Modelling water-related ecological responses to coal seam gas ...

54 downloads 0 Views 3MB Size Report
the Commonwealth Scientific and Industrial Research Organisation (CSIRO) for the .... 2.3 Addressing issues of scale and uncertainty in ecological conceptual ..... ecological responses to coal seam gas extraction and coal mining. Glossary ..... Another advantage of this approach arises when trying to weigh the benefits and.
Modelling water-related ecological responses to coal seam gas extraction and coal mining This report was commissioned by the Department of the Environment on the advice of the Independent Expert Scientific Committee on Coal Seam Gas and Large Coal Mining Development (IESC).

January 2015

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Copyright © Copyright Commonwealth of Australia, 2015.

Modelling water-related ecological responses to coal seam gas extraction and coal mining is licensed by the Commonwealth of Australia for use under a Creative Commons By Attribution 3.0 Australia licence with the exception of the Coat of Arms of the Commonwealth of Australia, the logo of the agency responsible for publishing the report, content supplied by third parties, and any images depicting people. For licence conditions see: http://creativecommons.org/licenses/by/3.0/au/ This report should be attributed as ‘Commonwealth of Australia 2015, Modelling water-related ecological responses to coal seam gas extraction and coal mining, prepared by Auricht Projects and the Commonwealth Scientific and Industrial Research Organisation (CSIRO) for the Department of the Environment, Commonwealth of Australia’. The Commonwealth of Australia has made all reasonable efforts to identify content supplied by third parties using the following format ‘© Copyright, [name of third party] ’. Enquiries concerning reproduction and rights should be addressed to: Department of the Environment, Public Affairs GPO Box 787 Canberra ACT 2601 Or by email to: [email protected] This publication can be accessed at: www.iesc.environment.gov.au

Acknowledgements This report was commissioned by the Department of the Environment on the advice of the Independent Expert Scientific Committee on Coal Seam Gas and Large Coal Mining Development (IESC). The report was prepared by Auricht Projects (Christopher Auricht and Sarah Imgraben) with input from Adjunct Professor Andrew Boulton (University of New England), Dr Justine Murray (CSIRO), Dr Carmel Pollino (CSIRO) and Dr Moya Tomlinson (Office of Water Science, Department of the Environment). The report was peer reviewed by Dr Martin Andersen (University of New South Wales), Professor Angela Arthington (Griffith University), Dr Bruce Chessman (ecological consultant), Dr Alexander Herr (CSIRO), Professor Ray Froend (Edith Cowan University) and Dr Anthony O’Grady (Ecology Lead, Bioregional Assessments Programme). Dr Jennifer Firn (Queensland University of Technology) reviewed Table 4.1 and Table 4.2 on Melaleuca irbyana and Dr Keith Walker reviewed the silver perch case study.

Disclaimer The views and opinions expressed in this publication are those of the authors and do not necessarily reflect those of the Australian Government or the Minister for the Environment or the IESC. While reasonable efforts have been made to ensure that the contents of this publication are factually correct, the Commonwealth and IESC do not accept responsibility for the accuracy or completeness of the contents, and shall not be liable for any loss or damage that may be occasioned directly or indirectly through the use of, or reliance on, the contents of this publication.

page ii

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Contents Summary ................................................................................................................................................. v Abbreviations .......................................................................................................................................... vii Glossary................................................................................................................................................... ix 1 Introduction ......................................................................................................................................... 1 1.1

Project context............................................................................................................................ 1

1.2

Purpose and outline of this report .............................................................................................. 1

1.3 Limitations in current assessment of water-related ecological responses to coal seam gas extraction and coal mining .................................................................................................................. 3 1.4

Potential water-related stressors associated with coal seam gas extraction and coal mining .. 4

1.5

Expected project outcomes ........................................................................................................ 6

2 Using models to predict water-related ecological responses to coal seam gas extraction and coal mining ................................................................................................................................................. 7 2.1

Using ecological conceptual models to represent complex ecosystems ................................... 7

2.2

Two broad types of conceptual models ..................................................................................... 8

2.3

Addressing issues of scale and uncertainty in ecological conceptual models ......................... 10

2.4

A framework for assessing vulnerability coal seam gas extraction and coal mining activities 12

3 Project methodology ......................................................................................................................... 17 3.1

Overview .................................................................................................................................. 17

3.2

Control and stressor models .................................................................................................... 19

3.3

Expert workshop assessment of some worked examples of ecological conceptual models ... 21

4 Results: case study and worked examples ...................................................................................... 23 4.1

Ecological conceptual models for Purga Nature Reserve ....................................................... 23

4.2

Bayesian network session........................................................................................................ 37

4.3

Gunnedah Basin case study: conceptual model for silver perch ............................................. 41

5 Discussion ........................................................................................................................................ 50 5.1 The role of ecological modelling in assessment of proposals for coal seam gas extraction and coal mining ........................................................................................................................................ 50 5.2

Ecological conceptual models in coal seam gas extraction and coal mining proposals .......... 51

5.3 Challenges in generating ecological conceptual models for proposals for coal seam gas extraction and coal mining ................................................................................................................ 52 5.4

Feasibility of the proposed approach as a desktop exercise ................................................... 54

5.5

Bayesian networks within an EIS application ........................................................................... 55

5.6

Conclusion................................................................................................................................ 56

6 References ....................................................................................................................................... 57 Appendix A - Case study: conceptual model for Silver Perch ............................................................... 62 Appendix B - Bayesian network models ................................................................................................ 66 Appendix C - Workshop agenda ........................................................................................................... 74 Appendix D - Workshop participants ..................................................................................................... 78

page iii

Modelling water-related ecological responses to coal seam gas extraction and coal mining Appendix E - Abstracts of presentations ............................................................................................... 80 Appendix F - Case study: Purga Nature Reserve ................................................................................. 88

Tables Table 4.1 Narrative table to accompany the control model ................................................................... 25 Table 4.2 Narrative table to accompany the stressor model ................................................................. 30 Table 4.3 Scenario construction for the Purga Nature Reserve, with type of stressor and frequency of occurrence ................................................................................................... 38 Table 4.4 Narrative table listing drivers, stressors, water-related ecological effects and hypothesised ecological effects on silver perch (SP) ....................................................... 43

Figures Figure 1.1 Hydrological stressors from coal seam gas extraction and coal mining ................................ 5 Figure 2.1 An integrated framework to assess the vulnerability of species to climate change ............. 13 Figure 2.2 Sensitivity assessment ......................................................................................................... 14 Figure 2.3 Conceptual model for brook trout ......................................................................................... 16 Figure 3.1 Flow-chart of ecological conceptual model development .................................................... 17 Figure 4.1 Location of Purga Nature Reserve ....................................................................................... 23 Figure 4.2 Box-and-arrow diagram of the control model for Melaleuca irbyana ................................... 29 Figure 4.3 Purga Nature Reserve (wet phase)...................................................................................... 32 Figure 4.4 Purga Nature Reserve (dry phase) ...................................................................................... 33 Figure 4.5 Conceptual model of a coastal and subcoastal floodplain tree swamp (Melaleuca and Eucalyptus spp.) ........................................................................................................ 34 Figure 4.6 Box-and-arrow diagram of the stressor model for Melaleuca irbyana ................................. 35 Figure 4.7 Landscape setting of Purga Nature Reserve ....................................................................... 36 Figure 4.8 Influence diagram developed in the workshop showing interactions between hydrological stressors and endpoints................................................................................ 39 Figure 4.9 Example of a small Bayesian network ................................................................................. 40 Figure 4.10 Conceptual model for silver perch...................................................................................... 42 Figure 4.11 Conceptual model of how fish are influenced by aspects of the riparian zone .................. 49

page iv

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Summary Ecological conceptual models are rarely used in Environmental Impact Statements (EISs) for coal seam gas extraction and coal mining proposals in Australia. In contrast, hydrological and hydrogeological conceptual models are well-established tools for identifying and assessing potential impacts of development projects. There is a need to integrate current hydrological and hydrogeological conceptual models with ecological ones to provide a complete picture of the likely water-related ecological impacts of coal seam gas extraction and coal mining. These combined models should then be used in EISs to support statements of likely ecological responses to coal seam gas extraction and coal mining, and to illustrate mechanisms by which proposed mitigation strategies would operate to reduce potential impacts. This report presents the findings of a project exploring an approach to ecological conceptual modelling aimed at improving the assessment of water-related ecological impacts of coal seam gas extraction and coal mining. The approach, presented as a series of consecutive steps and illustrated with worked examples, could assist those preparing and reviewing EISs to construct ecological conceptual models and associated narrative tables that specify hypothesised responses, and document supporting evidence. By using this approach, assumptions about ecological impacts in assessment of development proposals are made explicit, response pathways are identified and illustrate interactive and cumulative effects, and there is a transparent and consistent framework for design of monitoring programmes to test the implicit hypotheses. The approach to ecological conceptual modelling in this report follows that described by Gross (2003) for constructing ‘control’ and ‘stressor’ models, except for the modification that the ‘control’ model includes not only natural drivers and stressors but also anthropogenic ones not related to coal seam gas extraction and coal mining. Thus, the ‘control’ model conceptualises ecosystem components and interactions within the area of project impact before coal seam gas extraction and coal mining, whereas the ‘stressor’ model includes the hypothesised ecological responses to drivers and stressors associated only with such activities. Comparing the ecological conceptual models of the ‘before’ and ‘after’ states illustrates hypothesised ecological responses to coal seam gas extraction and coal mining. Pictorial conceptual models, influence diagrams and a Bayesian network were developed as a ‘proof-of-concept’ trial, and refined during an expert workshop that was informed by a field visit to a case study area. Pictorial conceptual models showing the components and processes in an area of interest help to make response pathways explicit. Models illustrating components, processes and responses developed at a hierarchy of spatial scales (e.g. groundwater-fed pools in river reaches nested in catchments) aim to portray spatial and temporal variability in ecological responses in an EIS. The temporal scale should take into account the time lags in hydrological and ecological responses to stressors such as groundwater extraction, which may extend for decades. Careful consideration of spatial and temporal scales is only one of the challenges in the assessment of ecological responses in EISs. Other challenges include gaps in data and site-specific knowledge, constraints in extrapolating short-term measurements to predict long-term responses, difficulty in demonstrating or quantifying causality, and the need to consider likely effects of stressors on various life-history stages as vulnerabilities may differ between recruitment/seedling establishment and adult stages.

page v

Modelling water-related ecological responses to coal seam gas extraction and coal mining

A key conclusion of the report is that the approaches to modelling and conceptualisation of hydrology and hydrogeology currently used in EISs should be extended to incorporate ecological components to produce ecohydrological models capable of illustrating likely water-related ecological responses to coal seam gas extraction and coal mining. Given that nearly all stressors interact, they should not be treated independently when assessing likely responses. Despite the challenges, the approach outlined in this report seeks to provide proponents of development proposals with the tools to better portray and understand the hydrology-ecology relationships in areas of planned coal seam gas extraction and coal mining, and to clearly articulate hypothesised stressor and response pathways, supported by reference to scientific and other credible literature. However, it is important to note that compiling conceptual models and the supporting narrative tables is not the final step. The purpose is to provide a transparent rationale, referenced to the scientific literature, for the ecological responses and proposed mitigation actions and monitoring strategies identified in an EIS. Application of the proposed approach is expected to: •

enhance capability in the resources industries to identify and predict the water-related impacts of coal seam gas extraction and coal mining, through uptake of the approach to ecological conceptual modelling and integration of the ecological modelling approach with hydrological and hydrogeological modelling and conceptualisation



improve identification and understanding of the potential water-related ecological responses to coal seam gas extraction and coal mining in Australia, achieved through assisting the Independent Expert Scientific Committee on Coal Seam Gas and Large Coal Mining Development (IESC) in its evaluation of EIS documentation for coal seam gas and coal mining proposals and provision of advice to regulators



provide a framework for ecological conceptual modelling that could be drawn upon in the bioregional assessments.

page vi

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Abbreviations General abbreviations

Description

ANAE

Australian National Aquatic Ecosystem

As

Symbol for arsenic

BA

Bioregional Assessment

BN

Bayesian network

BOD

Biological oxygen demand

CPT

Conditional probability table

CSG

Coal seam gas

CSGCM

Coal seam gas and coal mining

CSIRO

Commonwealth Scientific and Industrial Research Organisation

DDT

Dichlorodiphenyltrichloroethane

DO

Dissolved oxygen

ECD

Ecological Character Descriptions

EHNV

Epizootic Haematopoietic Necrosis Virus

EIS

Environmental Impact Statement

EM

Expectation Maximisation

EPBC Act

Environment Protection and Biodiversity Conservation Act 1999

Fe

Symbol for iron

GDE

Groundwater dependent ecosystem

Govt

Government

IESC

Independent Expert Scientific Committee on Coal Seam Gas and Large Coal Mining Development

IPCC

Intergovernmental Panel on Climate Change

MI

Melaleuca irbyana

Mn

Symbol for manganese

NSW

New South Wales

OWS

Office of Water Science

PVA

Population viability analysis

Qld

Queensland

page vii

Modelling water-related ecological responses to coal seam gas extraction and coal mining

General abbreviations

Description

RCI

River Condition Index

RO

Reverse osmosis

SP

Silver perch

TDS

Total dissolved solids

US

United States

WAIT

Water Asset Information Tool

y

Year

page viii

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Glossary Term

Description

Bioregion

As defined in the bioregional assessment methodology (Barrett et al. 2013): ‘…the land area that constitutes a geographic location within which are collected and analysed data and information relating to potential impacts of coal seam gas or coal mining developments on receptors identified for key water-dependent assets’

Bioregional assessments

A bioregional assessment (BA) is a scientific analysis of the ecology, hydrology, geology and hydrogeology of a bioregion, with explicit assessment of the potential direct, indirect and cumulative impacts of coal seam gas and coal mining development on water resources. The central purpose of BAs is to analyse the impacts and risks associated with changes to water-dependent assets that arise in response to current and future pathways of coal seam gas and coal mining development (Barrett et al. 2013)

Baseflow

The groundwater contribution to stream flow (Fetter 2001)

Coal seam gas development

Any activity involving coal seam gas extraction that has, or is likely to have, a significant impact on water resources either in its own right or when considered with other developments, whether past, present or reasonably foreseeable (IESC 2014)

