water for

    environmental flow

    release

    Extended period of low/zero

    flow

    Loss of longitudinal

    hydrologic connectivity

    Reduced quantity and quality of physical

    instream hydraulic habitat

    Reduced abundance and/or diversity of

    aquatic biota

    Flow delivery impeded

    by council weir pool

    Extended period of low/zero

    flow

    Reduced quantity and quality of physical

    instream hydraulic habitat

    Reduced abundance and/or diversity of

    aquatic biota

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    35/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 27

    management (eg. to

    maintain full weir pools

    for recreation)

    Loss of longitudinal

    hydrologic connectivity

    Flow delivery absorbedby water users with

    private entitlements

    Extended period of low/zeroflow

    Loss of longitudinal

    hydrologic connectivity

    Reduced quantity and quality of physicalinstream hydraulic habitat

    Reduced abundance and/or diversity ofaquatic biota

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    36/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 28

    Table 3: Evaluation of the selected endpoints sustainable populations of Freshwater Catfish and macroinvertebrate community

    diversity with respect to congruence with required attributes of assessment endpoints

    Endpoint Attribute Sustainable populations of Freshwater Catfish Macroinvertebrate Community Diversity

    Social importanceand relevance

    Highly-valued native species; listed under the Flora andFauna Guarantee Act 1988; classified as Endangered

    in Advisory List of Threatened Vertebrate Fauna in

    Victoria (DSE 2003a).

    Although non-endemic in the region, it has established

    self-sustaining populations in the Lower Wimmera

    River which now constitutes a stronghold for the species

    in Victoria.

    Recreationally important.

    Macroinvertebrate community diversity is relatively wellappreciated by the community due to national awareness

    programs such as Waterwatch.

    Macroinvertebrate communities have been judged to have

    sufficient credibility to be used in federal and state

    environmental monitoring frameworks for reference

    condition monitoring (eg. in National River Health

    monitoring and the Victorian Index of Stream Condition).

    Indices of macroinvertebrate community are a required

    component of ecological objectives in the StateEnvironmental Protection Policy(Waters of Victoria)

    (SEPP WoV).

    Ecological

    importance and

    relevance

    Species is of high conservation value.

    Occupies bottom waters prone to water quality problems

    which managers hope to address through environmental

    flow management.

    Macroinvertebrate community represents a critical

    pathway for the transport and utilization of energy and

    aquatic matter in aquatic ecosystems (Carmargo 2004)

    Susceptible to

    hazards/stressors

    Species has declined in abundance and distribution

    throughout its natural range in Victoria, but specific

    stressors which might account for this have not been

    established (Clunie and Koehn 2001). Monitoring is

    required to establish response to flow regime and water

    quality changes effected by environmental flow

    strategies.

    There is considerable evidence that macroinvertebrate

    assemblages can be useful indicators of high nutrient

    input or pollutants and contaminants (Mebane 2001;

    Chessman 2003). However, there is little information on

    how invertebrates might respond to changing flow

    regimes (SKM 2003a). Monitoring is required to establish

    response to flow regime and water quality changes

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    37/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 29

    effected by environmental flow strategies.

    Operational

    definition

    Can be framed clearly and unambiguously. Can be framed clearly and unambiguously.

    Measurementfeasibility

    Sampling of Freshwater Catfish to estimatedemographic parameters for quantifying viability may

    be time-consuming and costly. Monitoring of fish

    communities has not occurred in any regular program,

    but programs are under development (DNRE 2002b).

    Some quantitative survey data exists from previous

    studies. Angler reports and annual results from fishing

    competitions held along the Lower Wimmera River are

    a potential source of useful data.

    Ubiquitous and relatively easy to collect. Standardizedsampling and processing protocols are well-developed

    and taxonomic keys are available to identify most

    macroinvertebrates.

    Component of routine river health monitoring programs

    in Victoria.

    Good, baseline data exists from monitoring carried out by

    the EPA.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    38/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 30

    Selection of assessment endpoints was a difficult exercise because the Lower Wimmera

    River is highly degraded and relatively depauperate in terms of socially important aquatic

    and riverine biota. Uncertainty regarding the susceptibility of each assessment endpoint torelevant hazards and stressors is an important limitation and arises from incomplete scientific

    understanding. Susceptibility is a function of exposure to the hazard/stressor and sensitivity

    to direct as well as indirect adverse effects of the hazard/stressor and consideration of

    susceptibility itself requires a qualitative or simple quantitative hazard assessment (Suter,

    1995).

    An anticipated difficulty in interpreting measurement of the endpoints is that the natural

    variability of physical and biological components of semi-arid, intermittent rivers is very

    high (Boulton, 1999). This variability has been documented for the macroinvertebrate

    community in the Wimmera River (Metzeling, 2002). The amount of variability could make

    it difficult to discern what effects, if any, environmental flows might have.

