DUNDONNELL WIND FARM

BROLGA ASSESSMENT

Trustpower Australia Pty Ltd

Suite 5 61 - 63 Camberwell Road, Hawthorn, VIC 3123 P.O. Box 337, Camberwell, VIC 3124 Ph. (03) 9815 2111 Fax. (03) 9815 2685 October 2014 Report No. 9184 (4.12) Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

CONTENTS

1. EXECUTIVE SUMMARY ...... 1 2. INTRODUCTION ...... 11 2.1. Policy and planning for wind farms ...... 11 2.2. This report ...... 11 2.3. Species biology...... 12 2.3.1. Species description ...... 12 2.3.2. Species status ...... 12 2.3.3. Brolga distribution and movement in Victoria...... 13 2.3.4. Brolga population size over time ...... 13 2.4. The Brolga Guidelines ...... 14 2.4.1. Investigating Brolga movements ...... 18 3. LEVEL ONE ASSESSMENT ...... 19 3.1. Step One: Undertake desktop studies into known and potential habitat ...... 19 3.1.1. Methods ...... 19 3.1.2. Results ...... 21 3.2. Step Two (Part 1): Field inspections ...... 23 3.2.1. Methods ...... 23 3.2.2. Results ...... 26 3.3. Step 2 (Part 2): Local community consultation and Brolga use survey ...... 28 3.3.1. Overview ...... 28 3.3.2. Methods ...... 28 3.3.3. Results and discussion ...... 31 3.4. Triggers for a Level Two Assessment...... 45 4. LEVEL TWO ASSESSMENT ...... 46 4.1. Overview of methods ...... 46 4.2. Breeding season surveys...... 50 4.2.1. Survey methods ...... 50 4.2.2. Survey results ...... 54 4.3. Flocking season surveys ...... 58 4.3.1. Survey methods ...... 58 4.3.2. Survey Results ...... 60 4.4. Migration season surveys ...... 67 4.4.1. Survey methods ...... 67

Page | i Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

4.4.2. Migration season results ...... 70 4.5. Brolga utilisation survey ...... 70 4.5.1. Methods ...... 70 4.5.2. Results ...... 71 4.6. Transmission line ...... 76 4.6.1. Methods ...... 76 4.6.2. Results ...... 76 4.7. Summary of Results ...... 79 4.8. Triggers for a Level Three Assessment ...... 82 5. LEVEL THREE ASSESSMENT ...... 84 5.1. Step One: Avoiding and mitigating potential impacts ...... 85 5.1.1. Background ...... 85 5.1.2. Breeding sites ...... 85 5.1.3. Flocking sites ...... 87 5.2. Step Two: Collision risk modelling ...... 92 5.3. Step Three: PVA Model ...... 96 5.4. Step Four: Compensation to achieve zero net impact ...... 97 5.4.1. Transmission line marking ...... 98 5.4.2. Protection and enhancement of breeding sites ...... 98 5.5. Cumulative Impacts ...... 100 5.6. Summary and Conclusions ...... 101 6. REFERENCES ...... 107

TABLES Table 1: Brolga population estimates, south-west Victoria ...... 13 Table 2: Three level assessment of wind farm impact on Brolgas: current investigation 15 Table 3: Level 1 Assessments: triggers and steps ...... 19 Table 4: Summary of 2009 wetland quality assessment and Brolga use data ...... 26 Table 5: Summary of 2011 wetland quality assessment and Brolga use data ...... 26 Table 6: No. property sub-areas with different levels of Brolga activity vs land use ...... 33 Table 7: No. property sub-areas with different levels of Brolga activity vs land type ...... 33 Table 8: Level 2 Assessments: triggers and steps ...... 46 Table 9: Summary of field methods and survey periods: Level Two Assessment ...... 47 Table 10: Summary of known breeding records and likelihood of future breeding attempts at wetlands ...... 54

Page | ii Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Table 11: Flight heights recorded during the 2013 flocking season survey ...... 66 Table 12: Results of the 2012 Brolga utilisation survey ...... 73 Table 13: Results of the 2013 Brolga Utilisation survey ...... 74 Table 14: Historical/recent Brolga breeding/flocking records for transmission line RoI 76 Table 15: Historical and recent Brolga observations in wind farm RoI and nearby area . 79 Table 16: Level 3 Assessments: triggers and steps ...... 84 Table 17: Flight speed of Cranes ...... 95 Table 18: Results of Brolga CRM ...... 96 Table 19: Results of Brolga PVA ...... 97 Table 20: Summary of Brolga risks, responses and outcomes for the Dundonnell Wind Farm ...... 102

FIGURES Figure 1: Wetland quality results for 2011 ...... 27 Figure 2: Historical Brolga count density (smoothing length 2500 m) ...... 34 Figure 3: Reported land use on farms in the broader region ...... 35 Figure 4: Reported land type on farms in the broader region ...... 36 Figure 5: Breeding occurrence on farms; as reported by landowners ...... 37 Figure 6: Frequency of large brolga groups/flocks on farms; as reported by landowners 38 Figure 7: Frequency of small brolga groups/flocks on farms reported by landowners .... 39 Figure 8: Reported land use on farms; detail of wind farm area ...... 40 Figure 9: Reported land type on farms; detail of wind farm area ...... 41 Figure 10: Breeding occurrence on farms reported by landowners: wind farm area ...... 42 Figure 11: Frequency of small brolga groups/flocks reported by landowners: wind farm area ...... 43 Figure 12: Frequency of large brolga groups/flocks reported by landowners: wind farm area ...... 44 Figure 13: Aerial survey transect lines and Brolga sightings ...... 53 Figure 14: Male Brolga displaying in a flock at wetland 585 ...... 61 Figure 15: Observed Brolga flight paths during flocking seasons ...... 63 Figure 16: Wetlands surveyed during the Brolga migration season survey 2011 ...... 69 Figure 17: Location of utilisation survey points ...... 75 Figure 18: Brolga records along proposed transmission line route ...... 78 Figure 19: Historical and recent Brolga records ...... 83 Figure 20: BL&A method for establishing turbine-free buffers around breeding sites .... 88 Figure 21: Constraints map (turbine-free buffers)...... 91

Page | iii Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

APPENDICES Appendix 1: Habitat assessment during the 2011 flocking season investigation ...... 109 Appendix 2: Flight paths recorded during 2012 flocking survey...... 113 Appendix 3: Home Range mapping methodology comparison report ...... 115 Appendix 4: Brolga activity spatial analysis – Symbolix ...... 116 Appendix 5: CRM Report for proposed Dundonnell Wind Farm and transmission line .. 117 Appendix 6: PVA Model Report by Dr McCarthy ...... 118

Page | iv Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

1. EXECUTIVE SUMMARY This report assesses the impact of the proposed Dundonnell Wind Farm on the FFG Act-listed Brolga. This assessment has been undertaken consistent with the requirements of the ‘Interim guidelines for the assessment, avoidance, mitigation and offsetting of potential wind farm impacts on the Victorian Brolga population’ (DSE 2012a), referred to hereafter as the Brolga Guidelines.

Overview Consistent with the Brolga Guidelines, this investigation has fundamentally informed the layout of the proposed Dundonnell wind farm through the definition of turbine-free buffers based on a number of independent lines of evidence about Brolga activity in the affected area. These buffers have been designed to reduce risk to the Brolga from collision with turbines and indirect disturbance to acceptable and manageable levels. During over four years of detailed Brolga studies in the area, a combination of accepted methods was used to gather information on the occurrence of the Brolga on and around the proposed wind farm site to a distance of up to 10 kilometres. This information covered both breeding and flocking seasons and was combined with historical information from a range of sources, including extensive consultation with local landholders. The investigations found that the region supports a small number of breeding pairs that use a variety of wetland sites, a small proportion of which occurred within a sensitive distance of the wind farm (i.e. 3.2 km) and were taken into account in defining turbine-free buffers. Flocking activity in the region was more consistent and involved larger numbers of birds. The investigations were able to document the movements and habitat choices of the Brolga during the investigation period and historically based on landholder observations and historical records of occurrence. Turbine-free buffers were defined to avoid significant risk to the species from collision and disturbance. In response to this extensive information, the wind farm layout has evolved significantly. Initially, a much larger proposal was tabled for consideration but based on initial advice from BL&A on the historical occurrence of breeding and flocking Brolgas, a smaller area was targeted for more detailed design and feasibility study that appeared to lack breeding and flocking records. The project boundaries and the turbine layout within this reduced extent have also been subject to considerable amendment. This has included the removal of turbines from identified flocking site and breeding site home range buffers where Brolgas have been observed during the detailed investigations reported herein or are predicted to occur consistently in the future based on a detailed understanding of the species’ habitat choice and historical activity. The vast majority of the potential impacts of wind farms on birds, including Brolgas, can be avoided through informed, careful siting of turbines to avoid significant risk of bird-turbine interaction. The extensive investigations of Brolgas

Page | 1 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) for the Dundonnell Wind Farm have informed the wind farm layout, which has been adjusted to avoid significant risk to the species. Consistent with the Brolga Guidelines, the proponent will support a program of breeding site management that will enhance breeding productivity of the species to the extent that the project will have zero net impact on the Victorian Brolga population.

Scope of Investigation The full investigation area for this assessment encompassed the proposed wind farm site and transmission line corridor, as well as a 10 km zone around these areas (referred to as the ‘radius of investigation’ (RoI)). Site-specific investigations of Brolga breeding and flocking behaviour were undertaken for any breeding or flocking events over the four-year study period within 3.2 km and 5 km respectively of the proposed wind turbines. The results of this Brolga assessment, combined with extensive historical information, are presented in accordance with the three-step approach prescribed by the Brolga Guidelines, namely: . Level One Assessment: Initial Risk Assessment . Level Two Assessment: Impact Assessment . Level Three Assessment: Design, Mitigation and Offset Although the investigations commenced in 2009, before the Brolga Guidelines were issued, they commenced following the release of the draft Brolga Guidelines (DSE 2009) and the AusWEA (2005) protocols, the latter of which the 2012 Brolga Guidelines explicitly acknowledge as the appropriate assessment framework to use. During the Level One Assessment, historical records of the Brolga in the RoI were compiled and analysed. To further assess the likelihood of interaction between local Brolgas and the proposed Dundonnell Wind Farm site, specialised analytical support was provided by Dr Elizabeth Stark from Symbolix Pty. Ltd. Existing records from a range of public database sources and landholder questionnaire surveys were used to compile historical information on the activity of the species in the wind farm RoI, to determine if there was a general spatial use pattern across the RoI and on the wind farm itself, and to see whether Brolga activity correlated with historic and current land use patterns in the wind farm RoI. Based on the findings of the level one assessment, the extent of the proposed wind farm was significantly reduced and turbine layouts were proposed for a more limited area where the data indicated Brolga activity was comparatively low. This area was then subject to the more detailed, Level Two assessment. The Level Two Assessment comprised breeding, flocking and migration surveys of Brolga between 2009 and 2014 undertaken by BL&A staff and two local residents, who were trained by and worked under the supervision and direction of BL&A. The survey effort from 2011 to 2013 by the field staff totalled 2,075 hours, and a total of 16,099 km was driven in the RoI in search of Brolgas over this period implementing the methods required by the Guidelines and described below. In addition, through their local network of landholder contacts, field staff were regularly informed of Brolga activity, including breeding and flocking events,

Page | 2 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) by landholders in and near the RoI. Fortnightly roaming surveys were undertaken, informed by landowner feedback and local knowledge of wetland condition to search for breeding and flocking birds, from 2011 to 2014. Any breeding or flocking event within 3.2 km or 5 km respectively triggered more detailed observations, including site-specific flight path mapping, of the birds concerned. Brolga utilisation surveys based on repeated point counts were undertaken during the flocking seasons in 2012 and 2013, when flocking events occurred within the RoI and the maximum numbers of Brolgas occurred in the region. These point counts were used to derive the number of flights of concern over the proposed, reduced wind farm site for the purpose of collision risk modelling. Roaming surveys were used to determine the spatial distribution of Brolgas on and around the proposed wind farm site. The combination of estimated number of flights and geographic distribution of those flights generated inputs for the collision risk model in the Level 3 assessment. The results of the level two assessment informed detailed mitigation measures, and in particular wind farm layout refinements, as described below. The Level Three Assessment involved four steps to identify suitable mitigation measures for the proposed development to achieve a zero net impact on the Victorian Brolga population: (1) avoid or mitigate potential impacts through identification of turbine-free buffers around Brolga breeding and flocking sites, (2) collision risk modelling to quantify residual impacts and risk (CRM), (3) Population Viability Analysis (PVA) model to identify population-wide consequences of residual impacts; and (4) definition of compensation and offset options to cancel our residual impacts and ensure a zero net impact outcome, as required by the Guidelines. Impacts were avoided by adjusting the positions of wind turbines to avoid turbine free buffers, defined based on the detailed Level Two assessment results combined with the historical information assembled for the Level One assessment. This resulted in a final wind farm layout that further avoided areas of estimated higher Brolga activity based on spatial distribution information. The results of the investigations undertaken during the three levels of assessment are summarised below.

Analysis of Brolga activity and land use/type Brolgas were noted by landholders more frequently in aquatic/wetland or arable (cultivated) areas than on stony ground. Additionally, infrequent or no Brolga use was consistently reported from stony land used for grazing. This pattern is mirrored by the historic data showing that Brolgas prefer aquatic/wetland habitat and are very rarely observed in drier, stony areas - which form the majority of the proposed wind farm site. This pattern was corroborated by subsequent, Level Two assessment investigations and informed the initial reduction in extent of the proposed wind farm.

Page | 3 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Breeding sites The historical databases, landholder records and observations prior to 2009 when this study commenced indicated that 39 wetlands have been used by Brolgas for breeding within the wind farm RoI. Three wetlands are unlikely to be used in the future due to permanent draining. During the Level Two investigations, eight additional breeding sites were found: This increased the total number of known breeding sites in the RoI to 47. Ten breeding sites were located within three km of the proposed transmission line. Five breeding sites within 3.2 km of the wind farm turbines in the final turbine layout assessed here (wetlands 118, 137, 254, 533 and 602) were used by Brolgas for breeding during the investigation period1. Two of these breeding events continued for a full breeding season and were completed successfully; the remaining attempts were not successful due to the loss of eggs or chicks before completion. Breeding attempts greater than 3.2 kilometers from the wind farm site (the distance within which turbine free buffers need to be defined) were not subject to such detailed observations and were monitored as part of the fortnightly roaming surveys. The number of active nests at breeding sites within the RoI is likely to vary between years depending on the condition of wetlands within the RoI and elsewhere. Although numerous pairs attempted to build nests and incubate, it was observed over the duration of investigations that no more than five pairs laid eggs and undertook incubation at any one time within the RoI. No breeding sites have historically been recorded within the final proposed wind farm turbine layout. One breeding attempt close to the proposed wind farm site between 2011 and 2014 (within 3.2 km) was successful in wetland 602. Flight path mapping provided information on the extent of movements of breeding birds when they were present but the short duration of most (unsuccessful) breeding attempts limited the amount of site-specific data that could be collected. The five breeding sites were surveyed over 150 hours between 2011 and 2013. Less than 20 flights away from these wetlands were recorded during these observations. During investigations, no more than two breeding pairs of Brolga were found at any time using breeding sites within 3.2 km of the proposed wind farm.

Flocking surveys The review of the databases and landholder information in the Level One assessment and the fieldwork for the Level Two assessment indicated that 20 traditional flocking sites were located within or near the wind farm RoI. Historical one-off flocking events had occurred at an additional 24 wetland sites. At a further eight non-wetland sites, flocks have been observed either once or more regularly but overnight roosting by these birds would not occur and they would return to a traditional or one-off flocking wetland site to roost for the night. One

1 The Brolga Guidelines recommend a 3.2 km radius turbine-free buffer from breeding sites.

Page | 4 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) unconfirmed traditional flocking site from the 1980s and early 1990s is located within three km of the transmission line (wetland 301). During the 2011 surveys, the RoI was found to support two distinct flocking site groups (eastern and western), including behaviour consistent with both traditional and one-off flocking activities. Birds from the western flocking site used habitats situated immediately west of the proposed wind farm, whilst birds from the eastern flocking site used habitats at least 3.3 km from the proposed wind farm. At any one time, the total surveyed numbers at all nearby flocking sites ranged between 10 and 52 individual Brolgas, while the most frequently sighted number of individuals was 29. The maximum number of birds recorded at the western site (which was within five kilometres of the wind farm) on one occasion was 32 birds, compared with 20 at the eastern site. In 2012, two flocking events were found: northern and southern. The northern flock comprised a maximum number of 47 adults and was found over six km from the proposed wind farm site (wetlands 240, 522)). The infrequency of records and lack of prolonged records during the flocking season, and the lack of historical flocking records at these wetlands indicated that this was most likely a one-off flocking event. The southern flock of 32 birds consistently used wetlands over four and up to seven km from the nearest proposed turbine (wetlands 248, 252 and 235). From April to August 2013, observations were undertaken of flocking Brolgas at a flocking site found by local observers. This flock was concentrated in and around Lake Sheepwash, 2.9 km south-east of the nearest proposed turbines of the final layout with a maximum number of 46 birds observed. They foraged in country to the south of this flocking site. In March 2014, a flock of 12 Brolgas was observed south-east of the wind farm They roosted in the salt lake south of the Woorndoo-Dundonnell Road (wetland 139). The birds were feeding on grain trails (stock feed) near wetlands 313 and 602, to the south-east and east of the current wind farm layout and for a period within the proposed turbine layout on grain put out as stock feed. Grain feeding of stock in this area will cease after wind farm construction commences.