Conceptual model

A conceptual model is ‘…a descriptive and/or schematic hydrological, hydrogeological and ecological representation of the site showing the stores, flows and uses of water, which illustrates the geological formations, water resources and water-related assets, and provides the basis for developing water and salt balances’ (IESC 2014). Ecological conceptual models show linkages among drivers, stressors, processes and components to represent known and hypothesised ecological responses to one or more stressors; a powerful way to communicate complex interactions among processes and components deemed important in an ecosystem with defined bounds and scope (after Gross 2003)

Confidence

A qualitative estimate of the quality of evidence and agreement among sources about a given situation, statement or hypothesis. This approach, used by the IPCC (2013) in efforts to predict the effects of future climate change, is used in this report as a surrogate partial measure of the uncertainty associated with support for hypothesised ecological responses to coal seam gas and coal mining development. However, ‘confidence’ is not the same as ‘uncertainty’, and these two terms should not be used interchangeably

Control conceptual model

A model that represents key processes, interactions and feedbacks (Gross 2003). In the context of this project, we define the control conceptual model as representing key processes, interactions and feedbacks in response to natural and anthropogenic activities not related to coal seam gas and coal extraction. This definition differs from the one by Gross (2003) that explicitly excludes stressors

page ix

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Term

Description

Drivers

The major external driving forces that have large-scale influences on natural systems. Drivers can be natural or anthropogenic forces (Jean et al. 2005)

Ecological endpoints

Ecological endpoints are a selected subset of the physical, chemical and biological elements and processes of natural systems that are selected to represent the overall health or condition of the system, known or hypothesised effects of stressors, or elements that have important human values (adapted from Gross 2003)

Ecosystem

Organisms and the non-living environment, all interacting as a unit

Groundwater

Water occurring in the saturated zone and the capillary fringe

Groundwater dependent ecosystem (GDE)

Natural ecosystems which require access to groundwater on a permanent or intermittent basis to meet all or some of their water requirements so as to maintain their communities of plants and animals, ecological processes and ecosystem services (Richardson et al. 2011). The broad types of GDE are (Eamus et al. 2006): • • •

ecosystems dependent on surface expression of groundwater ecosystems dependent on subsurface presence of groundwater subterranean ecosystems

Hyporheic

Associated with the saturated sediments below and alongside rivers and streams where surface water and groundwater exchange

Spring

A natural discharge of water from the ground (modified from Barrett et al. 2013)

Stressors

Physical, chemical, or biological perturbations to a system that are either foreign to that system or natural to the system but applied at an excessive or deficient level. Stressors cause significant changes in the ecological components, patterns and processes in natural systems (Gross 2003)

Stressor conceptual model

A model that represents relationships among stressors (or drivers), ecosystem components and effects (Jean et al. 2005). In the context of this project, we define the stressor conceptual model as representing the relationships between coal seam gas and coal mining-related stressors and their ecological effects. The control and stressor models are combined to conceptualise water-related ecological responses to natural and anthropogenic (including coal seam gas extraction and coal mining) drivers and stressors

Stygofauna

Aquatic fauna living in groundwater

Uncertainty

A partial or total lack of understanding or knowledge of an event, its consequence, or its likelihood (modified from Barrett et al. 2013). This definition is derived from the Standards Australia and New Zealand Risk Management Guidelines (AS/NZS ISO 31000:2009)

Water-related asset

‘A defined value or public benefit with a dependence on surface or groundwater, including water dependent ecosystems (as defined by the Water Act 2007 (Cwth)), drinking water, public health, recreation and amenity, Indigenous and cultural values, fisheries, tourism, navigation, agriculture and industry values’ (IESC 2014)

page x

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Term

Description

Water resources

Defined by the Water Act 2007 (Cwth) as: ‘…surface water or groundwater; or a watercourse, lake, wetland or aquifer (whether or not it currently has water in it); and includes all aspects of the water resource, including water, organisms, and other components and ecosystems that contribute to the physical state and environmental value of the resource’ (IESC 2014)

page xi

Modelling water-related ecological responses to coal seam gas extraction and coal mining

1 Introduction 1.1 Project context In 2012, the Australian Government established an Independent Expert Scientific Committee on Coal Seam Gas and Large Coal Mining Development (IESC) to provide scientific advice to government regulators on the impacts that coal seam gas (CSG) extraction and large coal mining development may have on Australia’s water resources. The IESC is supported by the Office of Water Science (OWS) within the Australian Government Department of the Environment. The OWS conducts research areas under three priority themes: •

hydrology: changes in dynamics and aquifer interconnectivity



ecosystems and water: environmental tolerances, responses and mitigation



chemicals: water-related risks to environmental health.

Monitoring, assessment and evaluation of cumulative impacts is a cross-cutting theme across the three priority themes. This project, modelling water-related ecological responses to coal seam gas extraction and coal mining, provides the theoretical basis for subsequent projects within the second theme above, Ecosystems and water (hereafter referred to as the ‘Ecology theme’). The aim of this project was to explore the development of tools specifically to assess the water-related ecological impacts of coal seam gas extraction and coal mining projects in Australia. The IESC also provides advice to the Australian Government on bioregional assessments (BA). In this context, a bioregional assessment is a collation of baseline information on the ecology, hydrology, geology and hydrogeology of a designated region, termed a bioregion, with explicit assessment of the potential direct, indirect and cumulative impacts of coal seam gas extraction and coal mining on water resources. Bioregional assessments and other research aim to improve the knowledge base regarding the potential water-related impacts of coal seam gas extraction and coal mining. The Bioregional Assessment Programme targets regions with significant coal deposits. Assessments are currently being undertaken in 13 subregions within six bioregions across central and eastern Australia, including the Clarence-Moreton Basin. As part of the bioregional assessments, the direct, indirect and cumulative impacts on receptors representing ecological, economic and socio-cultural water-dependent assets will be reported. This Ecology theme project and an expert-panel workshop (section 3.3) explored approaches for developing ecological conceptual models that portray likely ecological water-related responses to coal seam gas extraction and coal mining, providing a framework that could be drawn upon by related work, such as the bioregional assessments.

1.2 Purpose and outline of this report The purpose of this project was to examine how ecological conceptual models could be used to improve current methods of assessment of the water-related ecological impacts of coal seam gas extraction and coal mining. Specifically, the project aimed to find the most feasible

page 1

Modelling water-related ecological responses to coal seam gas extraction and coal mining

approach for developing ecological conceptual models to support this assessment process, trial the approach as a ‘proof-of-concept’ using a case study in the Clarence-Moreton Basin, and discuss the models and results with scientific experts at a facilitated workshop that included a field visit to the case study area. This report begins (Chapter 1) with a brief description of the project’s context with the current BAs, followed by a review of the limitations of the present approach to assessing water-related ecological responses to coal seam gas extraction and coal mining in Australia and some examples of ecological assumptions derived from recent EISs. The potential water-related stressors associated with coal seam gas extraction and coal mining are briefly reviewed, supported by a diagram showing how they likely interact with each other. This chapter concludes with a list of the expected project outcomes. Chapter 2 gives some theoretical background to the approach taken in this project. Ecological conceptual models are defined and their advantages and uses are listed, followed by a description of ‘control’ and ‘stressor’ models (Gross 2003) and their combination in the current project to represent likely water-related ecological responses to coal seam gas extraction and coal mining. This chapter concludes with a brief review of the issues associated with scale and uncertainty in ecological conceptual models; both are major considerations in using these models to assess water-related ecological responses. Chapter 3 outlines the project methods, describing the seven-step approach to deriving an ecological conceptual model that combines ‘control’ and ‘stressor’ models to illustrate the likely pathways by which one or more stressors associated with coal seam gas extraction and coal mining would affect specific ecosystems, habitats, species populations or life history stages at different scales. It also lists the main types of information needed to compile the models and accompanying narrative tables. The methods used in the case studies (including a test of the Bayesian Network (hereafter BN) approach) are presented, along with a brief summary of the expert workshop procedure. Chapter 4 describes the results of the case studies, and presents the control and stressor ecological conceptual models for the ‘wet’ and’ dry’ phases of the Swamp Tea-tree (Melaleuca irbyana) population in the Purga Nature Reserve in the Bremer River catchment, Clarence-Moreton bioregion. The full narrative tables for the control and stressor models are provided. This chapter also presents a BN derived at the workshop to predict likely responses of the Swamp Tea-tree population in the Purga Nature Reserve to hypothesised water-related stressors of coal mining. This derivation was done to test the feasibility of the BN approach for indicating potentially important mechanism(s) by which the stressors elicit ecological responses (i.e. a ‘proof-of-concept’). Chapter 5 begins by discussing the roles of ecological modelling in assessing water-related ecological responses to coal seam gas extraction and coal mining, recommending that the approaches to modelling and conceptualisation of hydrology and hydrogeology currently used in EISs be extended to incorporate ecological components to produce ecohydrological models capable of predicting likely water-related ecological responses to coal seam gas and coal mining development. As virtually all models rely on a conceptual framework, the rest of the discussion focuses on ecological conceptual modelling, especially the benefits and challenges involved in deriving the conceptual models. After discussing the ‘lessons learned’ from the various case studies and the BN analysis, this chapter concludes by listing the principal specific questions that should be addressed by future ecological conceptual modelling, including the approach proposed in this project.

page 2

Modelling water-related ecological responses to coal seam gas extraction and coal mining

1.3 Limitations in current assessment of water-related ecological responses to coal seam gas extraction and coal mining The assessment of water-related ecological impacts of development proposals for coal seam gas extraction and coal mining is challenged by our incomplete understanding of ecological responses to hydrological alteration, particularly interactive and cumulative effects at multiple spatial and temporal scales. Currently, analysis of ecological impacts in development assessments is largely qualitative, disregards ecological processes, is poorly integrated with hydrogeological conceptualisation and hydrological modelling, and lacks robust and transparent consideration of multi-stressor impacts and cumulative effects. There is a pressing need to improve the sophistication of ecological assessment by improving the capacity to predict ecological responses, incorporating consideration of ecosystem processes such as nutrient cycling and organic matter decomposition (Bernhardt & Palmer 2011), and better integrating ecological and hydrogeological conceptualisation and modelling. These predictions (hypotheses) need to be clearly stated and their assumptions validated with explicit reference to relevant scientific literature, empirical data and other credible evidence. Environmental assessment documentation for coal seam gas extraction and coal mining projects in Australia reveals a number of assumptions regarding water-related ecological impacts. These assumptions are seldom supported by data or a scientific rationale. Examples include: •

Vegetation in the study area is drought-tolerant and has low physiological sensitivity to water availability (i.e. is resistant to hydrological change).



Instream fauna is tolerant of turbidity, poor water quality and flow variability, and therefore will be unaffected by any hydrological impacts of coal seam gas or coal mining.



The ecology of the area is already impacted by clearing and grazing so any further impacts will be insignificant.



Brigalow is relatively tolerant of periodic inundation, so impacts of subsidence are considered minimal.



Ponds created by subsidence will provide enhanced habitat for aquatic species.



There are no cumulative impacts of subsidence, groundwater drawdown and loss of stream flow.



Groundwater in the study area is too deep to be accessed by vegetation.



There is no surface water-groundwater interaction in the project area.



There is limited connectivity between the coal seams and the source aquifers for springs, and therefore there will be no significant impact on the springs.



Impacts on springs can be mitigated by piping water to the spring.



Groundwater contribution to flow in non-perennial rivers is ecologically insignificant.



Fracturing of stream beds may lead to drainage of overlying pools, loss of aquatic habitat and associated biota and loss of connectivity between pools. Such losses would

page 3

Modelling water-related ecological responses to coal seam gas extraction and coal mining

not be important in non-perennial drainage lines, as aquatic habitat would be present only during flow events and for a short time thereafter. This project explores the use of ecological models in making these assumptions explicit, identifying causal pathways (i.e. how a stressor might elicit an ecological response), investigating interactive and cumulative effects and providing a framework for testing the implicit hypotheses. To put these hypotheses into an appropriate context, a logical starting place is the preparation of credible ecological conceptual models, tailored to appropriate scales of space and time, to complement the current hydrological and hydrogeological conceptual models in many EISs. By integrating these three forms of conceptual models, robust hypotheses can be derived about the likely water-related ecological responses to one or more impacts of coal seam gas extraction and coal mining. These, in turn, could lead to the development of quantitative decision-support tools that would enable more transparent and defensible decisions and facilitate ecologically sustainable water management (Arthington et al. 2010).

1.4 Potential water-related stressors associated with coal seam gas extraction and coal mining Activities associated with coal seam gas extraction and coal mining typically lead to a range of water-related stressors. Most of these stressors interact and their effects can seldom be separated. Indeed, assessing individual effects is inappropriate because it is the collective suite of effects and their interactions (Figure 1.1) that are responsible for water-related ecological changes caused by coal seam gas extraction and coal mining. The two principal types of stressors are those associated with water regime and those with water quality, and these also interact. Surface water regime, as presented in Figure 1.1, refers to where, when and how much water is present. In standing waters, this regime would include water levels, extent and permanence, whereas in running waters, discharge characteristics (volume, seasonal pattern, variability) and velocity are also relevant aspects of the water regime. Groundwater regime includes water table fluctuations and groundwater flux, pressure, and residence time. Water quality is defined here as the physical and chemical features of either surface water or groundwater, affecting ecological processes, the distribution of biota and human uses. Stressors that alter surface water and groundwater regimes result from activities that directly remove or add water (e.g. water extraction for mining, disposal of co-produced water) or activities that indirectly affect water regimes by impounding stream flow and altering catchment and floodplain runoff, infiltration and recharge (Figure 1.1). Stressors that alter surface water and groundwater quality arise from direct contamination (e.g. runoff from stockpiled mine waste) or activities that indirectly affect water quality such as when barriers alter water regimes in rivers. Some stressors (e.g. those that alter water quality) may be caused by multiple activities, and these are likely to have cumulative effects that interact in a complex way. Some stressors give rise to a cascade of related stressors. For example, groundwater drawdown may reduce water availability for deep-rooted and riparian vegetation, change surface water-groundwater connectivity regimes and baseflow volumes leading to increased duration and spatial extent of cease-to-flow periods, change extent and quality of habitat for stygofauna and hyporheic fauna, change environmental conditions that support biogeochemical processes in the hyporheic zone and in aquifers, and reduce spring discharge. The key point here is that most ecological responses to water-related stressors result from cumulative effects of a suite of interacting stressors rather than from a single stressor. Ecological conceptual models strive to portray this complex cumulative interaction as simply as possible – seldom an easy task.

page 4

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Note. Interactions among stressors (bold black type) from coal seam gas extraction and coal mining (blue type); dashed lines represent possible linkages.

