Environmental Flows Monitoring and Assessment Framework ... · Cover photographs courtesy of Robert...

Environmental FlowsMonitoring and AssessmentFramework

Version 1.1

P. CottinghamCRC for Freshwater Ecology, Victoria

G. QuinnMonash University, Victoria(Now at Deakin University, Victoria)

A. KingDepartment of Sustainability and Environment, Victoria

R. NorrisUniversity of Canberra

B. ChessmanDepartment of Infrastructure, Planning and Natural Resources, NSW

C. MarshallDepartment of Natural Resources and Mines, Qld

Cooperative Research Centre for Freshwater Ecologyii

The Cooperative Research Centre for Freshwater Ecology is a national research centrespecialising in river and wetland ecology. The CRC for Freshwater Ecology provides theecological knowledge needed to help manage rivers in a sustainable way. The CRC wasestablished in 1993 under the Australian Government’s Cooperative Research CentreProgramme and is a joint venture between:

ACTEW CorporationCSIRO Land and WaterDepartment of Environment and Conservation, NSWDepartment of Infrastructure, Planning and Natural Resources, NSWDepartment of Natural Resources and Mines, QueenslandDepartment of Sustainability and Environment, VictoriaDepartment of Water, Land and Biodiversity Conservation, South AustraliaEnvironment ACTEnvironment Protection Authority, VictoriaGoulburn-Murray WaterGriffith UniversityLa Trobe UniversityLower Murray Urban and Rural Water AuthorityMelbourne WaterMonash UniversityMurray-Darling Basin CommissionSydney Catchment AuthorityThe University of AdelaideUniversity of Canberra

Please cite this report as:

Cottingham P., Quinn G., Norris R., King A., Chessman B. and Marshall C. (2005).Environmental Flows Monitoring and Assessment Framework. Technical report. CRC forFreshwater Ecology, Canberra.

ISBN 0-9751642-05

© Cooperative Research Centre for Freshwater Ecology 2005

Cover photographs courtesy of Robert Ashdown (floodplain near Cooper Creek),John Hawking (Philopotamidae) and Murray-Darling Freshwater Research Centre

Further copies are available from:CRC for Freshwater EcologyTel: 02 6201 5168Fax: 02 6201 5038Email: [email protected]: http://freshwater.canberra.edu.au

Design and typesetting: TechType, Giralang ACTPrinting: Elect Printing, Fyshwick ACT

Environmental flows: monitoring and assessment framework iii

Table of Contents

Preface .................................................................................................................................... v

Acknowledgements ............................................................................................................... v

1. Monitoring and assessment of environmental flows .................................................. 1

1.1. Key steps of the monitoring and assessment framework ............................................................ 1

1.2. Define the scope of the monitoring and assessmentprogram and its objectives .......................................................................................................... 3

1.3. Define the conceptual understanding of flow–ecologyrelationships and the questions (hypotheses) to be tested .......................................................... 9

1.4. Select the variables to be monitored ......................................................................................... 10

1.5. Determine the study design ...................................................................................................... 11

1.6. Optimise study design ............................................................................................................... 15

1.7. Implement the study design ...................................................................................................... 18

1.8. Have the environmental flows met their specific objectives? .................................................. 19

2. Other issues ................................................................................................................... 20

2.1. Levels of evidence and causal inference .................................................................................. 20

2.2. Analysis of monitoring data ...................................................................................................... 21

2.3. Priorities for monitoring ........................................................................................................... 21

3. Further reading ............................................................................................................. 22

Appendices

Appendix 1. Overview of the Multiple Lines and Levels of Evidence(MLLE) approach ................................................................................................................ 24

Appendix 2. Wimmera-Glenelg environmental flows monitoring program ............................................. 26

Appendix 3. Potential study designs ......................................................................................................... 31

Boxes

Box 1. Environmental flows and river management .................................................................................. 2

Box 2. Environmental flow objectives ...................................................................................................... 4

Box 3. Assessing predicted responses to environmental flows: the Queensland andNew South Wales approaches ...................................................................................................... 16

Cooperative Research Centre for Freshwater Ecologyiv

PrefaceThe Cooperative Research Centre for Freshwater Ecology currently has research programs on:

1. flow-related ecological processes

2. restoration ecology

3. conservation ecology

4. water quality and ecological assessment.

These programs are anticipated to provide valuable new information on environmental wateringrequirements of river systems and on assessing the performance of environmental flow regimes,whether for the protection or the rehabilitation of river systems.

This report is intended to be a ‘live’ document, because it will be updated as new insightsemerge on environmental flows and how to measure their performance.

AcknowledgementThank you to everyone who contributed to this report by attending workshops and commentingon drafts.

Environmental flows: monitoring and assessment framework 1

1. Monitoring andassessment ofenvironmental flows

Environmental flow or streamflowmanagement plans to meet environmentalwater requirements have been prepared formany regulated and unregulated streamsacross Australia. The emphasis of theseplans has been to maintain or protectenvironmental or ecological values byensuring sufficient water is available (e.g. byrestrictions on diversions) for plant andanimal communities or ecosystem functionsto remain viable, or by returning water toflow-stressed streams as a rehabilitationmeasure (see Box 1, page 2). Theimplementation of the flow regimesrecommended by these plans can represent alarge investment in river protection andrehabilitation. Increasingly, stakeholderswith an interest in water resourcemanagement will expect to see evidence ofthe environmental or ecological response ofrivers to environmental flow regimes.Monitoring and assessment is thereforeessential if we are to confirm and understandhow environmental flows result in thepredicted outcomes, and apply any lessonslearnt to future management.

Environmental flows may focus on changesto the whole flow regime (e.g. changes tomean annual flow), on specific short-termevents (e.g. targeted pulses) or both, or onthe protection of components of the flowregime critical to the protection ofecological ‘assets’ — species, communitiesor ecological functions that haveenvironmental values that are to beprotected. As there are likely to be multipledrivers (i.e. factors other than modificationto the flow regime) of river condition formost systems, other complementarymanagement actions (e.g. provision ofpassage past barriers to migration, improvedwater quality, protection or reintroduction ofphysical habitat) may also be required if theanticipated ecological responses toenvironmental flows are to occur.

Monitoring and assessment is a majorcomponent of an adaptive managementcycle (e.g. IEPEF 2002, Bosch et al. 2004),which includes steps such as the:

• establishment of managementobjectives

• review of resource condition

• formulation of management questions(hypotheses) to be tested

• implementation of management actions

• monitoring and assessment of collectedinformation

• review of management objectives andwhether they have been met, andrevision of management objectives andactions in the light of new evidence.

The scope and goals of a monitoring andassessment program are best consideredfrom the outset of an environmental flowproject. This will help to ensure that theprogram is aligned with the ecologicalobjectives of the environmental flow regimeand can be included in managementplanning.

1.1. Key steps of the monitoring andassessment framework

This framework has been produced as aguide for water and catchment managementagencies and authorities with responsibilityfor delivering and assessing theeffectiveness of specific environmental flowregimes. The framework also serves as achecklist for scientists involved withdesigning and implementing monitoringprograms. The framework focuses onassessments of particular environmentalflow projects and can be applied to assessthe ecological responses resulting from:

• releases of water from a storage such asa reservoir (regulated rivers),

• modified water extraction directly froma river (regulated and unregulatedrivers),

• water allocation to specific sections of asystem (e.g. allocation of water to iconsites in the Living Murray).

Cooperative Research Centre for Freshwater Ecology2

Box 1. Environmental flows and river management

The following definition has been adopted for this framework:

An environmental flow results from a management intervention that protects ormodifies the flow regime of a river to achieve an ecological or environmentaloutcome.

Such a definition can be applied equally to unregulated and regulated rivers.

In many instances, meeting a stream’s environmental water requirements means theprotection of existing components of the flow regime that are ecologically important. Inunregulated streams, for example, this may be to ensure sufficient water remains in astream so that critical habitat remains for biota during periods of low or zero flow, or toprotect flow pulses that provide important biological cues. In such situations, limits may beapplied to the volume or timing of water that can be diverted from a stream.

In regulated streams, environmental flows are usually designed to return some aspect ofthe volume, timing or frequency of flow components that may have been lost or modifiedby the presence of dams, weirs and associated infrastructure. Manipulating the flow regimeto achieve an ‘ecological or environmental outcome’ means that environmental flows canbe considered as a rehabilitation measure, guided by the science of restoration ecology(e.g. Lake 2001, Palmer et al. 1997, Bradshaw 1996). River rehabilitation implies thereturn of attributes such as community structure (e.g. fish or macroinvertebratepopulations) or function of the original (e.g. production, respiration, nutrient cycling) butwithout a complete return to pre-disturbance condition (Figure 1). In many instances,factors such as widespread change to land-use or water management will mean thatrehabilitation will not be possible. Environmental flows may then serve as a form ofremediation where the stream moves to some state that represents a ‘new’ ecosystem.

A good conceptual understanding of how a river may recover from disturbance (in thiscase changes to the flow regime) is important to any monitoring and assessmentframework. For example, there may be time lags before recovery becomes evident, thesystem may not respond as desired and progress to some alternative state, or may beunstable (Bradshaw 1996, Lake 2001). Monitoring and assessment programs have toaccount for such possibilities in order to answer questions such as ‘What was done?’ ‘Didit work?’ ‘Why did it work or not work?’ and ‘Will it work in other situations?’ (Michener1997).

New ecosystem

Remediation

2

3

1

Original ecosystem

Rehabilitation

Restoration

Ecosystem structure: species richness

Ecos

yste

m fu

nct

ion

: bio

mas

s

Degraded ecosystem

Degradation Figure 1. Potential ecosystem response todisturbance (from Bradshaw 1996). Note thatrehabilitation targets will be less than fullrestoration of the original condition.