    The effect of multiple stressors interacting in complex and often non-linear ways over

    different spatial and temporal timescales makes it difficult to untangle and definitively

    establish causal relationships for observed impacts. In addition to these uncertainties,environmental flow management in the context of a complex distribution system like the

    Wimmera Mallee system needs to take into account logistical constraints. Consequently, it

    would be advantageous if we could develop a knowledge base of how equivalent

    environmental outcomes might be achieved (within limits of acceptable risk) for different

    patterns of flow deployment. This would provide valuable flexibility in the operational

    management of the EWR.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    39/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 31

    5. Development of the Bayesian network model for Freshwater Catfish

    in the Lower Wimmera River

    5.1 Bayesian network modelling

    Bayesian networks (BNs) (also known as Bayes nets, Bayesian belief networks, causal

    probability networks and probability networks) are well suited to the task of modelling a

    situation in which causality plays a role, but where our understanding is incomplete, so we

    need to describe events probabilistically (Charniak, 1991). Bayesian methods provide a

    formalism for reasoning about beliefs under conditions of uncertainty. In this formalism,propositions are given numerical parameters signifying the degrees of belief accorded them

    under some body of knowledge and the parameters are combined and manipulated according

    to the rules of probability (Pearl, 1991). Good introductions to BNs are available in

    Neapolitan (1990), Pearl (1991), Jensen (1996) and Pearl (2000).

    The use of BN models is well-developed in the fields of knowledge engineering, software

    engineering and artificial intelligence (Neilet al

    ., 2000; Fenton and Neil, 2001). Recentapplications in ecological management include modelling a restoration strategy for a

    temperate lake in Finland, modelling salmon fisheries management in the Baltic Sea (Varis

    and Kuikka, 1999), evaluating fish and wildlife population viability under land management

    alternatives (Marcot et al., 2001), evaluating land management alternatives on salmonid

    populations and habitat in the Columbia River Basin (Rieman et al., 2001) and modelling

    estuary eutrophication in the Neuse River in North Carolina, USA (Borsuk et al., 2004).

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    40/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 32

    presence of an arc (Pearl, 2000; Borsuk et al., 2004). Submodels may be derived from any

    combination of process knowledge, statistical correlation or expert judgement depending onthe type and scale of information available (Borsuk 2004; Borsuk et al. 2004). Similarly,

    CPTs can be constructed using empirical data, output from process models, theoretical

    insight, probabilistic or deterministic functions, ancillary data from empiric studies

    independent of the constructed system, and expert judgements (Reckhow, 1999b; Lee, 2000;

    Borsuk et al., 2001; Cain, 2001. Generation of CPTs from functional relationships uses

    Monte Carlo simulation to draw representative samples from a nodes predecessors, to build

    up a discrete conditional distribution of the node based on the functional relationship.

    With the model fully specified and validated, the BN model can be used to produce

    probability distributions (or risk profiles) for model endpoints and any other variables of

    management interest for both inference and prediction. This is basically performed by

    altering the states of some nodes in accordance with observations/findings or proposed

    interventions while observing the effect this has on model endpoints. The impact of changing

    any variable is transmitted throughout the network in accordance with the relationships

    encoded in the CPTs and the joint probability distribution of the entire network conditioned

    on these observations is inferred or calculated for other variables using Bayes Theorem

    (Pearl, 1991).

    BN modelling is useful because the process of formal identification of system variables and

    interactions forces model builders to articulate, structure and thereby clarify their

    understanding, not only of the relevant influences on management objectives, but of the

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    41/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 33

    5.2 Overview of the development and construction of the Freshwater Catfish BN

    The two main tasks in BN modelling are construction of a graphical structure and

    construction of CPTs for each node.

    Developing the graphical structure involves formal, systematic identification of system

    variables and interactions and well-founded decisions for what variables and values to

    include and omit from the network (Batchelor and Cain, 1999; Coup et al., 2000). For the

    Freshwater Catfish BN key variables and the nature of their dependencies were identified

    and refined through:

    the workshop process of hazard/threat identification, the workshop process of conceptual model construction, a review of the environmental flow recommendations documented in the Stressed

    Rivers Project: Environmental Flows Study (SKM, 2002a),

    a comprehensive survey of the relevant literature, and consultations with fisheries managers and experts in hydrology, freshwater ecology and

    Freshwater Catfish biology.

    The sources above were used along with practical BN modelling guidelines provided in Cain

    (2001) to develop a model linking Freshwater Catfish population characteristics with the

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    42/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 34

    BN modelling was carried out using the software Netica (Norsys Software Corp. 1997-

    2003). Netica uses junction tree algorithms to perform probabilistic inference (Norsys,1997). Details on computation and algorithms used in Netica are available in Neapolitan

    (1990) and Spiegelhalter et al. (1993).