Figure 1.1 Hydrological stressors from coal seam gas extraction and coal mining

Furthermore, responses to hydrological stressors are likely to interact over space and time, including beyond the period of coal seam gas extraction or coal mining. Factors such as the flow regime, shapes of the channels and drainage networks, and the effects of stressors unrelated to coal seam gas extraction and coal mining (e.g. agriculture or urbanisation) are likely to determine the ecological responses at different points along a river (cf. McCluney et al. 2014) whereas seasonal factors and temporal changes in land use may govern ecological responses at different points in time. The cumulative effects of some stressors, such as those from the discharge of mine-affected water or co-produced water from coal seam gas operations, may ameliorate with increasing distance from the point of discharge if the system has the capacity to assimilate the impacts through dilution by inflows from unaffected tributaries (Dunlop et al. 2013). Perceptions of the extent and severity of these ecological responses and their cumulative interactions are also strongly influenced by the physical scale of the modelling (discussed in Chapter 2).

page 5

Modelling water-related ecological responses to coal seam gas extraction and coal mining

1.5 Expected project outcomes The expected outcomes from the project were: •

enhance capability in the resources industries to identify and predict the water-related impacts of coal seam gas extraction and coal mining, through uptake of the approach to ecological conceptual modelling and integration of the ecological modelling approach with hydrological and hydrogeological modelling and conceptualisation



improve identification and understanding of the potential water-related ecological responses to coal seam gas extraction and coal mining in Australia, achieved through assisting the IESC in its evaluation of EIS documentation for coal seam gas and coal mining proposals and provision of advice to regulators



provide a framework for ecological conceptual modelling that could be drawn on in the bioregional assessments.

page 6

Modelling water-related ecological responses to coal seam gas extraction and coal mining

2 Using models to predict water-related ecological responses to coal seam gas extraction and coal mining 2.1 Using ecological conceptual models to represent complex ecosystems Natural ecosystems are incredibly complex, comprising numerous components and interactions that constantly change at multiple temporal and spatial scales. For most ecosystems, understanding of responses to natural and anthropogenic disturbances is limited. However, it is acknowledged that these responses are often unexpected ‘ecological surprises’ (Gordon et al. 2008), especially when multiple interacting stressors are involved. Many ecological responses to stressors in ecosystems are nonlinear, frequently resulting in dramatic and rapid changes in species abundances or community composition or even switches between alternative states (Scheffer & van Nes 2007). These changes may be irreversible (e.g. for some aquatic ecosystems in salinised parts of the Western Australian Wheatbelt [Davis et al. 2010]), extinguishing natural biodiversity and producing ecosystems that no longer deliver desired goods and services. To predict the risk of irreversible changes and undesirable outcomes in response to human activities, ecological models are commonly used (Lindenmayer et al. 2010). Ecological models range from verbal descriptions and pictorial graphics to mathematical descriptions and computer-aided models that seek to quantify outcomes and their probability (Jean et al. 2005). This report uses verbal descriptions and pictorial graphics as a means of conceptualising interactions among drivers, stressors, components and processes in an ecosystem, and refers to these as ‘ecological conceptual models’. There is seldom time to determine experimentally the responses of natural ecosystems to different types of disturbances, especially in assessment of likely environmental impacts of a given development such as coal seam gas extraction or coal mining in which stressors and responses may occur over large spatial scales that are difficult or impossible to replicate experimentally. Therefore, models need to be based on the best available science (Ryder et al. 2010) to help identify likely important pathways of cause and effect, how these would be influenced by activities associated with coal seam gas extraction and coal mining, and what might be the water-related ecological responses. These models aim to integrate hydrological and hydrogeological models (e.g. Wondzell et al. 2010; Gondwe et al. 2010), predict and compare likely outcomes from various management actions and enhance communication between scientists and representatives of resource-extracting industries (Westgate et al. 2013). Ecological conceptual models are especially powerful for this last goal. The many advantages of using ecological conceptual models in ecosystem science and environmental monitoring (Lindenmayer & Likens 2010) include: •

specifying the scope and scales of the system of interest



illustrating the main components, processes and interactions at a given scope and scale

page 7

Modelling water-related ecological responses to coal seam gas extraction and coal mining •

generating explicit hypotheses about particular interactions and outcomes



integrating input from different experts into a formalised shared understanding



facilitating rapid communication among scientists, managers and the public about the complexity of the diverse ecosystem components, interactions and responses to multiple stressors



revealing likely responses to one or more stressors so that potential management strategies to minimise impacts can be devised.

Ecological conceptual models are universally used as an essential component of effective environmental science, monitoring and impact assessment (Noon 2003; Jean et al. 2005; Harwell et al. 2010). Without a proper scientific framework based on one or more reliable conceptual models, predictions lack credibility or consistency and costly errors result (Lindenmayer & Likens 2010). Therefore, an excellent investment of time at the start of any project is to develop conceptual models using expert advice, relevant scientific literature and other credible information (Chapter 3), making successive refinements as more information and understanding is achieved by monitoring and research (Westgate et al. 2013). Many ecological conceptual models of complex ecosystems are used to explore and portray how the interactions among different components of the ecosystems influence some particular component or process of interest. In this context, the component or process is an ‘ecological endpoint’ of the ecological conceptual model (Section 2.2) and might be selected because it represents the overall health or condition of the system, known or hypothesised effects of stressors or elements that have important human values (Gross 2003). This definition and uses of ‘ecological endpoint’ closely resembles those of ‘ecological indicators’, and many of the desirable attributes are the same: they must be easily measured, be sensitive to relevant stressors, respond to these stressors in a predictable manner and have a known response to natural disturbances and anthropogenic stressors (Cairns et al. 1993; Dale & Beyeler 2001). Consequently, literature from the research discipline exploring the uses and constraints of ecological indicators is a valuable source of information and examples when selecting appropriate ecological endpoints for use in ecological conceptual modelling. A good starting place is the review by Niemi and McDonald (2004) about the use of ecological indicators. This review deals explicitly with the importance of clearly stated objectives, the recognition of spatial and temporal scales, assessments of statistical variability, precision and accuracy, and establishing linkages with specific stressors.

2.2 Two broad types of conceptual models Gross (2003) recognises two fundamentally different types of conceptual models: control models and stressor models. He defines a control model as a: ‘…conceptualism of the actual controls, feedback, and interactions responsible for system dynamics…’ (p.6). This is probably what most ecologists would think of as a typical ecological conceptual model. He defines a stressor model as one: ‘…designed to articulate the relationships between stressors, ecosystem components, effects, and (sometimes) indicators…’ (p.7).

page 8

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Stressor models typically contain only a subset of system components and aim to illustrate sources of stress and the ecological responses of some attribute(s) of interest. These models are based on known or hypothesised ecological relationships, often derived from control models (Gross 2003). The two types of models are distinguished from each other because they have different goals. Control models portray the most complete and accurate picture of the ecosystem components and interactions whereas stressor models illustrate direct linkages between stressors, ecological responses and ecological endpoints. A key goal of this project was to support the IESC in providing advice on water-related ecological responses to coal seam gas extraction and coal mining (Chapter 1). This goal led to an important modification of the approach described by Gross (2003). As the intention was to portray potential ecological responses to coal seam gas extraction and coal mining in landscapes that are often already modified by other human activities, control and stressor models had to be combined so that the ‘control’ model included natural and anthropogenic drivers and stressors not related to coal seam gas extraction and coal mining. This model represents the state before extraction and mining. The ‘stressor’ model incorporates the hypothesised drivers and stressors associated with coal seam gas extraction and coal mining, and the resulting potential ecological responses. Comparing the ecological conceptual models of the ‘before’ and ‘after’ states illustrates hypothesised ecological responses to coal seam gas extraction and coal mining at a given spatial and temporal scale. The two types of models and the approach to conceptual modelling described by Gross (2003) were adopted for this project because this method is currently used by many other major Australian programs in natural resource management (e.g. Ramsar site ecological descriptions [Butcher & Hale 2010]) and has underpinned the management of national parks in the US for over a decade (Gross 2003; Jean et al. 2005). Another advantage of this approach arises when trying to weigh the benefits and environmental costs of allowing a development to proceed (i.e. setting the two types of conceptual models in the context of society’s values). One way of representing these values is to consider them in terms of ‘ecosystem services’. Ecosystem services are the benefits that people derive from the components and processes of natural ecosystems (Millennium Ecosystem Assessment 2005), including pollination of crops, water filtration in river beds, and atmospheric oxygenation by plants. All ecosystem management, including the management of water-related ecological responses to coal seam gas extraction and coal mining, should explicitly address ecosystem services as well as intrinsic values such as biodiversity (Dudgeon 2014). In a recent paper, Keble et al. (2013) argue that ecological conceptual models should explicitly identify relevant ecosystem services. This approach would help shift the perspective from a narrow one, looking at the impacts of single issues and often short-term economic gains, to a broader one that considers longer-term social benefits by optimising provision and protection of diverse ecosystem services. These authors describe the application of this ecosystem-service perspective to ecological conceptual modelling of the Florida Keys and Dry Tortugas ecosystem. This approach may be relevant for application of ecological conceptual models to coal seam gas extraction and coal mining. Although this report describes the components and processes in the conceptual models in ecological terms, many of these could also be communicated as ecosystem services.

page 9

Modelling water-related ecological responses to coal seam gas extraction and coal mining

2.3 Addressing issues of scale and uncertainty in ecological conceptual models Two issues deserve special discussion because they influence all conceptual models and their development and interpretation. The first of these is scale: how to choose appropriate scales in space and time, how to represent temporal changes (or differences in time scales) on a two-dimensional spatial conceptual model, and how to integrate models representing ecosystems and their components and processes at different scales. In developing ecological conceptual models, choice of scale is dictated by the goals of the model and the bounds of the system (Gross 2003). For example, if a model aims to represent the ecological processes that influence the persistence of a species population, the spatial and temporal scales at which these processes operate are relevant. Drivers and processes affecting the species population operate at different spatial and temporal scales. Further, a particular driver or process will often occur at multiple spatial and temporal scales, with its influence varying accordingly. For example, drivers such as climate and landform may operate at the landscape scale down to the scale of microhabitats. Although the spatial bounds might be specified in an EIS as the mine site and an area of groundwater drawdown, it is likely that stressors and processes (e.g. species dispersal or recruitment) affecting the relevant ecological endpoints operate at broader landscape scales. This wide range of spatial and temporal scales means it is unlikely that a single conceptual model could ever capture their full span, obliging the modeller to decide on one or more scales of space and time that best represent the main ecological pathways and responses in the context of the goal of the model and the bounds of the system being examined. Potentially, two models could be developed: a broad-scale one (landscape to catchment) that includes longer-term processes (decades to centuries) and a series of nested ones at finer spatial and temporal scales that focus on particular locations (e.g. a spring complex, riparian zone or river reach) at seasonal to annual scales. As an example of this approach, Ogden et al. (2005) present a ‘total system’ ecological conceptual model of the Everglades, supplemented by a series of ‘regional’ conceptual models such as that of the southern marl prairies (Davis et al. 2005) within the Everglades. In this approach, the accompanying narratives describing each ecological interaction (Chapter 3) are crucial because they specify the spatial and temporal scales of effect and response. Two-dimensional pictorial models are good for showing static, spatial arrangements of ecosystem components but are unable to effectively illustrate temporal trends in a simple way. One solution is to generate several pictorial models to represent the system at different times (e.g. wet season compared with dry season; immediately after an impact compared with a decade later). Another solution might be to supplement the two-dimensional conceptual models with accompanying plots of expected changes in the state of a variable over time. A third, better suited for computer presentations, could employ animations to show changes over time. Incorporation of multiple spatial and temporal scales in two-dimensional ecological conceptual models should complement the spatial and temporal scales of hydrological and hydrogeological conceptual models currently presented in many EISs. Consideration should also be given to integrating models describing ecosystems and their components and processes at different scales so that they capture the interactions among these scales. One challenge is matching hydrological and hydrogeological conceptual models, which are usually presented at the regional or landscape scale, with ecological models where some of the processes may be operating at much finer scales (e.g. fish feeding on macroinvertebrates in a river pool). Another challenge is adequately representing

page 10

Modelling water-related ecological responses to coal seam gas extraction and coal mining

the effects of stressors that operate at multiple interacting scales. For example, alteration of flow regime by adding co-produced water to a usually dry streambed may have particle-scale effects on biofilm dynamics, reach-scale effects on algal productivity and aquatic invertebrate population sizes, and catchment-scale influences on channel shape and form, potentially assisting dispersal of invasive fishes. All of these effects potentially interact. The second issue to consider is uncertainty: its definition, sources in ecological modelling and implications for deriving hypotheses from ecological conceptual models. Uncertainty is defined following the Standards Australia and New Zealand Risk Management Guidelines (AS/NZS ISO 31000:2009) as: ‘…the state, even partial, of deficiency of information related to understanding or knowledge of an event, its consequence, or likelihood’. This definition was chosen because it accords with the risk-based assessment approach endorsed by the IESC and is adopted in the bioregional assessment methodology (Barrett et al. 2013). Estimates of causes and relative magnitudes of uncertainty are especially important because the bioregional assessments include risk analyses (Component 4, Barrett et al. 2013). These risk analyses combine information from the BA’s risk register (prepared for each bioregion) with the likelihood of an event occurring and an understanding of the uncertainties associated with the impacts. Ecological conceptual models can inform this process by: 1. portraying the predicted pathways of ecological effects and responses resulting from particular events 2. indicating the degree of uncertainty associated with these predictions, as explained in more detail below. Inevitably, every modelling effort is plagued by uncertainty (Tartakovsky 2013). In ecological conceptual models, uncertainty has multiple causes ranging from poorly understood interactions of nonlinear responses that generate ‘ecological surprises’ (Gordon et al. 2008) through to the unknown effects of different scales of impact and response, the often limited availability of data, and inherent uncertainty surrounding all assumptions underpinning all modelling approaches (Lindenmayer & Likens 2010; Westgate et al. 2013). Panels of experts are often used when ecological conceptual models are being developed and potentially introduce further uncertainty as motivational and/or cognitive bias in their input; a rich literature describes these issues and approaches to address them (reviewed in Krueger et al. 2012). Therefore, every prediction from a model must involve rigorous uncertainty quantification (Tartakovsky 2013). This process involves estimates of the effects of structural uncertainty (uncertainty about the validity of a particular model) and parametric uncertainty (uncertainty about the parameters and driving forces in a model). These two sources are sometimes termed epistemic uncertainty because they can be reduced by collecting more data in contrast to irreducible uncertainty, which arises from ‘inherently random phenomena’ (Tartakovsky 2013), exemplified by uncertainty resulting from the interactions of many ecological processes. In the current project, both sources of uncertainty are relevant and, in the absence of further data, there is heavy reliance on expert input and robust ecological conceptual models that record the supporting science and specify the sources and relative magnitudes of uncertainty.

page 11

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Irreducible uncertainty is especially prevalent in ecological modelling, and like that encountered in efforts to predict the effects of future climate change (IPCC 2013), arises from uncertainty about starting conditions, response pathways and model approximations. The IPCC expressed this uncertainty in a qualitative manner, based on the extent of agreement between evidence from different sources (low, medium and high) and the quality and consistency of this evidence (limited, medium and robust). Combining the agreement and quality of the evidence resulted in five grades of confidence (used here as a partial surrogate for uncertainty): 1. Very low: low agreement, limited evidence 2. Low: low agreement, medium evidence; medium agreement, limited evidence 3. Medium: low agreement, robust evidence; medium agreement, medium evidence; high agreement, limited evidence 4. High: high agreement, medium evidence; medium agreement, robust evidence 5. Very high: high agreement, robust evidence. A similar approach to that of the IPCC (2013) could be used to qualitatively estimate irreducible uncertainty in ecological models, accepting that experts will differ in their judgements within these categories of agreement and evidence quality. An example of this application is illustrated in a narrative table accompanying an ecological conceptual model for silver perch (Appendix A).