This framework:

• assumes that development ofenvironmental flow recommendationshas isolated the role of the flow regimein maintaining or improving rivercondition and considered otherappropriate non-flow managementactions;

• focuses predominantly on situationswhere (i) maintenance of flow or waterlevels is required as an ecosystemprotection measure, and (ii) changes tothe flow regime are required as arehabilitation measure (althoughassessing the impact of further waterresource development is not excluded);

• considers aspects of conditionassessment, compliance monitoring andcausal links (dose–response);

• considers the need to detect if there hasbeen a response to the intervention (i.e.the direction of the response — e.g.increased or decreased abundance ofbiota) and the level of the effect (i.e. thestrength of the response);

• recognises that stakeholder input will berequired, particularly on agreeing on thesize of the ecological or environmentalresponse that will be the basis of anassessment program.

This monitoring and assessment frameworkis based on the following key steps:

1. Define the scope of the program and itsobjectives

2. Define the conceptual understanding offlow–ecology relationships and thequestions (hypotheses) to be tested

3. Select variables to be monitored

4. Determine the study design, accountingfor the specific activities and location

5. Optimise the study design and identifyhow data are to be analysed

6. Implement the study design

7. Assess whether the environmental flowshave met the specific objectives andreview the conceptual understandingand hypotheses.

The framework is consistent withapproaches recommended by the AustralianGuidelines for Water Quality Monitoringand Reporting (ANZECC & ARMCANZ2000), and the steps recommended byDownes et al. (2002). The framework alsoresponds to the recommendation of King etal. (2003), who argued that ‘a consistent andrigorous approach to the design ofmonitoring would result in greaterconfidence about links between ecologicalresponse and flow change.’ The key steps ofthe framework are summarised in Figure 2and explained in the following sections.

1.2. Define the scope of the monitoringand assessment program and itsobjectives

It is important to reflect on the objectives oroutcomes that are the basis of theenvironmental flow recommendations for ariver system and the objectives that are setfor a monitoring and assessment program.Monitoring and assessment programs aremost effective and informative whendesigned to answer clear and precisemanagement and scientific questions.

This framework assumes that therelationship between flow regime and rivercondition has been examined in arriving atenvironmental flow recommendations andthat the need to protect or modify the flowregime has been established (see Box 2).This step is thus one of revisiting orrestating the existing objectives and ofdefining those objectives that will form thebasis of the monitoring and assessmentprogram.

Often, an environmental flow regime for ariver is presented as a package of recom-mendations related to various flowcomponents, such as low flows, bank-fullflows, flow pulses, overbank flows and ratesof rise and fall (e.g. DNRE 2002). Eachrecommendation should be related to an envi-ronmental or ecological objective or outcome.


Box 2. Environmental flow objectives

Most environmental flow studies make clear statements on the timing, duration andmagnitude of flow events predicted to achieve some desired outcome or condition (seeTable 1). Environmental flow objectives that describe specific elements are preferred tobroader statements of outcome such as ‘… improved river health’, which can be difficultto define.

A challenge is then to move from stated objectives to the quantifiable targets that are thebasis of a monitoring and assessment program. Developing quantifiable targets requiresconsideration of appropriate variables to measure, and of the amount of evidence (effectsize) needed to convince stakeholders that the environmental flows had the desiredoutcome. These issues will be considered in greater detail later in the framework.

Heron et al. (2002) proposed that environmental flow objectives should contain fivedistinct elements, which provide a useful checklist when integrating monitoring andassessment within an adaptive management cycle:

1. the component of the environment that is addressed (e.g. individual species,communities, process)

2. the ‘event’ that needs to be protected (e.g. fish spawning, fish migration, communitydiversity)

3. the target. This is essentially the purpose or aim of the objective. It may be a valuethat the event should reach, or how far it may deviate from natural, or some targetcompared to the current condition .

4. the ‘Success Criteria’, detailing what conditions need to be achieved to ensure that theobjective is met. The success criteria always relate directly to the ‘event’.

5. the ‘Measure of Success’, or the variable that needs to be measured, and what value itmust attain. Often, the Success Criteria cannot be easily measured directly, so theMeasure may be some other factor that can be used as a surrogate.

Table 1. Example objectives and environmental flow recommendations for the Wimmera River, Reach 2 (Huddleston-McKenzie River); Compliance point = Faux Bridge, Gauge no. 415240 (SKM 2002)

Flow

Season Magnitude Frequency Duration Objective/Rationale

Summer 0 ML/d Annually 17–30 days Natural stress to promote macroinvertebrate biodiversity

Minimum flow 6 ML/d

Annually Dec–May

Maintain quality and quantity of habitat for native fish, macroinvertebrates and aquatic vegetation

>16 ML/d 3 annually 7–15 days Enhance recruitment of short-finned eels and river blackfish

Spring (Jul–Nov)

>164 ML/d 2–3 annually Minimum 14 days

Maintain riparian vegetation and habitat for native bird species

Minimum flow 60 ML/d

Annually July–Nov Inundate snags and other elements that provide habitat for native fish, macroinvertebrates and aquatic vegetation, and maintain longitudinal connectivity

Annual 6,000 ML/d Annual Minimum 2 days

Provide high flows to cue and enhance recruitment of golden perch, Murray cod and Macquarie perch, maintain riparian vegetation and habitat for native bird species and facilitatechannel-forming processes.


Figure 2. Generic environmental flow monitoring framework — summary outline. Note that variables to be measured will be thosehypothesised to respond to flow change.

Understand the system:

• Confirm the environmental flow objectives • Confirm the scale at which environmental

• flows apply (spatial and temporal)

Confirm the objectives for the monitoring program:

• Includes definition of measurable• end points

Develop conceptual models based on:

• Literature

• Hypotheses

• Expert opinion

Consider the levels of evidence that support the conceptual models and indicator variables that respond to flow

Select response variables based on strengths of association identified in conceptual modelling and reviews of previous studies and their outputs

High strength of association variables

Low strength of association variables orlittle data with which to draw conclusions

‘Communicable’‘iconic’ or ‘flagship’ variables required?

Continue with monitoring design

Further investigations or alternative approaches and review are required

Compliance monitoring — delivery of the agreed flow regime

Define conceptual understanding

and hypotheses to be tested

Select variables to be

monitored based on

conceptual strength of

association

Define the scope of the project and its objectives

Coo

perative Research

Cen

tre for Fresh

water Eco

log

y6

Continued

Can measure change in variable (time trend) at the intervention location (s). Inference that change is causally linked to flow is weaker without spatial ‘controls’ and/or ‘before’ data. See Appendix 3.

Yes

2. Reference– Intervention design

Can measure change in variable at the intervention location(s) relative to reference or target condition. Inference that change is causally linked to flow is weaker without spatial ‘controls’ and/or ‘before’ data. See Appendix 3.

Are ‘before’ data available (i.e. information on the pre- environmental flow release condition of the river)?

Are spatial ‘control’ locations available?

Are reference locations available?

Yes

1. Intervention-only design

No No

No

Are reference locations available?

Yes

No

3. Control–Intervention design

Can measure change in variable at the intervention location(s) relative to ‘control’ location. Inference that change is causally linked to flow is weaker without ‘before’ data. See Appendix 3.

4. Control–Reference–Intervention design

Can measure change in variable at the intervention location(s) relative to ‘control’ location(s) and reference condition. Inference that change is causally linked to flow is weaker without ‘before’ data. See Appendix 3.

Model reference condition?

Yes

Determine study design,

accounting for the specific

activities and location

from page 5

Environ

mental flo

ws: m

on

itorin

g an

d assessm

ent framew

ork

7

Continue with framework

Can measure change in variable before versus after intervention at intervention location(s) only. Inference that change is causally linked to flow is weaker without spatial ‘controls’. See Appendix 3.

Determine study design,

accounting for the specific

activities and location

Yes

5. Before–After–Intervention design

No NoAre reference sites available?

Are spatial ‘control’ sites

available?

6. Before–After Reference–Intervention (BARI) design

Can measure change in variable from before to after the intervention at intervention location(s) and compare with reference condition. Inference that change is causally linked to flow is weaker without spatial ‘controls’. See Appendix 3.

Yes

Are reference sites available?

No

7. Before–After Control–Intervention (BACI) design

Can measure change in variable from before to after the intervention at intervention and ‘control’ location(s). Strongest inference about causal links between flow and ecologic al response at the intervention site. See Appendix 3.

8. Before–After Control–Reference–Intervention (BACRI) design

Can measure change in variable before versus after the intervention at intervention and ‘control’ location(s), and assess ifdirection of change is towards the desired reference condition. Strongest inference about causal links between flow and ecological response at the intervention site. See Appendix 3.

Yes

from page 6: 'before' data are available


For each variable, agree on the effect size (size of the ecological response to be detected) and the duration and spatial extent of the sampling design (potentially an iterative process).

This step requires stakeholder input, potentially as a 3-step process: 1. Stakeholder group to examine effect size (evidence required). 2. Undertake pilot study (feasibility of establishing monitoring sites or

measuring variables). 3. Revisit effect size with stakeholder group.

Statistical analysis may be required to inform stakeholders of the implications of effect size adopted.

Develop a contingency plan. Undertake risk assessment of:

•

unacceptable change due to implementation of environmental flow regime (e.g. carp breeding and distribution) and

• risk to the system if environmental flows are not delivered.