    5.3 Stakeholder considerations in BN model development

    The objective of constructing the BN model was to capture understanding of how flow

    conditions are believed to affect key in-stream physicochemical and habitat variables

    governing Freshwater Catfish population structure and viability. The temporal focus of the

    model is the summer period when environmental conditions are expected to be most stressful

    for in-stream aquatic biota and when catchment managers are typically required to make

    decisions about the deployment of environmental flows.

    In the course of the risk assessment workshops, during which a large number of hazards

    were considered, it became clear that the factors that stakeholders believed were mostimportant with respect to maintaining self-sustaining populations of fish were the quantity

    and quality of in-stream habitat as determined by the magnitude, timing and duration of

    flows. With respect to water quality, the importance of turbidity, pH and concentrations of

    nitrogen and phosphorus were acknowledged, but it was believed that ultimately the two

    most important aspects were dissolved oxygen (DO) and salinity.

    There was an expectation that providing (minimum) flows over summer would increase the

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    43/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 35

    collated and reviewed. In general, Freshwater Catfish appear to be relatively long-lived,

    hardy and extremely tolerant of poor physicochemical conditions. They favour slow,sluggish waters and spawn and develop in carefully tended nests constructed in shallow

    waters. Detailed results from the review are summarized in Chee (2005).

    The phenomenon of density stratification, bottom water deoxygenation and conclusions from

    previous work done on salinity and thermal stratification in the Lower Wimmera was also

    reviewed. The detailed review is available in Chee (2005). A summary of the review

    follows.

    In the Lower Wimmera River, both thermal and salinity stratification occur. Thermal

    stratification is extensive and affects a much greater proportion of the river than salinity

    stratification (Anderson and Morison, 1988b). It is persistent during summer, but also occurs

    for shorter periods in autumn (Western and Stewardson, 1999). Salinity stratification

    develops most often as a consequence of saline groundwater seeping directly into river

    pools, and less frequently as a consequence of saline inflows to a relatively fresher pool from

    an upstream source (Anderson and Morison, 1988b; Western, 1994). Salinity stratification istherefore largely dependant on the presence of sites of saline groundwater intrusion, which

    may be very localized. Sites of groundwater intrusion characteristically have well-developed,

    persistent stratification of the water column and where significant groundwater intrusion

    occurs, salinity stratification will persist most of the year, particularly under low flow

    conditions (Anderson and Morison,, 1988b; Western, 1994; Western et al., 1996; SKM,

    1997) but also under moderate flow events (Ryan et al., 1999).

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    44/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 36

    5.5 Graphical structure of Freshwater Catfish BN

    The spatial boundary of the model encompasses an approximately 62 km reach of the Lower

    Wimmera River, from the junction of the Wimmera and McKenzie Rivers to Big Bend. This

    reach was chosen because it is known to be inhabited by Freshwater Catfish, it contains

    known sites of saline pools and cross-section profile elevation data are available for

    developing flow, depth and cross-section area relationships. The model is focussed on

    environmental conditions during summer and the model endpoint is the viability of

    Freshwater Catfish populations within the specified river reach. The full graphical Bayesian

    probability network model linking summer flow regime to Freshwater Catfish populations in

    the Lower Wimmera River is shown in Figure 5.

    The graphical BN model was structured according to the main conceptual ideas as follows:

    The viability of Freshwater Catfish populations in the focal reach (Viability of CatfishPopn) is determined by the abundance of the breeding population (Abundance

    Breeding Popn) and the recruitment of juvenile and larval catfish (Recruitment

    Juvenile Catfishand Recruitment Larval Catfish). These population characteristicsare in turn dependent on the quantity and quality of physical hydraulic habitat for living

    (Hydraulic Habitat) and spawning (Spawning Area Availability) and on stochastic

    elimination via adverse slug events (Severity of Slug Event), and in the case of

    larval catfish, via high flow velocities (Peak XS Velocity) as well. This is depicted by

    the graphical structure in Box A in Figure 5.

    The quantity and quality of physical hydraulic habitat (Hydraulic Habitat) is

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    45/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    46/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 38

    Table 4: Summary of model variable definitions, states and sources of information and/or data used to discretize variables and construct

    probability models

    Model

    Variable

    Model

    Variable

    Definition

    Variable States: Names &

    Descriptions

    Node Parents Purpose/Usage of Variable in the Model Info source for

    discretizing

    variable and/or

    specifying

    probability

    model

    Median

    Mean

    Daily

    Flow

    (MDF)

    Median mean

    daily flow over

    summer

    MDF0: 0 ML/day

    MDF5: 5 MlL/day

    MDF10: 10 ML/day

    MDF20: 20 ML/day

    MDF30: 30 ML/day

    None Characterizes and summarizes the

    minimum flow conditions experienced

    in the nominated reach over summer.

    Determines stream depth distribution

    over the reach. Also used to assess the

    maximum interval between mix events.