Brolga utilisations surveys During the 2012 and 2013 Brolga flocking seasons, a Brolga utilisation survey was undertaken on the wind farm site to characterise Brolga activity across the project area during periods of higher Brolga numbers in the RoI. The purpose of these surveys was to generate movement rate information for input to the collision risk model (see later). In 2012, seventeen Brolga observations (14 during 40 hours of formal counts and three outside the formal survey periods) were recorded during the utilisation survey. Of those, seven were flight observations (four formal, and three outside formal observation periods). Over ten observations of Brolgas counted were observed in arable (cropping) land. The sightings involved birds neither breeding nor flocking. The majority of flight observations were distributed to the north-west of the final turbine layout in arable land near Chinaman’s Swamp that was part of a turbine-free buffer area.

Page | 5 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

In 2013, 11 Brolga observations were recorded during the 40 hours of formal utilisation survey. Of those, six were flight observations. Over half of Brolga observations were observed in arable land and the remainder of observations were in stony outcrops, on a spring or in unknown habitat. The majority of flight observations were distributed in the southern section of the study area in all types of habitat, but mostly in arable land or in a spring fed drainage line. The information gathered from historical data sources, landholder consultation and detailed breeding and flocking season field studies over a four year period provide a comprehensive and robust set of data to inform the assessment of potential wind farm impacts, the development of mitigation measures and the assessment of residual impacts and associated compensation and offset measures. The information is sufficient to inform application of the Brolga Guidelines and to achieve an outcome for the project that results in a zero net impact on the Victorian Brolga population.

Avoiding and mitigating potential impacts Turbines and associated infrastructure have been excluded from turbine free buffers, which have been defined by combining breeding and flocking site home range buffers and an additional 300 metre disturbance buffer, consistent with the Brolga Guidelines. Breeding site home range buffers have been defined based on predicted habitat use around all known breeding sites within 3.2 kilometres of the final turbine layout. The limited Brolga breeding activity within 3.2 km of the proposed wind farm limited the number of observations of Brolga movements around active nests at breeding sites. Therefore, a habitat-modelling approach was adopted based on observations of the movements of Brolgas around breeding sites in the past. The validity of the habitat modelling approach was tested on the flight path mapping of breeding Brolgas at Dundonnell and by comparison with detailed home range mapping results at the Penshurst Wind Farm. This indicated that turbine-free buffers based on the habitat modelling approach would be effective in avoiding the breeding site home range buffers and associated 300 metre disturbance buffer at any of the five breeding sites within 3.2 kilometres of the final turbine layout. This is consistent with the objective in the Guidelines of not disrupting breeding outcomes for the species. For this project, a suitable flocking site home range buffer has been set in a way that avoids disrupting Brolgas moving from flocking sites to surrounding foraging habitats, and indirect effects on them, such as disturbance. Turbine-free buffers for flocking sites have been identified based on the results of site-specific field observations of Brolga behaviour while using flocking sites, as well as historical information on distribution from existing data and habitat choice information from landholder surveys. The buffers for flocking sites were defined by: . Excluding from development any area where flocking Brolgas were observed foraging except where it is clearly and directly related to an on-site land use practice, such as grain feeding of stock, that can be avoided in the future;

Page | 6 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Excluding any area where regular flights by flocking Brolgas were observed except where it is clearly and directly related to an on-site land use practice, such as grain feeding of stock, that can be avoided in the future; . Excluding any area where the land type was arable and where there had been previous sightings of Brolgas; and . Adding the 300 m disturbance buffer to these areas as required in the Brolga Guidelines.

Collision risk modelling (CRM The CRM predicts that the preferred, final wind farm layout will lead to a long term, annual average of between 0.09 (99% avoidance) and 0.91 (90% avoidance) Brolga collisions with wind turbines. The 98% avoidance rate result is considered to represent the rate most likely to occur in nature based on a review of current literature. However, erring on the conservative side of the collision risk estimate, the 95% avoidance rate result is presented in detail as it has been used in the past to inform wind farm development decisions in Victoria. Under this scenario, on average, the final wind farm layout would result in one bird being lost from the Victorian Brolga population every two years.

Population Viability Assessment (PVA) The PVA is a widely accepted modelling method that attempts to predict the trajectory of the population of a plant or animal and calculates the quasi- extinction probability. The purpose of the Victorian Brolga PVA is to model population scenarios given impacts on Brolga survival rates of wind farm development (using collision risk modelling results) and to set targets for compensation and offset measures to ensure zero net impact on the Victorian Brolga population. Based on an expected minimum population of 809 birds (the expected population without the impacts of the Dundonnell wind farm project), the PVA predicts that after 25 years (the usual life of a wind farm project), the impacts of the final turbine layout will result in a population size of between 808 (99% avoidance rate) and 795 birds (90% avoidance rate), a reduction of between one and 13 birds over 25 years compared with baseline conditions. The predicted result for 95% avoidance rate, which is considered the most realistic conservative collision rate, is a population size of 805 birds, representing a reduction of four birds in the Brolga population over 25 years.

Compensation to achieve zero net impact Measures to protect or improve breeding sites will be implemented and will include: . Restoration of the natural flooding regime of wetlands by closing drains. . Increasing inundation frequency of breeding wetlands through artificial flooding.

Page | 7 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Creating new potential breeding habitat by damming or modifying existing wetlands or dams, including blocking previously installed drains. . Management of wetland vegetation condition through controlled grazing (or stock removal) to improve suitability as a breeding site. . Addition of nesting material to potential breeding wetlands to facilitate nest building. . Fox control at key breeding habitats. The number of Brolgas that need to be added to the south-west Brolga population over the 25 year life of the project to meet the required zero net impact objective of the Brolga Guidelines was estimated. With an average of 0.5 extra brolga deaths per year with a 95% avoidance rate, this translates to the loss of an extra 13 Brolgas over 25 years, resulting in a net decline in the population over that period of four birds. Therefore, mitigating this loss by increasing breeding success would require an extra 13 birds being raised to adulthood over the 25 year period, thereby preventing the net decline of four birds. The difference between the population reduction (four birds) and the required compensation number (13 birds) is due to the loss of some of the 13 added birds from the combination of natural and wind farm related mortality during the 25 year project life. Managing Brolga breeding sites to boost the production of young, thereby adding more birds to the state’s population is considered feasible as it has been achieved before and it has occurred naturally during wetter years when breeding habitat quality and availability has improved. It is possible based on the workings of the PVA model to set a priori a compensation target that can be translated readily into a target for the production of fledged birds. This target, based on the mortality rates in the PVA, is the production of an additional fledged young every second year from pro-actively managed Brolga breeding sites. Given that not all breeding sites produce young every year, a number of breeding sites will have to be managed to meet the required target. The number required each year will have to be determined adaptively based on the success of breeding outcomes at managed breeding sites over time with reference to the above target. The proponent has committed to implementing, for the life of the project, a program of Brolga breeding site enhancement and management to boost the production of fledged young to the extent required.

Conclusions The application of the Brolga Guidelines to inform the layout of the proposed Dundonnell Wind Farm and additional management commitments by the proponent has resulted in an assessment that provides assurance that the project will have zero net impact on the Victorian Brolga population. The assessment has reduced the possible impacts of the Dundonnell wind farm on Brolgas to acceptable and manageable levels as it has: . resulted in significant reductions in the extent and refinements to the layout of wind turbines in the final layout by the adoption of well-researched turbine- free buffers around all known breeding and flocking sites;

Page | 8 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. provided estimates of residual population impacts through collision risk modelling and population viability assessment; and . identified a corresponding scale of residual impact compensation, though enhancing the breeding success of the species for the life of the project that is readily achievable.

Acknowledgements This report has benefitted greatly from a peer review by Ian Smales and Dr Joanne Potts of Biosis Pty Ltd and reviews by the Department of Environment and Primary Industries (DEPI). Their time and feedback is gratefully acknowledged.

Page | 9 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Glossary

AusWEA Australian Wind Energy Association DEPI Department of Environment and Primary Industries (Vic.) DSE (former) Department of Sustainability and Environment (Vic.) EPBC Act Environment Protection and Biodiversity Conservation Act (C'th) FFG Act Flora and Fauna Guarantee Act (Vic.) RoI Radius of investigation; defined in the Brolga Guidelines as below:

Here used as:  wind farm RoI: the area within 10 km of the wind farm boundaries; where 'RoI' is used without description in this document it refers to wind farm RoI; and  transmission line RoI: the area within 10 km of the new transmission line (note that slightly over the first 10 km of the latter area will be within the wind farm RoI RSA Rotor Swept Area RSA height Above 23 m and below 165 m above the ground VBA Victorian Biodiversity Atlas

Page | 10 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

2. INTRODUCTION

2.1. Policy and planning for wind farms The policy and planning guidelines for wind farms in Victoria require that the potential impacts of wind farms on species listed under the Commonwealth Environment Protection and Biodiversity Conservation Act 1999 (EPBC Act) or the Victorian Flora and Fauna Guarantee Act 1988 (FFG Act) be assessed (DPCD 2012). One such species is the Brolga, which occurs in the region within which the proposed Dundonnell Wind Farm is located. The species is listed as threatened under the FFG Act and, consistent with Clause 12.01 of the State Planning Policy Framework and DPCD (2012), planning authorities must consider the impacts of wind farm developments on this species before making decisions on permit applications. In February 2012, the Department of Environment and Primary Industries (DEPI, formerly DSE) issued a document titled ‘Interim guidelines for the assessment, avoidance, mitigation and offsetting of potential wind farm impacts on the Victorian Brolga population.’ (DSE 2012a), referred to hereafter as the Brolga Guidelines.2 These Brolga Guidelines provide a three stage method for assessing and mitigating the potential impacts of wind farms on the species.3

2.2. This report In accord with the Brolga Guidelines, this report assesses the potential impact of the proposed Dundonnell Wind Farm on the Brolga, which occurs in the region, and proposes relevant mitigation actions. Work commenced on this Brolga assessment in 2009, three years before the final 2012 Brolga Guidelines were issued. However, as those guidelines parallel the process set out in the AusWEA (2005) document ‘Wind farms and birds: interim standards for risk assessment’, and this assessment followed the steps in those earlier foundational standards, the work reported herein conforms with the requirements of the Brolga Guidelines. The focal investigation area encompassed the proposed wind farm site and the new transmission line corridor, as well as a 10 km zone around these areas, referred to as the ‘radius of investigation’ (RoI), as defined in the Brolga Guidelines (p. 13). Less frequent observations were also made beyond this 10 km distance, to provide context for nearby observations within the RoI, and these are also reported. The results of this Brolga assessment are presented in accordance with the three- level approach prescribed in the Brolga Guidelines, namely:

2 An earlier form of this document was available before that date (DSE 2009). 3 They were prepared with input from a Scientific Advisory Panel to a joint government – industry working group on Brolgas and wind farms in south west Victoria, of which Brett Lane, principal consultant and project director for the Dundonnell Wind Farm flora and fauna impact assessment reported herein, was a member.

Page | 11 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Level One Assessment: Initial Risk Assessment; . Level Two Assessment: Impact Assessment; and . Level Three Assessment: Design, Mitigation and Offset. In this report, the methods and sources of information are described followed by the results and conclusions for each level of investigation. The biology of the Brolga in Victoria and the Brolga Guidelines are discussed briefly below.

2.3. Species biology

2.3.1. Species description The Brolga (Grus rubicunda) is listed as a threatened species under the FFG Act. Brolgas belong to the family Gruidae (cranes), of which two species (including the Brolga) occur in Australia (Marchant and Higgins 1993). Cranes are generally large-bodied, long-legged and long-lived, with Brolgas being very similar to other cranes in general ecology and biology. Adults can range in weight between four and eight kg, and are approximately one m tall with a wingspan of two m. During the non-breeding season, Brolgas can form large flocks (occasionally as large as 200 birds) but typically are seen in small groups (10 - 20 individuals). Breeding pairs can form long-term bonds and, if one of the pair dies, the remaining individual can take several seasons to find another mate (Marchant and Higgins 1993). Typically, pairs only produce one or two offspring per breeding season and therefore recruitment into the population is low. The Brolga’s annual cycle is divided into two principal parts: . the breeding season, from July to December, during which territorial pairs nest in shallow freshwater wetlands that are often ephemeral, holding water reliably only in winter and spring; and . the non-breeding (or flocking) season, from December to June, when Brolgas disperse from drying breeding wetlands to larger, often permanent wetlands to congregate with others to form flocks that roost at the wetland and move out to forage in adjacent terrestrial and wetland habitats. (DSE 2012a) In between dispersing for breeding and returning to non-breeding flocking sites, the Brolga moves about the landscape during two migration periods that can overlap with the months above.

2.3.2. Species status The Brolga is a secure species nationally, numbering in the tens of thousands across northern Australia (Marchant and Higgins 1993). However, in Victoria the range of the Brolga has contracted since European settlement as a consequence of wetland drainage, loss of habitat due to agricultural development and predation of eggs and young by the introduced Fox. Its former range included northeast Victoria, Gippsland and the wetlands of the Melbourne region. Currently birds are found in the south-west and in the north of the state in parts of the Murray River basin (Du Guesclin 2003).

Page | 12 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

2.3.3. Brolga distribution and movement in Victoria The distribution of the Brolga in the main part of its Victorian range, the south- west, varies seasonally. In the breeding season adult pairs disperse to small and moderately sized seasonal or semi-permanent wetlands to breed as territorial pairs. At this time, small numbers of non-breeding birds can form flocks on larger wetlands. In the flocking season, birds congregate in larger wetlands as the smaller, seasonal wetlands dry out over summer. Brolga movements in south-west Victoria are not yet completely understood. Seasonal movements, referred to as migration movements, occur in south-east Australia between flocking and breeding sites. Local movements can also take place when birds are moving between roosting and feeding sites, in response to rainfall and the availability of food. Long distance movements may take place in very dry years and populations may move from dry inland wetlands to wetlands associated with the Murray River (Marchant and Higgins 1993). In very wet seasons, birds may remain at breeding sites throughout the year and not move to flocking sites. Therefore, Brolga movements and distribution are heavily dependent on climate and foraging opportunities. Consistent Brolga flocking sites in south-west Victoria that account for a significant proportion of the population occur in the locations listed below, based on information compiled by Sheldon (2004) and provided by the then-DSE: . the Grampians region; . Strathdownie; . Cressy; . Streatham (mainly on Lake Wongan and in the Skipton area); . Hamilton, Dunkeld and Penshurst areas; . Edenhope area; . Toolondo; . Willaura and Stavely areas; and . Darlington area (this last area is relevant to the current wind farm proposal).

2.3.4. Brolga population size over time The 1984 estimate of the Victoria brolga population was 600-650 birds, with approximately 550-600 of these birds (c. 92%) in south-west Victoria (Arnol et al. 1984). This and subsequent estimates are noted in Table 1. Table 1: Brolga population estimates, south-west Victoria

Month/ Est. no. No. ≤2 % ≤2 Same Source yrs. old yrs. old day year counts http://bird.net.au/bird/index.php?title=Brolga#20 4/2013 907 - 17 Yes 13_Update_- _provided_by_Richard_Hill.2C_DEPI.2C_Casterton (viewed August 2014) 2012 448 - 16 Yes http://bird.net.au/bird/index.php?title=Surveys#S

Page | 13 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

urvey_results_summary (viewed January 2014) 2011 250 50 20 No

2010 401 39 10 No

2004 675 - - No Sheldon (2004)

2002 402 - - No P. Du Guesclin, DSE (pers. comm., 2007)

1984 550- - - No Arnol et al. (1984) 600

The April 2013 count reported above was organised by DEPI and was conducted at Dundonnell, Penshurst, Willaura, Strathdownie, Lake Bolac, Streatham, Boole Lagoon (S.A.) and Lake Wongan. The largest flock recorded was 320 birds at Strathdownie. The counts were undertaken systematically by having different sites counted on the same day across the state, to avoid re-counting flocks. The 2012 count was also conducted at multiple sites on the same day. Earlier non-simultaneous counts (from the 1980s to 2011) are not be directly comparable to the counts from 2012 and 2013, as counts conducted over multiple days may result in over-estimation of the number of birds due to multiple counting of individuals. That so many young were observed in flocks compared with the previous drought years (BL&A, unpubl. data) indicates how effective improving the availability of breeding habitat can be in increasing the Brolga population. Recent counts show a substantial increase in successful breeding, represented by significantly higher numbers of young birds (see Appendix 6). This is almost certainly related to an increase in the availability of wetland habitat from average and above-average winter-spring rainfall since 2010 and the resultant likely longer duration of inundation of these wetlands. This has ensured habitat availability for adult and young birds for the entire breeding cycle until young fledge.

2.4. The Brolga Guidelines The objective of the Brolga Guidelines is to ensure that each wind farm development has at a minimum a zero net impact on the Victorian Brolga population (DSE 2012a, p.6). To meet this objective, three levels of investigations must be conducted. Information is gathered at each investigation level to inform the impact assessment and mitigation strategies. Each level also informs the next and all three levels are applied if there is potential for a significant impact that requires informed mitigation and offset. Further details on the Brolga Guidelines and their application in this work is summarised in Table 2.

Page | 14 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Table 2: Three level assessment of wind farm impact on Brolgas: current investigation

Level Step Assessment triggers* Current investigation - outcomes and actions

The proposed wind farm site is situated within the Victorian range of the Brolga The presence of Brolgas within the radius of investigation (i.e. within 10 km of the proposed wind farm boundary) Trigger for Level 1 The presence of potential Brolga habitat within Level One Assessment triggered and conducted the radius of investigation OR The location of the proposed development is within an area that may be used by Brolgas during seasonal movements between breeding and flocking habitats

Undertake desktop studies into known and All available historical and recent Brolga records within the RoI have been collated 1 potential habitat areas for Brolga and reviewed to identify the extent of Brolga occurrence in the RoI.

1 A site inspection was undertaken to identify potential Brolga habitat on and around Initial field inspection and local community the proposed wind farm site. 2 consultation Extensive landholder consultation within the RoI has been undertaken to identify potential Brolga sites that may not be in the available data.