Implement monitoring program

Revisit study environmental flow objectives and conceptual models within an adaptive management framework

Quantify conceptual models in the light of monitoring results(see later sections of this report)

Revisit study environmental flow objectives and conceptual models within an adaptive management framework (after delivering and assessi ng delivery of flows) — analysis, conclusions, feedback.

Optimise study design

Implement the study design

Assess whether the

environmental flows have met

specific objectives, and review

hypotheses

from page 7: study design determined


These can include geomorphology and waterquality attributes if linked conceptually tothe interaction of organisms with theirenvironment. The temporal and spatial scaleat which the objective or outcome will applyis likely to vary, depending on the nature ofthe flow component and the biota orecological processes that are predicted torespond. A package of environmental flowrecommendations may represent a largechange to the management of a river system,and the water that may be available forconsumptive or agricultural purposes. Somerecommendations may be implementedquickly (e.g. those that pose little risk to thesecurity of urban or agricultural supply),while other recommendations may not bedelivered for some time (e.g. whileenvironmental water rights are secured, or ifenvironmental flow releases pose a high riskto infrastructure), if at all. Confirming whichenvironmental flow recommendations are tobe delivered and their spatial and temporalbounds is an important consideration, as itinfluences the scope and realism of themonitoring program objectives.

1.3. Define the conceptualunderstanding of flow–ecologyrelationships and the questions(hypotheses) to be tested

Conceptual models are useful tools forexplicitly defining interactions in a riversystem, in this case the relationship betweenthe flow regime and potential ecologicalresponses. The models can be used to:

• highlight the relationships betweenbiota and the physical and chemicalenvironment,

• show how a river might respond todisturbances or events such as alteredflow regimes,

• provide the basis for hypotheses thatcan be tested in a monitoring andassessment program.

Ideally, the relevant conceptual models thatwere the basis for setting environmental

flow objectives will also provide the basisfrom which to design a monitoring andassessment program.

Conceptual models are best developed froma broad knowledge base of the study regionincluding biological, chemical, hydrological,geological and geomorphological attributes.This can include knowledge extrapolatedfrom similar systems, the scientificliterature, general hypotheses and modelsrelevant to that type of river system, such asthe Flood Pulse Concept (Junk et al. 1989),and considerations from experiencedscientists and managers, such as thoseappointed to ‘scientific panels’ (Cottinghamet al. 2002). The different components andlinks in a model are likely to have varyinglevels of uncertainty. However, a review ofenvironmental flow monitoring programs(King et al. 2003) found that the underlyingassumptions and uncertainty associated withconceptual models are rarely statedexplicitly. The level of uncertainty and thetemporal scale of predicted ecologicalresponses to changes in the flow regime areimportant considerations for a monitoringand assessment program.

Figure 3 is an example of a conceptualmodel that can be used to develophypotheses to be tested by a monitoring andassessment project. For example, the modelsuggests that decreased low flows and areduced frequency of flushing have led to anincreased retention of nutrients and finesediment, resulting in conditions favourablefor the growth of filamentous algae andbiofilm that is unpalatable for macro-invertebrates, increased armouring of thestream bed and a reduction in habitatavailability and quality for macro-invertebrates and small fish. A set ofenvironmental flow hypotheses might thenbe that:

a flow pulse (e.g. equivalent to bankfulldischarge) with a duration of 3–4 dayswill

• mobilise and flush fine sedimentsfrom the bed substrate,


• scour filamentous algae andbiofilm from the bed,

• increase habitat diversity andavailability and, ultimately,increase macroinvertebrate andfish diversity and abundance.

Such hypotheses are an important basis forthe selection of variables to be measured aspart of the monitoring and assessmentprogram, in this case suggesting that flow,sediment grain size, filamentous algae andbiofilm cover, macroinvertebrate and fishcommunities should all be measured.

1.4. Select the variables to bemonitored

The selection of appropriate variables is avery important component of monitoring

and assessment program design. Factors toconsider when selecting variables include:

• the specific environmental flowobjectives and hypotheses to beexplored by the monitoring andassessment program,

• the degree of confidence that changes ina variable imply that there are causallinks between flow changes andenvironmental or ecological response,

• information that may be required toassess and manage risks to the system(e.g. if the system does not receive therequired environmental flows, or if theenvironmental flows result in someundesirable outcome),

• information to assist communicationand foster community engagement (e.g.icon species).

Figure 3. Ecological responses expected when flow is reduced in the Cotter River, Australian Capital Territory (R.Norris, CRCFE, pers. comm.).

1. Riparian vegetation encroaches into the channel and reduces channel capacity. 2. Unpalatable filamentous algae accumulate. 3. Reduced flow results in armouring, reduced flushing of detritus, nutrients, fine sediment. 4. Habitat space for macroinvertebrates and fish in the substratum is reduced because of armouring and infilling with fine sediments. Also, some parts of the bottom may be exposed. 5. Sediment and organic matter may enter the channel directly from adjacent valley slopes and may not be flushed by low flows in the main channel.

Cotter River conceptual model

Reduced flow

1

2

4

P < R

5

Natural low flow3


There is an enormous amount of literatureon potential variables of a wide range ofdifferent stressors in river systems (Downeset al. 2002). Watts et al. (2001) used anextensive range of criteria to select whichvariables would be used when studyingenvironmental flows in the MurrumbidgeeRiver, including:

• responsive to changes in flow at spatialand temporal scales relevant to rivermanagement,

• responsive within the timeframe of theproject,

• have scientific justification,

• represent important structural and/orfunctional component of the riverineecosystem,

• easily measured and quantitative,

• responses easy to interpret,

• can determine and measure directions ofchange,

• respond differently to backgroundvariability,

• cost-effectiveness,

• relevant to policy and managementneeds,

• cover a range of habitats and trophiclevels, several measures of biodiversity,a range of organisational levels and arange of spatial and temporal scales.

King et al. (2003) examined the variablesused in existing environmental flowmonitoring programs in Australia. Theyfound that only some of these variables haveas yet been causally linked to changes inflow regime and respond in a predictablemanner (Table 2). New variables, with directand predictable responses to flows, will nodoubt emerge in time and monitoringprograms should be flexible so that newvariables can be incorporated as ourunderstanding of relationships between flowchange and ecological responses improves.

This framework does not include instructionon how to measure selected variables andmanage the data collected. Guidance onthese and other related issues can be

obtained from resources such as theAustralian Guidelines for Water QualityMonitoring and Reporting (ANZECC &ARMCANZ 2000) and RecommendedMethods for Monitoring Floodplains andWetlands (Baldwin et al. 2004). Thesereferences contain detailed descriptions ofwater quality and biological measures,including information on the spatial scale atwhich to monitor. Importantly, a qualityassurance/quality control (QA/QC) programis recommended as an essential step incollecting high quality and reliable data.

1.5. Determine the study design

This framework assumes that we areinterested in establishing causal linksbetween environmental flows and ecologicalresponses in a specific river system. Theterm ‘location’ is used to represent a sectionof river that is the target of environmentalflow recommendations. A ‘location’ may bea whole stream, a reach or a localised pool–riffle sequence or wetland. The spatial scaleof a ‘location’ will be determined by theenvironmental flow objectives. The term‘intervention’ is used to describe theenvironmental flow regime.

Australian rivers are often highly variable innature, in terms of both hydrology andecological response (Puckridge et al. 1998).Separating changes in ecological conditiondue to environmental flows from othernatural or human induced variabilityrequires an understanding of conditions bothbefore and after environmental flows aredelivered. Conditions at the location wherean environmental flow regime isimplemented (preferably assessed bothbefore and after the intervention) can thenbe compared with conditions at locationsthat represent ‘control’ and/or ‘reference’conditions (Downes et al. 2002). ‘Control’locations are as similar to the interventionlocation as possible, except that there is nointervention (environmental flow) there. Forexample, if an environmental flow were tobe released from a large dam on a regulated


Table 2. Environmental variables* with established causal links with changes to the flow regime (adapted from King et al. 2003)

Ecosystem components

Response variables Where used Comments

River productivity

Benthic production/respiration, water column production, bacterial activity

Mitta Mitta Short-term responses to specific flow events as predicted

Biofilm Total/algal/organic biomass, productivity

Murrumbidgee, NSW IMEF†, Mitta Mitta

Consistently responded as predicted to flow events in the Murrumbidgee and Mitta Mitta

Composition Mitta Mitta Structural and functional responses of biofilm were evident immediately following peak flows

Macroalgae Filamentous algae abundance Mersey, Snowy Preliminary results suggest good re-sponse to flow events in Mersey study

Macro-invertebrates

Community structure and abundance Snowy, Mitta Mitta, Mersey

Several attributes measured in cobble habitats responded rapidly to variable flow releases in Mitta Mitta study. Pre-liminary results suggest good response to flow events in Mersey study.

Number of families, SIGNAL scores Mitta Mitta Responded rapidly to variable flow release

Community structure, relative abundance and species occurrence on snags

Campaspe Preliminary results suggest good response to flow stress

Mayfly larvae (abundance, species richness and diversity)

Murrumbidgee Responded predictably to flow events in upper reaches

Abundance and composition of shrimp fauna


Community structure and abundance of wetland macroinvertebrates

NSW IMEF†, Barmah-Millewa wetlands

Some evidence that shows good responses to flow events at Barmah-Millewa wetlands

Vegetation Riverbank understorey vegetation (species composition, distribution, abundance, survival, growth, reproduction)

Murrumbidgee Survival and total biomass responded predictably to flow events in lower reaches

Wetland vegetation NSW IMEF†, Barmah-Millewa

Successful for Barmah-Millewa project

Fish Larval fish (occurrence, relative abundance, community composition)


Recruitment Snowy, Mersey Preliminary results suggest potentially good response to flow stress for the Mersey

Waterbirds Abundance, diversity and breeding occurrence in wetlands

Barmah-Millewa wetlands

Easily communicated and can assess effect of watering quickly

Frogs Abundance, diversity and breeding occurrence in wetlands

Barmah-Millewa wetlands

Easily communicated and can assess effect of watering quickly

*It is assumed that hydrological variables such as mean daily flow will be automatically included. †IMEF = Integrated monitoring of environmental flows


river then a ‘control’ would be a similarriver where flow is regulated via a dam, butwithout an environmental flow release.‘Control’ locations are always more useful ifthey are in rivers separate from the rivershaving intervention, although occasionallyupstream versus downstream comparisonsmight be applicable, such as upstream versusdownstream of a storage from which flowsare released.