    Time series flow

    data measured at

    Horsham gauge

    415200 (1945-

    2001 inclusive):

    (10)

    Depth Distribution of

    stream depths

    over the reach

    VShal: < 1.0 m

    Shal: 1.0m 2.0 m

    Mod: 2.0 m 2.5 m

    Deep: >2.5 m

    Median Mean

    Daily Flow

    (MDF)

    Depth is used as a variable for

    predicting the status of bottom DO and

    the availability of spawning area for

    Freshwater Catfish over the reach.

    Data, hydraulic

    model, empirical

    probability

    model: (9)

    XS Area Distribution of

    flow cross-

    sectional area

    over the reach

    A_0to20: 0-20 m2

    A_20to40: 20-40 m2

    A_40to80: 40-80 m2

    Median Mean

    Daily Flow

    (MDF)

    Cross-sectional area is used as index of

    the quantity of physical hydraulic

    habitat. It is also used to calculate flow

    velocities for the peak cross-sectional

    velocit variable and to determine

    Data, hydraulic

    model, empirical

    probability

    model: (9)

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    47/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 39

    A_80to120: 80-120 m2

    A_120to160: 120-160 m2

    A_160to200: 160-200 m2

    A_mt200: >200 m2

    whether or not mixing occurs due to

    median MDF and freshes.

    Extent of

    Groundwa

    ter

    Intrusion

    Extent of

    saline

    groundwater

    intrusion

    Low: 0-5% of river section

    affected

    Mod: 5-10% of river section

    affected

    High: > 10% of river section

    affected

    None Estimates the extent of groundwater

    intrusion as a precursor to assessing

    bottom water salinity.

    Data: (2), (8)

    Obs Range

    BottomSalinity

    Observed

    range ofbottom salinity

    (as measured

    by electrical

    conductivity

    below the

    halocline)

    Good: !3,000 EC

    Okay: 3,000-12,000 EC

    Poor: 12,000-20,000 EC

    Very Poor: > 20,000 EC

    None Characterizes the observed range of

    bottom salinity values within aparticular river section. Used in

    conjunction with the extent of

    groundwater intrusion to assess bottom

    salinity.

    Data: (1), (3),

    (4), 5), (8)

    Bottom

    Salinity

    Salinity as

    measured by

    electricalconductivity of

    water below

    Good: !3,000 EC

    Okay: 3,000-12,000 EC

    Poor: 12,000-20,000 EC

    Very Poor: > 20,000 EC

    Extent of

    Groundwater

    Intrusion

    Obs Range

    Bottom

    Characterizes bottom salinity conditions

    over summer within the reach of

    interest. Bottom salinity is an importantfactor in the assessment of bottom water

    quality.

    Analyst

    judgement

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    48/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 40

    the halocline Salinity

    Bottom

    DO

    Bottom

    dissolved

    oxygen

    concentration

    (% saturation)

    Stressful: !10% sat.

    Okay: > 10% sat.

    Depth Characterizes bottom DO conditions

    over summer within the reach of

    interest. Bottom DO is an important

    factor in the assessment of bottom water

    quality.

    Data, empirical

    relationship: (1),

    (3), (5)

    Fresh

    Magnitude

    Magnitude of

    summer

    freshes

    F20: 20 ML/day

    F40: 40 ML/day

    F80: 80 ML/day

    F115: 115 ML/day

    F150: 150 ML/day

    F190: 190 ML/day

    F240: 240 ML/day

    None Allows exploration of the effects of

    employing freshes of different

    magnitudes. Needs to be used in

    conjunction with Fresh Freq node.

    User-defined

    Fresh Freq Frequency of

    occurrence of

    freshes over

    summer

    NoFr: 0 freshes

    Fr_x1: 1 fresh

    Fr_x2: 2 freshes

    Fr_x3: 3 freshes

    Fr_x4: 4 freshes

    None Allows exploration of the effects of

    deploying freshes at different

    frequencies. Needs to be used in

    conjunction with Fresh Magnitude

    node.

    User-defined

    Mixing

    due to

    Median

    Occurrence of

    mixing as a

    result of

    Yes

    No

    Median Mean

    Daily Flow

    (MDF)

    Mixing events are expected to have an

    ameliorating effect on bottom water

    ualit over summer. Mixin of

    Data, functional

    relationship,

    em irical

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    49/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 41

    MDF median MDF

    for any given

    cross-sectional

    area

    (MDF)

    XSArea

    thermally stratified waters occurs due to

    flow-related turbulence when flow

    velocities are sufficiently high. This

    variable characterizes the probability

    that mixing occurs due to median MDFover summer. The flow velocity criteria

    used to determine the probability of

    mixing are discussed in greater detail in

    Chee (2005).

    relationship: (6),

    (11)

    Mixing

    due to

    Freshes

    Occurrence of

    mixing as a

    result of

    freshes for

    any given

    cross-sectional

    area

    Yes

    No

    Fresh

    Magnitude

    XSArea

    Distribution

    Mixing events are expected to have an

    ameliorating effect on bottom water

    quality over summer. Mixing of

    thermally stratified waters occurs due to

    flow-related turbulence when flow

    velocities are sufficiently high. This

    variable characterizes the probability

    that mixing occurs due to freshes over

    summer. The flow velocity criteria used

    to determine the probability of mixing

    are discussed in greater detail in Chee

    (2005).