Records of breeding or flocking habitats within the radius of investigation The proposed development is located in an Trigger for Level 2 area which may be used by Brolgas moving Level Two Assessment triggered and conducted seasonally between breeding and foraging sites, and may potentially create a barrier reducing movements between these habitats OR

Page | 15 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Level Step Assessment triggers* Current investigation - outcomes and actions The proposed location of new powerlines associated with the development may create new collision risks for Brolgas

Site-specific investigations determined that the Brolga utilised the landscape surrounding the proposed Dundonnell Wind Farm and was recorded breeding and flocking in a number of wetlands in the wind farm RoI. The Level 2 Assessment collects comprehensive data about the location, nature Extensive site-specific field investigations were undertaken during breeding and and extent of Brolga habitats, and patterns of non-breeding periods from 2009 to 2014, commencing at a much higher level of 2 - habitat use and behaviour at breeding, flocking detail in 2011, to document the extent of Brolga activity in the RoI, as well as flight and foraging sites within the radius of paths, and current and historical spatial patterns of activity and relationship with investigation land use patterns. Brolga utilisation surveys undertaken that generated quantitative information on activity on the proposed wind farm site suitable for collision risk modelling.

Qualitative risk assessment (AusWEA 2005) of Trigger for Level 3 project following site design is greater than Level Three Assessment triggered by assessment of the project and conducted "low"

Turbine layout has been adjusted based on the identification of turbine-free buffers Avoid or mitigate all potential impacts to Brolga that incorporate historical breeding and flocking records and estimated areas of breeding and flocking home ranges within the 1 activity around such sites, breeding and flocking site linked flight paths observed radius of investigation with turbine-free buffer during site-specific investigations and habitat likely to be used by the species, areas together with a 300 m disturbance buffer.

Develop a site-specific collision risk model for Residual risk has been ascertained as less than one bird colliding with turbines per 3 2 Brolgas utilising or moving through the radius year based on a conservative 95% avoidance rate for birds that fly towards of investigation turbines.

PVA modelling has shown that this scale of impacts reduces the expected minimum Use PVA to estimate the impact of the 3 population size by an estimated four birds over a 25 year project life (out of a proposed development current expected minimum population size of 809 birds).

4 Identify appropriate compensation strategies to Compensation will focus on readily quantifiable outcomes, and in particular the

Page | 16 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Level Step Assessment triggers* Current investigation - outcomes and actions ensure a zero net impact on the Victorian production of additional fledged young at additional restored and protected brolga population breeding sites at a rate that results in an extra four adult birds being present in the population no later than 25 years after operations commence. With an average of 0.5 extra brolga deaths per year with a 95% avoidance rate, this translates to the loss of an extra 13 Brolgas over 25 years. Therefore, mitigating this loss by increasing breeding success would require an extra 13 birds being raised to adulthood over the 25 year period.

* Information in this column is taken verbatim from DSE (2012a)

Page | 17 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

2.4.1. Investigating Brolga movements Brolgas are difficult to study due to low population density, wide spatial distribution and the difficulty of identifying them individually without causing stress to individual birds (as would occur in capture and marking studies). Rather than (unsuccessfully) attempting to collect absolute population numbers requiring individual bird identities, data collection was conducted using a range of sources and methods compliant with the Brolga Guidelines to provide data on which to base wind farm design and mitigation decisions. These data collection methods required the use of different, well-accepted, statistical techniques than would be used with absolute population numbers. The aim of the investigations was to determine if and what impact this wind farm would have on the Brolga population within the RoI and how this could be mitigated. From April 2011, fortnightly roaming surveys within the RoI were used to detect flocking or breeding events. The local observers' network of landholder contacts and permission to access land was a significant means of overcoming the limitations posed by private land access restrictions. For this reason, no further aerial surveys were undertaken. Before then, less frequent surveys were undertaken. To maximise understanding of the behaviour of breeding and flocking Brolgas, whenever a breeding or flocking event was detected during a fortnightly roaming survey, or by a landholder who communicated the fact to the local observers, detailed behavioural observations were triggered. These comprised observations to ascertain the usual routine of the birds (observations showed that Brolgas were birds of routine), involving both the local observers and early participation of BL&A personnel to assist in establishing the particular routine. This was followed by flight path mapping observations, to document movements away from breeding and flocking sites, mostly undertaken by the local observers. This approach to observations ensured that a very comprehensive set of data on Brolga behaviour and flight paths was assembled. This was of sufficient detail and fine scale to inform application of the Brolga Guidelines.

Page | 18 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

3. LEVEL ONE ASSESSMENT The Level One Assessment provides a preliminary determination of whether a proposed wind farm development represents any level of risk to the Victorian Brolga population (DSE 2012a). The triggers and steps for Level 1 Assessments are noted in Table 3. Table 3: Level 1 Assessments: triggers and steps

Level Step Assessment triggers

The proposed wind farm site is situated within the Victorian range of the Brolga The presence of Brolgas within the radius of investigation (i.e. within 10 km of the proposed wind farm boundary) Trigger for Level 1 The presence of potential Brolga habitat within the radius of investigation OR The location of the proposed development is within an area that may be used by Brolgas during seasonal movements between breeding and flocking habitats

Undertake desktop studies into known and potential habitat 1 areas for Brolga 1

2 Initial field inspection and local community consultation

All four assessment triggers apply to the proposed Dundonnell Wind Farm. This assessment is described in this section.

3.1. Step One: Undertake desktop studies into known and potential habitat

3.1.1. Methods Records of the Brolga in the RoI were compiled from the sources described below.

. A list of the Brolga records in the study area was obtained from the Victorian Biodiversity Atlas (DSE 2012b) . the south-west Victorian flocking site database (complied by Sheldon 2004 and provided by the then-DSE)) . The Atlas of Australian Birds and Birdata (BirdLife Australia) accessed 2011/2012 . Handbook of Australian, New Zealand and Antarctic Birds – Volume 2 (Marchant and Higgins 1993) . Action Statement No. 119. Brolga Grus rubicunda (DSE 2003) . DSE (Philip Du Guesclin) unpublished records (2006 – 2007) . DSE unpublished records (2010) . DSE (Richard Hill) unpublished records (1992) . Arnol et al. (1984) unpublished records from the Victorian Biodiversity Atlas (up to 1984)

Page | 19 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Community groups such as BirdLife Australia (formerly Bird Observers Club of Australia and Birds Australia) . Previous BL&A records published in the Mortlake Wind Farm planning application documents (BL&A 2007 – 2009). . Discussions with local landholders to obtain information on recent and historical flocking and breeding sites and movements within the study area. These sources provided the most comprehensive available Brolga records that currently exist, apart from those collected during the targeted investigations described in the Level Two Assessment (see Section 4 of this report).

Flocking site definitions The Brolga Guidelines state that a flock roost site must meet the criteria below.

For initial analysis and short-listing of possible flocking sites, including during the landowner surveys, sites that had supported five or more birds (i.e. more than the maximum family party, representing sites where more than one family party came together) were located based on existing records. These sites were divided into two categories: . Traditional flocking sites: are not specifically defined in the Brolga Guidelines, but are referred to as sites to which Brolgas return 'year after year' or 'each year'; and . One-off flocking sites: are defined in the Brolga Guidelines as sites where a flock of Brolgas is observed or has been observed previously on a single occasion, but the site is not a traditional and regularly-used site. This includes single records of a flock or repeat records once within a month or less, and flocks observed foraging during the day away from wetlands. Traditional flocking sites are considered to have much greater value for Brolgas than one-off flocking sites, as they represent a key habitat used for safe overnight roosting after a day of foraging in the surrounding landscape. Movements to and from one-off sites are more likely to resemble the movements Brolgas make in the migration season, movements that the Brolga Guidelines state can be considered in determining the residual risk of the project to the Victorian Brolga population. One-off flocking site records may also correspond to an observation of a flock foraging during the day away

Page | 20 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) from its traditional flocking site (i.e. the wetland to which it returns each night to roost) and can often be of birds not using a wetland (e.g. see Table 15).

3.1.2. Results The results from this section of the work are included in Table 15 and Figure 19, along with data from the Level 2 Assessment.

Historical breeding sites The historical databases, landholder records and observations prior to 2009 when this study commenced indicated that 39 wetlands have been used by Brolgas for breeding within the wind farm RoI (see Table 15 and Figure 19).4. During the Level Two investigations, eight additional breeding sites were found: one in a wetland located north of the wind farm site (wetland 533) 0.75 km from the nearest wind turbine, a second in a wetland approximately 2.0 km east of the nearest wind turbine (wetland 254), a third approximately 800m south of Woorndoo-Darlington Road (wetland 605), approximately 7.6 km south-west of the nearest turbine, two (wetlands 241 and 587) east of Lake Gellie (wetland 239), approximately 4.1 and 5.3 km respectively north- east of the nearest wind turbine, one at wetland 304, approximately 8.0 km south-east of the nearest wind turbine and one at wetland 320, approximately 8.5 km north-west of the nearest wind turbine. These are included in Table 15 and bring the total number of known breeding sites in and near the wind farm RoI to 46. Of these, three sites (wetlands 303, 607 and 608) have been permanently drained and no longer have the potential to support breeding Brolgas. The number of active nests within the RoI is likely to vary between years depending on several factors, including rainfall. No nests have historically been recorded within the final proposed turbine layout.

Historical flocking sites The review of the databases, landholder records and observations prior to 2009 when this study commenced identified 19 traditional flocking sites within the RoI and 24 one- off flocking sites, including seven sites where flocks were recorded away from wetlands where they would not have roosted for the night (see Table 15 and Figure 19). During the current investigations (i.e. since 2009), an additional two traditional flocking sites were identified (wetlands 252 and 585). Observations of flocks away from traditional flocking sites were not included as one-off flocking sites and have only been included from the historical data for completeness. The distance of a traditional flocking site from turbines if within five km is also provided in Table 155. A brief description of known traditional flocking sites within ten km of the proposed wind farm site is provided below. . Wetland 110 (Lake Sheepwash): This wetland has been a popular flocking site in recent years (Du Guesclin pers. comm. re 2007, local landholder, pers. comm. re 2009, 2010). In 2011, a flood breached the dam wall that holds water in this wetland. This site is located 2.9 km south-east of the nearest proposed turbine.

4 The Brolga Guidelines recommend a 3.2 km radius turbine-free buffer from breeding sites. 5 The Brolga Guidelines recommend a 5 km radius turbine-free buffer from flock roost sites.

Page | 21 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Wetland 112 (south of Chinaman's Swamp): Local landholders reported that Brolgas used this area in the late 2000s. Brolgas used it as well as Chinaman's Swamp and wetland 585 during the 2011 breeding season flocking event. This wetland is located approximately 1.4 km west of the nearest proposed turbine. . Wetland 118 (Chinaman’s Swamp): Local landholders reported that Brolgas flew from Lake Sheepwash (wetland 110) indirectly to Chinaman’s Swamp (wetland 118) in 2009 and 2010 on a regular basis to roost and forage. They used the site as a flocking site during the 2011 breeding season (possibly involving juvenile, non- breeding birds). This site has not been reported as a flocking site in either the Victorian Biodiversity Atlas (DSE 2012b) or the Sheldon database (2004). This wetland is located approximately 2.6 km north-west of the nearest proposed turbine. . Wetland 146 (south-east of Lake Bolac): This wetland is located at the northern edge of the RoI, just east of Pagels Lane. Local landholders reported occasional use by Brolgas in the late 1990s and in 2007. It is located over 10 km from the proposed wind farm site. . Wetlands 235 and 304 (Terrinallum South): Flocks have been observed here in recent years. These two large wetlands provide suitable flocking habitat, as adjacent areas are cropped, making the area attractive for foraging. These sites are approximately seven km south-east of the nearest proposed turbine. . Wetland 239 (Lake Gellie): This wetland is comparatively extensive and the main flocking aggregation on the lake lies approximately two km north-east of the nearest proposed turbine. There are several historical records of flocking Brolgas: o 120 Brolga April 1983 o 32 Brolga in 1993 o 30 Brolga in 2002 (Sheldon 2004). . Wetland 240 (south side of the western end of the Nerrin-Pura Road): Landholders have reported Brolga flocking in some years at this wetland. Flocks were reported again in 2011 and 2012. This site lies over 5 km from the proposed wind farm site. . Wetland 244 (Lake Terrinallum): This site is well documented as being a regular flocking site (Sheldon (2004), DSE (2012b), local landowners (pers. comm.)). It is located approximately 2.6 km east of the nearest proposed turbine. . Wetland 248 (Lake Barnie Bolac): This site is well documented and well known by locals. It is located eight km south-south-east of the nearest proposed turbine. . Wetland 301: This site is located along Woorndoo-Darlington Road. This site is large ephemeral grassy swamp. Flocks of 30 to 50 Brolgas were reported here in the late 1980s to early 1990s (Landowner, pers. comm). It is located approximately five km south-west of the proposed nearest wind turbine and within five km of the proposed power transmission line. . Wetland 303: A local landowner reported that small groups of 12 to 15 Brolgas occasionally flock at this grassy swamp most years, particularly when sheep are being hand fed (Landowner (pers. comm.)). This site lies approximately 7.1 km south-east of the nearest proposed turbine. . Wetland 323: This is located along Nerrin Nerrin Estate Road and is a vegetated freshwater wetland. Brolgas travel between this wetland and wetland 153, which is

Page | 22 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

one km north of it. There have been reported observations of up to 13 Brolgas utilising this site over a number of years (Landowner (pers. comm.)). The site is approximately 8.5 km north of the nearest proposed turbine. . Wetland 511 (south-east of Lake Gellie): This wetland is listed once from 2001 in the Sheldon (2004) database and from 1983 in the VBA. No recent flocking records have been recorded from it, although it formed part of the foraging habitat used by Brolgas roosting at Lakes Gellie and Terrinallum during the current investigations. It lies 1.6 km north-east of the nearest proposed wind turbine . Wetland 526: This wetland is located along Trawalla South Estate Road. Flocks of Brolgas have been sighted over a number of years in nearby cropping paddocks surrounding this wetland and wetland 600. It has been suggested that they may travel to and from Pink Lake, which is a traditional flocking site. These sites are located approximately eight km north-east of the nearest proposed turbine. . Wetland 585 (south of Chinaman's Swamp): This extensive, seasonally dry wetland was used by Brolgas to roost during the 2011 flocking event over several months. The wetland is 1.4 km from the nearest proposed wind turbine. . Wetland 603 (Long Dam): This wetland contains water all year round and may provide suitable habitat for flocking (Landowner evidence to Mortlake Wind Farm Planning Panel). Due to limited access, this site could not be visited. The landowner has provided occasional updates on Brolga activity at this site to Trustpower Pty Ltd and the previous developer, New En Australia Pty Ltd, and these records have been included in Table 15. This site is located approximately ten km south-south-east of the nearest proposed turbine.

3.2. Step Two (Part 1): Field inspections

3.2.1. Methods

Wetland quality assessment and Brolga wetland use To provide information on the status, distribution and possible occurrence of Brolgas on lakes and swamps in the larger region (including the RoI), an initial habitat assessment was undertaken by two BL&A observers during November 2009. The survey area for the initial field inspection included the RoI and some wetlands located beyond that area in the surrounding region. As far as possible, all wetlands not subject to private land access limitations were visited and surveyed. In order to ensure that the knowledge of local landholders was accessed, the proponent retained the services of two local residents at Dundonnell from 2011 to 2014. These residents (field staff) were trained in Brolga survey methods by BL&A and undertook extensive field work under the guidance and supervision of BL&A. Given their knowledge of the local community, they were able to discuss the occurrence of Brolgas with local landholders across the majority of the RoI for the duration of the investigations. Dates, locations and estimated numbers of Brolgas historically recorded on each landholder’s property were documented in one-on-one interviews. Only three of the 76 landholders did not provide information or allow access to their properties for field staff:

Page | 23 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Information on land type but not land use or Brolgas was provided by Mount Fyans immediately to the south of the proposed wind farm site (south of the Woorndoo - Dundonnell Road and West of the Darlington - Dundonnell Road) and access was denied . No data or access was provided at the property north of Hamilton Highway, south of Lake Barnie Bolac and west of the Darlington – Dundonnell Road (about seven km south of the proposed wind farm site) . No data or access was provided for the property immediately to the north-west of the proposed wind farm site, both sides of Ennerdale Lane.

Data issues Data on Brolgas has been collected from various sources over the last 30 or more years, and there are possible issues re data comparability and accuracy. There are two key issues: location specification and record keeping. Location specification: the accuracy of some data is only to the nearest latitude/longitude minute (equivalent to approximately 1.85 km), five minutes (nine km) or ten minutes (18 km). Record keeping: some information provided, such as some landowner data, was based on memory as many would not have reported their sightings at the time to an institution (e.g. DEPI) and did not have a hard copy of their observations to verify their recollections. Consequently, some breeding records were mapped in locations in the middle of paddocks which do not currently support a wetland. In such cases, the breeding site record was relocated to the nearest wetland, where possible to do so accurately based on local knowledge of Brolga breeding activity in the region. The location of sightings (i.e. records of birds neither breeding nor flocking) was not altered in this way as it is not uncommon for Brolgas to be found some distance from wetlands when foraging. Brolga breeding behaviour and locations were recorded during a wetland survey and quality assessment undertaken from 5th to 10th November 2009 (see section 4). This survey was undertaken in spring but after a prolonged drought. Brolgas were observed with binoculars and all effort was made to avoid interrupting normal behaviour by remaining at least 300 m from birds and, where observation location permitted it, remaining in vehicles to avoid disturbance to Brolgas. Surveys began at sunrise and finished after sunset and any Brolgas observed were recorded. The wetland survey and quality assessment was repeated in 2011. This survey involved the same methodology and was undertaken in April 2011 (autumn but after above average rainfall). During this survey, some additional wetlands were surveyed that were not surveyed in 2009. The numbering of wetlands in Figures 1 and 19 generally reflects the order in which they were documented, not necessarily a geographical order.