Reference locations are those that are, asnearly as possible, in the condition of anenvironment undisturbed by human activity.Having both ‘control’ and referencelocations allows us to determine if anenvironmental flow causes an ecologicalresponse, and if the condition at theintervention location changes towards adesired future state. Reference conditionshelp to describe what a river system mightbe in the absence of disturbance (e.g. flowregulation or diversion) and so provideuseful comparison with which to gaugerecovery at the intervention location. It isimportant to distinguish between ‘natural’and ‘target’ reference condition (see Box 1).Returning a modified river system to a‘natural’ or pre-disturbance state is usuallyunachievable. A ‘natural’ referencecondition provides a useful basis againstwhich river condition can be compared, butit should not be confused with the ‘target’condition/s upon which the environmentalflow objectives have been set and will beassessed. Reference locations, like‘controls’, may not be readily available.However, it may be possible to modelreference condition based on a desiredfuture condition or conditions where theinfluences of flow regulation or waterdiversion have been removed. For example,environmental flow studies often model thepre-disturbance flow regime by removingthe influence of impoundments and waterextraction.

In addition to spatial ‘controls’ (i.e. ‘control’locations), temporal ‘controls’ also increaseour confidence that an observed response isdue to the environmental flow regime.

Temporal ‘controls’ simply mean measuringthe selected variables at the intervention sitebefore the environmental flow regimecommences. Collecting information beforeand after an intervention, at both ‘control’and intervention locations, allows us to useBACI (Before–After Control–Intervention)designs that can be very powerful forinferring causality between a managementaction and an ecological response (Downeset al. 2002). Designs 7 and 8 in Figure 2 areBACI designs and provide the strongestinference about causal links betweenecological responses and flowmodifications, assuming that variables beingmeasured have a strong conceptual basis.These more ‘experimental’ designs arebased on the BACI designs for monitoringthe effects of spatially-explicit humanactivities in the environment (traditionalimpact assessment) (Downes et al. 2002).

It is often the case that ‘before’ data, and/or‘control’ and reference locations are notavailable, meaning that traditional BACIdesigns cannot be implemented. This isparticularly likely for large regulated rivers,where comparable rivers withoutenvironmental flows do not exist. Forexample, what river might be an appropriate‘control’ for environmental flows on theRiver Murray? In addition, funding andpersonnel constraints will almost alwaysmean there is a limit to what can bemonitored, at what spatial scale and for howlong (Michener 1997). BACI designsincluding ‘before’ sampling plus ‘control’(and even reference) rivers can beexpensive, and will usually only be used inhigh priority cases, rather than generallyapplied across a broad scale to assessresponses to environmental flows. Thiscreates the dilemma of making tradeoffs indesigning monitoring programs. Is it betterto sample a limited number of variables at afew locations for a long period of time, or tosample a larger number of variables over ashorter period of time? Is it better to samplea limited number of variables at a largenumber of locations? The use of


environmental flows as a river protectionand rehabilitation tool is a relatively newpursuit and only a small number ofenvironmental flow monitoring andassessment programs have been establishedin Australia (King et al. 2003). It isrecommended that an emphasis be placed,where possible, on measuring fewer high-quality variables within a scientificallysound study design. It is likely that thisinvestment will be more informative to rivermanagement and restoration ecology thantrying to measure a large number ofpotentially less informative variables.

In situations where ‘controls’ are notavailable, then monitoring designs arerestricted to simply assessing the responsesto environmental flows at interventionlocations. Two types of designs might beused. If before-intervention data areavailable, because the environmental flowregime had a clear starting date, then‘before’ versus ‘after’ contrasts are possible(designs 5 and 6 in Figure 2). If ‘before’data are not possible, then monitoring canonly assess responses at interventionlocations through time (design 1 in Figure2). There may also be situations where‘before’ data are not possible but ‘control’(or reference) rivers are available, sointervention versus ‘control’ (or reference)contrasts through time are possible (designs2–4 in Figure 2). Advice from experiencedstatisticians will be helpful whenconsidering the inferences that may bedrawn from the study designs available andhow best to proceed with data analysis (seelater sections).

Clearly, designs that focus on assessingphysical, hydrological and ecologicalresponses to changed flow regimes in riverswithout ‘before’ data or ‘control’ rivers (i.e.only at intervention locations) will becommonly used. This might be because‘control’ rivers or ‘before’ data are notavailable (e.g. Wimmera and GlenelgRivers; Sharpe and Quinn 2004). It may alsobe because the goal of monitoring is toprovide an assessment of environmental

flow regimes more regionally and BACI-type designs are not feasible economicallyfor so many rivers. Queensland and NewSouth Wales have encountered bothproblems when trying to adopt BACIdesigns for monitoring environmental flowsand have responded by evaluatingpredictions, based on an understanding ofhow rivers respond hydrologically orecologically to modified flows (includingenvironmental flows). Both jurisdictionsmonitor numerous intervention locations tomeasure whether the predictedenvironmental or ecological outcomes holdtrue. The approach used by Queensland istermed ecological performance monitoringand focuses on measuring hydrological andhydraulic conditions required (or critical)for identified ecological assets. NSW hasmodelled conditions that would existwithout the environmental water allocationand also the natural condition, so that threescenarios can be tested: ‘environmentalflows’, ‘full development’ and ‘natural’.This enables a BACI-type experiment, albeitwith modelled, rather than physical,‘controls’ and benchmark conditions. TheNSW method is an integrated approach tomonitoring environmental flows, measuringboth physical and ecological responses.These two programs are summarised inBox 3.

Designs without spatial and/or temporal‘controls’ make it harder to determinewhether the observed ecological orenvironmental responses are caused byenvironmental flows, i.e. it is more difficultto rule out alternative explanations.Location-specific BACI-type designs allowus to test predictions about ecologicalresponses to environmental flow responsesmore formally, and provide greaterconfidence when inferring a causal linkbetween responses and environmental flows.Understanding causal links betweenobserved responses and environmental flowsis critical for future predictions and adaptivemanagement. Where environmental flowscan be treated as a management experiment,


and before-intervention data and/or spatial‘control’ rivers are available, the BACIdesigns outlined in this framework should beadopted. Experience to date in Australiasuggests that opportunities to apply BACIdesigns will be relatively rare. Wherepossible, we should take advantage of suchopportunities, as they will provide thestrongest inference that an environmentalflow causes the predicted environmental orecological response and will alsocomplement studies where evaluation ofpredictions at intervention locations is theonly option available (a ‘levels of evidence’approach, see Appendix 1).

The ‘study design’ section in Figure 2provides a decision tree to help identify thedesign/s that may be applied at a particularlocation, taking into account the availabilityof ‘before’ data, ‘control’ and referencelocations. The inferences that may be drawnfrom the various study designs are discussedin more detail in Appendix 3.

1.6. Optimise study design

Arriving at the optimal study design willoften be an iterative process. It is notunusual for aspects of the preferred studydesign to be confounded with each other —for example, due to logistical constraintssuch as difficult study site access, or due tounforeseen factors such as localiseddisturbance (e.g. localised pollution).

A critical step in the optimisation process isgetting agreement on the evidence that willconvince stakeholders that theenvironmental flows delivered the predictedresponse. This requires consideration ofeffect size, which is the size of theecological response that is to be detected bythe monitoring and assessment program.Effect size is, therefore, closely linked tospecific targets that should be the measureof the set environmental flow objectives. Forexample, if an environmental flow objectiveis to protect or reinstate native fishpopulations, then measurable targets might

include the species of interest, targets ofabundance (e.g. 50% increase over 3 years),frequency of successful recruitment (e.g. 2out of 5 years) and spatial extent (range)over which recruitment is expected. Recentreviews (e.g. Lloyd et al. 2003) havehighlighted the non-linear nature of manyecological responses to changes in the flowregime; in some instances large ecologicalresponses have resulted from relativelysmall changes to the flow regime, while inother cases relatively large changes to theflow regime were required before anecological response was detected. Thepotential for such hysteresis effects shouldbe considered when evaluating a suitableeffect size.

The smaller the effect size to be detected,the greater the sampling intensity andtherefore resources required. As mostmonitoring and assessment programs arelikely to have limited resources, thechallenge will be to minimise the effect sizewith the given resources. Any trade-offbetween sampling intensity for a giveneffect size and budgets will be determined,at least in part, statistically (Downes et al.2002). Statistical advice should beconsidered for informing stakeholders aboutthe implications of trade-offs between thedesired effect size and study design.

A pilot study is very valuable as it helps todefine the spatial and temporal variation thatexists within the study system. Theinformation collected during a pilot studymay be used to refine the study design ifappropriate locations are not available, or ifit is not possible to measure the desiredvariables, or if variability means it isunlikely that the desired effect size can bedetected.