    Data, functional

    relationship,

    empirical

    relationship: (6),

    (11)

    Max

    Interval

    Betw MixEvts

    Maximum

    interval

    betweenmixing events

    for an iven

    Int_0to21: 0-21 days

    Int_22to28: 22-28 days

    Int_29to35: 29-35 days

    Mixing due to

    Median MDF

    Fresh Freq

    Mixing due to

    Mixing events are expected to have an

    ameliorating effect on bottom water

    quality over summer. The effectivenessof this depends upon the length of time

    between mix events and is characterized

    Functional

    relationship

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    50/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 42

    cross-sectional

    areaInt_mt35: > 35 days Freshes by the maximum interval between

    events.

    BottomWater

    Quality

    Descriptiveindicator of

    bottom water

    quality

    Poor: habitable by adultFreshwater Catfish but low in

    terms of food productivity and

    likely to be harmful to eggs

    and larval catfish

    Okay: habitable by Freshwater

    Catfish (& other aerobic

    biota), moderate in terms of

    food productivity and not

    detrimental to eggs and larvalcatfish

    Good: habitable by Freshwater

    Catfish (& other aerobic

    biota), good in terms of food

    productivity and provides

    favourable physicochemical

    conditions for eggs and larval

    catfish

    BottomSalinity

    Bottom DO

    Max Interval

    Betw Mix Evts

    Synthetic variable that characterizesbottom water (in the terms defined), on

    the basis of bottom salinity, bottom DO

    and the maximum interval between mix

    events. Used in conjunction with cross-

    sectional area to rate hydraulic habitat.

    Processknowledge,

    analyst

    judgement

    HydraulicHabitat

    Descriptiveindicator of

    hydraulic

    LowMod

    XSAreaBottom Water

    Quality

    Synthetic variable that characterizeshydraulic habitat on the basis of

    quantity (as indicated by cross-sectional

    Processknowledge,

    analyst

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    51/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 43

    habitat High area) and quality (as indicated by

    bottom water quality).

    judgement

    Flush

    Magnitude

    Magnitude of

    the peak flow

    event over

    summer

    F0to300: 0-300 ML/day

    F300to600: 300-600 ML/dayF600to1200: 600-1,200

    ML/day

    F1200to3000: 1,200-3,000

    ML/day

    F3000to6000: 3,000-6,000

    ML/day

    Fmt6000: > 6,000 ML/day

    None Characterizes the distribution of annual

    summer peak flow events. In other

    words, the probability of occurrence of

    peak flow events of varying magnitudes

    each summer (ie. on an annual basis).

    Time series flow

    data measured at

    Horsham gauge

    415200 (1945-

    2001 inclusive)

    and empirical

    probability

    model: (10)

    Peak XSVelocity

    Peak cross-sectional flow

    velocity

    experienced

    over summer

    Low: < 0.05 m/s

    Mod: 0.05-0.08 m/s

    High:> 0.08 m/s

    FreshMagnitude

    Flush

    Magnitude

    XSArea

    Characterizes the peak cross-sectionalflow velocity experienced over summer,

    whether as a result of freshes or a

    flush event. Used in conjunction with

    bottom water quality to assess the

    severity of slug events. Also an

    important factor in assessment of the

    recruitment success of larval catfish.

    Functionalrelationship,

    expert

    judgement: (13)

    Severity of

    SlugEvent

    Descriptive

    indicator of theseverity of an

    adverse slug

    Low

    Mod

    High

    Bottom Water

    QualityPeak XS

    Velocity

    Synthetic variable that characterizes the

    severity of a slug event on the basis ofbottom water quality status and mean

    cross-sectional peak velocity.

    Analyst

    judgement

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    52/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 44

    event

    Abundanc

    e Breeding

    Popn

    Descriptive

    indicator of

    abundance of

    Freshwater

    Catfish

    breeding

    population

    within the

    reach at the

    end of summer

    Low

    Mod

    High

    Hydraulic

    Habitat

    Severity ofSlug Event

    Synthetic variable that characterizes the

    abundance of the Freshwater Catfish

    breeding population within the reach on

    the basis of hydraulic habitat and

    mortality impact of slug events. Is a

    key factor in assessing the viability of

    Freshwater Catfish within the reach.

    Expert

    judgement: (13)

    Recruitme

    nt Juvenile

    Catfish

    Descriptive

    indicator of

    recruitmentsuccess of

    juvenile

    catfish within

    the reach at the

    end of summer

    Poor

    Mod

    Good

    Hydraulic

    Habitat

    Severity of

    Slug Event

    Synthetic variable that characterizes the

    recruitment success of juvenile catfish

    within the reach on the basis ofhydraulic habitat and mortality impact

    of slug events. Is an important factor

    in assessing the viability of Freshwater

    Catfish within the reach.