2.4 A framework for assessing vulnerability coal seam gas extraction and coal mining activities Several frameworks have been proposed for assessing vulnerability of species to climate change, especially where uncertainty is high about what species, habitats and ecosystem are most vulnerable, what aspects of species’ ecological and evolutionary biology determine their vulnerability, and how this information can be used to minimise the potential impacts. The framework by Williams et al. (2008) is especially appealing because it integrates insights from the disciplines of ecology, physiology and genetics into assessing which ecological traits dictate vulnerability of a given species or group of taxa. Vulnerability, defined as the susceptibility of a system to a negative impact (Smith et al. 2000), is the outcome of the extrinsic factors that determine exposure to a stressor and the intrinsic factors that govern sensitivity to it (i.e. ecological traits). Williams et al. (2008) portray exposure at two scales in their framework (regional and microhabitat; orange boxes in Figure 2.1), and then go on to show how these features of exposure interact with changes in habitat (induced by external drivers) and ecology (e.g. habitat use; pale yellow box in Figure 2.1) as one component of vulnerability. The other component, species sensitivity, arises from adaptive capacity and resilience (bright yellow boxes in Figure 2.1) and resistance that, in turn, arise from aspects of the species’ ecology, physiology and genetics.

page 12

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Source: Williams et al. (2008). See main text for details.

Figure 2.1 An integrated framework to assess the vulnerability of species to climate change

Once an estimate of vulnerability is derived and appropriate pathways of exposure and ecological response portrayed in the ecological conceptual models, management strategies can be recommended that would reduce or remove actual or potential impacts of coal seam gas extraction and coal mining. The framework by Williams et al. (2008) also includes feedbacks (blue box in Figure 2.1) whereby changes in ecological interactions and ecosystem processes caused by existing anthropogenic stressors potentially feed back into the knowledge of species’ ecology, physiology and genetics. Walker (2010) modified this framework in a project assessing vulnerability of species in the South Australian River Murray corridor to climate change. He grouped and simplified some of the features of the model by Williams et al. (2008) and used this framework to identify ecological, physiological and genetic traits that an expert panel could consider to address 12 propositions (hypotheses) about the extent to which the regional population of a given

page 13

Modelling water-related ecological responses to coal seam gas extraction and coal mining

species might tolerate climate change. Degree of impact was presented on a qualitative scale (minor, moderate or major, with a fourth option of ‘unknown’) and colour-coded in ‘RAG’ format (Figure 2.2). Thus, for the 12 propositions for 10 very different species in the study area, a range of sensitivities could be portrayed, and summed for an overall indication of sensitivity (Figure 2.2), which also could be adjusted for null assessments.

Source: Walker (2010).

Figure 2.2 Sensitivity assessment

Provisional assessments of sensitivity for 10 selected species of flora and fauna from the River Murray study area (Table 4.3 in Walker 2010). In response to the question “To what extent does this trait constrain the ability of the regional population of this species to withstand exposure to climate change?”, experts’ responses (null: unknown, 1: minor, 2: moderate, 3: major) have been colour coded in “RAG” format for easy reference (Red 3, Amber 2, Green 1, null blank). Initial outcomes (the numbers within each category) are shown in the three right-hand columns, and imply the hardyhead is most sensitive and the yabbie is the least. This framework and sensitivity-scoring approach may be useful in assessing water-related ecological responses by various species to coal seam gas extraction and coal mining development, although Walker (2010) warns that choice of the traits and wording of the propositions must be careful. Perhaps this approach would be most useful where a number of species are to be considered in the EIS for a given area and some effort is being made to determine which ones are most vulnerable and therefore deserve most attention. It will also reveal where information is lacking as well as where groups of species may share parallel responses and, hence, some redundancy in selection of species to model in more detail. A further example of ecological conceptual modelling is an examination of impacts of hydraulic fracturing on eastern brook trout (Salvelinus fontinalis) in the Marcellus Shale region of the eastern US (Figure 2.3). The approach used was a causal conceptual model,

page 14

Modelling water-related ecological responses to coal seam gas extraction and coal mining

wherein life-cycle components of the trout were used as the endpoints. This ecological conceptual model portrays how different stages of the life cycle of the trout vary in their vulnerability to different stressors associated with hydraulic fracturing, and emphasises the complexity of assigning vulnerabilities at the species level when multiple life stages are involved. Unfortunately, this information is seldom available for species that are likely to be affected, especially for their juvenile stages (e.g. seedlings, larvae) which tend to be the most sensitive to most stressors.

page 15

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Conceptual model of relationships between drilling and hydraulic fracturing activities and the life cycle of eastern brook trout (Salvelinus fontinalis). UIC = Underground Injection Control; TDS = total dissolved solids. Source: Weltman-Fahs and Taylor (2013).

Figure 2.3 Conceptual model for brook trout

page 16

Modelling water-related ecological responses to coal seam gas extraction and coal mining

3 Project methodology 3.1 Overview Activities associated with coal seam gas extraction and coal mining occur or are predicted in environments ranging from arid and semi-arid inland areas to temperate or subtropical regions near the coast (hence, the range of bioregions described in section 1.1). These environments have diverse geology, soils, topography, hydrogeology, surface drainage, land use, vegetation cover, and communities of plants and animals. To identify water-related ecological impacts arising from coal seam gas extraction and coal mining in these different areas, ecological conceptual models are needed that include the appropriate drivers, stressors, components and processes for the linked terrestrial and aquatic ecosystems in each area. This section describes seven steps (Figure 3.1) in developing an ecological conceptual model, which were followed during the expert workshop. These steps may be a useful sequence for similar models when preparing an EIS. The first two steps in the process are to agree on the goals of the conceptual models and specify the bounds of the system of interest (Figure 3.1). These are related issues because the goals dictate the selection of the bounds (spatial and temporal scales) of the conceptual model (discussed in section 2.3).

Figure 3.1 Flow-chart of ecological conceptual model development

page 17

Modelling water-related ecological responses to coal seam gas extraction and coal mining

These steps in ecological conceptual model development were followed in this project (following recommendations by Gross (2003) with modifications described in section 2.2). Although the final output is the full model including stressors associated with coal seam gas extraction and coal mining, comparison with the control model excluding these (created in Step 4) helps to predict ecological effects of coal seam gas extraction and coal mining. Feedback loops exist among most steps in this process because successive refinements of the conceptual models will occur as more information becomes available. The resulting conceptual models can then be further analysed using approaches such as BNs. The third step (Figure 3.1) is to identify the drivers, stressors, ecological effects and likely ecological endpoints. A logical starting place is to compile a table of likely drivers and stressors associated with natural and anthropogenic perturbations, the latter both related and unrelated to coal seam gas and coal extraction. Although attention focuses on coal seam gas extraction and coal mining, their ecological effects have to be predicted in the context of natural and pre-existing anthropogenic factors as well. This table can then be used as a checklist to ensure that region-specific conceptual models include all the principal drivers and stressors. Later, this table can be extended as a ‘narrative table’ to include explicit reference to relevant literature on water-related ecological responses to these different stressors and drivers, especially where region-specific information exists (e.g. that from BAs or new research in the area), and is expanded in the fourth step when the control model is constructed. The fourth step is drawing up a control model to portray the main interactions among relevant ecosystem components in a given area at a given temporal scale. Only the main interactions and components should be selected so that the control model is tractable (Gross 2003); it is too cumbersome to show every possible interaction and component. The result is a pictorial representation of the main drivers, stressors, processes, components and interactions (including feedbacks) of the linked terrestrial and aquatic ecosystems in that area, except drivers, stressors and responses associated with coal seam gas extraction and coal mining. Typically, this pictorial representation is either a ‘box-and-arrow’ diagram (also termed an ‘influence diagram’, see Appendix B) or an illustration that represents the landscape using cross-sectional diagrams and icons. On the ‘box-and-arrow’ diagram, the boxes represent drivers, stressors and ecosystem components and the arrows portray pathways of influence. The ‘box-and-arrow’ diagram is unable to convey information about, for example, the relative locations of stressors and water-related assets at a given site. In contrast, the landscape illustration is able to show geographic proximity of stressors and assets, and uses icons to represent drivers, stressors and ecosystem components. Although arrows can be included in a landscape illustration to show movements of water, materials such as sediments or nutrients, and biota, it is seldom possible to portray the pathways of ecological responses to one or more stressors as clearly as on the box-andarrow diagram. Further, the box-and-arrow diagram is a more useful starting point for BNs than the landscape diagrams (section 4.2). As each graphic has its own advantages and strengths in illustrating different aspects of the ecological responses, both are often presented. These diagrams are supplemented by a matching narrative table (often presented as a legend at the bottom of the landscape figure) that states the hypothesised or known ecological responses to a given stressor. Where possible, relevant scientific and other credible literature is cited in support of each hypothesis or statement. It will seldom be possible to present adequate detail for all the components in a single model (Ogden et al. 2005). Therefore, nested within this general control model are likely to be submodels dealing with specific ecosystems (e.g. springs, riparian zones), whose linkages

page 18

Modelling water-related ecological responses to coal seam gas extraction and coal mining

are shown in the broader-scale general model. The general control model and the specific control submodels nested within it serve to illustrate broad assumptions about how drivers unrelated to coal seam gas extraction and coal mining influence ecosystem components, processes and interactions. Where possible, these control models would indicate the magnitude and direction of the effects. The fifth step is to identify anthropogenic stressors associated with coal seam gas extraction and coal mining (Figure 3.1). Again, this step should be accompanied by a detailed narrative, with reference to relevant literature, that describes the stressors and likely spatial and temporal scales. Linking this stressor model (and narrative) to the control model developed in Step 4 is the sixth step and yields the final conceptual model of hypothesised water-related ecological responses to coal seam gas extraction and coal mining. This conceptual model and its associated narrative table(s) are used in Step 7 to predict the likely ecological outcomes of interacting stressors (Figure 3.1). The combination of these two models, along with their associated narratives, now allows the generation of hypotheses about likely water-related ecological responses to different scenarios of coal seam gas extraction and coal mining in a given biophysical region. Of course, these hypotheses will include varying degrees of uncertainty (section 2.3) because ecosystems are dynamic and ecosystem components usually interact in nonlinear ways and change over time.

3.2 Control and stressor models One of the main goals of this project was to trial the process of constructing one or more ecological conceptual models, following the steps described in section 3.1. The intention was a ‘proof-of-concept’ to ascertain whether it was feasible to produce useful control and stressor models based on information from sources that would be accessible to proponents seeking to assess the likely water-related ecological responses to coal seam gas extraction and coal mining at a given site. This information would include relevant details from the bioregional assessments (Chapter 1) as well as the site- and region-specific information identified by the IESC (2014) Information Guidelines: •

geological information, including names and descriptions of formations with accompanying data on surface and subsurface geology (e.g. as cross-sectional diagrams) and information on structures (e.g. faults, strata of high hydraulic conductivity) that may affect movement and connectivity of water, especially flow, recharge and discharge of groundwater



hydrogeological information on hydraulic features (e.g. hydraulic conductivity and storage characteristics) of each hydrogeological unit, the varying depths to these units (including standing water levels or potentiometric heads) and their hydrochemical features, and the likely recharge and discharge pathways and volumes for each unit, especially those likely to be affected by the proposed development



geomorphological information (e.g. drainage patterns, channel features, floodplain development) matched with relevant information on the hydrological regime (e.g. temporal trends in stream flow and/or water levels, flood regimes and areas inundated at a range of flows exceeding bankfull discharge), sediment regime (e.g. turbidity and sources of sediment production and deposition), and geochemical features and processes that would affect water quality (e.g. alkalinity, salinity, ionic

page 19

Modelling water-related ecological responses to coal seam gas extraction and coal mining

composition, and concentrations of organic chemicals, radionuclides and other potentially harmful materials) •

hydrological information not covered above, including timing, volumes and directions of surface water-groundwater exchanges, connectivity among aquifers, and connectivity with sea water



information on the water resources of the site and surrounding region, including aspects of the water balance (e.g. seasonal and annual variations in precipitation, evapotranspiration, surface water permanence and exchange with groundwater) and other relevant hydrological and hydrogeological data, including water quality for surface and groundwater (e.g. alkalinity, salinity, ionic composition, and concentrations of organic chemicals, radionuclides and other potentially harmful materials)



information on the water-related assets (e.g. surface waters, springs and other groundwater dependent ecosystems) of the site and surrounding region, including data from surveys of relevant habitats and their biota, especially details of their reliance on surface water and groundwater resources, and the associated ecological processes



information about the natural and pre-existing anthropogenic drivers and stressors at the site and in the surrounding region (to be used for the control model) and about the drivers and stressors likely to be associated with coal seam gas extraction and coal mining (to be used for the stressor model).