Wetland mapping All maps in this report are based on the DEPI wetland layer. The 2011 survey was undertaken after two years of average or above-average rainfall, when lakes and many small seasonal wetlands were holding water. However, the 2009 survey was undertaken at the end of a prolonged drought and habitat quality changes were observed between 2009 and 2011. For this reason, the surveys are considered to be representative of a range of likely future habitat availability.

Page | 24 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Habitat quality Habitat quality of each wetland was assessed using the criteria detailed below, which are based on current information on Brolga habitat requirements in western Victoria (Marchant and Higgins 1993, Du Guesclin 2003). High: Habitat components listed below are usually all present. . Permanent, or largely permanent, shallow freshwater body guaranteed to hold water . Shallow freshwater marsh or shallow freshwater meadow less than 0.5 m deep . Waterbody with some aquatic and emergent vegetation (e.g. rushes, tussock grass) . Little or no signs of changed water regimes (e.g. drained wetlands) . Little or no signs of disturbance (e.g. cultivation, native vegetation removal, grazing).

Moderate: Some fauna habitat components are often missing although wetlands still provide some characteristics to provide flocking opportunities. . Waterbody likely to hold water long enough for flocking or breeding (i.e. permanent or largely permanent) . Waterbody with some aquatic and emergent vegetation (e.g. rushes, tussock grass) . Some changes to water regime may have occurred (drainage lines) . Wetland shows some signs of disturbance (such as some limited access to stock, cultivation).

Low: Many habitat elements have been lost. Wetland habitats that: . Are likely to be ephemeral or drained (only hold water for limited time of the year) . Have little or no aquatic or emergent vegetation . Have a changed water regime, little water present . Show signs of disturbance (such as being heavily grazed by stock, being cultivated, or feral predators). Any wetland found to have been permanently drained was not considered suitable for Brolgas, in line with the Brolga Guidelines. All other historical records of breeding or flocking were assumed to indicate sites where breeding or flocking could occur in the future, except for permanently drained sites6.

6 For example, wetland 607 & 608 (see Figure 19), had historical breeding records from the local landholders. However, the wetland had been drained and, based on inspection of it by BL&A, was considered no longer suitable for Brolga breeding.

Page | 25 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

3.2.2. Results During the initial field inspection of the wind farm RoI in 2009, 113 wetlands were assessed using the methods described above. Many wetlands marked on the 1:50 000 topographic map sheets were dry during the survey, or had been dry for many years, as they had been converted for agricultural use. These sites no longer held water and were unsuitable as breeding habitat for the Brolga. Predictably, no birds were seen on these sites or reported by landholders. A summary of the results from the 2009 wetland assessment is presented in Table 4 Table 4: Summary of 2009 wetland quality assessment and Brolga use data Habitat quality No. of wetlands % of wetlands No. of Brolgas Low 115 79 0 Moderate 11 8 3 High 19 13 13 Total 145 100 16

The majority of wetlands (79%) were of low quality. Moderate quality wetlands accounted for 8 % of the surveyed wetlands and 13 % were of high quality. A total of 184 wetlands were assessed during the 2011 investigation (Table 5). These included most of the 189 functional7 wetlands in the RoI and its surrounds. These included all of the wetlands assessed in 2009. Details of these wetlands, including their habitat quality, presence of Brolgas and site descriptions are presented in Appendix 1 and their locations are presented in Figure 1 and Figure 19. Table 5: Summary of 2011 wetland quality assessment and Brolga use data Habitat quality No. of wetlands % of wetlands No. of Brolgas Low 64 35 4 Moderate 70 38 28 High 50 27 19 Total 186 100 51

In 2011, the majority of wetlands were full at the time of the field assessment. Some 38% of the wetlands were categorised as moderate quality habitat and 27% as high quality, using the criteria outlined earlier. The differences in wetland quality between the two years were substantial. The improvement in wetland quality was considered to be in response to the intervening years of average to above average rainfall. This meant that more wetlands had held water more recently and wetland vegetation had recovered, leading to an improvement in Brolga habitat quality. It is likely that such an improvement occurred across the Brolga’s range in south-western Victoria, a fact reflected in the significant improvement in breeding success across its range in the state (see Appendix 6).

7 Functional: capable of holding water.

Page | 26 JAGO ROAD 158 PARUPA PARK ROAD

547 546

LAKE ROAD 572 571 148 574 206 149 147 573 569 541 545 548 565566 146 563 TRAWALLA SOUTH ESTATE ROAD 568 152 542 543 SOUTH BEACH ROAD567 562 544 553 564 153 WHITES LANE 79 570 575 530 590 315 78 560 323 224 317 557 561 558 559 314 528 208 MORTLAKE-ARARAT ROAD 316 532529 556 527 531 225 125 526 WARINGS LANE 126583576 525 75 582 504 600 552 PAGELS LANE 240 312 82 223 209 522 322 NERRIN-PURA ROAD 127 523521 320 MCKAYS LANE 524 601 123 519 130 128 516 124 517 222 518 520 550 85 129 129 122 503 549 321 86 133 120 121 539 502 501 241

ENNERDALE LANE 327 239 587 245 SMITHS LANE 535 DARLINGTON-NERRIN ROAD BLOMLEYS242 LANE 533 238 555 VITE VITE ROAD 87 536 534 326 506 510511 88 243 221 WOORNDOO-STREATHAM ROAD 118 505 507 91 512 515 508 509 LYONS LANE 513514 117 DARCYS LANE 246 92 90 538 310 537 FASHAMS LANE DOHERTYS LANE 607 309 606 585 CARLONS ROAD 111 244 POST OFFICE LANE

BOLAC PLAINS ROAD 112 DARLINGTON-CARRANBALLAC ROAD DAWES LANE 324 VEALS LANE 307 308 MORTLAKE-ARARAT ROAD 584 325 254 LADES LANE 311 114113 602

141 WOORNDOO-DUNDONNELL ROAD 137

313 DUNDONNELL-DERRINALLUM ROAD 97 138 605 139 306 110 236

328 235 CHATSWORTH ROAD 247 145 301 TERRINALLUM ROAD

WOORNDOO-DARLINGTON ROAD 101 144 318 9 MOUNT FYANS LANE 304 305 252 NINE MILE LANE 109 586 11

103 253 BARNIE BOLAC ROAD 303 248 DARLINGTON-SKIPTON ROAD 302 5

603

MORTLAKE-ARARAT ROAD NORTH STATION 604

DUNDONNELL-MORTLAKE ROAD 251

DARLINGTON ROAD HAMILTON HIGHWAY JELLALABAD ROAD DARLINGTON ROAD

KURWEETON ROAD NORTH STATION

SIX MILE LANE

CASTLE CAREY ROAD PRICES LANE

DARLINGTON ROAD

Legend Metres 01,500 3,000 6,000 Wind Farm Site SIX MILE LANE Area of Investigation (10km) Figure 1: Wetland quality results for 2011

Wetland Quality Project: Dundonnell Wind Farm High Client: TrustPower Pty Ltd Moderate Low Project No.: 9184 Date: 12/03/2014 Created By: M. Ghasemi Not accessible ¯ Roads Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

3.3. Step 2 (Part 2): Local community consultation and Brolga use survey

3.3.1. Overview The results of detailed landholder surveys, when combined with historical data, enabled a more complete and longer term picture to be assembled of Brolga activity in the RoI to supplement and provide context for the four years of detailed, targeted investigations reported in the next Section (i.e. the Level Two Assessment). These data were used to determine if there was a general spatial use pattern across the wind farm component of the RoI and on the wind farm itself.

3.3.2. Methods

Probability density map using historical data To determine the spatial distribution of Brolgas across the broader landscape within the wind farm RoI and predict where they are more likely to occur, individual historic Brolga records were transformed into a probability density map. This used the Brolga records from the DEPI (DSE) database (VBA) and the BirdLife Australia database (Birdata). The probability density map illustrates the relative probability of a Brolga observation occurring at a given location (inside and outside of the wind farm). Relative probability has been used as it is impossible to provide absolute probabilities, that is, a true/accurate location for future Brolga records cannot be 100% confidently predicted based on historical records. Relative probability was calculated by including a “smoothing length” of 2.5 km. This “smoothing length” corresponded to an approximate 95% probability that the observation’s true/accurate location was within five km of the recorded location (i.e. within 2.5 km in any direction of the location in the collected data). The chosen distance (2.5 km) is a balance between increasing the likelihood of including the true location and reducing bias (i.e. false positives).

Data issues This analysis, however, does not account for Brolga behaviour and assumes that Brolgas have an equal probability of occurring in any location. This is a limitation of this analysis, as Brolgas are less frequently seen far from their preferred habitat (i.e. high quality wetlands and arable land). Therefore the results of this spatial statistical analysis must be interpreted within the context of an understanding of Brolga habitat use, discussed later in this section. Brolga observations have been collected over many years and submitted to the available databases by land owners, professionals and amateur observers. However the data cannot be scaled for survey effort and there were no absence data available (i.e. locations visited where Brolgas were searched for but not observed). Therefore the data provided a general overview of Brolga spatial use. These historical data can be used together with other data collected during targeted surveys between 2009 and 2013, as well as information from landholders, to understand Brolga activity in the wind farm RoI.

Page | 28 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Landholder surveys Detailed landholder surveys were undertaken within the RoI to further assess the likelihood of interaction between Brolgas and the proposed Dundonnell Wind Farm site. 8 Location information from these surveys is included in Table 15 and Figure 19 and the summary of Brolga occurrence in the RoI presented earlier includes landholder information.

Surveys of landholders within the RoI Information sought from landholders is noted below. In addition to presence and location, landholders were queried about historical Brolga use on their properties and the historical and current land use/s and land type/s in these locations. Participating landholders were asked to rank and categorise information regarding their property as described below. Brolga frequency: Small/family group sightings (< 5 individuals): 0 - none sighted 1 - very infrequent 2 - infrequent 3 - sightings (usually when no frequency was recorded) 4 - frequent 5 - very frequent N/A - not available Large groups/flocks (>5 individuals): ranked as above Presence/absence of breeding: 0 - No breeding/nesting 3 - Breeding/nesting recorded (rank number consistent with above category) N/A - not available Land use: . Cropping . Grazing . Mixed Land type:

8 The information collected on land use also feeds into the conclusions about Brolga use of the landscape in the Level Two Assessment. For readability, all landowner results are presented in this section.

Page | 29 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Arable . Stony . Aquatic (includes wetlands, dams and other water sources) . Mixed . Cleared/semi cleared (of rock) . N/A. A Brolga frequency score was calculated to take into account the frequency of all types of Brolga activity (nesting, small groups and flocking). This score was then used to identify areas where Brolgas were more frequently engaging in these activities. Brolga frequency score = Small group score + Large group/flock score + Breeding score This equation assumes that each behaviour is equally impacted by the wind farm. If future evidence suggests that a behaviour poses a higher risk of causing a negative impact on Brolgas, the equation can be altered to reflect this.

Additional survey of landholders within the wind farm site Landholders within the proposed wind farm site were surveyed in more detail to provide a more accurate picture of their land uses in relation to Brolga activity. Each of the properties was subdivided to provide more spatial detail. Note that this survey pre-dates the addition of properties to the north-eastern part of the wind farm, so the wind farm boundary in this mapping represents approximately 70 per cent of the current proposed area of the wind farm. The survey method was similar to that for the wind farm RoI but with the following changes: Brolga small/family groups (<5 individuals): Infrequent (every two or more years) Frequent (annually for the last 5 years) Large groups (≥ 5 individuals): Ranked as above Land use: . Cropping . Grazing . Cropping alternating with grazing . Parts cropped, parts grazed Land type: . Aquatic (including wetlands) . Cleared (of rocks) . Semi-cleared (of rocks) . Stony . Arable (subject to past or recent rock removal and/or cultivation) . Mixed

Page | 30 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Combined analysis A standard chi-squared test compared each of the Brolga behaviours with land use and land type data. Significant correlations (P<0.05) between Brolga activity and land use or type have been used. All analysis used the data that was collected. Some landholders were unable to provide answers to some questions and this has been shown in the maps as ‘no data’. Where gaps existed, these have been documented in the text or in the maps. Each data set was analysed separately because each had unique strengths and limitations (as noted below). Additionally, the data were not in the same format (statistically) to combine into one analysis and therefore there is no single formal assessment of Brolga landscape use presented here. Although each set was analysed and presented separately, patterns that were evident in all data sets have been used to further refine the understanding of past and likely future spatial patterns of Brolga activity in the RoI and on the wind farm site. Similarly, patterns that are evident in one or two sets may be artefacts of limitations and therefore may be unreliable. The potential for bias in the landholder information was explored and none was found: the patterns of Brolga activity in relation to land use and land type, and the historical information on Brolga location, did not differ between landholders inside and outside the wind farm boundary or from historical data from other sources.

Data issues All willing landholders within the wind farm RoI were interviewed. As noted in Section 3.2.1, three of 76 property owners did not participate in the survey (as was the case for other Brolga data collection) and therefore there are gaps in the maps, the most significant of which involve a property to the immediate north- west of the wind farm (encompassing about two per cent of the area covered by the landholder surveys (i.e. some 1,397 ha out of a total 61,330 ha assessed for land type) and another property representing a further 12.6 per cent of the area surveyed. The quality of this data is likely to vary due to landholder interest and length of residency; however the information obtained has added (independent) information to the overall picture of Brolga activity and has provided an indication of preferred land use/s and land type/s.

3.3.3. Results and discussion

Probability density map using historical data The probability density map generated from the historical data indicated a comparatively high likelihood of Brolga activity in the south-east of the RoI (Figure 2). Most of the data from this area was contributed by a single observer and therefore the high activity is likely biased due to the higher observer effort there. Since historical records can be biased based on observer effort they are not reliable on their own, but can be used together with other data collected, for example breeding and flocking surveys undertaken for the Level Two Assessment (see Section 4).

Page | 31 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

The distribution of likelihood also indicates a lack of expected Brolga activity in, and just beyond, the eastern parts of the proposed wind farm site.

Landholder surveys

Surveys of landholders within the wind farm RoI The results of the landholder survey illustrate that, within the RoI, the central region and areas to the south, east and north of the wind farm boundary were predominantly stony ground used for grazing (Figures 3 and 4). Cropping on arable and mixed land types was more prevalent to the north and west of the wind farm boundary and well to its east. The frequency of landowner breeding records in the RoI suggests that most nests are located in the eastern half of the RoI, with some in the north and south (Figure 5). Generally, small groups correspond with the location of breeding (Figure 5 and Figure 7), and the large groups/flocks appear to use areas not frequented by small groups and pairs (Figure 6).

Surveys of landholders within the wind farm site boundary An additional survey of landholders was undertaken at a greater level of detail within the wind farm boundary. This found that the site is dominated by stony ground and nearly all of this has been used for grazing (Figure 8 and Figure 9). The paddocks immediately west of the proposed wind farm site have been used for cropping on arable and mixed land types (Figure 3 and Figure 4). Similarly some sections in the southern portion of the wind farm site comprise arable and semi-cleared land that has historically been cultivated. The frequency of breeding reported by landholders in this wind farm site survey indicated that the proposed wind farm site has been used once before by nesting Brolgas (wetland 608). As this wetland is now drained and no longer supports habitat suitable for breeding, it is unlikely that the Brolga will breed on the site in the future. Other reports of breeding birds in the northern and north-western parts of the wind farm relate to birds breeding in the 1980’s at wetland 607, which is now drained (northern) and breeding birds from a wetland to the west (wetland 118) (north-western). There is also little use of the wind farm by large and small groups (Figure 11 and Figure 12), with small groups more likely to use the paddocks immediately west of the wind farm site than either large groups or areas within the proposed wind farm site.

Analysis of Brolga activity and land use/type Some categorical data were combined to obtain sufficient counts to allow a reliable correlation analysis (Table 6 and Table 7). There was a significant difference between land use and frequency of Brolga activity for large or small groups (Small groups: χ26 = 34.5; p<0.001; Large groups: χ24 = 13.0; p<0.05; Table 6). Similarly there was a relationship between small and large groups and land type (Small groups: χ29 = 39.0; p<0.01; Large groups: χ26 = 20.5; p<0.001; Table 7). Both small and large groups were noted by landholders more frequently in aquatic or arable areas than on stony ground. Additionally, the incidence of infrequent or

Page | 32 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) no Brolga use in stony land used for grazing was high and therefore it is more likely for Brolgas to be absent from than present in these areas. This pattern is mirrored by the historic data, and also supports anecdotal evidence that Brolgas prefer aquatic habitat and are very rarely observed in stony areas. Table 6: No. property sub-areas with different levels of Brolga activity vs land use Land use Cropping Grazing Mixed Total Small groups Frequent/very frequent 17 5 15 37 Sightings - unknown frequency 5 2 7 14 Infrequent/very infrequent 24 15 23 62 None recorded 17 40 9 66 Total 63 62 54 179 Large groups/flocks Frequent /sightings 10 2 7 19 Infrequent/very infrequent 10 5 13 28 None recorded 42 55 34 131 Total 62 62 54 178

Table 7: No. property sub-areas with different levels of Brolga activity vs land type Mixed/semi- Land type Aquatic Arable Stony Total cleared Small groups Frequent/very frequent 14 17 8 2 41 Sightings - unknown frequency 10 5 3 2 20 Infrequent/very infrequent 21 17 14 18 70 None recorded 9 15 3 31 58 Total 54 54 28 53 189 Large groups/flocks Frequent/very 11 9 5 2 27 frequent/sightings Infrequent/very infrequent 5 15 4 4 28 None recorded 38 30 19 46 133 Total 54 54 28 52 188 Records from mixed and cleared/semi-cleared land were excluded from correlation analysis due to numbers of records being inadequate to meet the assumptions of the statistical testing. Frequent and very frequent observations have been combined with sightings for large groups/flocks for statistical reasons.