Ideally, optimisation of the study designrequires stakeholder input, potentially as a3-step process:

1. Get stakeholders to examine the effectsize required (evidence required fromthe monitoring program).


Box 3. Assessing predicted responses to environmental flows: the Queensland and NewSouth Wales approaches

Managers sometimes require a regional-scale (even statewide) assessment of ecologicalresponses of rivers to environmental flows. This might involve assessing a range of rivertypes with quite different environmental flow objectives, ‘control’ rivers will not alwaysbe available, and the environmental flow regime may be initiated gradually, precludingsimple before–after intervention comparisons. Even if ‘before’ data and/or ‘control’ riverswere available, it would be too expensive to implement full BACI designs at so manyrivers to provide a regional assessment. One option for monitoring in such circumstances isto focus on assessing outcomes predicted by hypothesised flow–ecology relationships onlyat intervention locations (i.e. rivers receiving environmental flows). The predictedoutcomes might be hydrological (with implied ecological consequences) or a mixture ofphysical and biological responses. This approach has been adopted for regional-scalemonitoring of environmental flows in Queensland and New South Wales.

In Queensland, each water resource plan (WRP) outlines a number of ‘ecologicaloutcomes’ relevant to that catchment. To assess the performance of a water resource planwith respect to meeting its environmental or ecological outcomes, the monitoring programwill aim to isolate the effects of flow from all other effects, in achieving these outcomes.River flow is only one of the many stressors that need to be managed effectively to ensurethat environmental or ecological outcomes are met, but the scope of a WRP is to manageonly flow. For this reason, assessment of a WRP is not based on information aboutecological condition, because it is recognised that condition cannot be directly andunequivocally attributed to management of water. Managing river flow alone cannotguarantee ecological outcomes, but provision of appropriate river flow is one importantmanagement action contributing towards achieving environmental and ecologicaloutcomes.

Monitoring to assess plan performance will be based on highly valued components of thenatural environment (‘ecological assets’) that reflect the ecological outcomes of that WRP.An ecological asset may be a species, group of species, biological function, particularecosystem or place of value for which the provision of water (flow) is directly critical. Theterm critical means that certain aspects of the way water is provided are necessary tomaintain the biological integrity of the asset. The intention is not to manage river flow tobenefit one asset in particular (i.e. not fish farming). Rather, the process involvesidentifying valued components of the ecosystem that have a critical link to differentattributes of the natural flow regime and then determining if flow management has thepotential to impact upon these attributes.

The scope of the ecological performance monitoring is therefore to measure whether watermanagement is providing flow related conditions (such as velocity, depth, connectivity torequired habitat, appropriate timing and duration etc.) that are critical for the identifiedecological assets. Critical ecological responses of assets to flow conditions may include:breeding/spawning of particular species of aquatic plants and animals, completion of lifestages/recruitment of particular species of plants and animals, or movement of particularspecies. By examining the direct link between aspects of ecology and their water-relatedcritical needs, it is expected that the effect of water management can be isolated andassessed.

(continued next page)


2. Undertake a pilot study to establish thefeasibility of establishing monitoringsites and evaluate the variability andsuitability of the variables to bemeasured.

3. Revisit the effect size with stakeholders,considering the variables to be includedand the benefit–cost tradeoffs ofsampling with spatial limits, temporallimits or limited frequency.

If there is no opportunity to undertake a pilotstudy due to time or resource constraints,

then the initial stages of the monitoringproject can be used as a ‘pilot’, recognisingthat the project will require review andpossibly further refinement after 1–2 years.

The Australian Guidelines for Water QualityMonitoring and Reporting (ANZECC &ARMCANZ 2000) provide a usefulchecklist from which to assess the finalmonitoring study design:

1. Has the study type been made explicitand agreed upon?

Box 3 (continued)

In taking this approach, monitoring responsibilities will be two-fold. First, it is necessaryto measure physical (hydraulic) variables such as water depth, water velocity, area ofinundation and timing of inundation, which are uniquely influenced by managing waterand are critically linked to the biological water requirements of identified assets. This willform the basis of the assessment criteria. Second, targeted research programs will bedesigned to improve our understanding and better quantify the critical water requirementsfor the asset of interest.

Ecological performance will be assessed by examining how the hydraulic variables wereprovided in space and time under current management arrangements, compared withassessment criteria for providing sustainable water requirements of variables of ecologicalassets. Assessment criteria will state the conditions for an acceptable risk to the sustainablefuture of an ecological asset. It is acknowledged that the ‘best available’ information aboutsome ecological assets and related variables is limited and this adaptive managementapproach will allow new information from the targeted programs to be incorporated intothe assessment process, as it becomes available.

New South Wales has also adopted a predictive approach to assessing responses toenvironmental flows in regulated rivers across the state. They had considered BACI-typedesigns for specific rivers but recognised that control and/or reference rivers were almostnever available. As the environmental flows were being implemented gradually, beforeversus after contrasts were also difficult. They developed the Integrated Monitoring ofEnvironmental Flows (IMEF) program for six rivers that are regulated by large storages.IMEF is based around 16 generic and valley-specific hypotheses of how these riversystems should respond to environmental flows. These hypotheses include aspects of waterquality (especially algal blooms), providing or improving habitat, maintaining orimproving the condition of estuaries and wetlands, temperature changes, wetting/dryingcycles, riparian vegetation and channel geomorphology. These hypotheses were prioritisedfor each river valley. The monitoring program was designed to evaluate these predictionsusing appropriate field sampling methods.

The IMEF program is important because the hypotheses being evaluated arose from NSWriver flow objectives and their subsequent environmental flow rules. The hypothesesrepresent predictive relationships between changes to river flows and ecological responsesthat can be tested not only by the type of sampling regime used with IMEF but also bycase-specific monitoring designs as described with this framework.


2. Have the spatial boundaries of the studybeen defined?

3. Has the scale of the study been agreedto?

4. Has the duration of the study beendefined?

5. Have the potential sources of variabilitybeen identified?

6. Are there sufficient sampling stations toaccommodate variability?

7. Are the sites accessible and safe?

8. Can sites be accurately identified?

9. Has spatial variation in sites beenconsidered, and have options tominimise this variation beenconsidered?

10. On what basis is the frequency ofsampling proposed?

11. Have decisions been made about thesmallest differences or changes thatneed to be detected?

12. Is replication adequate to obtain thedesired level of precision in the data?

13. Have the measurement parameters beenchosen?

(a) Are they relevant?

(b) Do they have explanatory power?

(c) Can they be used to detect changesand trends?

(d) Can they be measured in a reliable,reproducible and cost-effectiveway?

(e) Are the parameters appropriate forthe time and spatial scales of thestudy?

14. Has the cost-effectiveness of the studydesign been examined?

15. Have the data requirements beensummarised?

1.7. Implement the study design

Implementing the adopted study designshould be relatively straightforward,

particularly if a pilot study has beenundertaken to help avoid or resolve potentialproblems (e.g. location of suitablemonitoring sites; assessing the suitability ofpotential variables). Some additionalplanning will help ensure that high qualitydata and information are collected as part ofthe study, and help provide flexibility toadapt the program in the light of newinformation or changed circumstances.

A QA/QC program is a wise investment thatwill help to minimise the sampling errorsand detect and correct problems that mayarise in a sampling program. The AustralianGuidelines for Water Quality Monitoringand Reporting (ANZECC & ARMCANZ2000) outline quality assurance (QA/QC)considerations for field sampling, laboratorytesting and data handling that serve as auseful guide for those designingenvironmental flow monitoring programs.

The delivery of a package of environmentalflow recommendations can represent asignificant change to the managementregime of a river. Circumstances, such asprolonged drought or changed managementpriorities, can mean that intendedenvironmental flow releases are notdelivered. It is recommended that acontingency plan be prepared that outlinessteps that would be taken in response tochanged circumstances. Such a plan shouldconsider the implication of, and response tosuch issues as:

• the risk to the river system ifenvironmental flows are not delivered;

• the rationale of the study design andpotential statistical analyses, and ifthese are likely to be compromised;

• the risk of an unacceptable change dueto implementation of an environmentalflow regime (e.g. carp breeding andincreased dispersal of introducedspecies; blackwater events).


1.8. Have the environmental flows mettheir specific objectives?

River rehabilitation or protectionexperiments, such as the delivery ofenvironmental flows, should be reviewedwithin an adaptive management framework.It is important that the findings aredisseminated quickly and efficiently tostakeholders, so that managers can use thenew information in their decision-making.As the assessment of large-scalerehabilitation projects is a relatively newpursuit in Australia, it is recommended thatthe results of such experiments be externallyreviewed and made widely available. Thismay be facilitated in the future viarepositories such as State agency websitesand databases such as the Victorian DataWarehouse.

Once the environmental flows have beendelivered and relevant data have beencollected and analysed, it is time to revisitthe environmental flow objectives,conceptual models and hypotheses that formthe basis of the monitoring and assessmentprogram. This ‘learning’ step helps tostrengthen or modify hypotheses, and guide

the refinement of the monitoring program inthe light of an improved understanding offlow–ecology relationships. It is alsoessential information for managers, who willoften have to compare the potential benefitsfrom managing flow and from othermanagement actions (e.g. protection orreinstatement of physical habitat), and setpriorities accordingly.

Assessment of monitoring results can alsoinform or assist the development andapplication of models that help explainbroad-scale processes, such as theinteraction of flow and other factors drivingriver condition or ecological processes.Modelling approaches such as Bayesiannetworks and artificial neural networks arebecoming more widely used in naturalresource management (e.g. Lek and Guegan1999, Borsuk et al. 2001), particularly interms of scenario testing and developing apredictive capability that can help setpriorities for action. Having monitoring andassessment results readily available willmake it easier to develop and adopt suchtools in the future.