Obviously, the amount and quality of this information largely dictate the level of detail that can be provided by the resulting ecological conceptual models and, in turn, their appropriateness for judging the likely water-related responses to coal seam gas extraction and coal mining at a particular site. It is also preferable to use relevant scientific and other credible literature to support the assumptions made when attributing various ecological responses to the drivers and stressors presented in the ecological conceptual model. Finally, it is likely that there will be one or more field surveys, conducted according to relevant protocols, to gather site-specific data on environmental conditions and the biota, and to inspect the water resources and their catchments or recharge zones to infer likely influences from natural and pre-existing anthropogenic drivers and stressors on relevant ecosystem components and processes. All of these activities generate information that can be used in the construction of the control and stressor versions of the ecological conceptual models and their accompanying narratives for the site and surrounding region. A hypothetical case study from the Clarence-Moreton bioregion was used to determine the feasibility of constructing several ecological conceptual models at varying spatial scales with sources of information described above that were readily available. A field site was visited during the expert workshop (section 3.3) but without ecological sampling. The primary goal of this part of the project was a ‘proof-of-concept’ to derive some ecological conceptual models in as complete a form as possible and to discuss the ‘lessons learned’ during the process. In an ecological assessment as part of EIS, ecological conceptual models must be informed by site-specific data collected at appropriate temporal and spatial scales. As a case study in developing a species-specific ecological conceptual model, the EPBC-listed silver perch (Bidyanus bidyanus) in a section of the Mooki River (Gunnedah Basin) was chosen (Appendix A). A desktop survey of relevant literature was used to generate a narrative table that listed hypotheses about inferred ecological responses by silver perch to various stressors. For each hypothesis, the table also presented qualitative estimates of evidence, agreement and confidence (following the IPCC 2013 approach

page 20

Modelling water-related ecological responses to coal seam gas extraction and coal mining

described in section 2.3) as a surrogate means of expressing uncertainty. The narrative table was used to help generate an influence diagram portraying the main natural and anthropogenic drivers and stressors (and their interactions) likely to affect the persistence of silver perch populations in a section of the Mooki River. The validity of the table and hypotheses was subsequently confirmed by an independent expert to test whether reliable narrative tables could be derived from desktop surveys of the literature.

3.3 Expert workshop assessment of some worked examples of ecological conceptual models One major aim of this project was to trial the development of ecological conceptual models at varying spatial scales and to supplement the current hydrogeological conceptual models, using the sources of information described in section 3.2 that were readily available. To do this in as realistic a way as possible, Auricht Projects was commissioned to generate several ecological conceptual models to represent hypothesised ecological responses to plausible coal mining scenarios in the Clarence-Moreton Basin as a case study. The case study focused on the potential ecological responses to coal mining of the Swamp Tea-tree (Melaleuca irbyana) population in Purga Nature Reserve in the Bremer River catchment, south-east Queensland. This species was chosen because “Swamp Tea-tree (Melaleuca irbyana) Forest of South-east Queensland” is listed as a Critically Endangered Ecological Community under the Environment Protection and Biodiversity Conservation Act 1999 (Commonwealth) and as an Endangered Regional Ecosystem under the Vegetation Management Act 1999 (Queensland). The case-study site was appropriate because coal mining has occurred in the area, the Swamp Tea-Tree Forest is of conservation interest, there was very little data on the region and the species of interest (i.e. a realistic situation), and the field visit observations helped generate the influence diagram used for exploring the potential for applying Bayesian modelling (section 4.2). Two ecological conceptual models illustrated the likely effects of coal mining on the Swamp Tea-tree Forest at two phases of its water regime: the ‘wet phase’ when the wetland is inundated and aquatic processes would be expected to be at their peak, and the ‘dry phase’ when surface water is absent. A third ecological conceptual model drawn at the landscape scale revealed the geographic context of the Purga Nature Reserve in the Bremer River catchment. The next step was to validate the veracity and usefulness of these ecological conceptual models in representing the likely water-related ecological responses to potential coal seam gas extraction and coal mining. This was done during a three-day workshop with scientists with expertise across hydrology, hydrogeology, biogeochemistry, freshwater ecology, groundwater dependent ecosystems and water resource management. Expert advice and information was sought about: •

the suitability of the conceptual framework of the project, such as the use of the control and modified stressor models (section 3.2)



current understanding of hydrology-ecology relationships at several spatial and temporal scales for selected taxa or communities in habitats likely to be affected by coal seam gas extraction and coal mining



the scientific accuracy and usefulness of the conceptual ecological models



the appropriateness of applying BNs (Appendix B) to supplement the use of the conceptual ecological models.

page 21

Modelling water-related ecological responses to coal seam gas extraction and coal mining

The workshop agenda is given in Appendix C, brief biographies of participants in Appendix D, and abstracts of presentations in Appendix E. Details of the case study area (Purga Nature Reserve) are given in Appendix F. Discussion in the field presented by experts with different backgrounds generated valuable insights into the types of information needed when compiling conceptual ecological models. These insights, rather than determining specific features of Purga Nature Reserve, were the focus of this exercise. Attention in the workshop focused on the Purga Nature Reserve case study and the trial application of the BN. The veracity and usefulness of the conceptual ecological model for the silver perch example from the Gunnedah Basin was also assessed at the workshop and, later, by an independent expert (Dr Keith Walker) familiar with the relevant ecological literature on this species. The conceptual models and accompanying text were also reviewed by four technical advisors with expertise in hydrogeology (Dr Martin Andersen), aquatic ecology (Dr Bruce Chessman), plant ecology (Prof. Ray Froend) and landscape ecology (Dr Alexander Herr), and by a theoretical and applied ecologist (Dr Jennifer Firn) with a research interest in M. irbyana, Dr Anthony O’ Grady (ecology lead, BAs) and Prof. Angela Arthington, the IESC ecologist.

page 22

Modelling water-related ecological responses to coal seam gas extraction and coal mining

4 Results: case study and worked examples 4.1 Ecological conceptual models for Purga Nature Reserve As the intention of this project was to provide a ‘proof-of-concept’ of the process for developing ecological conceptual models by following the seven steps in Figure 3.1, the results for each step are presented in sequence. The primary goal of the ecological conceptual model for this case study (Step 1) was to portray the main drivers, stressors and pathways of likely water-related ecological effects of coal mining affecting the principal ecosystem components that support the persistence of the Swamp Tea-tree Melaleuca irbyana population in the Purga Nature Reserve. The horizontal bounds in space of the ecological conceptual model (Step 2) were set as the episodically filled basin and fringing margins of the wetland within the 140-hectare Purga Nature Reserve (Figure 4.1). A field visit to the site indicated that the basin of this wetland lacks any distinct edge defined by either geomorphology (e.g. bank or sediment strand-line) or bordering semi-aquatic vegetation (e.g. a fringe of reeds or rushes). This is common for many shallow seasonally filled wetlands but does not prevent development of ecological conceptual models. The aquatic-terrestrial transition zone is an important ecological component and must be included in all models of aquatic ecosystems. Although the wetland may be perched above the regional water table, the swamp tea-trees were hypothesised to access groundwater occasionally, so the vertical spatial bound of the ecological conceptual model was set to encompass the likely annual range of groundwater fluctuation below the wetland and its fringing vegetation.

Dotted line encloses approximate bounds of target population of Swamp Tea-tree. Data sources: World_Imagery - Source: Esri, DigitalGlobe, GeoEye, i-cubed, Earthstar Geographics, CNES/Airbus DS, USDA, USGS, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community.

Figure 4.1 Location of Purga Nature Reserve

page 23

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Natural and anthropogenic drivers and stressors not associated with coal mining were identified (Step 3) and are listed in the first two columns of Table 4.1. This step also entailed listing the key ecosystem components that could have a bearing on the persistence of the Melaleuca irbyana population, the ecological endpoint. These ecosystem components are presented in the third column of Table 4.1, along with their hypothesised ecological effects (Step 4). Information to support these hypotheses was derived from published and ‘grey’ literature (Table 4.1), discussion among experts during the field site visit and scientific advice from Prof. Ray Froend and Dr Jennifer Firn. Step 4 also entailed drawing up a box-and-arrow diagram (Figure 4.2) to illustrate the ecological interactions among the relevant ecosystem components in the wetland and their hypothesised ecological responses to the drivers and stressors presented in Table 4.1. This diagram was used as the basis for the two control models illustrating the ‘inundated’ and ‘dry’ phases of the seasonal wetland (upper panels of Figure 4.3 and Figure 4.4). A conceptual model developed by the Queensland Department of Environment and Heritage Protection (DEHP) for coastal and subcoastal floodplain tree swamps with Melaleuca and Eucalyptus species (Figure 4.5) was also drawn upon to assist the conceptualisation of the components and processes of Purga Nature Reserve. This model has been verified and reviewed by experts, and therefore is a robust starting point. It is not, however, specific to a site or species (i.e. Melaleuca irbyana), and so not all information is directly transferable (refer to section 2.3). The main natural drivers affecting the persistence of the Melaleuca irbyana population at the study site are hypothesised to be climatic, hydrological and hydrogeological, geomorphological (landform) and geological ones affecting stressors such as fire regime, water and nutrient availability, and soil pH and salinity (Table 4.1, Figure 4.3 to Figure 4.5). Anthropogenic drivers and stressors not associated with coal mining include the effects of historical and current land clearance, weed invasion and grazing by non-native animals (Table 4.1, Figure 4.2 to Figure 4.4).

page 24

Modelling water-related ecological responses to coal seam gas extraction and coal mining Table 4.1 Narrative table to accompany the control model This table lists the natural and anthropogenic drivers and stressors (excluding those associated with coal mining) and their hypothesised ecological effects on the persistence of the Melaleuca irbyana (MI) population in Purga Nature Reserve. Numbers in the third column correspond to arrows in the box-and-arrow diagram (Figure 4.2). Climate change, although an important stressor, is not included in this table or in the conceptual ecological models. References relevant to each hypothesis are included (where available), along with qualitative estimates of evidence, agreement and confidence (following the IPCC 2013 approach presented in section 2.3). E = evidence, A = agreement, C = confidence.

Driver

Stressor

Hypothesised ecological effects on the persistence of the Melaleuca irbyana (MI) population

References

Climate

Maximum air temperature

1. Sustained high air temperatures probably stress adult MI and kill seedlings.

Dept. of the Environment (2014)

Fire frequency and intensity

2. MI plants are likely killed by frequent burning and/or very hot fires (and recruitment is probably especially vulnerable because very frequent fire inhibits regeneration).

Dept. of the Environment (2014)

1

1

1

Low humidity and high evaporation (including wind)

3. MI populations cannot persist if evaporative losses (accelerated by low humidity, high air temperatures and warm winds) are too high for too long (exact limits unknown but seedlings are likely to be high vulnerable).

Logan City Council (n.d.)

1

1

1

Amount and timing of annual rainfall

4. MI population persistence probably requires a ‘window’ of inundation that occurs at the right time of the year and is long enough to supply the species’ needs but not so long that it kills MI plants or enables competitors to succeed.

DEHP (2013)

1

1

1

Exposure to solar radiation

5. MI, especially seedlings, are likely to be harmed by excessive exposure to solar radiation (e.g. edge effects of fragmentation; loss of overstorey shading).

Logan City Council (n.d.)

1

1

1

Soil fertility

6. MI plants tolerate low-nutrient soils, potentially giving them a competitive advantage over other species.

Dept. of the Environment (2005)

1

1

1

Soil pH

7. MI plants grow on vertosols that are alkaline, potentially giving them a competitive advantage over other species.

Dr Jenn Firn, pers. comm.

1

Soil salinity

8. It is likely that excessive and/or sustained soil salinity impairs the species’ population persistence (although most Melaleuca species are quite salt-tolerant).

DEHP (2013)

1

Landform and geology

page 25

E

A

C

1

1

1

1

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Driver

Hydrology

Agriculture bordering the Nature Reserve

Stressor

Hypothesised ecological effects on the persistence of the Melaleuca irbyana (MI) population

References

E

A

C

Soil features (e.g. heavy, grey, cracking clays)

9. MI plants require seasonally cracking, grey clay soils that are heavy, coarse, and poorly drained, which likely gives them a competitive advantage over many other species.

Dept. of the Environment (2005)

1

1

1

Cracking characteristics

10. The cracking characteristics of the soils alter microtopography, may provide microclimates for germination of MI, and also trap organic matter and other nutrients that support the plants’ growth.

1

Topography

11. Basin shape, drainage and microtopography likely create important microclimates for germination and persistence of the MI population, especially in terms of water regime and inundation.

1

Groundwater quality (including salinity)

12. As MI plants are thought to have a deep root system and may have ‘some reliance on groundwater supplies’ (Logan fact sheet), poor groundwater quality (including high salinity) may impair MI population persistence.

Logan City Council (n.d.)

1

1

1

Seasonal water table fluctuations

13. As MI plants are thought to have a deep root system and may have ‘some reliance on groundwater supplies’ (Logan fact sheet), excessive or sustained groundwater drawdown may impair the species’ population persistence.

Logan City Council (n.d.)

1

1

1

Flow regime of Purga Creek

14. Assuming overbank flows from Purga Creek are relevant to the wetland and the MI forest, altered flow regimes may change inundation patterns, impairing MI population persistence.

1

Extraction of groundwater

15. Groundwater extraction for agricultural use may lower the water table, reducing access by MI plant roots and impairing the species’ population persistence.

1

page 26

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Driver

Historical and current land clearance

Stressor

Hypothesised ecological effects on the persistence of the Melaleuca irbyana (MI) population

References

Extraction of surface water

16. Surface water extraction from the Bremer River and Purga Creek for agricultural use may reduce the very occasional flooding from the river into the wetland, altering the inundation regime and impairing MI population persistence.

1

Increased contamination risk from chemicals

17. Agricultural chemicals carried in runoff are likely to impair MI population persistence.

1

Increased edge effects from land clearance, tracks, fence lines, etc. for agriculture.