    Analyst

    judgement

    Spawning

    Area

    Availabilit

    y

    Descriptive

    indicator of the

    availability of

    spawning areaover the reach

    Low

    Mod

    High

    Depth Characterizes availability of spawning

    area within the reach over summer. Is

    an important factor in the assessment of

    the recruitment success of larval catfish.

    Data, expert

    judgement: (7),

    (12)

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    53/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 45

    Recruitme

    nt Larval

    Catfish

    Descriptive

    indicator of

    recruitment

    success of

    larval catfishwithin the

    reach at the

    end of summer

    Low

    Mod

    High

    Spawning

    Area

    Availability

    Hydraulic

    Habitat

    Severity of

    Slug Event

    Peak XS

    Velocity

    Synthetic variable that characterizes the

    recruitment success of larval catfish

    within the reach on the basis of

    spawning area availability, hydraulic

    habitat and mortality impacts of slugevents and high flow velocity

    conditions. Is an important factor in

    assessing the viability of Freshwater

    Catfish within the reach.

    Analyst

    judgement

    Viability

    of Catfish

    Popn

    Descriptive

    indicator of the

    viability of

    FreshwaterCatfish

    populations

    within the

    reach at the

    end of summer

    Poor

    Good

    Abundanace

    Breeding Popn

    Recruitment

    JuvenileCatfish

    Recruitment

    Larval Catfish

    Model endpoint. Synthetic variable that

    characterizes the viability of Freshwater

    Catfish populations within the reach.

    Analyst

    judgement

    (1) Anderson and Morison (1988b); (2) Western (1994); (3) Metzeling (1995); (4) SKM (1997); (5) Ryan et al . (1999); (6) Western and

    Stewardson (1999); (7) Clunie and Koehn (2001); (8) SKM (2003b); (9) Cooperative Research Centre for Catchment Hydrology; (10) Grampians

    Wimmera Mallee Water (formerly Wimmera-Mallee Water; (11) A.Western, unpublished data; (12) P.Clunie, pers.comm., 2005; (13) T.Ryan,

    pers.comm., 2005.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    54/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 46

    5.6 Reviewing BN model structure and node relationships

    The graphical and probability structure of the BN was checked and reviewed using methods

    described in Cain (2001). With respect to overall graphical structure, checks were performed

    to ensure that all parentless nodes represented either controlling factors or interventions and

    that the childless node described the stakeholder management objective. Expert review (by

    P. Clunie, Arthur Rylah Institute for Environmental Research, ARI, and T. Ryan, formerly of

    ARI) was used to ensure that all important variables had been included and that any

    omissions were justifiable/defensible. Node names and states were checked in consultation

    with experts to ensure that they were adequately and appropriately defined in terms of spatial

    and temporal scale and specific problem context.

    Node connections, relationships and probability structure were reviewed in consultation with

    experts and by altering node states to observe how the model behaved. Node relationships

    were systematically reviewed according to Cains (2001) methodology. Full details of the

    review process and outcomes are available in Chee (2005).

    5.7 Sensitivity to findings

    Sensitivity to findings analysis was used to identify network variables which have the

    greatest influence on the variables of management interest. These variables, denoted as

    query nodes, include: Bottom Water Quality, Hydraulic Habitat, Severity of Slug

    Event, Abundance Breeding Popn, Recruitment Juvenile Catfish, Recruitment Larval

    Catfishand Viability of Catfish Popn.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    55/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 47

    where qis a state of the query variable, Q,fis a state of the findings variable, F, and the

    summations refer to the sum of all states qof fof variables Qor F(Pearl, 1991).The entropy reduction or mutual measures can be used to rank variables according to the

    capacity for entered evidence at these variables to change the posterior probability of the

    query node. As entropy reduction describes the reduction in uncertainty in a query node

    when information is available for a findings node, it can be used to help identify areas of

    research priority by indicating which variables to target to enable optimal learning.

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    56/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 48

    6. Management Implications

    6.1 General

    The pattern of allocation of available water over the suite of flow components, comprising

    cease-to-flow periods, minimum flows and freshes constitutes the environmental flow

    release strategy over summer. Exploration of flow scenarios or strategies using the BN

    model was carried out by entering findings at the appropriate nodes and examining the

    resultant changes in probability distributions of the variables of management interest. This

    allows a model-user to assess the sensitivity of outcomes to different managementinterventions. Two basic scenarios were used to provide baselines for comparative

    evaluation of environmental flow strategies:

    a worst case scenario (called Dry) in which the focal reach receives no flow at allthroughout summer, and

    the SKM (2002a) environmental flow recommendations (Table 5) which consisted of acease-to flow period of 21 days; median MDF=5 ML/day and four evenly distributedfreshes over summer of 20 ML each. The implementation of this strategy in the full

    BN model is shown in Figure 6.