2. Other issues

2.1. Levels of evidence and causalinference

A Multiple Lines and Levels of Evidence(MLLE) approach (Figure 4) is described inAppendix 1. MLLE can contribute tomonitoring programs that are designed todetect ecological responses to managementinterventions (see also Beyers 1998, Downeset al. 2002). There is increasing recognitionthat strong ‘experimental’ designs (e.g.BACI) will often not be possible for manymonitoring programs and that othersupporting evidence will be needed tostrengthen the inference of causal linksbetween the intervention (e.g. environmentalflows) and the response. Downes et al.(2002) proposed that a ‘levels of evidence’approach could be used to provide furthersupport for conclusions that an observedecological response in a monitoring programwas due to the intervention being monitored.This general concept of using a range ofdifferent types of evidence when drawingconclusions has broad acceptance in thescientific community. However, a consistent

way of combining the different forms ofevidence in a formal, quantitative, way hasnot yet been devised.

Norris et al. (2004) proposed MLLE as away of formalising and refining theconceptual understanding of ecologicalresponses to interventions such as environ-mental flows. This application of MLLE ispartly based on a system for rankingdifferent forms of evidence and henceweighting their contributions to an overallconclusion about causal links. While only inthe early stages of development, the MLLEapproach should be considered as a potentialtool that can help us reach conclusions aboutcausal relationships and appropriateresponse variables in situations in which:

• additional information can add toimproved conceptual understanding atthe location of interest and help directthe collection of new data,

• natural variability makes it difficult toreach a conclusion about a causalrelationship,

• monitoring designs incorporating‘before’ data and/or ‘control’ locations,

Question(s) & conceptual model

Relevant lines of evidence

• Literature review• Assemble and analyse local data

Additional lines of evidence

Weight all literature and local data relative to qualityVerdict

Characteristics of human activityCharacteristics of impact location

Figure 4. Steps in the MLLE process (after Downes et al. 2002; restated in Norris et al. 2004)


which provide good ‘experimental’evidence for causality, are not possible.

MLLE is currently being trialled to examinerelationships between ecological attributesand flow regime in the Cotter River in theAustralian Capital Territory (Norris et al.2004). Whether or not the proposedweighting system is broadly applicable tocombining different forms of evidence inmonitoring programs will be evaluated asthe project unfolds.

2.2. Analysis of monitoring data

There are two broad types of statisticalanalysis that would be applicable to themonitoring data collected from the variousdesigns in Figure 2.

First, linear models relating the variable ofinterest to either spatial (intervention versus‘control’) or temporal (‘before’ versus‘after’, or trends through time) comparisonsare appropriate for single response variables(e.g. species richness, ecological health,abundance of key taxa). The linear modelsare sometimes known as regression orANOVA models, although more flexibleversions include generalised linear modelsand generalised additive models (Quinn andKeough 2002). A range of methods isavailable for assessing the fit of variousmodels to the monitoring data. Whiletraditional frequentist methods that produceconfidence intervals and P-values forrejecting null hypotheses can be useful,there is increasing application of Bayesianmethods that assess model parameters moredirectly.

Second, multivariate methods are valuableto find patterns when many variables areconsidered together (e.g. abundances ofmany taxa). These analyses are oftensummarised graphically (ordination plots orcluster diagrams), but complex hypothesesabout multivariate responses can also betested (Quinn and Keough 2002).

The critical issue is that the analysis must beformally linked to the monitoring design and

the specific hypotheses of interest. If themonitoring is well designed, the statisticalanalysis, whether traditional or Bayesian,will be robust and interpretable. Theinvolvement of advisors with statisticalexpertise is essential in the design andanalysis of the monitoring.

2.3. Priorities for monitoring

The range of designs in Figure 2 suggeststhat criteria for prioritising which to use inindividual situations is required. Clearly,applying full BACI (or BAC(Reference)I)designs for all environmental flowmonitoring will not be economically viable.These designs should only be used in thosesituations where ‘before’ data and/or valid‘control’ rivers are available, and theexpected outcomes from the monitoring willhave broad conceptual value (i.e. contributeto our understanding of flow–ecologyrelationships) and be applicable to otherriver systems. In other cases, especiallywhen an assessment of responses toenvironmental flows at a regional (evenstate-wide) scale is required, predictions canbe evaluated by monitoring only atintervention sites. The combination offocused BACI experiments with assessmentsof responses at other intervention sites,where ‘before’ data or spatial ‘controls’ areunavailable, should provide the best mix ofcausal understanding and spatial generalityto inform river managers.

There may be some situations where there islittle justification for investing in anymonitoring. In particular, if the plannedchange to flow regime is very small andbefore-intervention data and/or spatial‘control’ locations are not available, it willbe difficult to design a cost-effectivemonitoring program that has a reasonablechance of detecting responses to flowchange. Pilot data on spatial and temporalvariability of chosen variables will be veryvaluable for deciding whether or not tomonitor, as well as for designing the mosteffective monitoring program if monitoringgoes ahead.


3. Further reading

ANZECC and ARMCANZ (2000). AustralianGuidelines for Water Quality Monitoring andReporting. Australian & New ZealandEnvironment and Conservation Council and theAgriculture and Resource Management Councilof Australia & New Zealand.http://www.deh.gov.au/water/quality/nwqms/pubs/mg-contents.pdf

Baldwin D., Nielsen D., Bowen T. and Williams J.(2004). Recommended Methods for MonitoringFloodplains and Wetlands. Murray-DarlingFreshwater Research Centre report to theMurray-Darling Basin Commission.

Beyers D. (1998). Causal inference in environmentalimpact studies. Journal of the North AmericanBenthological Society 17(3), pp. 367–373.

Borsuk M., Higdon D., Stow C. and Reckhow K.(2001). A Bayesian hierarchical model to predictbenthic oxygen demand from organic matterloading in estuaries and coastal zones.Ecological Modelling, 143, pp. 165–181.http://www.iemss.org/iemss2002/proceedings/pdf/volume%20due/11_borsuk.pdf

Bosch O., Ross H., Witt B. and Smith C. (2004).Adaptive management: outcomes of the OzAM2003 workshop, Brisbane. University ofQueensland and the Murray-Darling BasinCommission.

Bradshaw A. (1996). Underlying principles ofrestoration. Canadian Journal of Aquatic Science51(1), pp. 3–9.

Cottingham P., Thoms M. and Quinn G. (2002).Scientific panels and their use in environmentalflow assessment in Australia. Australian Journalof Water Resources 5(1), pp. 103–112.

DNRE (2002). A method for determiningenvironmental water requirements in Victoria.Catchment and Water Division, Department ofNatural Resources and Environment, Melbourne,Victoria.

Downes B., Barmuta L., Fairweather P., Faith D.,Keough M., Lake P.S., Mapstone B. andQuinn G. (2002). Monitoring EcologicalImpacts: Concepts and Practice in FlowingWaters. Cambridge University Press, UK.

Heron S., Doeg T. and Sovitslis A. (2002).‘Maribyrnong River Flow Restoration Plan: Management Options for Ameliorating FlowStress.’ Report for the Port Phillip Catchmentand Land Protection Board and the Departmentof Natural Resources and Environment,Melbourne.

IEPEF (2002). ‘An Adaptive Management Approachto the Implementation of Environmental Flows inthe Hawkesbury-Nepean River.’ Report of theIndependent Expert Panel on EnvironmentalFlows for the Hawkesbury Nepean, Shoalhavenand Woronora catchments.http://www.dlwc.nsw.gov.au/care/water/sydneys_future/pdfs/adaptive_mngmnt_approach_enviroflows.pdf

Junk W., Bayley P. and Sparks R. (1989). The floodpulse concept in river-floodplain systems.Canadian Special Publication of Fisheries andAquatic Sciences 106, pp. 110–127.

King A., Brooks J., Quinn G., Sharpe A. andMcKay S. (2003). ‘Monitoring Programs forEnvironmental Flows in Australia — ALiterature Review.’ Department of Sustainability& Environment, Sinclair Knight Merz and theCRC Freshwater Ecology.http://www.dpi.vic.gov.au/dse/nrenari.nsf/

Lake P.S. (2001). On the maturing of restoration:linking ecological research and restoration.Ecological Management & Restoration 2(2),pp. 110–115.

Lek S. and Guegan J. (1999). Artificial neuralnetworks as a tool in ecological modelling, anintroduction. Ecological Modelling 120,pp. 65–73. http://www.clas.ufl.edu/users/mbinford/geo5159/GEO5159/Lek1999.pdf

Lloyd N., Quinn G., Thoms M., Arthington A.,Gawne B., Humphries P. and Walker K. (2003).Does Flow Modification Cause Geomorpho-logical and Ecological Response in Rivers?A Literature Review from an AustralianPerspective. Technical report 1/2004, CRC forFreshwater Ecology, Canberra.http://freshwater. canberra.edu.au> Publications>Technical reports > 2003.

Michener W. (1997). Quantitatively evaluatingrestoration experiments: research design,statistical analysis, and data managementconsiderations. Restoration Ecology 5(4),pp. 324–337.

Norris R., Liston P., Mugodo J., Nicols S., Quinn G.,Cottingham P., Metzeling L., Perriss S.,Robinson D., Tiller D. and Wilson G. (2004).Multiple lines and levels of evidence fordetecting ecological responses to managementinterventions. Proceedings of the FourthAustralian Stream Management Conference,Launceston (to appear).