18. Land clearance for agriculture around the Nature Reserve causes edge effects:

Altered rates of sedimentation

19. Sedimentation from agricultural runoff may smother seedlings, impairing MI population persistence.

Increased risk of agricultural weed invasion

20. Exotic pasture grasses and other weeds are likely to invade from agricultural areas, restricting germination and competing with MI seedlings for resources such as water, nutrients and space (also may alter fuel loads, affecting local fire regimes).

Dept. of the Environment (2005)

1

1

1

Fragmentation by land clearance (including for fire breaks and tracks)

21. Fragmentation of populations causes edge effects (see below) and leads to loss of genetic diversity because of disruption to natural gene flow, impairing long-term population persistence of MI.

Dept. of the Environment (2005)

1

1

1

Altered rates of sedimentation

22. Sedimentation may smother seedlings and erosion may expose roots, impairing MI population persistence.

1

Removal of native plant cover

23. Clearing plant cover alters rainfall interception and infiltration patterns, affecting runoff and soil moisture, impairing MI population persistence.

1

Logan City Council (n.d.)

E

1

A

1

C

1

“MI communities are likely to be negatively impacted by edge effects such as weed invasion, increase in wind and evaporation, and changes to solar radiation and temperature changes”.

page 27

1

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Driver

Tourism and other human activities in Purga Nature Reserve

Stressor

Hypothesised ecological effects on the persistence of the Melaleuca irbyana (MI) population

Increased salinity in saltprone areas

24. Salinity (‘secondary salinity’) in salt-prone areas may rise through evapoconcentration in seasonal wetlands, impairing MI persistence.

1

Increased risk of weed invasion

25. Clearance increases the risk of spread of invasive plants (weeds) that compete with MI, especially seedlings.

1

Altered rates of sedimentation

26. Sedimentation from altered runoff and cleared pathways and carparks may smother seedlings, impairing MI population persistence.

1

Tourist pressure

27. Compaction by vehicles and trampling likely alter local surface water and rainfall runoff and infiltration patterns, potentially impairing MI population persistence.

1

Infrastructure

28. Construction and maintenance of tourist facilities such as boardwalks, carparks and pathways may fragment MI populations, accentuate problems associated with edge effects, and alter runoff and infiltration patterns.

1

Illegal wood collection

29. Removal of dead timber illegally from the Nature Reserve reduces stocks of organic matter and nutrients, affecting natural decomposition processes and altering carbon cycling in a way that may impair MI population persistence.

1

Illegal rubbish disposal

30. Dumping of household or industrial rubbish illegally in or near the Nature Reserve may physically impair MI seedling establishment and growth, poison adult and young MI plants, and alter runoff and infiltration patterns, potentially impairing MI population persistence.

1

Grazing by non-native animals

31. Non-native animals such as rabbits, hares and other vertebrates impair MI population persistence (and probably recruitment) by grazing, especially on new growth and seedlings.

(e.g. boardwalks, pathways)

page 28

References

Dept. of the Environment (2005)

E

1

A

1

C

1

Modelling water-related ecological responses to coal seam gas extraction and coal mining

This diagram shows the hypothesised ecological effects on the persistence of the Swamp Tea-tree Melaleuca irbyana population in Purga Nature Reserve influenced by the natural and anthropogenic drivers and stressors (excluding coal mining) listed in Table 4.1. Numbered arrows refer to specific hypotheses in Table 4.1.

Figure 4.2 Box-and-arrow diagram of the control model for Melaleuca irbyana

page 29

Modelling water-related ecological responses to coal seam gas extraction and coal mining

The fifth step was to list the water-related stressors associated with a scenario of coal mining near Purga Nature Reserve, and the results of this step are given in Table 4.2. As before, the hypothesised effects of these stressors were also tabulated (together with relevant references) and used to generate the stressor models shown in the lower panels of Figure 4.3 and Figure 4.4. Combining the stressor and control models (Step 6), resulted in the box-and-arrow diagram in Figure 4.6, ultimately used for the assessment of the BN approach. The principal stressors associated with the scenario of coal mining near Purga Nature Reserve were hypothesised to be alterations to overland flow, runoff and inundation regimes of the wetland, topographic changes through subsidence, and weed invasion. Table 4.2 Narrative table to accompany the stressor model This table lists the drivers and stressors associated with coal mining (the case study scenario) and their hypothesised water-related ecological effects on the persistence of the Melaleuca irbyana (MI) population in Purga Nature Reserve. Numbers in the third column correspond to arrows in the box-and-arrow diagram (Figure 4.6). References relevant to each hypothesis are included, along with qualitative estimates of evidence, agreement and confidence (following the IPCC 2013 approach presented in section 2.3). E = evidence, A = agreement, C = confidence.

Driver

Stressor

Hypothesised water-related ecological effects on the persistence of Melaleuca irbyana (MI)

References

E

A

C

Coal mining

Groundwater drawdown

32. As MI plants appear to have a deep root system and may have ‘some reliance on groundwater supplies’ (Logan fact sheet), excessive or sustained groundwater drawdown may impair the species’ persistence.

Logan City Council (n.d.)

1

1

1

Altered ground water quality

33. Changes in pH and concentrations of nutrients and salt of subsurface water may impair MI population persistence either physiologically, by favouring competitors, or in both ways.

1

Subsidenceinduced topographic change

34. Topographic changes caused by subsidence may alter floodplain inundation and/or overland flow, in turn altering the ‘window’ of inundation, either reducing it to being insufficient to supply the species’ needs or increasing it so that it kills MI plants or enables competitors to succeed.

1

Altered surface water quality

35. Changes in pH and concentrations of nutrients and salt of surface (and infiltrated) water may impair MI population persistence either physiologically, by favouring competitors, or in both ways.

1

Altered floodplain inundation and/or overland flow

36. Increased or decreased floodplain inundation and/or overland flow may alter the ‘window’ of inundation, either reducing it to being insufficient to supply the species’ needs or increasing it so that it kills MI plants or enables competitors to succeed.

page 30

DEHP (2013)

1

1

1

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Driver

Stressor

Hypothesised water-related ecological effects on the persistence of Melaleuca irbyana (MI)

References

E

A

C

Altered flow regime in Purga Creek

37. Changes in the flow regime of Purga Creek that either increase or reduce river water inputs to the wetland may alter the ‘window’ of inundation, either reducing it to being insufficient to supply the species’ needs or increasing it so that it kills MI plants or enables competitors to succeed.

DEHP (2013)

1

1

1

Altered rates of sedimentation

38. Increased sedimentation rates may smother seedlings whereas erosion may expose roots, impairing MI population persistence.

1

Increased spread of exotic species

39. Weed invasion, especially of pasture grasses, may restrict germination and weeds may compete with MI seedlings for resources such as water, nutrients and space (also may alter fuel loads, affecting local fire regimes).

1

40. Non-native animals such as rabbits, hares and other vertebrates impair MI population persistence (and probably recruitment) by grazing, especially on new growth and seedlings.

1

page 31

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Figure 4.3 Purga Nature Reserve (wet phase)

page 32

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Figure 4.4 Purga Nature Reserve (dry phase)

page 33

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Source WetlandInfo, Department of Environment and Heritage Protection, Queensland, Accessed 12th July 2014, . For explanation of symbols, see website.

Figure 4.5 Conceptual model of a coastal and subcoastal floodplain tree swamp (Melaleuca and Eucalyptus spp.)

page 34

Modelling water-related ecological responses to coal seam gas extraction and coal mining

This diagram shows the combined control and stressor model showing the hypothesised water-related ecological effects on the persistence of the Swamp Tea-tree Melaleuca irbyana population in Purga Nature Reserve influenced by the drivers and stressors listed in Table 4.1 and Table 4.2. Numbered arrows refer to specific hypotheses in Table 4.1 and Table 4.2.

Figure 4.6 Box-and-arrow diagram of the stressor model for Melaleuca irbyana

page 35

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Figure 4.7 Landscape setting of Purga Nature Reserve

page 36

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Step 7, the final one, uses the narrative tables (Table 4.1 and Table 4.2) and the ecological conceptual models (Figure 4.2 to Figure 4.4, Figure 4.6) to predict and compare likely ecological responses to coal mining in the study area. In this example, the responses to stressors by the Melaleuca irbyana population in the Purga Nature Reserve are likely to vary between the seasonal wetting and drying phases. For example, a fire occurring during the inundated phase will probably be less intense than one that occurs during the dry phase when there is likely to be a large fuel load of dry organic matter, resulting in a hotter fire potentially intense enough to kill Melaleuca irbyana seedlings and impair regeneration. Consequently, dry-season fires are likely to be more harmful to the persistence of the Melaleuca irbyana population. Temporal factors such as these seasonal differences are important aspects to include in ecological conceptual models, often requiring multiple-panel diagrams such as Figure 4.3 to Figure 4.5. Another important aspect of understanding the water-related ecological effects of natural and anthropogenic drivers is knowledge about the landscape context of the Melaleuca irbyana population in the Purga Nature Reserve. This information is not captured by the box-and-arrow diagrams or the site-scale models. Figure 4.7 presents a catchment-scale pictorial ecological conceptual model that portrays the geographic proximity of the various natural and anthropogenic drivers and stressors potentially affecting Melaleuca irbyana population persistence in the Purga Nature Reserve within the Bremer River sub-catchment. This diagram shows how the wetland is likely to be more influenced by flooding in Purga Creek than by the main stem of the Bremer River which receives inputs such as sediments and excessive nutrients from abattoirs, irrigated pastures and croplands (Figure 4.7). Therefore, water quality in the Bremer River is unlikely to be relevant to conditions in the wetland or Melaleuca irbyana population persistence in Purga Nature Reserve.

4.2 Bayesian network session This section describes the outcomes from the facilitated workshop session held in July 2014. The purpose was to trial the development of a Bayesian network as a potential method for use in EISs. Bayesian networks are causal networks with predictive capabilities that can be used to explore knowledge gaps and potential stressors, their interactions, and their strengths in influencing an endpoint. Note, all outcomes are hypothetical, with the purpose being to demonstrate potential of the approach rather than developing a robust model. Endpoint The Bayesian network was focused on the Swamp Tea-tree (Melaleuca irbyana) population at the Purga Nature Reserve. Two endpoints were identified: •

persistence of the Melaleuca irbyana population (where persistence may be measured by adult reproduction and seedling establishment and maturation)



composition of the overall vegetation community, representing potential for change to a more terrestrial vegetation type.

The intended outcome was a model that can be used to better explore the role of hydrology in supporting ecological values, and the potential interactions of coal mining stressors on the system. The outcome was not intended to be a modelling tool to inform management. Scale The spatial scale for the model is the Purga Nature Reserve, and the timeframe considers 20 years of mining operation.

page 37

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Scenario of change The workshop constructed a scenario from expert opinion of how nine stressors could affect the Swamp Tea-tree (Melaleuca irbyana) Forest at the Purga Nature Reserve (Table 4.3). Table 4.3 Scenario construction for the Purga Nature Reserve, with type of stressor and frequency of occurrence Stressor

Scenario construction

Stressor type and frequency

1. Groundwater drawdown

1 metre

Chronic

2. Change in river flow regime - assume groundwater connectivity, mine de-watering

Decreased duration of baseflows, increased magnitude and duration of high flows, altered timing, linked to Stressor 9

Event: multiple/season

3. Altered floodplain inundation

Increased duration of drying, reduced frequency of inundation events, altered timing, linked to Stressor 9

Event: 1 in 10 years

3. Altered rainfall runoff patterns (overland flow) - local inundation

Increased duration of inundation after rainfall

Event: wet season

4. Altered sedimentation from river

Increased deposition in very wet years, levee failure

Event but cumulative 1 in 30 years

5. Subsidence-induced topographic change

Localised, 10s of cm

Chronic

6. Altered surface water quality

Changes to cation concentrations - soil structures, organic acids - pH, inorganics, organics

Cumulative, event and cumulative, event spills

7. Altered groundwater quality

As above

Cumulative, event and cumulative, event spills

8. Spread of exotic species

Weed invasion, drying sediments terrestrialisation

Chronic, chronic

9. In-stream barriers, diversions and levees

Decrease in hydrological connectivity

Chronic

Influence diagram An influence diagram was used to explore the interactions between stressors and endpoints (Figure 4.8). Potential pathways and major knowledge gaps were explored as part of model

page 38

Modelling water-related ecological responses to coal seam gas extraction and coal mining

development. The influence diagram was informed by available information on the species listing and coal resource development, and a site visit. Note that the influence diagram is largely hypothetical, and is for demonstration purposes only. The experts acknowledged a potential for increased groundwater drawdown, leading to increased local subsidence from the loss of groundwater pressure, and the regime of overbank flows inundating the Reserve (information on the precise nature of the hydrogeology was not available at the workshop; the scenario explored is speculative only). The scenario focussed on local subsidence that may influence localised floodplain inundation, potential for groundwater drawdown, increased interception of rainfall runoff, extraction and disposal of surface water, presence of infrastructure (levees and barriers), and potential interactions with climate and pests.

Note - this influence diagram is for demonstration purposes only.

Figure 4.8 Influence diagram developed in the workshop showing interactions between hydrological stressors and endpoints

In terms of impacts, de-watering of the aquifer could lead to disposal of water into the local stream, which would decrease the duration of low-flow periods. An increased interception of rainfall and associated runoff from the mining development could lead to a decrease in overland flow. This decrease would lead to changes in aspects of floodplain inundation (drying) such as decreased duration of inundation, decreased frequency of inundation and decreased extent of inundation. These changes would affect soil water storage, which would affect the Melaleuca species’ population persistence through changes in both adult survival

page 39

Modelling water-related ecological responses to coal seam gas extraction and coal mining

and recruitment opportunities. For example, high rainfall extremes, levee failure and construction of impervious surfaces might also lead to increased sedimentation. Cumulative changes in hydrological processes could enhance the habitat for those native species that have increased tolerance for drier conditions as well as exotic species. These hydrological changes could influence the structure and composition of the swamp community. Bayesian network To demonstrate how a causal model can be framed in a Bayesian network, a model was developed in the workshop on the basis of the influence diagram (Figure 4.8), but not populated. Note that the states of the variables when populated would capture a control model versus a stressor model. A small Bayesian network example (Figure 4.9) was presented to allow workshop participants to gain an understanding of how a Bayesian network is developed, populated and used for running some scenario analyses. The participants were able to follow the Bayesian network example with a temporary set of chosen probabilities as an example of how the wet and dry phases could be captured within the model. For example, increased groundwater drawdown and increased temperature would stress most components of the vegetation community, under especially dry conditions, potentially leading to the elimination of some species.