Note: The site boundary on Figure 3 to Figure 12 has been updated for this report, whereas the original analysis undertaken in 2012 was based on a previously proposed site boundary (see Appendix 4).

Page | 33 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 2: Historical Brolga count density (smoothing length 2500 m)

Page | 34 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 3: Reported land use on farms in the broader region

Page | 35 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 4: Reported land type on farms in the broader region

Page | 36 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 5: Breeding occurrence on farms; as reported by landowners

Page | 37 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 6: Frequency of large brolga groups/flocks on farms; as reported by landowners

Page | 38 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 7: Frequency of small brolga groups/flocks on farms reported by landowners

Page | 39 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 8: Reported land use on farms; detail of wind farm area

Page | 40 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 9: Reported land type on farms; detail of wind farm area

Page | 41 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 10: Breeding occurrence on farms reported by landowners: wind farm area

Page | 42 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 11: Frequency of small brolga groups/flocks reported by landowners: wind farm area

Page | 43 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Figure 12: Frequency of large brolga groups/flocks reported by landowners: wind farm area

Page | 44 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

3.4. Triggers for a Level Two Assessment The Level Two Assessment in the Brolga Guidelines is designed to provide a comprehensive record of the location, nature and extent of Brolga habitats and behaviour within the wind farm and the surrounding RoI (DSE 2012a) to inform an impact assessment and detailed mitigation strategies. A Level Two Assessment is triggered by one or more of the conditions listed in Table 8.At least the first and third of those conditions apply to the proposed Dundonnell Wind Farm and the remaining condition may also apply. Therefore, a Level Two Assessment has been undertaken and is reported in the following section.

Page | 45 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

4. LEVEL TWO ASSESSMENT The Level Two Assessment provides a comprehensive record of the location, nature and extent of Brolga habitats and behaviour within the wind farm and the surrounding RoI to inform an impact assessment and detailed mitigation strategies (DSE 2012a). The triggers and steps for Level 2 Assessments are noted in Table 8.

Table 8: Level 2 Assessments: triggers and steps

Level Step Assessment triggers

Records of breeding or flocking habitats within the radius of investigation The proposed development is located in an area which may be used by Brolgas moving seasonally between Trigger for Level 2 breeding and foraging sites, and may potentially create a barrier reducing movements between these habitats OR The proposed location of new powerlines associated with the development may create new collision risks for Brolgas

The Level 2 Assessment collects comprehensive data about the location, nature and extent of Brolga habitats, 2 - and patterns of habitat use and behaviour at breeding, flocking and foraging sites within the radius of investigation

As noted, at least the first and third triggers apply to the proposed wind farm.

4.1. Overview of methods A variety of survey methods may be used to gather relevant information, including roaming surveys, aerial surveys, flight behaviour studies and gradient studies. The choice of methods varies with the nature of the questions (e.g. flocking versus breeding activities), the type of countryside and its accessibility. The following methods were applied in this assessment (see later in this section for detailed methods): . Ground-based roaming observational surveys (greatest effort); . Liaison with landholders in the RoI and field follow-up of reports of Brolga activity; . Wetland quality assessments (during and post drought); . One aerial survey for breeding birds (at DSE’s request in 2010); . Systematic flight path mapping during breeding and flocking events; . Incidental observation and flight path mapping; and . Formal, quantitative bird utilisation surveys of the wind farm site during flocking events in the RoI. These methods and relevant dates are summarised in Table 9.

Page | 46 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

In 2009 and 2010, investigations were undertaken by personnel from BL&A. From 2011 to 2014, the proponent appointed two local residents to act as field staff. They were trained by and worked under the supervision and direction of BL&A. All work they undertook was in accordance with methods established by BL&A and a member of the BL&A team travelled to the region to induct and brief the two field staff on the method to be used for each survey period or type. Data were returned to BL&A on a monthly basis and personnel from BL&A regularly travelled to the region to discuss progress and answer questions and discuss data collection issues with the field staff. BL&A staff also undertook continuous observations for periods of up to three days of breeding and flocking birds to confirm their daily routine, as reported by the field staff. Phone communications were also maintained regularly. The total survey effort from 2011 to 2013 by the field staff totalled 2,075 hours, and a total of 16,099 km was driven in the RoI in search of Brolgas over this period. Further details of survey activities are provided in Table 9. Note that these totals exclude time spent by the field staff incidentally observing Brolgas during non-formal survey periods, which assisted them in keeping track of Brolga activity in parts of the RoI, including breeding and flocking activity. Table 9: Summary of field methods and survey periods: Level Two Assessment Survey Dates Activities Brolga breeding season (July – December) Searches of 145 wetlands Breeding season ground in the RoI for breeding survey, including wetland 5th – 10th November 2009 Brolgas, recording Brolga quality assessment sightings and assessing wetland quality. Low level aerial survey of Breeding site aerial survey 5th and 7th October 2010 RoI and additional area to 2010 north and north-east. Searches of 186 wetlands within the RoI for breeding Breeding season roaming 23rd June to 11th November Brolgas, recording Brolga survey 2011 2011 sightings and assessing wetland quality. Field inspections fortnightly of all historical known breeding sites and other Breeding season roaming 14thAugust 2012 to wetlands that were suitable survey 2012 7thJanuary 2013 within the RoI, and follow up of landholder breeding reports (see below). Detailed surveys of seven 21st September to 22nd active nest sites to map Nest watch surveys 2012 December 2012 flight paths and monitor breeding activities. Regular advice received Landholder liaison 2012 All year round from some participating landholders about Brolga

Page | 47 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Survey Dates Activities activity that was immediately followed up with a field visit Field inspections fortnightly of all historical known breeding sites and other Breeding season roaming July to December 2013 wetlands that were suitable survey 2013 within the RoI, and follow up of landholder breeding reports (see below) Detailed surveys of seven active nest sites to map Nest watch surveys 2013 Early September 2013 flight paths and monitor breeding activities. Regular advice received from some participating landholders about Brolga Landholder liaison 2013 All year round activity that was immediately followed up with a field visit.

Flocking season surveys (December to June )

Field inspections fortnightly Flocking season survey 30th April to 11th November of all historical known 2011 (unseasonal) 2011 flocking sites that were suitable. Regular advice received from some participating landholders about Brolga Landholder liaison 2011 All year round activity that was immediately followed up with a field visit. Incidental flocking sightings by field observers and April to December 2012 sightings from landholders Flocking season survey were recorded 2012 Field inspections fortnightly 16th August 2012 to 1st of all historical known March 2013 flocking sites that were suitable. Fixed-point survey across 22 sites within the wind farm boundary to quantify Bird Utilisation Survey 6th April to 25th May 2012 and locate Brolga activity and usage during flocking events. Regular advice received Landholder liaison 2012 All year round from some participating landholders about Brolga

Page | 48 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Survey Dates Activities activity that was immediately followed up with a field visit. Field inspections fortnightly Flocking season survey of all historical known 27th January to March 2013 2013-14 flocking sites that were suitable. Detailed surveys of flocking Flight behaviour studies 11th April 2013 to 1st sites to map flight paths during Flocking Survey August 2013 and monitor flock 2013 movements. Fixed-point survey across 20 sites within the wind farm boundary to quantify Bird Utilisation Survey 9th to 25th May 2013 and locate Brolga activity and usage during flocking events. Regular advice received from some participating landholders about Brolga Landholder liaison 2013 All year round activity that was immediately followed up with a field visit. Detailed surveys of flocking Flight behaviour studies sites to map flight paths 1st January - mid-June 2014 during flocking season early and monitor flock 2014 movements. Other non-breeding season surveys Searches of 137 wetlands Detailed Migration Survey 5th January and 2nd March for migrating Brolgas, 2011 2011 recording sightings and movements. Assessment of wetlands Wetland Quality within the RoI for habitat 18th April to 12th May 2011 Assessment 2011 quality, in addition to recording Brolga sightings. Regular advice received from some participating landholders about Brolga Landholder liaison 2011 18thApril to 12thMay 2011 activity that was immediately followed up with a field visit. Field inspections fortnightly of all historical known Migration Survey 2012 All year round breeding and flocking sites and other wetlands that were suitable. Regular advice received Landholder Brolga use March 2012 from some participating

Page | 49 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Survey Dates Activities survey 2012 landholders about Brolga activity that was immediately followed up with a field visit Field inspections fortnightly of all historical known Migration Survey 2013 All year round breeding and flocking sites that were suitable.

In addition to those activities, field staff were regularly informed of Brolga activity through their local network of landholder contacts in and near the RoI, including breeding and flocking events. Such reports were immediately followed up with a field visit. This resulted in additional, new breeding sites being added to the inventory of such sites within the RoI, as noted in Section 3.1.2. The rest of this section is divided into breeding, flocking and migration surveys. These Brolga surveys were conducted over five years from 2009 to 2014. The methods applied comply with a Level Two Assessment as outlined in the Brolga Guidelines. Each section describes methods then presents results. Conclusions are provided at the end.

4.2. Breeding season surveys

4.2.1. Survey methods Work was undertaken over five breeding seasons: 2009, 2010, 2011, 2012 and 2013. Average annual rainfall for the area (based on 20 years rainfall records) is 577.9 mm (data extracted from the Mortlake weather station, BOM 2014). The annual rainfall was recorded over the duration of the study period: 2009 (532.4 mm), 2010 (830.8 mm), 2011 (702.4 mm), 2012 (547.6 mm) and 2013 (559.8 mm). Based on this information, 2010 and 2011 were years of above average rainfall, which resulted in the majority of wetlands and lakes filling in the region. Significantly, these two years had the highest rainfall in the 20 years of readily available Brolga records. A range of methodologies was used to maximise the detection of breeding Brolgas. In 2009, ground-based searches were undertaken for Brolgas and in particular for nesting attempts, as well as an initial assessment of wetland quality (methods and results presented in section 3). This was followed by an aerial survey in 2010, as advised at the time by DSE (R. Hill). Once local field staff were employed in 2011, the field effort significantly increased to at least fortnightly, with more detailed monitoring of breeding events when they occurred. In 2012 and 2013, fortnightly roaming surveys were undertaken by the local field staff visiting all known and potentially suitable wetlands, based on historical breeding records, local knowledge of field staff and advice from landholders on whether wetlands held water. Breeding was also observed during the flocking season of 2011 (see flocking season for methodology) and these observations have been

Page | 50 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) incorporated into this section. The survey methods and timing of these surveys are detailed below.

2009 Breeding survey Brolga breeding activity and locations were recorded during the wetland quality assessments from 5th – 10th November 2009. Brolgas were observed with binoculars and all effort was made to avoid interrupting normal behaviour by remaining at least 300 m from birds and, where possible, remaining in vehicles.

2010 Aerial breeding survey The aerial survey was carried out at the recommendation of DSE (R. Hill, pers. comm.), primarily as a means of overcoming the private land access limitations of ground-based surveys. The survey was undertaken during fine weather conditions on 5th and 7th October, 2010 and lasted six hours. The aerial survey was designed to identify Brolga breeding sites within the proposed wind farm site, part of the RoI and nearby wetlands in the wider region (Figure 13). Prior to undertaking the survey, north/south transect lines were defined throughout the study area. Transect length was dependent on transect location within the survey area boundaries. The survey was undertaken in a fixed-wing, four-seat Cessna 182 RG (retractable undercarriage) flying between 90 and 120 m above ground, at a speed ranging between 209 and 240 km/hour (105 knots). Variation in height and speed depended on flight safety requirements.9 The survey team comprised the pilot, a navigator and two observers. Transect details were provided to the observers by the navigator. One observer was located on each side of the plane. Observers scanned an area of approximately 250 m either side of each transect, using binoculars when necessary. When Brolgas were observed, their location was recorded on a GPS and transect information was noted. This included the transect number, the direction and distance of the birds from the observer and general description of habitat or wetland on which the Brolga was sighted. When flying over the birds, it was observed that the Brolgas looked up, but did not appear distressed or fly off. No ground-truthing was considered necessary as Brolgas were readily identified from the air and they were standing next to nests when observed.

Limitations of aerial surveys Aerial surveys can be inaccurate. Their speed means that some nests and birds may be missed; the distance at which aerial observers operate may miss birds hidden in vegetation. The combination of the initial on-ground assessment in 2009 and the aerial surveys in 2010 provided information consistent with historical records of breeding Brolgas. Therefore, the combination of ground and aerial survey results are considered to generate representative data on breeding locations. Notwithstanding this, from 2011, more thorough ground-based surveys, also

9 The methodology selected in discussion with DSE differed slightly from that subsequently recommended by DSE in the (2012) Brolga Guidelines.

Page | 51 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) involving extensive liaison with participating landowners, were undertaken as described below.

2011 - 2013 - breeding season surveys and nest watches Formal roaming observational surveys were undertaken by two local observers retained by Trustpower. The observers visited wetlands and lakes within, but not limited to, the RoI on a fortnightly basis. Each fortnight, between approximately 35 and 85 wetlands were visited of the 189 mapped in Figure 18. Wetlands visited depended on suitability as breeding sites during time of survey. Time spent at wetlands varied between a few minutes (when no birds could be found using optical equipment from vantage points) up to two hours when breeding pairs were present. The choice of wetland to visit during the roaming surveys was based on them being potentially suitable given historical records, or whether local knowledge of field staff and information from landholders indicated that they held water. Dry wetlands were not visited. The observers were in regular contact with other local residents and recorded any incidental sightings. Any sightings reported to them or to Trustpower were followed up by the observers for corroboration. The observers have an excellent local knowledge of past and present Brolga locations and the area in general. Hence the amount of information gathered by observers from formal and incidental recordings was substantially more than a more structured scientific study (i.e. wholly conducted by ornithologists). This provided excellent data on Brolga activity in the region. Targeted flight path observations (see below) were undertaken outside the formal fortnightly roaming surveys. Considering the observers were local residents, they were also able to record Brolga records daily as incidental records in addition to the fortnightly formal searching. Some nests were also monitored regularly to record the fate of breeding attempts. This was undertaken at 13 nest sites within ten km of the proposed wind farm from 2012 to 2013.

Limitations of breeding surveys Three of the 76 properties that may support Brolgas were inaccessible (as noted in Section 3). Further, at the request of landholders, some wetlands were avoided during particular times of the year due to farming practices such as lambing season, which is a vulnerable time for disturbance. The contacts of the observers with local residents overcame some of these limitations through regular exchange of knowledge about Brolga sightings and breeding sites in the RoI.

Page | 52 (! (!

Kilometres 0 1.5 3 6 Legend Figure 13: Aerial survey transect lines and Brolga sightings

Study Area Project: Dundonnell Wind Farm Flight line Client: TrustPower Pty Ltd Project No.: 9184 Date: 25/09/2013 Created by: I. Kulik / M. Ghasemi (! Brolga sightings Wind Masts ¯ Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

4.2.2. Survey results The number of breeding sites recorded during the studies, and based on historical information, is provided in Table 10. A total of 47 wetlands within the RoI has been used for breeding since 1980; these are tabulated below, along with the likelihood that a particular wetland will be used in the future for breeding based on existing information on the current condition of the wetland. As indicated in the Brolga Guidelines, historically recorded breeding sites can be ruled out of further consideration if their capacity to hold water has been compromised by wetland drainage for agriculture in the intervening period. Table 10: Summary of known breeding records and likelihood of future breeding attempts at wetlands Note: italicised rows refer to wetlands that are permanently drained and no longer able to support breeding Brolgas.

Wetland Likelihood of future Observations Number breeding attempts

The wetland was a partly drained lake in approx. Sub-optimal, however 5 2007. Wetland assessed as being of moderate quality may be occasionally in 2011. used

Approx. 15 ha seasonal wetland. Wetland assessed 79 Likely as moderate quality in 2011. Landholder record. Large freshwater swamp with abundant aquatic vegetation. Was recorded as breeding during the field 82 Likely assessments. Wetland assessed as being of high quality in 2011. Old historical record. Swamp supports some aquatic Sub-optimal, however 90 vegetation, though it is grazed by stock. Wetland may be occasionally assessed as being of moderate quality in 2011. used Old historical record. Freshwater tussock wetland Sub-optimal, however 103 which is shallow. Supports some vegetation. Wetland may be occasionally assessed as being of low quality in 2011. used Large shallow freshwater swamp with emergent 118 vegetation. Wetland assessed as being of high quality Likely in 2011, but has been dried out more recently. Shallow freshwater swamp with some aquatic 120 vegetation. Wetland assessed as being of moderate Likely quality in 2011. Large shallow freshwater wetland with emergent and fringing aquatic vegetation. Recent breeding record 127 Likely in 2011. Wetland assessed as being of high quality in 2011. Shallow freshwater swamp with abundant emergent 130 vegetation. Recent breeding record. Wetland Likely assessed as being of moderate quality in 2011. High quality vegetated wetland. Recent breeding 137 Likely record.