Palmer M., Ambrose R. and Poff N.L. (1997).Ecological theory and community restorationecology. Restoration Ecology 5(4), pp. 291–300.


Puckridge J., Sheldon F., Walker K. and Boulton A.(1998). Flow variability and the ecology of largerivers. Marine and Freshwater Research 49,pp. 55–72.

Quinn G.P. & Keough M.J. (2002). ExperimentalDesign and Data Analysis for Biologists.Cambridge University Press.

Sharpe A.K. & Quinn G.P. (2004). MonitoringEnvironmental Flows in the Wimmera andGlenelg Rivers. Sinclair Knight Merz and CRCfor Freshwater Ecology, Melbourne.

SKM (2002). Stressed rivers project environmentalflow study: Wimmera River. Sinclair KnightMerz, Melbourne.

SKM (2003). Wimmera bulk entitlementenvironmental investigation. Sinclair KnightMerz, Melbourne.

Watts R., Ryder D., Chisholm L. and Lowe B. (2001).Assessment of environmental flows for theMurrumbidgee River: Developing biologicalindicators for assessing river flow management.Final report to the NSW Department of Land andWater Conservation. Johnson Centre, CharlesSturt University, Wagga Wagga, Australia.


Appendix 1. Overviewof the Multiple Lines andLevels of Evidence(MLLE) approach

MLLE is proposed as a logical way oforganising evidence to make a causalinference (e.g. Beyers 1998, Downes et al.2002). A MLLE framework can helpresearchers and managers reach conclusionsabout causal relationships in situationswhere:

• additional information can add toimproved conceptual understanding atthe location of interest and help directthe collection of new data,

• natural variability makes it difficult toreach a conclusion about a causalrelationship,

• monitoring designs incorporating‘before’ data and/or ‘control’ locations,which provide good ‘experimental’evidence for causality, are not possible.

A line of evidence is:

• a type of evidence; for example, anecosystem attribute that is investigatedin relation to a stressor or intervention(e.g. fish abundance, macroinvertebratespecies richness, macrophyte biomass).

A level of evidence is:

• the value of one of a number of criteriaused to determine the case for inferring(i.e. strength of evidence) that a given

human activity causes a givenecological change (Table 3).

Norris et al. (2004) adapted the stepsrecommended by Downes et al. (2002) whenconsidering ecological responses to changesin the flow regime of the Cotter River, ACT(Figure 4). The steps outlined (i.e.characterising the activity at the interventionlocation; exploring the conceptualunderstanding of the system in order topredict responses to the intervention (e.g.environmental flow); and confirming thelines of evidence (variables) to consider) areall consistent with aspects of this frameworkfor monitoring and assessing environmentalflows.

Downes et al. (2002) presented their ‘levelsof evidence’ approach as a method forascribing causal links when theirrecommended BACI designs could not beapplied and causal inference from amonitoring program was weak. Using otherlevels of evidence was proposed tostrengthen conclusions that a response wascaused by an intervention, such as anenvironmental flow regime. However,Downes et al. (2002) highlighted that we donot yet have a method for combining theselevels of evidence in a robust way to drawconclusions about strength of inference.

In their use of MLLE to examine flow–ecology relationships in the Cotter River,Norris et al. (2004) trialled a formalprocedure for weighting the quality ofscientific papers based on aspects such asthe type of study design, the number of

Table 3. Examples of causal criteria to be applied when evaluating levels of evidence (Norris et al. 2004)

Causal criterion Description

Biological plausibility Biological mechanism that could explain the relationship

Biological response Evidence of the biological response following the stressor

Dose–response relationship with the stressor

Evidence of a dose–response relationship between the stressor and the biological response

Consistency of association Expected biological response always occurs in the presence of the stressor


‘control’ or reference sites and the numberof impact (intervention) sites. The resultssupported the inference that changes tomacroinvertebrate community structure canbe causally linked to changes in the flowregime in rivers comparable to the Cotter.This would suggest that macroinvertebratecommunities should be included whenassessing predictions about ecologicalresponses to changes in the flow regime ofthe Cotter River. While Norris et al. (2004)found that causal inferences about theresponse of other river attributes (e.g. fishand vegetation communities) to changes inthe flow regime were weaker than formacroinvertebrates, this does not mean thatthese attributes were unimportant or couldbe ignored. If such results were used as thesole basis for selecting variables to monitor,then this would mean that we only acceptedevidence for well-studied attributesregardless of whether they were the mosteffective means for drawing causalinferences.

The trial of MLLE by Norris et al. (2004)focused on selection of variables to bemonitored, but this will be governed in largepart by the conceptual understanding of thesystem. So for variables where directevidence is not strong, a well-designed and

powerful scientific design will be requiredto establish causal links between theintervention and the environmental orecological response (in this case changes tothe flow regime and ecological responses).

MLLE can potentially be used in distinctways when developing a monitoring andassessment program:

• reviewing the existing literature forevidence of a general proposition (e.g.that change from natural flowregime reduces macroinvertebratespecies richness in upland streams);

• using evidence for such a generalproposition in design of a localmonitoring program to test a specificproposition (e.g. that change from thenatural flow regime has reducedmacroinvertebrate species richness);

• interpreting the data from a localmonitoring program to assess theevidence for the specific proposition.

The use of MLLE to support environmentalor ecological assessment is an area ofongoing research and has the potential to bea valuable tool in the future. New insightson its application will be considered infuture reviews of this framework.


Appendix 2. Wimmera-Glenelg environmentalflows monitoringprogramMore details on the Wimmera-Glenelgenvironmental flows monitoring programcan be obtained from Sharpe and Quinn(2004). The key steps involved aresummarised in the following sections. Notethat the monitoring and assessment programwas developed without specific informationabout the package of environmental flows tobe delivered. Thus the program wasdesigned to be flexible enough toaccommodate more specific objectives inthe future.

A2.1. Define the scope of the projectand its environmental objectives

Six broad environmental objectives hadpreviously been identified for both theWimmera (SKM 2002) and the Glenelgrivers and formed the basis forenvironmental flow recommendations. Theobjectives were based on maintaining orreinstating components of the flow regimethat: (i) contribute to channel-formingprocesses, (ii) maintain or improve habitatconditions for biota such as fish, and (iii)control nuisance growth of algae and aquaticplants. For example, the broad objectives setfor the Wimmera River catchment were to:

1. Provide an environmental flow regimethroughout the year that includes:

• periods of no flow comparable infrequency and duration to thosethat would have occurred duringpre-water resource developmentconditions;

• minimum environmental flowsduring low flow periods; and

• flows of a sufficient magnitude tomaintain water quality andfacilitate geomorphologicalprocesses.

2. Maintain, and where possible restore,longitudinal connectivity by:

• providing minimum environmentalflows during low flow periods;

• ensuring farm dam development inthe upper catchment does notimpact upon flow magnitude andvariability in downstream reaches;and

• improving the frequency, durationand magnitude of floods in theterminal lakes.

3. Maintain, and where possible improve,stream habitat condition by providingenvironmental flows that can facilitatechannel-forming processes.

4. Manage flows for 24 threatened, flowdependent, flora species.

5. Maintain self-sustaining populations ofendemic native fish including riverblackfish, southern pygmy perch andmountain galaxias.

6. Manage flows to minimise algal bloomsand the development of Azolla mats.

Management objectives set for various flow-dependent assets (e.g. threatened fishspecies, riparian vegetation communities)and links with components of the flowregime were reviewed (e.g. Table 4).

A2.2. Define the conceptualunderstanding of flow–ecologyrelationships and the questions(hypotheses) to be tested

Flow recommendations for specific reachesof the Wimmera (SMK 2003) and Glenelgrivers were reviewed (e.g. Table 5).Information on the timing, duration andmagnitude of flow events and the biologicalor geomorphic outcomes expected are basedon the conceptual understanding (model) ofthe river system. The conceptualunderstanding of the flow-dependency ofecological assets and their response tochanges in the flow regime were describedon a reach-by-reach basis. This conceptual


understanding also underpins the monitoringand assessment program.

A2.3. Select the variables to bemonitored

The following criteria were used to selectvariables to be measured as part of themonitoring program:

• Links to the environmental flowobjectives.

• There is an established causal linkbetween the variable and the stressor orrehabilitation activity.

• The variables include those of highsocio-economic or ecologicalimportance.

• The variables are efficient (i.e. cost-effective) to sample.

• The availability of baseline data tocomplement ‘before–after’comparisons.

Information from previous studies andhistorical data can provide valuableinformation on the condition of flow-dependent assets under different flowregimes. Information on water quality,hydrology, biological and physicalconditions was reviewed, and knowledgegaps were identified. Because much of theavailable data and information had beencollected for purposes other than detectingecological responses to environmental

Table 4. Example environmental management objectives for the Wimmera River catchment (from SKM 2002)

Environmental objective Target feature Relevant flow component

Maintain self-sustaining populations of river blackfish and short-finned eel

• Habitat for subsistence • Recruitment/breeding

• Seasonal low flows throughout the year • Spring/summer freshes

Restore self-sustaining populations of Murray cod, golden perch and Macquarie perch

• Habitat for subsistence • Recruitment/breeding • Movement

• Seasonal low flows throughout the year • Winter/Spring freshes • Winter/Spring high flows

Table 5. Example objectives and environmental flow recommendations for the Wimmera River, Reach 2 (Huddleston-McKenzie River); Compliance point = Faux Bridge, Gauge no. 415240 (SKM 2003)

Flow

Season Magnitude Frequency Duration Objective/Rationale

Summer 0 ML/d Annually 17–30 days Natural stress to promote macroinvertebrate biodiversity

Minimum flow 6 ML/d

Annually Dec–May

Maintain quality and quantity of habitat for native fish, macroinvertebrates and aquatic vegetation

>16 ML/d 3 annually 7–15 days Enhance recruitment of short-finned eels and river blackfish

Spring (Jul–Nov)

>164 ML/d 2–3 annually Minimum 14 days

Maintain riparian vegetation and habitat for native bird species

Minimum flow 60 ML/d

Annually July–Nov Inundate snags and other elements that provide habitat for native fish, macroinvertebrates and aquatic vegetation, and maintain longitudinal connectivity

Annual 6,000 ML/d Annual Minimum 2 days

Provide high flows to cue and enhance recruitment of golden perch, Murray cod and Macquarie perch, maintain riparian vegetation and habitat for native bird species and facilitatechannel-forming processes.


flows, it was not surprising that data suitablefor assessing baseline conditions or topredict ecological responses to changes inthe flow regime were limited.