This example focuses on impacts of groundwater drawdown on a Melaleuca community in wet and dry phases. Note this model is for demonstration purposes only to depict how a BN would predict an outcome. The underlying probabilities are only an example. Thus the model does not portray a realistic scenario of a system.

Figure 4.9 Example of a small Bayesian network

page 40

Modelling water-related ecological responses to coal seam gas extraction and coal mining

The next steps for this model would be to: •

carefully define a specific, measurable ecological endpoint (e.g. recruitment, adult survival)



define states for all nodes



determine whether data and/or expert knowledge are available to populate the model or any of its component variables. The Bayesian network can consist entirely of empirical evidence or expert opinion or can be a mixture of these or other knowledge sources (e.g. output from another model).



do an initial parameterisation by populating the conditional probability tables from a range of inputs, including expert knowledge and field and simulated data.

4.3 Gunnedah Basin case study: conceptual model for silver perch Silver perch (Bidyanus bidyanus) was chosen as a case-study example to explore the process of developing a species-specific ecological conceptual model and its accompanying narrative table for a given species at a specific location. The chosen endpoint was the persistence of a silver perch population in the section of the Mooki River at its confluence with Quirindi Creek. A desktop survey of relevant literature (Appendix A) was used to generate the control conceptual model of the natural drivers and stressors and anthropogenic ones not related to coal seam gas extraction and coal mining (upper part of Figure 4.10), and its supporting narrative table (Table 4.4). This table lists specific hypotheses for various inferred pathways in the ecological conceptual model, and presents qualitative estimates of evidence, agreement and confidence (following the IPCC (2013) approach described in section 2.3) as a surrogate of uncertainty. Hypothetically assuming that longwall mining for coal might occur in the vicinity, a list of likely drivers and stressors associated with this form of coal mining was added to the narrative table, and expected ecological responses were hypothesised. This process guided the next step of adding the stressor model (lower part of Figure 4.10) to produce an influence diagram representing the main water-related ecological responses of silver perch to natural and anthropogenic drivers and stressors at this site. Not all stressors listed in Table 4.4 turned out to be relevant to the study area (e.g. presence of instream barriers) or pertinent to silver perch (e.g. effects of legal fishing because in NSW, fishing for this species is illegal) so they were omitted from the final ecological conceptual model. The ecological endpoint for this ecological conceptual model could be a commonly measured characteristic of silver perch such as physical condition or abundance (Figure 4.10).

page 41

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Conceptual model of hypothesised water-related ecological effects of natural drivers (yellow boxes) and anthropogenic drivers excluding those associated with longwall mining (green boxes) drivers on abundance of silver perch [upper part of figure, control conceptual model] and to the coal-mining-associated driver of longwall mining [grey box, lower part of the figure, stressor conceptual model] in the Mooki River at the Quirindi Creek confluence. Red text denotes stressors listed in Table 4.4. Thick arrows indicate the major pathways of ecological effects. The dashed lower box encloses groups of the predicted principal determinants of silver perch population size in the case study area. Note that some processes (e.g. sedimentation) are listed several times on the diagram for simplicity of representation; different stressors affecting these processes would interact.

Figure 4.10 Conceptual model for silver perch

page 42

Modelling water-related ecological responses to coal seam gas extraction and coal mining Table 4.4 Narrative table listing drivers, stressors, water-related ecological effects and hypothesised ecological effects on silver perch (SP) Relevant references and qualitative estimates of evidence, agreement and confidence (following the IPCC (2013) approach presented in section 2.3) are included. E = evidence, A = agreement, C = confidence.

Driver

Stressor

Water-related effects

Hypothesised ecological effects on SP

References

E

A

C

Climate

Air temperature

Water temperature

Extremes of water temperature over 38°C are harmful to SP.

NSW DPI (2006)

2

1

2

Spawning will occur when water temperatures exceed 23°C (21.6°C in Thurstan & Rowland 1995).

Lake (1967a); Frawley et al. (2011)

3

2

4

Warm water temperatures (20-25°C) promote invertebrate secondary production, increasing food resources for SP.

Boulton et al. (2014) and references therein

2

2

3

Landform and geology

Rainfall (volume and timing)

Runoff and river flow

Prolonged low-flow and cease-to-flow conditions, especially during periods of normal migration (OctApr), likely reduce SP dispersal and recruitment.

Mallen-Cooper et al. (1995)

2

1

2

Rainfall variability

Variability in runoff and river flow regime

Natural variability in flow regime favours native fish species such as SP, especially for migration and spawning.

DSE (2005); NSW DPI (2006) and references therein

3

3

5

Topography

River channel morphology

Flat topography and lowland meandering rivers and floodplains provide suitable habitat for SP.

Cadwallader & Backhouse (1983); Rowland (1995); NSW DPI (2006)

3

3

5

Soil features

Turbidity

SP can tolerate naturally high turbidity.

NSW DPI (2006); McNeil et al. (2013)

2

2

4

Soil features

Dissolved nutrient concentrations

Background concentrations of nutrients entering the waterway under normal conditions (i.e. pre-clearing and fertilisation) would not affect SP (e.g. via algal blooms) except when natural peaks in nutrient

NSW DPI (2006)

1

1

1

page 43

Modelling water-related ecological responses to coal seam gas extraction and coal mining Driver

Stressor

Water-related effects

Hypothesised ecological effects on SP

References

E

A

C

Boulton et al. (2014) and references therein

2

2

3

1

1

1

concentrations promote invertebrate secondary production, increasing food resources for SP.

‘Hydrogeology’

Agricultural land use (for beef cattle, dryland cropping)

Soil fertility

Allochthonous organic matter (OM) production

Fertile catchments (and riparian zones) favour inputs of allochthonous OM, which support prey of SP (food web link). Allochthonous OM as leaf litter and wood from trees in the riparian zone provide habitat for SP prey.

Soil pH

Water pH

The pH of river water under natural conditions is unlikely to fall outside of tolerances of SP adults. There may be sub-lethal effects on SP eggs and/or larvae.

Soil salinity

Water salinity

The salinity of river water under natural conditions is unlikely to exceed adult SP tolerances (salinity LC50 = 16 g/L).

McNeil et al. (2013)

2

1

2

Groundwater regime

Baseflow

Natural variability in flow regime favours native fish species such as SP, especially for migration and spawning.

DSE (2005); NSW DPI (2006) and references therein

3

3

5

Groundwater salinity

Water salinity

The salinity of river water under natural conditions is unlikely to exceed adult SP tolerances (salinity LC50 = 16 g/L).

McNeil et al. (2013)

2

1

2

Native vegetation clearance (incl. riparian zone veg.)

Sedimentation

Excessive fine sediments may smother eggs and prey of SP.

Clunie & Koehn (2001)

2

2

3

Allochthonous OM inputs*

Removal of native vegetation from catchment and riparian zone may alter the quantity and quality of allochthonous detritus entering the river, potentially

Boulton et al. (2014) and references therein

1

1

1

page 44

Modelling water-related ecological responses to coal seam gas extraction and coal mining Driver

Stressor

Water-related effects

Hypothesised ecological effects on SP

References

E

A

C

constraining food supply and habitat for SP prey.

Agricultural chemicals

Water extraction from river

Shading*

Unshaded water may be warmer but given high natural turbidity and tolerance of adult SP to high water temperature, effects may be minimal.

Boulton et al. (2014) and references therein

1

1

1

Reduced instream wood*

Reduced inputs of instream wood may reduce habitat for SP prey.

DSE (2005)

1

2

2

Reduced bank stability*

Removing riparian zone vegetation may cause banks to slump, resulting in impacts of sedimentation on SP (see earlier).

Clunie & Koehn (2001)

2

2

3

Secondary salinity

Unless salinisation is severe, river salinity is unlikely to exceed adult SP tolerances (salinity LC50 = 16 g/L).

McNeil et al. (2013)

2

1

2

Inputs of pesticides and herbicides

Inputs of agricultural chemicals are unlikely to directly harm SP but may reduce invertebrate prey populations.

Sunderam et al. (1992)

2

2

3

Inputs of fertilisers

Inputs of nutrients entering the waterway from poorly managed fertilisation would not affect SP (e.g. via algal blooms) except when natural peaks in nutrient concentrations promote invertebrate secondary production, increasing food resources for SP.

NSW DPI (2006)

1

1

1

Reduced river flow

Prolonged low-flow and cease-to-flow conditions, especially during periods of migration (Oct-Apr), likely reduce SP dispersal and recruitment.

Mallen-Cooper et al. (1995)

2

1

2

Altered flow regime

Natural variability in flow regime favours native fish species such as SP (especially for migration and spawning) over species of exotic fishes that cannot tolerate wide variation in flow regime.

DSE (2005); NSW DPI (2006) and references therein

3

3

5

page 45

Modelling water-related ecological responses to coal seam gas extraction and coal mining Driver

Instream barriers (incl. weirs, road crossings)

Translocation of species by humans

Stressor

Water-related effects

Hypothesised ecological effects on SP

References

E

A

C

Altered flooding regime

Spawning of SP appears to be related to flooding. However, it does not seem essential (Mallen-Cooper & Stuart 2003) for spawning success of SP.

DSE (2005)

2

2

3

Removal of water

Pumping from weir pools may remove SP eggs and larvae.

Gilligan & Schiller (2003)

1

2

2

Groundwater extraction

Reduced river flow

Prolonged low-flow and cease-to-flow conditions, especially during periods of normal migration (Oct-Apr), likely reduce SP dispersal and recruitment.

Mallen-Cooper et al. (1995)

2

1

2

Physical barrier

Altered flow regime

Natural variability in flow regime favours native fish species such as SP, especially for migration and spawning.

DSE (2005); NSW DPI (2006) and references therein

3

3

5

Altered flooding regime

Spawning of SP appears to be related to flooding. However, it does not seem essential (Mallen-Cooper & Stuart 2003) for spawning success of SP.

DSE (2005)

2

2

3

Impede instream movement of biota

Impeding normal migration likely reduces SP dispersal and recruitment. Barriers may also cause physical injury and/or mortality to drifting eggs and larvae of SP.

Mallen-Cooper et al. (1995); Clunie & Koehn (2001)

2

2

3

Alter sediment regime

As SP prefer sandy beaches, sediment supply to restore beaches downstream may be impaired by barriers that retain the sediments.

J. Koehn, unpubl. data cited in DSE (2005)

2

1

2

Clog waterways

Dense infestations of translocated water plants such as Typha may constrain waterways and restrict fish migration.

Boulton et al. (2014) and references therein

1

1

1

Exotic water plants

page 46

Modelling water-related ecological responses to coal seam gas extraction and coal mining Driver

Stressor

Exotic fishes

Water-related effects

Hypothesised ecological effects on SP

References

E

A

C

Compete with native water plants

Although native water plants provide habitat for the prey of SP, and SP have been found in stands of Phragmites, it is likely exotic plants could play equivalent roles as habitat. Therefore, this is likely not a serious threat to SP.

Cadwallader (1979)

1

1

1

Carp

Carp may threaten SP through competition for food resources and by increasing sedimentation through their feeding habits.

DSE (2005)

1

2

2

Gambusia

Gambusia are not considered a major threat to SP.

NSW DPI (2006)

1

2

2

Fish diseases

Exotic fishes such as carp and gambusia may be major vectors transmitting diseases to SP. Epizootic Haematopoietic Necrosis Virus (EHNV) is a particular concern in NSW (NSW DPI 2006).

Langdon (1989); Glazebrook (1995); Whittington et al. (1995); Dove et al. (1997)

2

2

3

Fishing

Commercial and recreational fishing

Removal of SP

As commercial and recreational fishing for SP are illegal, this stressor is unlikely to be serious. Illegal fishing may deplete SP stocks at low flows or in remnant pools.

NSW DPI (2006)

2

2

3

Longwall mining

Subsidence

Reduced surface runoff to river

Prolonged low-flow and cease-to-flow conditions, especially during periods of normal migration (OctApr), likely reduce SP dispersal and recruitment.

Mallen-Cooper et al. (1995)

2

1

2

Altered flow regime

Natural variability in flow regime favours native fish species such as SP, especially for migration and spawning.

DSE (2005); NSW DPI (2006) and references therein

3

3

5

Altered flooding regime

Spawning of SP appears to be related to flooding. However, it does not seem essential (Mallen-Cooper & Stuart 2003) for spawning success of SP.

DSE (2005)

2

2

3

page 47

Modelling water-related ecological responses to coal seam gas extraction and coal mining Driver

Stressor

Water-related effects

Hypothesised ecological effects on SP

References

E

A

C

Groundwater drawdown

Reduced baseflow

Prolonged low-flow and cease-to-flow conditions, especially during periods of normal migration (OctApr), likely reduce SP dispersal and recruitment.

Mallen-Cooper et al. (1995)

2

1

2

Spoil piles and processing

Input of toxicants

Toxicants associated with mine waste may be sublethal to SP eggs, larvae and adults.

1

1

1

Salinity

Unless excessive, inputs of salt from mining activity into the river is unlikely to exceed SP tolerances (salinity LC50 = 16 g/L).