    Under conditions of low median MDF, the most likely state of cross-sectional area

    distribution (XSArea) is 0-20 m2with a probability of 0.83 and the most likely states of

    Bottom Water Quality are Okay and Good with probabilities of 0.44 and 0.35

    respectively (Figure 6). These conditions in turn result in Hydraulic Habitat being most

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    57/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    58/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    59/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    60/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    61/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 53

    Table 6: Results of sensitivity to findings analysis for the seven variables of management interest. The rank of each network

    variable (findings node) with respect to the query variable is given by the value in each cell (with a rank of 1 indicating the

    network variable with the greatest influence on the query variable). The entropy reduction value associated with each

    findings node, I, is given (to four decimal places) in parentheses within each cell.

    Model Variables Bott Water

    Quality

    Hydraulic

    Habitat

    Severity of

    Slug Event

    Abundance

    Breeding

    Popn

    Recruitmt

    Juvenile

    Catfish

    Recruitmt

    Larval

    Catfish

    Viability

    Catfish Popn

    Median MDF 15

    (0.0041)

    17

    (0.0024)

    17

    (0.0027)

    16

    (0.0004)

    16

    (0.0010)

    17

    (0.0003)

    16

    (0.0001)

    Depth 10

    (0.0287)

    10

    (0.0131)

    10

    (0.0127)

    10

    (0.0027)

    10

    (0.0060)

    8

    (0.0186)

    10

    (0.0008)

    XSArea 16

    (0.0037)

    11

    (0.0088)

    12

    (0.0074)

    11

    (0.0024)

    11

    (0.0030)

    12

    (0.0086)

    11

    (0.0005)

    Extent Grdwater

    Intrusion

    19

    (0.0014)

    19

    (0.0006)

    20

    (0.0006)

    19

    (0.0001)

    19

    (0.0003)

    19

    (0.0001)

    19

    (0.0000)

    Obs Range Bott

    Salinity

    6

    (0.1005)

    8

    (0.0348)

    9

    (0.0135)

    9

    (0.0088)

    9

    (0.0110)

    13

    (0.0060)

    9

    (0.0011)

    Bottom Salinity 3

    (0.2762)

    6

    (0.0932)

    6

    (0.0402)

    7

    (0.0215)

    6

    (0.0300)

    9

    (0.0159)

    8

    (0.0029)

    Bottom DO 5

    (0.1340)

    7

    (0.0592)

    5

    (0.0603)

    8

    (0.0119)

    8

    (0.0260)

    7

    (0.0237)

    7

    (0.0030)

    Fresh Magnitude 20 20 21 21 21 20 21

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    62/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 54

    (0.0001) (0.0001) (0.0002) (0.0000) (0.0000) (0.0000) (0.0000)

    Fresh Freq 13

    (0.0069)

    14

    (0.0030)

    16

    (0.0029)

    15

    (0.0005)

    15

    (0.0010)

    16

    (0.0005)

    15

    (0.0001)

    Mixing due to MDF 11

    (0.0156)

    16

    (0.0025)

    14

    (0.0032)

    18

    (0.0002)

    17

    (0.0009)

    21

    (0.0000)

    17

    (0.0000)

    Mixing due to

    Freshes

    17

    (0.0031)

    18

    (0.0010)

    19

    (0.0006)

    17

    (0.0003)

    18

    (0.0003)

    14

    (0.0015)

    18

    (0.0000)

    Max Int Mix Evts 9

    (0.0301)

    12

    (0.0059)

    13

    (0.0066)

    13

    (0.0007)

    12

    (0.0020)

    18

    (0.0002)

    14

    (0.0002)

    Bott Water Quality 1

    (0.4386)

    1

    (0.3344)

    3

    (0.0781)

    2

    (0.1500)

    2

    (0.0827)

    4

    (0.0150)

    Hydraulic Habitat 1

    (0.4386)

    2

    (0.1693)

    1

    (0.1839)

    1

    (0.3250)

    1

    (0.1584)

    2

    (0.0275)

    Flush Magnitude 21

    (0.0000)

    21

    (0.0000)

    18

    (0.0014)

    20

    (0.0000)

    20

    (0.0000)

    15

    (0.0005)

    20

    (0.0000)

    Peak XS Velocity 18

    (0.0020)

    13

    (0.0036)

    8

    (0.0246)

    12

    (0.0014)

    13

    (0.0020)

    10

    (0.0130)

    12

    (0.0004)

    Severity of Slug

    Event

    2

    (0.3344)

    4

    (0.1693)

    5

    (0.0383)

    3

    (0.0980)

    3

    (0.0769)

    5

    (0.0120)

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    63/84

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    64/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 56

    6.3Analysis/evaluation of environmental flow scenarios and strategies

    The predicted probability distributions for all seven variables of management interest show

    the relative likelihood of alternative states under the worst case scenario and the SKM flow

    strategy (Figure 8). If flows cease completely over summer and there is no management

    intervention, Bottom Water Quality is likely to be in a Poor or Okay state with a

    probability of 0.38 and 0.37 respectively. Similarly, Hydraulic Habitat is likely to be a

    Poor or Average state with almost equal probabilities of about 0.42. The probability that

    the severity of a slug event is High is 0.26.