Tables that summarised the variables,appropriate methods and spatial andtemporal attributes of the monitoringprogram were provided (e.g. Table 6), aswere location-based monitoring schedules(e.g. Table 7).

A2.4. Determine the study design

Two broad strategies for monitoring theeffects of a changed flow regime wereinitially considered. The first was acomparison between conditions at theintervention locations and ‘reference’condition (e.g. least disturbed) that is usedfor river health assessment. The secondstrategy was to use the traditional BACIdesigns often adopted for detectingecosystem response to human disturbance(impact assessment). There were difficultiesin applying both strategies in the Wimmeraand Glenelg catchments. Referencecondition monitoring does not easilyidentify causal links between changes in

river health and the flow regime. Thedifficulty in assigning a starting point for theenvironmental flow regime and how waterwould be allocated spatially made it difficultto define the ‘before’ and ‘control’ elementsof a BACI design. The monitoring approacheventually recommended was based ondetecting trends at key locations over timeand comparing the direction and magnitudeof these changes with the environmentalobjectives set for each system. Specificcontrasts between reaches (e.g. upstreamversus downstream comparisons) could beused to infer causal links between flowchanges and observed ecological responses.While not true ‘control’ versus impactcomparisons, they can contribute to a levels-of-evidence approach to linking changes tothe flow regime and ecological response.

Potential study locations were identified onthe basis of the sites used to develop reach-specific environmental flow recommend-ations, representativeness of the proposedlocation with respect to the reach, and sitesestablished as part of pre-existing programsthat complement the data to be collected bythe environmental flows monitoringprogram.

Table 6. Example of recommended response variables to be recorded at each location in the Wimmera and Glenelg rivers (from Sharpe and Quinn 2004)

Variable Methods Spatial design Temporal design

Water quality

Key: pH, DO, EC, temp

Second tier: TN and TP

Appropriate portable meter, collect water samples for laboratory analysis for any nutrient analyses. Nutrients probably not required for pools and VWQMN data will probably suffice for this.

Water quality will need to be collected in pools at each site. All Wimmera reaches except Burnt Creek and all Glenelg reaches except Chetwynd-Wannon have active VWQMN stations. These stations should be used and additional in-situ measures should be taken at other sites.

Water is sampled monthly at VWQMN sites. These data should be used. Key parameters should also be measured monthly at additional sites during summer when water quality is likely to be a problem.

Fish Backpack electrofishing by a qualified person. Fyke nets set out overnight, ensuring end out of water so don’t drown mammals or diving birds. Bait traps set overnight near snags or emergent vegetation.

Three replicate pools at each site.

Sampling in summer only, focusing on pools at each site.


A2.5. Identify potential statisticalanalyses

As this monitoring and assessment programwas developed without specific details onthe package of environmental flows to bedelivered, four broad analytical approacheswere outlined:

1. Detection of temporal trends in keyresponse variables at selected locations.Analyses include time-series and linearmodel methods, where the responsevariable is modelled against time.

2. Specific temporal contrasts betweensets of years or between before and aftera particular flow event. Such temporalcontrasts can be analysed usingANOVA designs.

3. Comparison of reaches (spatialcontrasts), which if incorporating‘before’ and ‘after’ comparisons mayalso be analysed using ANOVA designs.

4. Multivariate comparison of assemblagesof organisms such as macro-

invertebrates or fish. Ordinationmethods (e.g. multidimensional scalingusing dissimilarity indices such asBray–Curtis) with specific spatial(between reach) and temporal (betweenyears or before and after events)contrasts using ANOSIM (analysis ofsimilarity) or NPMANOVA (non-parametric multivariate analysis ofvariance).

Statistical advice should be sought on theassumptions and applicability of the aboveapproaches so the most appropriate methodsare selected.

A2.6. Implementation and assessmentof objectives

The Wimmera-Glenelg environmental flowmonitoring and assessment program iscurrently being implemented and results arenot yet to hand.

Table 7. Example of recommended monitoring sites for reaches in the Upper Wimmera catchment

Recommended monitoring sites

Variable Monitoring frequency

Key site: Glynwylln VWQMN site 415206 Will need to establish cross-sections at this site

Water Quality: Measure DO, EC, pH, Temp at the surface and depth in pools

Monthly Additional event monitoring to assess changes after freshes

Hydrology: Measure discharge and water levels Visually assess flow and habitat inundation

During flow events Only needs to be done once for each flow type, not repeated each year

Geomorphology: Measure pool dimensions, sediment deposition, distribution of debris (photopoints and/or direct measurement)

Short-term responses measured before and after specific flow events. Only needs to be done once for each flow event.

Geomorphology: Measure channel cross-sections and longitudinal sections, vegetation extent and vegetation composition

Every 3–5 years but should always be done in summer to accurately measure vegetation.

Macroinvertebrates: Standard EPA rapid bioassessment techniques

Autumn and Spring every 3–5 years.

Fish: Various sampling techniques

Early summer every 3–5 years, but will need to be done more frequently if trying to detect responses to specific flow releases such as spring freshes.


Appendix 3. Potentialstudy designsThe study designs identified in Figure 2have the following characteristics:

(1) Intervention-only design. Incircumstances where an environmentalflow regime has already beenimplemented (no before-interventiondata are possible) and there are nospatial ‘controls’ or reference systemsfor comparison, monitoring isconstrained to measuring changes inchosen variables in the interventionriver. These responses can be evaluatedagainst specific predictions based on theconceptual model. Causal links betweentemporal change in ecological responseand flow are difficult to determinebecause the change might have occurredwithout the environmental flow. Thisdesign is very common, especially forlarger rivers (no ‘controls’) and whenregional-scale (state-wide) assessmentis required (see Box 3).

(2) Reference–Intervention design. Amodification of (1) above, where thereare no before-intervention data but thesame variable(s) are measured throughtime in a reference system, i.e. one thatis much less flow-modified andrepresents the desired direction ofchange for the intervention system. Thisdesign provides slightly better evidencefor causal link between temporal changein response and flow, because naturalchanges through time can be measuredat reference sites. It is also possible toassess whether the trend of change atthe intervention location is towards thereference condition.

(3) Control–Intervention design. Like (2)above except that comparison is with a‘control’ system, i.e. a river systemsimilarly flow-modified to theintervention system but withoutenvironmental flows. This designprovides stronger inference about

causality because comparison with thespatial ‘control’ reduces the likelihoodof flow effects being statisticallyconfounded with natural change.

(4) Control–Reference–Intervention design.Combination of (2) and (3) above.Statistical analyses test for divergencein temporal trends between theintervention and the ‘control’, and forconvergence in temporal trends betweenthe intervention and the referencelocation. This design provides causalstrength similar to (3), with the addedadvantage of assessing whether thetrends are in the desired direction —towards reference condition.

(5) Before–After–Intervention design.Standard ‘intervention analysis’ designcomparing before versus afterintervention. ‘Before’ data act as abaseline or temporal ‘control’, ameasure of whether temporal trendsoccur naturally (although obviously at adifferent time to ‘after’ interventiondata). Evidence for causal links islimited by lack of spatial ‘controls’, soit is unclear whether or not the changeafter intervention would have occurredindependently of environmental flows.This design is also difficult to use if anenvironmental flow regime isimplemented gradually, because thenbefore–after comparisons are hard todefine.

(6) Before–After Reference–Intervention(BARI) design. As for (5) but with aspatial component; namely, a referencesystem that provides some measure ofwhether natural change coincides withintervention. This design allowsassessment of whether the trend of aresponse is towards the referencecondition. The test of interest is whetherany before–after difference at theintervention location is the same as atthe reference location. The causalinference associated with this design islimited because the reference systemand the intervention system are in


different conditions prior to theintervention. This makes it difficult torule out a response to some other factorat the intervention location coincidingwith the start of the environmental flow.

(7) Before–After Control–Intervention(BACI) design. As for (6), but using aspatial ‘control’ system instead of areference system. This design providesa strong inference about causalitybecause comparison with spatial andtemporal ‘controls’ reduces thelikelihood of confounding flow effectswith natural spatial and temporalchange, i.e. any change in the river afterintervention is more likely to be due toenvironmental flows.

(8) Before–After Control–Reference–Intervention (BACRI) design. Acombination of (6) and (7) that providesstrong evidence for causal linksbetween flow change and response andalso measures whether the change is inthe desired direction — towardsreference condition.

Note: Designs involving control–intervention contrasts are improved byhaving multiple ‘control’ streams (e.g.MBACI designs; see Downes et al. 2002) toreduce the likelihood that the changeobserved in the intervention stream mighthave happened anyway.

Date post:	21-Sep-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

Environmental Flows Monitoring and Assessment Framework ... · Cover photographs courtesy of Robert...

Documents