2

1

2

Acidification

Unless excessive ( 2 mg/L), turbidity (‘high’), cease-to-flow conditions (‘high’). Diet: Includes zooplankton (major component, NSW DPI 2006), crustaceans, aquatic insects and algae; the proportion of algae in the diet increases with age (Clunie & Koehn 2001). Adult SP are omnivorous. Larvae are obligate planktivores (McNeil et al. 2013).

page 63

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Threats (non- coal seam gas and non-coal mining): Instream barriers can prevent upstream migrations and alter flow and temperature regimes, affecting spawning success and the survival of eggs and juveniles (Koehn & Morrison 1990). Cold water pollution (from low level outlets on dams) may lead to localised extinctions downstream of large dams if water consistently fails to reach temperatures required for spawning (23°C). Upstream migration (triggered at temperatures above 20°C) (Mallen-Cooper et al. 1995) may also be affected, as may metabolic functioning and growth, feeding, maturation and food availability (Clunie & Koehn 2001; Ryan et al. 2004). Barriers to migration may limit or prevent adults and juveniles accessing upstream habitats, and consequently prevent their dispersal and access to feeding areas and their ability to compensate for downstream drift of eggs and larvae, resulting in the local extinction of SP in affected stretches of river. Furthermore, eggs and larvae may settle out in the low flow areas immediately above barriers, subjecting them to conditions that threaten their survival. Barriers may also cause physical injury and/or mortality to drifting eggs and larvae (Clunie & Koehn 2001). River regulation and water abstraction may affect spawning success because spawning is at least partially initiated by rises in water level. Adults move upstream prior to spawning, and adult movement patterns may also be affected. River regulation and abstraction may also alter both the quality and availability of floodplain habitats such as backwaters and billabongs in which SP have been recorded (Clunie & Koehn 2001). The recruitment of SP may be more localised and opportunistic than previously believed, and fish may spawn both during inchannel flows and during large floods (Clunie, pers comm. cited in DSE 2005). The NSW DPI (2006) report has a detailed discussion of specific aspects of likely effects of altered flows. Water diversions and pumping from weir pools may remove eggs and larvae (Gilligan & Schiller 2003). Competition for food from introduced cyprinids and predation by Redfin (Perca fluviatilis) may also represent a threat. While the exact impact of Carp (Cyprinus carpio) on SP is not clear, perceived problems include competition for food resources and increased sedimentation due to the feeding habits of Carp (DSE 2005). Gambusia are not considered a major threat to SP (NSW DPI 2006). Sedimentation: Deposited sediments may be detrimental to eggs and larvae of SP, particularly in still-water and depositional habitats such as backwaters, floodplains and weir pools. If depositional events occur when SP spawn and eggs and larvae settle in still waters, reproductive success may be reduced. Deposited sediment may reduce gas exchange and inhibit development of eggs, larvae and juveniles (Clunie & Koehn 2001). Sedimentation may also affect the abundance of food items such as phytoplankton, zooplankton and insects associated with aquatic macrophytes (Clunie & Koehn 2001). It is not known whether high suspended sediment levels affect respiration or feeding in SP (DSE 2005). Instream habitat losses: Although the significance of aquatic vegetation as a habitat component for SP is unknown, it is possible that aquatic vegetation provides nursery habitat for juveniles. Aquatic vegetation also supports assemblages of aquatic insects which are in turn a food source for SP (Clunie & Koehn 2001). The significance of woody debris as a habitat component (including habitat markers, refuges from high water velocity, protection from predators, or nursery sites for larvae and juveniles) for SP is unknown (DSE 2005). However, many food items of SP (e.g. chironomid larvae and small crustaceans) are found on woody debris (DSE 2005).

page 64

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Salinity: SP appear quite tolerant of high salinity levels, although (like most fish species) early life history stages are the most sensitive (DSE 2005). The effects of sub-lethal levels of salinity on SP (including stress which may make them more susceptible to infections) are unknown, as are the effects of elevated salinity levels on food sources such as invertebrates, algae and macrophytes. Impacts on habitat complexity and quality are also largely unknown (DSE 2005). Low dissolved oxygen concentrations are considered responsible for at least two recorded fish kills associated with sedimentation (NSW DPI 2006). Agricultural chemicals: Residues of DDT and endosulfan have been recorded from fish flesh in some rivers. In toxicity tests, SP were found to be one of the least sensitive species to endosulfan (Sunderam et al. 1992). Exposure to endosulfan and chlorpyrifos reduced the critical upper lethal water temperature of SP (Patra et al. 2007 cited in McNeil et al. 2013). Degradation and destruction of riparian vegetation: The specific impacts of these processes on SP have not been determined. Generally accepted adverse effects on instream habitat include loss of shading, loss of organic inputs, increased runoff, increased erosion, streambank slumping and sedimentation (DSE 2005). Such changes may have affected SP in relation to food sources, water quality and breeding success. Disease: Very little is known about the prevalence of diseases in SP. However, three diseases and one parasite have been identified as potential threats. These are: Epizootic Haematopoietic Necrosis Virus (EHNV) to which SP has been found to be highly susceptible; Viral Encephalopathy and Retinopathy which has been demonstrated to cause mortalities of SP in trials; Goldfish Ulcer Disease; and Asian Fish Tapeworm (Langdon 1989; Glazebrook 1995; Whittington et al. 1995; Dove et al. 1997). Native fish are generally believed to become infected with these diseases following contact with introduced fish species (which act as vectors). EHNV is a particular concern in NSW as it seems to be widespread (NSW DPI 2006). Angling pressure: Unknown. Bag limit of 5, size limit of >250 mm (DSE 2005) and fishing permitted only in stocked waters in NSW. SP in NSW rivers have been totally protected from angling since 1998. Commercial fishing in NSW for the species has collapsed and a total ban has been in place since 2001 (NSW DPI 2006). Algal and cyanobacterial blooms: It is not known whether algal and cyanobacterial blooms have played a significant role in the decline of SP, or whether associated water quality problems have had less obvious, sub-lethal effects. Five key threatening processes (listed in NSW DPI 2006): Degradation of riparian zone vegetation, removal of woody debris, introduction of fish outside their normal range, instream barriers and alteration of flow regime.

page 65

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Appendix B - Bayesian network models This appendix gives a brief overview of BNs and the process for building one. Bayesian networks (also referred to as Bayesian decision networks or Bayesian belief networks) are model-based decision-support tools that are ideal for environments where considerable uncertainty exists, and for diverse problems of varying size and complexity, where disparate issues require consideration. Their graphical model structure depicts the causal or correlative relationships between key factors and final outcomes. They provide clarity by making causal assumptions explicit and are often used to model relationships not easily expressed in mathematical notation. A Bayesian network is represented as a directional graph of connected variables (henceforth called nodes), wherein directed connections from terminal (parent) nodes to a child node indicate that the parent node is having a direct influence on the child node. The BN uses conditional probability distributions under each child node to define dependencies between the interacting parent nodes and their associated categories (henceforth called states) within the nodes (Murray et al. 2014). Probabilities, which describe the strength of relationships between variables, can be defined from: empirical data (observed data, monitoring data, etc.), input data from other models, other ‘parent’ models, expert knowledge or a combination of these sources. A conditional probability distribution (often defined as a conditional probability table) is used to describe the relative likelihood of the state of a child node, conditional on every possible combination of states in the parent(s). Bayesian models are particularly useful for rapidly reviewing alternative scenarios of system change, including change in response to management actions. Consultation through workshops and via one-on-one meetings is an integral part of building a BN. Workshops can assist in developing or refining model structure, identifying and refining model inputs, and reviewing model outputs.

Bayesian probabilities Bayesian probability interprets probability as "a measure of a state of knowledge", rather than as a frequency (as in frequentist statistics). The Bayesian interpretation of probability is seen as an extension of logic that enables reasoning with uncertain statements. To evaluate the probability of a hypothesis, a prior probability (which can also be uninformative or ‘flat’) is used that can be updated with new relevant data. BNs use the network structure, combined with the junction tree algorithm, to calculate how probable certain events are, and how these probabilities can change given subsequent observations, or predict change given external interventions. A prior (unconditional) probability represents the likelihood that an input node will be in a particular state; a conditional probability calculates the likelihood of the state of a child node given the states of input parent nodes affecting it; and a posterior probability calculates the likelihood that a node will be in a particular state, given the input parent nodes, the conditional probabilities, and the rules governing how the probabilities combine. The network is solved when nodes have been updated using Bayes’ Theorem: 𝑃𝑃(𝐴𝐴|𝐵𝐵) =

𝑃𝑃(𝐵𝐵|𝐴𝐴)𝑃𝑃(𝐴𝐴) 𝑃𝑃(𝐵𝐵) page 66

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Where P(A) is the prior distribution of parameter A. After collection of data B, P(A|B) represents the posterior (new) distribution of A given the new knowledge (B). P(B|A) is the likelihood function that links A and B. BNs use the network structure to calculate the probability certain events will occur, and how these probabilities will change given subsequent observations or a set of external (management) interventions. Probabilities can be updated as new information becomes available, using Bayes’ Theorem. Being probabilistic, BNs readily incorporate uncertain information, with uncertainties being reflected in the conditional probabilities defined for linkages. When analysing risk, communication of the sources and magnitudes of uncertainties is essential. Uncertainty sources can include imperfect understanding or incomplete knowledge of the state of a system, randomness in the mechanisms governing the behaviour of the system, or a combination of these factors. Major limitations of the approach are the: •

need to express conditional probabilities as discrete nodes with categorical states



inability to incorporate feedbacks or loops in models



difficulties in eliciting expert knowledge in complex models



potential for introduction of expert bias.

Table B1 shows the strengths and weaknesses of Bayesian networks (Hart & Pollino 2009). Table B1 An overview of strengths and weaknesses of Bayesian networks Criteria

BNs

Dynamic systems (loops)

Poor

Continuous distributions

Poor

Imprecise probabilities: Exact inference *

Poor

Transparency

Poor/Good

Multiple stressors

Good

Communication tool

Good

Integration tool: Across disciplines, data and knowledge

Good

Adaptive management: Model updating

Good

Scenario analysis: What if?

Good

* Exact inference refers to probabilities not being bounded (as in Bayesian statistical approaches).

page 67

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Constructing Bayesian networks To construct BNs, the following steps are followed: 1. Selection of endpoint/s 2. Development of influence diagram (a box-and-arrow conceptual model) 3. Creation of model structure from influence diagram 4. Discretisation of nodes (assigning states) and clarification of definitions for nodes and states 5. Specification of probabilities 6. Parameterisation of parent nodes using data (optional) 7. Compilation of model 8. Model evaluation 9. Identification of knowledge gaps and priority risks 10. Alternative scenario analysis (optional)

Selection of endpoints An endpoint is the output of the model being developed and investigated. It can be in the form of an endpoint that can be measured at one level of organisation (e.g. population birth rate and mortality of an individual) that could be incorporated in a model predicting effects on an endpoint at another organisational level (e.g. availability of habitat for a species in a stream). Endpoints need to be ecologically relevant, ideally representative of how the ecosystem is structured and functions, and sensitive enough to respond to the stressors within the ecosystem (Landis et al. 2005). Points of consideration in this study include assessing the relevant scale, the availability of suitable expertise, and the overarching objectives of the project. In a previous study (Jacobs SKM 2014) exploring direct and indirect impacts of coal seam gas development on peat swamps, the endpoint selected was change in the EPBC-listed community ‘Temperate Highland Peat Swamps on Sandstone’ (Figure B1). In this case there was one endpoint. This model was then adapted to explore the impacts on individual species.

page 68

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Source: Jacobs SKM (2014).

Figure B1 Example of an endpoint with direct and indirect effects from the impacts of longwall coal mining on peat swamps

Development of the influence diagram The next step is to develop the influence diagram leading to the endpoints. In the context of this study, an influence diagram is a series of working hypotheses connected together by arrows to indicate relationships (i.e. the box-and-arrow diagram described in section 3.1). It generally portrays how an ecological system functions in its current state as well as the potential effects of stressors on the ecosystem (Landis et al. 2005). Figure 2.3 is a good example of the detail that can go into an influence diagram. It captures the ecological components that are important for persistence of eastern brook trout populations and adds the ecological effects of drilling activities for hydraulic fracturing.

Creation of model structure from an influence diagram The next step is to develop a causal structure in a BN format (based on the influence diagram), with relevant nodes (variables) and dependencies. Important criteria for inclusion of variables in BNs are that the variable is either: (a) manageable, (b) predictable, or (c) observable at the scale of the management problem. This structure can be derived from conceptual models developed during a ‘problem formulation’ phase. See for a listing and comparison of BN software.

Discretisation of nodes (assigning states) and clarification of definitions for nodes and states States can be qualitative or quantitative, categorical (e.g. absent vs. present; 0 vs. 1) or discrete (continuous data can be represented as a set of discrete intervals), where numerical ranges are assigned (e.g. 0 to 3, 3 to 10). Nodes can be discretised according to guidelines, existing classifications or percentiles of data. The number of states is unlimited but as it increases, so does the number of probabilities to be estimated. Nodes and states need to be clearly defined to facilitate interpretation of the network.

page 69

Modelling water-related ecological responses to coal seam gas extraction and coal mining

Specification of probabilities After defining node states, the strengths of relationships between nodes need to be described. A probability distribution is required to describe the relative likelihood of the state of each child variable, conditional on every possible combination of parent variables. This relationship is defined with a conditional probability table (CPT). If a node has no parents, it can be described probabilistically by a marginal probability distribution. Figure B2 shows how CPTs work within a simple BN, where nodes A and B (parent nodes) represent the causal factors of node C (child node). This example was created with the programming shell Netica (http://www.norsys.com).

Figure B2 Bayesian network – simple example.

All nodes are discretely binomial, with the states defined as either true or false but the probability distributions unspecified. The parent nodes A and B can be defined by marginal probabilities, but the state probabilities for the child node C are conditional on how the states of A and B combine. The entries in a CPT can be ‘parameterised’ by a range of methods, including directly observed data (monitoring, research), probabilistic or empirical equations, results from model simulations, elicitation from expert knowledge, or any combination of these methods. The methodology used to parameterise variables and the sources of information for each variable are documented for each model. In Figure B3, direct expert elicitation is used.

Figure B3 Conditional probability table

Elicitation often takes the form of scenarios, which are described as they appear in the table (e.g. given A is true and B is true, what is the probability that C is true (here 100%)?) The

page 70

Modelling water-related ecological responses to coal seam gas extraction and coal mining

elicitation process can represent probabilities as bounds to capture uncertainties in knowledge. The method used for probability generation must be rigorously documented, including any assumptions and limitations. When the probability distributions of each node have been defined, the network can be ‘compiled’ or ‘solved’. After evaluation tests, the BN is complete and can be used for scenario analysis.

Parameterisation of parent nodes with data (optional) The quality of knowledge (Table B2) can also vary and this has implications on the robustness of an assessment. Probabilistic relations can be specified from data (organised as case files). Data sources can be entered into the network as a series of ‘cases’. Cases can represent data collected during a monitoring program or as part of a research study. Data can be used to specify probability distributions, via learning algorithms in Netica (e.g. the Expectation Maximisation or EM algorithm). Table B2 Narrative quality ranking for different inputs to the risk analysis BN Rank

Statistical analyses

Processbased model

Database

Literature

Expert

High

High calibration with data (≥95%)

Comprehensive validation using independent data set

Large samples, multiple sites and times

Published in peer reviewed forum

Multiple experts – high consensus

Moderately well calibrated with data (90 to