    The most likely state for Abundance Breeding Popnis moderate with a probability of 0.53and the most likely state for Recruitment Juvenile Catfishand Recruitment Larval

    Catfish is Low with a probability of 0.51 and 0.68 respectively. With respect to catfish

    viability (Viability of Catfish Popn), the most likely state is Good with a probability of

    0.85. Bottom Water Quality, Hydraulic Habitat, Severity of Slug Event and

    Recruitment Juvenile Catfishare all predicted to be substantially better under the SKM

    flow strategy, which provides a low level of minimum flows as well as four small, regularly

    spaced freshes. This is evident in the decrease in probability of the least desirable state foreach variable (particularly bottom water quality and hydraulic habitat) under the SKM

    strategy (Figure 8). Smaller improvements were observed in the abundance of catfish

    breeding population and recruitment of larval catfish and as expected there was little change

    in catfish population viability (Figure 8).

    Examination of the decision function for Median Mean Daily Flow (MDF)showed thatfor

    the utility function defined above, expected utility will be greatest when the median MDF is

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    65/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 57

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    Poor Okay Good

    B o t t o m W a t e r Q u a l i t y

    Dry

    SKM

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    Poor Ave Good

    H y d r a u l i c H a b i t a t

    Dry

    SKM

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0.7

    0.8

    0.9

    Low High

    S e v e r i t y o f S l u g E v e n t

    Dry

    SKM

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    Low Mod High

    A b u n d a n c e B r e e d i n g P o p n

    Dry

    SKM

    0.3

    0.4

    0.5

    0.6

    Dry

    SKM

    0.4

    0.5

    0.6

    0.7

    0.8

    Dry

    SKM

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    Poor Okay Good

    B o t t o m W a t e r Q u a l i t y

    Dry

    SKM

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    Poor Ave Good

    H y d r a u l i c H a b i t a t

    Dry

    SKM

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0.7

    0.8

    0.9

    Low High

    S e v e r i t y o f S l u g E v e n t

    Dry

    SKM

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    Low Mod High

    A b u n d a n c e B r e e d i n g P o p n

    Dry

    SKM

    0.3

    0.4

    0.5

    0.6

    Dry

    SKM

    0.4

    0.5

    0.6

    0.7

    0.8

    Dry

    SKM

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    66/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 58

    mixing events occur tends be high because the majority of pools have small cross-sectional

    areas see Figures 6 and 7), and the lower magnitude of such freshes reduces the risk of aslug event of high severity.

    40

    42

    44

    46

    48

    50

    52

    54

    56

    1 (240) 2 (120) 3 (80) 4 (60)

    No. of 'freshes' ('Fresh' volume in ML)

    E

    xpected

    UtilityValue

    MDF=0 ML/day

    MDF=10 ML/day

    MDF=30 ML/day

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    67/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 4 59

    environmental outcomes can be achieved for quite a wide range of environmental flow

    volumes as long as the maximum number of four freshes are provided over summer.

    Risk based Approaches for Managing Contaminants in Catchments Report 2 60

  • 8/12/2019 Use of a Bayesian Network Decision Tool to Mgt Environ Flow in River

    68/84

    Risk-based Approaches for Managing Contaminants in Catchments Report 2 60

    Table 7: Summary of the main characteristics and expected utility value of various flow strategies evaluated using the BN decision

    network. The two shaded rows highlight the base cases of a worst case Dry scenario and the SKM flow strategy that serve as

    benchmarks for evaluation.

    Scenario/Flow Strategy (ID,

    Description)

    Median MDF

    (ML/day)

    Fresh

    Magnitude (ML)

    Fresh

    Frequency

    Total Volume

    Required (ML)

    Utility Value

    1 Worst case Dry: no flows at all oversummer

    0 0 0 0 42.5

    2 SKM environmental flow

    recommendations

    5 20 4 765 52.4

    3 High sustaining flow strategy: high

    median MDF & no freshes

    30 0 0 2070 52.6

    4 Moderate sustaining flow strategy:

    low-mod median MDF & no freshes

    10 0 0 690 49.5

    5 Low sustaining flow strategy: lowmedian MDF & no freshes

    5 0 0 345 46.0

    6a

    b

    c

    d

    e

    f

    g

    Freshes only strategy:

    No MDF & 4 freshes

    0

    0

    0

    0

    0

    0

    0