Page | 54 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Likelihood of future Observations Number breeding attempts Drainage line prone to disturbance - close proximity to Sub-optimal, however buildings, heavily grazed, significant habitat quality 145 may be occasionally reduction likely since the 2007 breeding record. used Wetland assessed as being of low quality in 2011. Large shallow freshwater swamp with some emergent Sub-optimal, however 209 vegetation. Likely to be ephemeral. Wetland assessed may be occasionally as being of high quality in 2011. used Large permanent lake with little emergent vegetation. 223 Stony raised banks. Recent breeding record. Wetland Likely assessed as being of moderate quality in 2011. Shallow freshwater wetland with some aquatic 224 vegetation. Wetland assessed as being of high quality Likely in 2011. Shallow dam near a swamp. Some aquatic vegetation Sub-optimal, however 225 present. May be ephemeral. Wetland assessed as may be occasionally being of moderate quality in 2011. used Man-made lake with permanent water. Recent 235 breeding record. Wetland assessed as being of Likely moderate quality in 2011. Tussock swamp surrounded by volcanic barriers. Breeding record in 2011. Nest disturbed by cattle. 236 Likely Wetland assessed as being of high quality in 2011. Recent breeding record in 2012. Located along Smiths Lane, north-east of Lake Gellie. 241 Likely One pair recorded nesting in September 2012. Swamp fenced off from stock. Recent breeding 242 record. Wetland assessed as being of moderate Likely quality in 2011. Large permanent freshwater lake with some fringing vegetation. Surrounded by crop. Recent breeding 243 Likely record. Wetland assessed as being of moderate quality in 2011. Lake often seen with Brolgas. Recent breeding 248 record. Wetland assessed as being of high quality in Likely 2011. Historical nesting site. Permanent, shallow freshwater swamp with no signs of disturbance. Wetland 251 assessed as being of high quality in 2011. Nest Likely abandoned in 2012. Recently nested and hatched chick in 2013. Permanent water. Banked spring. Vegetated. Wetland assessed as being of moderate quality in 252 Likely 2011.Recent nesting record in 2012 and 2013. Two nests by two different pairs. Swamp with shallow water and tussock vegetation. 253 Likely Recent breeding record in 2011 and 2013 Ephemeral shallow wetland with emergent vegetation. 254 Likely Breeding occurred here in 2013.

Page | 55 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Likelihood of future Observations Number breeding attempts Permanently drained swamp. Disturbed and no 303 Unlikely aquatic vegetation. Man-made. Brackish semi-permanent lake. Recent 304 breeding record in 2012 and 2013. Wetland Likely assessed as being of high quality in 2011. Shallow freshwater swamp with tussocks and native 305 vegetation. Recent breeding record. Wetland Likely assessed as being of high quality in 2011. Shallow freshwater wetland. No signs of disturbance. 306 Recent breeding record. Wetland assessed as being Likely of high quality in 2011. Small semi-dry swamp with vegetation.Historical 311 Likely nesting site.

Permanent wetland with aquatic fringing and 320 emergent vegetation. Wetland assessed as being of Likely high quality in 2011. Man made permanent lake with some fringing 322 vegetation. Wetland assessed as being of moderate Likely quality in 2011. Dry marsh land. Likely to be ephemeral. Wetland 503 Likely assessed as being of low quality in 2011.

Swamp with shallow water and emergent vegetation. 504 Likely Wetland assessed as being of low quality in 2011. Large swamp which is now permanently drained. Far 533 south-western depression can hold water and has Likely been used for breeding. Sometime forms wet depressions in wet years. Lacks Sub-optimal, however 539 aquatic vegetation. Wetland assessed as being of low has been occasionally quality in 2011. used Nerrin Nerrin Swamp, c. 240 ha seasonal freshwater 546 swamp. Landholder record. Assessed as high quality Likely habitat in 2011.

568 Small seasonal freshwater wetland. VBA record. Likely

Located north of Blomleys Lane along Darlington- 587 Carranballac Road. Recorded nesting in October Likely 2012. Freshwater swamp with high emergent vegetation. Sub-optimal, however May be ephemeral. Recent breeding record. 601 may be occasionally Wetland was assessed as being of moderate quality used in 2011.

Page | 56 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Likelihood of future Observations Number breeding attempts Recent breeding record in 2013. Two chicks hatched but did not fledge. Wetland was not assessed in 602 2011, however observations during the 2013 Likely breeding survey indicated that the wetland was of high quality. Suitable wetland present, recent breeding record. 603 Likely Wetland was not assessed by BL& A in 2011. Swamp on Darlington Road. Recent breeding record, though disturbance may be high. Wetland was not 604 assessed by BL&A in 2011, however in 2007 the Likely wetland was assessed by BL&A as being of moderate quality with a likelihood of future breeding attempt. Man-made wetland. Dam with moat which provides 605 suitable nesting habitat. Assessed as high quality. Likely Nesting occurred here in late 2013.

606 VBA nesting record along Carlons Road. Likely

Landholder reported breeding in the 1980s, but 607 Unlikely wetland since permanently drained. Landholder reported breeding in 2000, but wetland 608 Unlikely since permanently drained.

The results presented in Table 10 indicate that, based on current knowledge, eight wetlands are sub-optimal, but could potentially be used in the future, and three are unlikely to be used in the future due to permanent draining. The remaining 36 wetlands have higher potential to support future breeding. Breeding sites within 3.2 km of proposed turbines were: . wetland 137 (three km south-west of the nearest proposed turbine); . wetland 602 (1.2 km east of the nearest proposed turbine) wetland 533 (approximately 0.8 km north of the nearest proposed turbine); . wetland 118 (approximately 2.6 km west of the nearest proposed turbine); and . wetland 254 (1.9 km south-east of the nearest proposed turbine). The pair on wetland 137 has regularly nested at this location and has successfully raised young in previous years (BL&A unpubl. data, landowner pers. comm.). This site was used in 2008, 2010, 2011 and 2013. Whilst the pair was not recorded nesting in 2009, a pair was sighted foraging there late in November 2009. During the 2012 surveys, a total of seven active nests sites were recorded and included wetlands 236, 241, 251, 252, 304, 587 and 602. Chicks hatched from wetlands 252, 304 and 602. A chick at 304 was observed to have successfully fledged. The nest located in wetland 602 was recorded as a breeding site for the first time during these investigations in 2012, although the landowner has historical observations of its past use as a breeding site (R. Hill, pers. comm.). The pair incubated two eggs which both hatched. One chick grew to a metre high before it disappeared.

Page | 57 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

During August 2013, Brolgas settled into wetlands and nest building commenced. Seven active nests were recorded during the 2013 breeding season survey in the RoI and included wetlands 137, 251, 252, 253, 254, 304 and 605. Four of these sites were also used in 2012. The Brolga pair nesting at Veals Lane (137) raised one chick, and Brolga pairs at wetlands 251 and 252 (well beyond 3.2 km from proposed turbines) successfully raised chicks during the 2013 breeding season. The number of active nests within the RoI is likely to vary between years depending on several factors, including rainfall. No nests have historically or recently been recorded within the proposed wind farm turbine layout. Flight paths of Brolgas during the 2011 to 2014 observation period are shown in Figure 15. Although numerous pairs attempted to build nests and incubate, it was observed over the duration of the study that no more than five pairs laid eggs and commenced incubation within the RoI at any one time. This is a lower breeding density than has been observed to the south (e.g. around the eastern part of the previously proposed Mortlake East wind farm site, south of the Hamilton Highway).

4.3. Flocking season surveys

4.3.1. Survey methods Observations of the Brolga population at Dundonnell indicated that there was not a clear delineation between the breeding and flocking seasons, as has been observed in other areas. The methods here therefore refer to periods when Brolgas were observed engaging in typical flocking behaviour in the RoI.

Observational surveys Flocking Brolga surveys within the proposed wind farm RoI were undertaken in four periods: . In 2011 an assessment of wetlands was undertaken and wetland habitat quality was again assessed well after the breaking of a long drought, after a period of above average rainfall (see Section 3 for methods and results). . Fortnightly roaming surveys from 30th April to 11th November 2011, including targeted unseasonal flocking flight path observations from 2nd June to 11th November 2011 (extending into the breeding season due to an unseasonal flocking event); . Fortnightly roaming surveys from 16th August 2012 to 1st March 2013 including flight path observations from 16th August to 30th December 2012; . Fortnightly roaming surveys from 27th January to March 2013, including flight path observations (south-east) between 11th April 2013 and 1st August 2013. . Fortnightly roaming surveys from 1st January 2014 until mid-June 2014 including targeted flocking behaviour observations on 20th February 2014 of a flock in the south-east of the wind farm site and from 9th March 2014 of a flock at Lake Gellie. A constant monitoring exercise was undertaken from 25th to 27th February 2014 recording flight paths of the flock in the south-east of the site. . Incidental flocking sightings from April to June 2012 were also recorded.

Page | 58 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Fortnightly roaming surveys were undertaken on any site that had previously supported flocks of Brolga (more than ten birds) as well as wetlands that held water during the time of the survey. Where the field staff obtained incidental observations of flocking Brolgas or news of flocks from local landholders, they included the areas in their fortnightly surveys and followed up landowner reports within a few days with a specific site visit. The number of sites surveyed per fortnight fluctuated according to water levels, and ranged from 64 to 138 wetlands within the RoI from 2011 to 2014. Where flocking was found at a wetland within five km of the proposed wind farm boundary, more detailed, targeted behavioural observations (including documenting flight paths) were triggered, to address the requirements of the Brolga Guidelines. Although no targeted behaviour observations were conducted in early 2012, incidental flocking sightings by the local observers and sightings by local landholders were recorded from April to December 2012, including at Lake Barnie Bolac and wetlands 522 and 240 located south of Nerrin-Pura Road. Total flocking season survey effort between 2011 and 2014 totalled a minimum of 325 days (excluding incidental observations), covering all wetlands within 10 km of the wind farm site over the three flocking seasons and one breeding season with a flocking event investigated. Details of the survey methods are provided below. Surveys were undertaken by vehicle and on foot. As previously mentioned the field staff also recorded incidental sightings when travelling in the region. The formal, fortnightly surveying focused on historical flocking sites, wetlands that held water and sites reported to the local observers as holding flocking birds within or near the wind farm RoI. Between 64 and 138 wetlands out of the 189 in the RoI were visited each fortnight, depending on whether they held water or not. Wetlands not visited were either dry or not accessible temporarily due to agricultural practices (e.g. lambing or crop harvesting). Time spent at wetlands varied between a few minutes (where no birds could be found using optical equipment from vantage points) up to two hours where flocks were present. Targeted flight path observations (see below) were undertaken outside the formal fortnightly roaming surveys. All Brolga sightings were recorded on topographic maps and information noted on behaviour and flights. If Brolga flocks were found during the fortnightly roaming surveys within five km of the proposed wind farm boundary, more detailed, daily observations were undertaken of their flights Observations were taken from no closer than 300m from the birds so that Brolga behaviour was not modified by the presence of the observer. Like other large-bodied birds with low manoeuvrability, Brolgas have very long flight initiation distances (i.e. distance from a human at which they flee; Price 2008). If possible, photographs were also taken. The information below was systematically recorded. . Date and time (beginning and end) of observation period; . The weather (including temperature, cloud cover, wind direction and strength, and rainfall) for the observation period;

Page | 59 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Habitat type and quality at the observation site and habitat type at nearby foraging sites; . Number of Brolgas present at the site and their age (i.e. adult or juvenile/fledgling, where possible); . Dominant behaviour at the site (flying, feeding, preening, resting, walking or displaying); . The height of any flight; . The direction of any flight; . The habitat to which the bird flew (lake, wetland, crop, grassland, other); and . The start and end locations and times of any observed flight. Flocking data were consolidated by tabulating flocks by location, date observed and numbers of Brolgas. Flocking flights observed at sites closer than five km from the proposed wind farm boundary were mapped on an aerial photograph of the region surrounding a flocking site, based on the observed flight path of the birds relative to obvious landscape features (e.g., roads, trees, houses, watercourses, etc.) and all flight paths were then consolidated onto one map (Figure 15).

Limitations and assumptions The study area covered a large area but access to three private properties (of 76) within the RoI was not permitted by the landowners. Every effort was made to gain access to areas frequented by Brolgas. Where access was not permitted, it is not possible to say what Brolga activity has occurred in recent years. For these areas, historical records were relied upon, some of which were extensive. It was assumed that flocks of the same number of individuals seen on the same day and location were the same flock, if seen by different observers. As it was impossible to determine or track individuals, observations cannot be considered independent.

4.3.2. Survey Results The results of the flocking season surveys are presented here. To summarise, the RoI has a total of 22 traditional flocking sites and 24 one-off flocking sites, including historical records at seven of these sites involving foraging birds in non-wetland habitats where they are unlikely to have roosted overnight (see Table 15 and Figure 19).

Page | 60 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Flocking Season 2011 During the 2011 surveys, the RoI was found to support two distinct flocking areas (eastern and western), encompassing behaviour consistent with both traditional and one-off flocking activities. The eastern flock used wetlands 235 and 304, with birds from these main sites moving to wetlands 244, 252, 302, 311, 324, 505, and 603. The western flock was found to use wetland number 585 and move to wetlands 87, 88, 90, 91, 112, 117, 118, 138, 513, 537, 538 and 584. Figure 15 shows the location of flocking activities, including all recorded flight paths of Brolgas from the main flocking sites. Birds from the western flocking site used habitats situated immediately west of the nearest proposed turbines, whilst birds from the eastern flocking site used habitats at least 3.3 km from the nearest proposed turbines. At any one time, the total surveyed numbers ranged between 10 and 52 individual Brolgas, while the most frequently sighted number of individuals was 29. The maximum number of birds recorded at the western site on any one occasion was 32 birds, compared with 20 at the eastern site.

Figure 14: Male Brolga displaying in a flock at wetland 585

Flights and habitat at destination During the 2011 surveys 37 flights were recorded, involving a total of 249 bird- flights. Three flights were recorded from the eastern site, 33 from the western site and one flight of 14 birds could not be attributed to either site. Habitats at the destination were predominantly wetlands and farmland. The birds roosted at wetlands and foraged in wetland and farmland habitats. Brolga were also recorded foraging in freshly cultivated paddocks and at sheep feeding areas (i.e. where stock were being fed grain). Based on these observations, Trustpower collaborated with two local landowners to alter their sheep feeding locations and Brolgas responded to this by changing their foraging area to the location of the sheep feeding stations. This was expected given that Brolgas are known to be opportunistic feeders.

Page | 61 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Flight distances, heights and paths Three flights were below 20 m, and hence below rotor swept area (RSA) height, with the majority of these (34 flights) being recorded within RSA height and none above this. Flight distances from the main flocking sites ranged from approximately 350 m to greater than two km. Half of the movements exceeded one km. The flight time ranged between less than one and six minutes, with half of the flights being a minute or less. Observations during the 2011 field assessments indicated that Brolga using the two flocking sites generally used the same flight paths, indicating regular movements to habitual foraging areas.

Page | 62 158 547

572 546 571 148 574 206 149 573 147 541 569 545 548 565 566 146 563 542 543 568 152 544 567 562 570 79 564 153 575 530 590 553 560 315 78 557 323 224 317 561 314 528 558 559 529 316 532 527 208 556 531 526 225 125 126 583 576 525 75 582 504 600 552 240 223 209 312 82 522 322 523 127 521 320 524 601 519 128 123 130 517 516 222 124 518 520 550 85 129 129 122 503 549 321 86 133 120 121 539 502 241 501 Lake Gellie 327 587 535 245 239 242 533 238 555 87 536 534 326 506 510511 Chinamans Swamp 88 243 221 118 505 507 512 91 508 515 509 513 514 117 246 90 92 538 310 537

607 309 585 Lake Terrinallum 111 244

112 324 307 308 584 325 254 311 113 602 114 Tiverton 141 137 313 97 138 605 139 Lake Sheepwash 306 236 110

328 235 247 145 301

101 144 304 305 9 252 109 586 11

103 Lake Barnie Bolac 253 303 248 302

5 Long Dam 603 604

251

Kilometers Legend Flight paths 2011 Flight paths 2013 0 1.5 3 6 Wind farm boundary Flocking event 1 Breeding Figure 15: Observed Brolga flight paths during Roads Flocking event 2 Flocking event flocking seasons Incidental Incidental Wetland Quality Project: Dundonnell Wind Farm BUS High Flight paths 2012 Client: TrustPower Pty Ltd Moderate Breeding event 1 Flight paths 2014 Project No.: 9184 Date: 12/09/2014 Created By: M. Ghasemi Low Breeding event 2 Flocking event 1 Not accessable Flocking event Flocking event 2 ¯ Incidental 251 Wetland Number Incidental BUS Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Flocking Season 2012 Five wetlands were used by flocks of Brolgas during the 2012 flocking season survey (wetlands 248, 252, 235, 240 and 522). Based on field observations, two flocking aggregations were found: northern and southern. Flight paths recorded during the 2012 survey are presented in Figure 15 and details are provided in Appendix 2.

Northern flocking site The northern flock was found roosting in wetlands 240 and 522 over six km from the proposed wind farm boundary and foraging elsewhere. The lack of frequent or prolonged records during the flocking season, and the lack of historical flocking records at wetland 522 indicated that this was most likely a one-off flocking event. Historical records of flocking from the landholder indicated that wetland 240 was likely to be a traditional flocking site. The maximum number observed on these wetlands and surrounding paddocks was 47 individuals that comprised mostly adults. This flock was thought by the landholder to have flown from a traditional flocking site located at Blue Lake (wetland 553), just beyond the RoI.

Southern flocking site The southern flock comprised birds that consistently used wetlands 248, 252 and 235, from six to 7.5 km from the nearest proposed turbines. A flock of 32 Brolga was recorded at Lake Barnie Bolac (wetland 248) and nearby wetlands 252 and 235. The use of flocking sites is likely to be dependent on wetland condition and it is possible that a varying number of these sites can be used in any one flocking season. This is corroborated by observations of the flock moving on as a result of Lake Barnie Bolac (wetland 248) drying out during the 2012 flocking season.

Flight paths and habitat at destination As no flocking events were within five km of the wind farm boundary as proposed in 2012, no intensive flight path observations were undertaken and only incidentally observed flight paths are mapped in Figure 15. During the 2012 surveys 42 flight paths were recorded involving 240 bird-flights. Six flights were recorded from a flock that was at neither the northern or southern flocking site, 26 from the northern site and 10 from the southern site. Flight paths recorded during the 2012 survey are presented in Figure 15 and details are provided in Appendix 2. Flight paths were not recorded for flocks recorded in wetlands 248, 252 and 235. Twenty-nine flights, involving 209 birds (67% of all flights), were less than 30 m above the ground, with 14 flight paths, involving a total of 31 bird-flights, between 30 and 80 m above the ground. This was a comparable height distribution to the observations during the 2011 survey.

Page | 64 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Incidental observations Incidental flocking observations were undertaken from April to June 2012 by local field staff to record Brolga activity in the region. In addition, any sightings made by local landowners during this time were also recorded. Observations are summarised below. Local observer sightings: . Wetland 235: This has been recorded as a traditional flocking site. Numbers between April and May 2012 fluctuated between 34 and 42 birds. . Wetland 110: No flocks were recorded at this wetland in early 2012. However, there was a pair recorded foraging at this site in late May 2012. . Wetland 118 (Chinaman’s Swamp): There was one sighting (of six Brolgas) at Chinaman's Swamp and surrounds in May 2012. One Brolga pair was present in this area during most of the flocking season. Records from landowners in the region: . Wetland 79: Located just outside the RoI. Brolgas are usually present in this area most years. From March to July 2012, numbers fluctuated between 30 and 60 birds. . Wetlands 153 and 323: Located in the north of the RoI. A flock of 40 Brolgas was recorded here over a few days in April 2012. This was a one-off flocking event. . Wetland 522: Between 20 and 50 Brolga were recorded early in the flocking season. Based on local landholder observations, it is suggested that the flock is part of the Blue Lake flock (wetland 553), located outside the RoI. . North-east of wetland 224: On 27th April 2012, a flock of 40 Brolgas were observed here. The flock split up and moved south, but no more than a few km. They returned soon afterwards to feed. (This was a foraging flock, not a roosting flock, and hence is not included in Table 15 or Figure 19.) As all of these observations were beyond five km from the wind farm boundary as it was proposed at the time, detailed flight path observations were not initiated.

Flocking Season 2013 A fortnightly roaming survey was undertaken by two local observers from January to December 2013. The majority of wetlands surveyed were dry or held little water. Several held more than 50% and a few were full towards the end of the winter in August after several rainfall events. During April to August 2013, detailed observations of flocking Brolgas were undertaken by local field staff and BL&A personnel. This was when flocking behaviour was first detected in the 2013 flocking season within five km of the proposed wind farm. This flock was concentrated in and around Lake Sheepwash (wetland 110), 2.9 km south-east of the nearest proposed turbine. Flight observations revealed that birds at this site had a distinct routine (Figure 15). Birds would roost at Lake Sheepwash (wetland 110), leave the site at first light and fly south to pasture and cultivated arable land surrounding wetlands 586, 303, and 252. In the evening they would return to Lake Sheepwash to roost on

Page | 65 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) dark. Numbers were consistently around 30 to 40 birds (maximum 46 birds observed) from April to June). Birds were also seen flying south towards wetland 603 (Long Dam – no access) and surrounding paddocks. Around mid-June, numbers declined, and field staff reported that the flock had moved east from Lake Sheepwash (wetland 110) to wetland 304. A new pattern emerged in which they would flock at wetland 304 and fly west to wetland 252 to forage in surrounding paddocks of pasture and cropping land. Numbers were consistently between 20 and 34 birds from June to July. By August 2013, numbers had dropped to less than ten birds, as the flock started to disperse in preparation for breeding. Two pairs were observed flying from wetlands 139 and 313 to wetland 602. These flights were recorded within the turbine exclusion zone to the south-east of the wind farm and occurred when the birds were not part of a flock. A landowner in the north-east of the RoI reported a flock of 37 birds over a two- week period during April 2013 at the southern end of Lake Gellie (wetland 239), two to three km north-east of the nearest proposed turbines. No formal flight path surveys were undertaken during this two-week period. This flock moved south to Lake Sheepwash where flight path surveys continued. Two Brolga pairs continued to reside at Lake Gellie over the following months (May to July 2013) and flight paths of these pairs were mapped and are presented in Figure 15. Flights were generally in a north-south direction, to mostly arable/pasture habitats.

Flights and habitat at destination As the flocking birds at Lake Sheepwash and wetland 304 occurred within five km of the proposed wind farm boundary, detailed flight path observations were made to ascertain whether birds regularly went to forage during the day in the proposed wind farm site. A total of 613 flight observations were recorded from April to August 2013 (see Figure 15). Approximately 37 per cent of flights by flocks of Brolgas (Table 11) were recorded at a height of less than 20 m. Remaining flights were above this height. Brolgas were in flight usually between one and two minutes. A total of 131 (21%) flights by Brolga flocks flew out of sight and their final destination could not be determined. A small number of incidentally observed flights at this time were observed generally within the south-eastern turbine exclusion zone (see Figure 15). Table 11: Flight heights recorded during the 2013 flocking season survey 80- Heights (m) 5-10 10-20 20-30 30-40 40-60 60-80 Total 100 Number of flights 28 198 130 109 116 28 4 613 Per cent of total 4.6 32.3 21.2 17.8 18.9 4.6 0.7 100.0 The majority of habitats to which Brolga flocks flew were wetlands (52%) followed by arable land (27%), pasture and grazing paddocks (7%), stony outcrops (3%) and unknown habitats (11%). Occasionally Brolgas were observed in other habitat types including grasslands, grazing paddocks and spring fed drains. If Brolgas flew out of sight, the habitat type was recorded as ‘unknown’. Observational effort had increased significantly during the 2013 flocking survey compared with previous years, and all flight paths were mapped. Figure 15 illustrates a large proportion of flights observed south-east of the wind farm site

Page | 66 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

(613 in total). A total of 1,246 person hours of work was undertaken during the 2013 flocking season survey compared with 829 person hours during the 2012 and 2011 surveys.

Flocking season 2014 From February to March 2014, a flock of 17 Brolgas was observed roosting at wetland 139 (located along Woorndoo-Dundonnell Road). This flock comprised seven resident birds (two adult pairs and a family group with two adult birds and an immature). The remaining ten birds moved around the landscape together and joined the resident birds to feed on grain trails (used for sheep feeding) 800m north of wetland 139, and three km north-east near wetland 602. The flock was also observed at Lake Sheepwash (110) and wetland 252. Most of the observed flights were within the southern part of the wind farm site (see Figure 15).

4.4. Migration season surveys

4.4.1. Survey methods Based on the wetland quality assessment in 2011, 94 wetlands and 11 dams were assessed as being suitable for migrating Brolga (see Figure 16). If wetlands held little to no water, showed evidence of disturbance (e.g. were drained) and held no previous Brolga records, they were excluded from the migration season surveys. The remaining wetlands were each surveyed for Brolga five times between 5th January and 2nd March 2011. Wetlands and all intervening dry land areas were accessed by vehicle and searched from the roadsides using binoculars and telescopes. The information below was recorded at each site. . Date and time (beginning and end) of observation period; . The weather (including temperature, cloud cover, wind direction and strength, and rainfall) for the observation period; . Number of Brolgas present at the site and their age (i.e. adult or juvenile/fledgling); . The height of any flight; . The direction of any flight; . The habitat to which the bird flew (lake, wetland, crop, grassland, other), if known and . The start and end location and time of any flight, where observable.

Limitations A vehicle-based roaming survey is dependent on accessibility of all wetlands suitable for Brolgas. It is essential that observers remain within the vehicle as Brolgas are more tolerant and less “flighty”, allowing closer observation. Access to some wetlands within the study area was not permitted by a small number of landowners (see earlier); however, some wetlands on these properties were visible from public roads.

Page | 67 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Due to the large number of wetlands covered by the survey, some Brolga flights may have been missed. Logistic constraints meant that only a proportion of wetlands could be visited in any one survey session. As Brolgas could move to and from any of the wetlands, they may have been missed. However, this was somewhat offset by visiting wetlands more than once. It is also possible that individual Brolgas were recounted within the same day, as individual identification was impossible.

Page | 68 PARUPA PARK ROAD PLANTATION ROAD 158 547 MCKENZIE LANE 546

LAKE ROAD 572 571 148 574 206 149147 573 569 541 545 548 565566 146 563 TRAWALLA SOUTH ESTATE ROAD 568 152 542 543 SOUTH BEACH ROAD567 562 544 553 564 153 WHITES LANE 79 570 575 530 590 315 78 560 323 224 317 557 561 558 559 314 528 316 529 208 NEVILLES LANE 532 527 556 531 526 225 MORTLAKE-ARARAT ROAD 125 WARINGS LANE 126583576 525 75 582 504 600 552 PAGELS LANE 240 312 82 223 209 522 322 NERRIN-PURA ROAD 127 523521 320 MCKAYS LANE 524 601 123 519 130 128 516 124 517 222 518 520 550 85 129 129 122 503 549 VITE VITE ROAD 321 86 133 120 121 539 502 501 241

ENNERDALE LANE 327 239 587 535 245 SMITHS LANE DARLINGTON-NERRIN ROAD BLOMLEYS242 LANE 533 238 555 87 534 326 536 506 510511 88 221 505 243 118 507 91 512 515 508 509 LYONS LANE WOORNDOO-STREATHAM ROAD 513514 117 DARCYS LANE 246 92 90 538 310 537 FASHAMS LANE DOHERTYS LANE 607 309 606 585 CARLONS ROAD 111 244 POST OFFICE LANE

BOLAC PLAINS ROAD 112 DARLINGTON-CARRANBALLAC ROAD DAWES LANE 324 VEALS LANE 307 308 584 325 254 LADES LANE 311 114113 602

141 WOORNDOO-DUNDONNELL ROAD 137

313 DUNDONNELL-DERRINALLUM ROAD 97 138 605 139 306 110 236

328 235 247 145 301 TERRINALLUM ROAD

WOORNDOO-DARLINGTON ROAD 101 144 318 MOUNT FYANS LANE 304 305 9 NINE MILE LANE 109 586 252

MORTLAKE-ARARAT ROAD 11 103 253 BARNIE BOLAC ROAD 303 248

302 DARLINGTON-SKIPTON ROAD 5

603 NORTH STATION 604

DUNDONNELL-MORTLAKE ROAD 251

SIX MILE LANE DARLINGTON ROAD HAMILTON HIGHWAY JELLALABAD ROAD DARLINGTON ROAD

NORTH STATION

KURWEETON ROAD

CASTLE CAREY ROAD PRICES LANE

DARLINGTON ROAD

STEELES LANE

SIX MILE LANE Metres Legend 01,500 3,000 6,000

FIVE MILE LANE Wetlands surveyed during Wind Farm Site the Brolga migration Figure 16: Wetlands surveyed during the Brolga migration Area of Investigation (10km) season survey 2011 season survey 2011

Wetland Quality Project: Dundonnell Wind Farm High Client: TrustPower Pty Ltd Moderate Low Project No.: 9184 Date: 20/06/2014 Created By: M. Ghasemi Not accessible ¯ Roads Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

4.4.2. Migration season results Usually Brolgas migrate to traditional flocking sites between December and February. Although there was one sighting of Brolgas flocking together in a small social group (a group of seven birds east of Blue Lake (wetland 553) on 31st January 2011), Brolgas remained in pairs or individuals, or family groups during the January – March 2011 migration season survey. This is their usual behaviour during the breeding season (July to November). It is thought that an abundance of wetland habitat in the region, after higher than average rainfall in previous months, may have contributed to this unexpected behaviour in early 2011. Brolgas were seen by observers on each survey and mostly in areas beyond the proposed wind farm boundary but within the RoI. One sighting along Ennerdale Lane was on the boundary of the wind farm and several were seen not far beyond the wind farm boundary at Chinaman’s Swamp. Two observations were on cropped land, two on wetlands, two on lakes and one in stony rise habitat.

4.5. Brolga utilisation survey During the 2012 and 2013 Brolga flocking seasons, a Brolga utilisation survey was undertaken across the wind farm site to characterise Brolga activity across the project area during periods of high Brolga numbers in the RoI. No breeding season utilisation survey was undertaken as numbers of breeding Brolgas in the RoI are comparatively low at this time. The purpose of the utilisation survey was to quantify the rate (number of birds/unit time) at which Brolgas flew across the proposed wind farm site, and the flight location, within a rigorous statistically based method that generates data that can be analysed using a variety of statistical methods for the purpose of testing hypotheses, such as collision risk modelling. The methods used and the results of this survey are documented below.

4.5.1. Methods The Brolga utilisation survey was conducted by two observers during the 2012 and 2013 flocking seasons. The 2012 survey commenced on 6th April 2012 and was completed by 25th May 2012. The 2013 survey was conducted between 9th and 25th May 2013. The survey was undertaken in 2012, when birds were flocking in the RoI (although not consistently within five km of the proposed wind farm boundary), and in 2013 when birds were flocking at Lake Sheepwash (wetland 110), 2.9 km from the proposed wind farm boundary. The locations of survey points were stratified based on habitat to ensure site coverage, then randomly selected within each stratum. Twenty-two fixed survey points were established across the proposed wind farm site; 15 were on stony ground and 7 were on arable/mixed use/semi-cleared land within the wind farm boundary. The survey point locations were chosen from a larger set of randomly located points and spaced as evenly as possible across the wind farm to ensure good coverage of the site. Figure 17 shows the location and habitat of the utilisation survey points.

Page | 70 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

During each survey, each point was visited eight times for a period of 15 minutes, with starting times being spread throughout the day for each point to account for time-of-day differences in Brolga activity. Any Brolga activity seen from the points was recorded. The main difference between the two years is that survey sites 21 to 24 were placed in the additional north-eastern section of the wind farm in 2013, and sites 23, 13, 19 and 20 in the west surveyed in 2012 were not surveyed a second time as this area had, by then, been excluded from the wind farm to avoid impacts on Brolgas. During formal counts, the number of Brolgas, as well as their activity, distance and direction from the observer were recorded. In addition, flight information (height, distance, direction, behaviour, habitat, flight start and end time) was also documented. Each flight path recorded was tabulated and plotted on a map. In addition to the observations during formalised, fixed-point counts, incidental observations of Brolgas were also made while moving about the wind farm site between point counts but these were excluded from the statistical analysis, although they were recorded. These point counts were used to determine the number of flights of concern. Roaming surveys were used to determine the spatial distribution of Brolgas, to estimate the number of active breeding pairs and to estimate the probability density (or utilisation map) of flights. The combination of the two sets of observations generated input into the collision risk model of the Level 3 assessment (see Section 5.2).

4.5.2. Results

2012 Survey Seventeen Brolga observations (14 during 40 hours of formal counts and three incidental observations) were recorded during the utilisation survey (Table 12). Of those, seven were flight observations (four formal, and three incidentals). Over ten observations of Brolgas counted were in arable (cropping) land. The sightings involved birds neither breeding nor flocking. The majority of flight observations were distributed in the north-western corner of the proposed wind farm site (survey sites 19 and 20 in 2012) in arable land near Chinaman’s Swamp (wetland 118, shown in Figure 15). This area is no longer part of the proposed wind farm. Additional incidental flights were recorded outside the formal Brolga utilisation survey times. These flights are included in Table 12 and illustrated in Figure 15. They were observed in similar locations to the flights recorded during the formal utilisation survey.

2013 Survey Eleven Brolga observations were recorded during the 40 hours of formal utilisation survey (Table 13). Of those, six were flight observations. Almost half of the Brolga observations were in arable (cropping) land and the remainder of observations were in stony rises, on a spring or in unknown habitat.

Page | 71 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

The majority of flight observations were distributed in the southern section of the study area (survey sites 2, 3 and 4) in semi-cleared, stony, and arable land respectively. No incidental observations were recorded outside the formal Brolga utilisation survey times in 2013.

Page | 72 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Table 12: Results of the 2012 Brolga utilisation survey

Initial Initial Survey Obs. Flight distance Survey End Obs. End No. Flight Flight Habitat at direction Date Site Start Obs. No. Start Activity Height Habitat at destination from Time Time Brolga Direction Behaviour Origin from time Time (m) observer observer (m)

18 7:30 7:45 incidental 7:31 7:32 2A Flying 5-10 W - - - >1km S

6/04/2012 19 8:15 8:30 1 8:16 8:30 2A Walking - - - Crop Crop 350 SW 2 9:00 9:02 3A Walking - - - Crop Unknown (out of view) 500 W 20 9:00 9:15 3 9:05 9:06 2A Flying 5-10 N flapping Crop Crop 400 E 4 9:08 9:09 3A Flying 5-50 N flapping Crop Crop 500 W 16/04/2012 19 14:15 14:30 1 14:15 14:18 2A Flying 40-80 E Flapping Wetland Arable 150 S

16 9:00 9:15 5 9:01 9:10 2A Walking - - - Crop Crop 2500 NE 17/04/2012

18 10:30 10:45 6 10:30 10:45 2A Walking - - - Crop Crop 900 W 19 11:15 11:30 7 11:15 11:30 2A Walking - - - Crop Crop 700 SE 19 8:15 8:30 2 8:15 8:30 2A Walking - - - Crop - 300 S 26/04/2012 20 9:00 9:15 3 9:10 9:15 1A Walking - - - Crop - 300 S 20 incidental 9:15 9:17 2A Flying 20-40 N flapping Crop Chinaman’s Swamp 300 S 8 18:00 18:06 2A Walking - - - Crop Crop 300 S 25/04/2012 20 18:00 18:15 9 18:06 18:07 2A Flying 20 N flapping Crop Swamp edge 300 S 10 18:07 18:15 2A Resting - - - Swamp edge Swamp edge 300 N 11 8:15 8:30 6 8:15 8:30 2A Walking - - - Dry Lake Lake 150 N 22/05/2012 12 incidental 8:30 out of sight 2A Flying 40-60 Nth Gliding Rubbish Tip Unknown (out of view) 150 NE Site numbers refer to the Brolga utilisation survey points (Figure 17).

Page | 73 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Table 13: Results of the 2013 Brolga Utilisation survey Initial Initial Obs. distance Survey Survey End Obs. End No. Flight Flight Flight direction Date Site Start Activity Habitat at Origin Habitat at destination from Start time Time Time Brolga Height (m) Direction Behaviour from Time observer observer (m) 5 8:15 8:30 8:15 8:30 2A walking, feeding - - - Stony Stony 400 SE 10/05/2013 15 15:00 15:15 15:00 15:15 2A walking, feeding - - - - - 700 NW 21/06/2013 11 17:25 17:40 17:27 17:28 2A flying 20-40 SE Flapping Arable Unknown 220 SW 10/05/2013 4 15:00 15:30 14:30 15:30 2 walking, feeding - - - Arable, near drain Arable, near drain 300 S 2 15:00 15:30 14:58 14:59 2 Flying 5 E, W Slow flapping unknown unknown 300 NW 13/05/2013 15:33 15:56 2 Walking - - - Arable Stony 200 NW 3 15:45 16:00 15:57 15:58 2 Flying 5 SW Gentle flapping Stony Water Spring 200 NW 14:54 14:59 2 walking - - - Arable arable 200 S 16/05/2013 4 15:00 15:30 14:58 14:59 2 flying 2 S flapping Arable unknown 250 S 3 12:45 13:00 12:46 12:48 2 flying 10 NE slow wing beat Spring, 200m W Site 2 Cultivated, Arable 600 W 17/05/2013 26 17:15 17:30 17:24 17:25 2 Flying 20 E Flapping unknown (west of 137) unknown (east of 137) 2km SW Site numbers refer to the Brolga Utilisation survey locations (Figure 17).

Page | 74 Site 19 [! Site 20 Site 18 [! [! Site 17 ^_ ENNERDALE LANE Site 22 ^_ Site 16 DARCYS LANE ^_ Site 21 Site 13 ^_ [!

Site 14 FASHAMS LANE Site 15 ^_ Site 12 [! DOHERTYS LANE Site 23 [! [! Site 11 ^_ Site 10 Site 25 ^_ ^_

POST OFFICE LANE Site 24 Site 8 ^_ ^_ Site 7 ^_ DAWES LANE Site 26 Site 9 Site 6 ^_ [! [!

Site 5 ^_

Site 4 [!

DARLINGTON-NERRIN ROAD WOORNDOO-DUNDONNELL ROADSite 1 Site 3 Site 2 ^_ ^_ [!

DUNDONNELL-DERRINALLUM ROAD

BARNIE BOLAC ROAD

Legend Figure 17: Location of utilisation survey points Project: Dundonnell Wind Farm Wind Farm Boundary Client: TrustPower Pty Ltd ^_ Random Sites - Stony Project No.:9184 Date: 12/09/2014 Created By: M. Ghasemi [! Random Sites - Arable/Mixed/Semi-cleared ¯ Kilometers 0 0.5 1 2 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

4.6. Transmission line

4.6.1. Methods Brolga observations along the proposed transmission line route between the proposed Dundonnell Wind Farm site and the Mortlake Power Station were obtained from the VBA, the BirdLife Australia Birdata database, and BL&A and landowner records. Note that an additional breeding season survey (within 3.2 km of the line) along that portion of the proposed transmission line route further than 10 km from the wind farm site boundary (and therefore not covered in earlier work) will be conducted during the 2014 breeding season. The results will be available for the forthcoming panel hearing.

4.6.2. Results Most historical records are situated in close proximity to the northern section of the transmission line route. All records within ten km of the proposed transmission line are displayed in Figure 18. Records additional to those reported in previous sections are listed in Table 14. Records within ten km of the wind farm boundary are shown in Table 15. These include breeding sites along Nine Mile Lane, Veals Lane, and other ephemeral wetlands within the RoI. Breeding sites located near the southern end of the transmission line route towards the Mortlake Power Station are located at the Mortlake Common and along Connewarren Lane. These records are shown in Table 14. Figure 18 illustrates these results and shows ten breeding sites within three km of the proposed transmission line and 33 between three and ten km of the proposed transmission line. Historical flocking records exist along Darlington-Nerrin Road, Lake Sheepwash, Chinaman’s Swamp and around Lake Barnie Bolac. One reported traditional flocking site is located along Woorndoo-Darlington Road (wetland 301), within one km of the proposed transmission line route. The landowner here reported flocks of 30 to 50 Brolgas in the late 1980s and early 1990s. It is not known if the birds roosted in this wetland overnight so it is designated as a potential traditional flocking site and has been treated as such for this impact assessment. A second traditional flocking site was located at Chinaman’s Swamp (wetland 118), also within five km of the proposed transmission line route. The remaining seven traditional flocking sites are located between five and ten km from the proposed transmission line route. All one-off flocking events occurred near the northern end of the transmission line route, with the exception of one record located in the southern end along Narong Lane (Table 14).

Table 14: Historical/recent Brolga breeding/flocking records for transmission line RoI Distance to Location* Recent Date No. Brolga transmission line Breeding Mortlake Common < 3 km 2008 2

Page | 76 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Distance to Location* Recent Date No. Brolga transmission line Coonewarren Lane < 3 km 2008 2 West Lake Connewarren < 3 km 2008 2 2km north Hexham 5-10 km 1987 2 Cnr Connewarren Lane & 3-5 km 2007 2 Hexham-Ballangeich Rod Cnr Hamilton Hwy and 6.5km 2007 2 Castel-Carey Rd Flocking – one-off sites Narong Lane 5-10 km 1987 35

* includes only records not in Table 15

Page | 77 C GEDDES ROAD

(! *# MCKENZIE LANE *# (!(! (!(! *# /" (! BERRAMBOOL ROAD LAKE BOLAC-WESTMERE#0 ROAD.! #0)" *# *# WHITES LANE /" *# *# (! # *# .! /") 0 NEVILLES LANE

MCINNES LANE KARMALA LANE .! *#/" WARINGS LANE /" ASTONS ROAD .!( )" )" )" PAGELS LANE CHATSWORTH-WICKLIFFE ROAD #0/" )" # *# )"MCKAYS LANE *)")" ")"/ " *# *# ).! MYUNA LANE )/" *# PATTISONS LANE )" )" )" )" #0 )"

(! SMITHS LANE " ENNERDALE LANE (!! /") CHATSWORTH-BOLAC ROAD )" (!. )" GALES LANE )" DARLINGTON-NERRIN ROAD *# )"(!)".! )" *# DARCYS LANE WOOLSTHORPE-STREATHA#0)" ROAD *# WOORNDOO-CHATSWORTH ROAD FASHAMS LANE HEARNS LANE *# (! (! POST OFFICE LANE *# .!(! .! DAWES LANE (! )" BOLAC PLAINS ROAD WOORNDOO-DARLINGTON ROAD GEDDES LANE " ) VEALS LANE *# *#)"*# LADES/" LANE WOORNDOO-DUNDONNELL ROAD *#)"/ CARAMUT-CHATSWORTH ROAD )"/")" *#0 MATHERSONS LANE )" (! )" )" HEXHAM-CHATSWORTH ROAD .!( /")" )" (!)"(! )" .! .!)" BOORTKOI ROAD MOUNT FYANS LANE (! ") .!()" )"/ )" NINE MILE LANE (! STREATHAM ROAD /")"

MORTLAKE-ARARAT ROAD /") (! /")" (! BARNIE BOLAC ROAD WORDLEYS LANE )" (!(!)" (!)")"(!)" )" ( .! )" (! BOORTKOI LANE )" ") CREARERS LANE )"(! *# NORTH STATION "))" (! HAMILTON HIGHWAY WOODCUTTERS LANE /" DARLINGTON ROAD )")")"")

FOUR MILE ROAD CASTLE CAREY ROAD NORTH STATION )" *# DARLINGTON ROAD BOONERAH ESTATE ROAD )"")

STEELES LANE HAMILTON HIGHWAY HARDYS LANE

SIX MILE LANE

FIVE MILE LANE

KINGS LANE )" CONNEWARREN LANE " )" EIGHT MILE LANE ) WAGGS LANE )" )" CEMETERY LANE IMMIGRANTS LANE HEXHAM-BALLANGEICH ROAD PURCELLS LANE CASTLE CAREY ROAD TERANG-MORTLAKE ROAD KILNOORAT ROAD REICHMANS LANE HOPKINS HIGHWAY NEW BRIDGE ROAD

HINKLEYS LANE MAHONEYS LANE MCRAES LANE

GORES LANE GORDONS LANE SMITHS ROAD

TEMPLETON ROAD RED LANE KOLORA LANE

MORTLAKE-FRAMLINGHAM ROAD

WILSONSCOORAMOOK LANE FOSTERS LANE TAPPS LANE LONDRIGANS LANE BLACKS ROAD

PURNIM LANE

DARLINGTON-TERANG ROAD

WALLACE LANE

MCCRAE ROAD

TUFFS LANE

CAMPBELLS LANE Legend Wind Farm Site VBA Sheldon Powerline Alignment *# One-off Flocking *# One-off Flocking 3 km Study Area for Breeding Sites Kilometers (! Traditional Flocking (! Traditional Flocking 0 2 4 8 5 km Study Area for Flocking Sites )" Breeding 10 km Study Area for Flocking Sites BL&A 2009-2013 Investigations Figure 18: Brolga records along proposed Powerline Route *# One-off Flocking Roads DSE Unpublished 2010 (! Traditional Flocking *# One-off Flocking Project: Dundonnell Wind Farm Wetland Quality )" )" Breeding High Breeding Client: TrustPower Pty Ltd Landowner Low BL&A 2007-2008 Investigations #0 One-off Flocking Project No.:9184 Date: 12/09/2014 Created By: I. Kulik / M. Ghasemi Moderate )" Breeding .! Traditional Flocking Not accessible /" Breeding Birdlife )" Breeding- wetland drained, ¯ )" Breeding not suitable as breeding site anymore Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

4.7. Summary of Results A full list of historical and recent breeding, flocking and one-off flocking sites in the vicinity of the wind farm, including the results from the current assessment, is provided in Table 15, which refers to the mapped wetland numbers in Figure 19.10 Note that one-off flocking sites also include non-wetland areas (mostly cropped and arable land) that have supported foraging flocks of Brolgas (i.e. when they have not been recorded roosting overnight). Table 15: Historical and recent Brolga observations in wind farm RoI and nearby area Note:  italicised rows are wetlands that have been permanently drained and are no longer able to support breeding Brolgas)  shaded wetlands are outside the wind farm RoI but in the nearby area (as shown in Figure 19) Wetland Year (s) of Observation Source Number* Breeding 5 2007 VBA 79 Historical records Landholder 82 Historical records, 2011 Landholder, field staff 90 1984 VBA 103 1984 VBA, Landholder, Field staff, BL&A 118 1984, 2011 VBA, field staff, BL&A (2.53) 120 1984 VBA 127 2010 DSE Unpubl. Records, field staff 130 Historical records, 2011 Landholder 137 2008 - 2013 Landholder, BL&A, field staff (2.90) 145 2007 VBA, field staff 209 2009 BL&A 223 2010 DSE Unpublished Records 224 Historical records, 2011 BL&A, 225 Historical records Landholder 235 2010 VBA, field staff 236 2002-2012 VBA, BL&A, field staff, Landholder 241 2012 field staff 242 1988, 2011 VBA, Landholder 243 2010 VBA 248 1992, 2002 VBA, field staff 251 1980 - 2013 VBA, BL&A, field staff, Landholder 252 1980 - 2013 VBA, BL&A, field staff, Landholder

10 Where several sources identified the same site, it was only counted once.

Page | 79 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Year (s) of Observation Source Number* 253 2007, 2008, 2011, 2013 VBA, field staff, landholder 254 2013 Field staff (2.0) 303 1980 - 2002 VBA 304 2012, 2013 Field staff 305 Historical records, 2011 Field staff, Landholder 306 1993, 2010 VBA 311 Historical records Landholder 320 2011 Field staff 322 Historical records VBA 503 2011 Field staff 504 2011 Landholder 533 2010 BL&A (0.75) 539 2002 VBA 546 Historical records Landholder 568 1992 VBA 587 2012 Field staff 601 1998, 2011 Landholder, Birds Austr., field staff 602 2000s, 2012 Landholder, field staff (1.4) 603 1980-2009 VBA, BL&A 604 1980-2007 VBA, BL&A 605 2013 Field staff 606 Historical record VBA 607 1980s Landholder 608 2000 Landholder Flocking - Traditional Sites 110 BL&A, Field Staff, Landholder, 2007, 2009, 2010, 2011, 2013 (3.2) Philip Du Guesclin 112 late 2000s, 2011 Field staff, Landholder 118 80s and 90s, 1988, 2001, 2003, Field Staff, Landholder, Sheldon (2.53) 2011 139 2014 Field staff 146 1998, 1999, 2007 Landholder 206 1988 - 1993 Sheldon 2003, 2011, 2012, 2013, Most 235 Field staff, Landholder, VBA Years Sheldon, VBA, Field staff, 239 1983, 2001, 2013 Landholder 240 some years, 2011, 2012 Field staff, Landholder 244 Most Years, 2009, 2011 BL&A, Field staff, Landholder, VBA Most Years, 1995, 2000-2002, BL&A, Field staff, Sheldon, VBA, 248 2004, 2008, 2009, 2012 DSE Unpublished pub.

Page | 80 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Year (s) of Observation Source Number* 252 2011, 2012, 2013 BL&A, Field staff 301 Late 80s - early 90s Landholder (unconfirmed) 302 Most Years Landholder 303 2001, 2013 BL&A, Field staff, Sheldon, VBA 304 2011, 2012, 2013, Most Years BL&A, Field staff, VBA, Landholder 323 Most Years Landholder 526 Most Years Landholder 546 1988, 1995 VBA, Sheldon 585 2011 Field staff 601 Historical records Landholder 603 2011, 2013 Field staff, VBA Flocking - One-off Sites** 78 2011 Landholder 82 1981, 1982 Landholder 90 early 2000s Landholder 91 2011 Field staff 127 2013 Field staff 148 2011 Landholder 152 1880, 1995 VBA, Sheldon 222 1993 Sheldon 225 2012 Field staff 254 2009 DSE Unpublished pub., VBA 513 2011 Field staff 519 2012 Landholder 522 2012 Field staff 544 1988 VBA, Sheldon 565 2000 - 2011 Landholder 568 2009 Landholder 584 2011 Field staff late 2000s, 2009, 2010, 2011, NE of 313 Field staff, Landholder 2014 NE of 316 1997 Sheldon NE of 516 recent DSE Unpublished pub. NW of 324 1988 Sheldon SW of 602 2011 Field staff SW of 90 1984 Sheldon NW of 534 2010 Landholder * Distance (km) from nearest turbine in brackets for breeding sites (within 3.2 km) and flocking sites (within 5 km). ** Includes non-wetland habitats visited by foraging birds, even if regular (i.e. sites not used for overnight roosting).

Page | 81 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

4.8. Triggers for a Level Three Assessment The Brolga Guidelines specify that a Level 3 Assessment is triggered if the qualitative risk assessment (AusWEA 2005) of the project following site design is greater than "low". Earlier proposed wind farm layouts included turbines in areas found to be used by Brolgas during the Level One and Level Two assessments. In addition, turbines were proposed within mapped breeding site home ranges areas. Significantly, turbines were also proposed to be within five km of two flocking sites [one to the west, (site 585) and one to the south-east (Lake Sheepwash, site 110)]. Based on this, the risk to Brolgas was assessed as not being low and a Level Three assessment was triggered. This assessment is described in Section 5.

Page | 82 NORBANK ROAD MAJOR MITCHELL ROAD STOCK PILE ROAD WHITES ROAD

W GEDDES ROAD YALLA-Y-POORA ROAD

WALDRONS ROAD WILLAURA-WICKLIFFE ROAD PARUPA ROAD BEAUFORT-CARRANBALLA ROAD

TARA ROAD

C GEDDES ROAD

PARUPA ROAD ELDERS ROAD STREATHAM-CARNGHAM ROAD BACK BOLAC ROAD TATYOON ROAD HINTONS ROAD

MCINTYRE ROAD PARTRIDGES ROAD

THISTLE LANE DOERYS ROAD HUCKERS ROAD

MAC INNES ROAD W GEDDES ROAD NERRIN NERRIN FLOODW ROAD

C GEDDES ROAD

MCMASTERS ROAD CASSANOVA ROAD SETTLEMENT ROAD NARRAPUMELAP ROAD WESTMERE SALT LAKE ROAD RUBBISH TIP ROAD

BAKERS ROAD

SWAMP ROAD KINNERSLEY ROAD

GLENELG HIGHWAY SAGO ROAD

PARUPA PARK ROAD WILLS ROAD

TATYOON ROAD

SERVICE ROAD 158

MCKENZIE LANE HOWLETTS ROAD CHATSWORTH ROAD 547 546

MCKENZIE LANE DARLINGTON-CARRANBAL ROAD MCMASTERS ROAD 572 571 148 LAKE ROAD *# 574 !(!( !(!(206 569 149 573 "/ 565 147 541 545 #0566 !. 563 548 (! *# 543 #0") 146 152 542 544 553 567 564 (! *# BASSETTS LANE "/ 568 562 153 575 *# 530 590 WHITES LANE 570 (! 78 79 560 557 !. "/") *# (! 224 MCINNES LANE 317 #0 561 558559 323 314 315 528 (!529 208 WALKERS ROAD 316 532(! MORTLAKE-ARARAT ROAD 556(! 527 KARMALA LANE 531 125 !. *#"/(!225

WARINGS LANE 583 KARMALA LANE 126 526 MOFFAT LANE 576 525 75 82 582 240 (!"/ PAGELS LANE (! ") 552 312 322 ") (! !.( 504 600 209") (! ATKINS ROAD BELLS ROAD HINTONS ROAD #"/ (! (! (! 223 0 127 *# 523 ") *#")(! 522 521 ")(! ") 123 524 "/(! 320 MCKAYS LANE 130 ! (! 222 ")")601 (! ( *# !. 124 *# (! (! ")"/ 517516 519 128 518 (! MYUNA LANE 122 520 *# 129 (! 503 550 85 (! (!549 321 86 133 ") (! 241 (! 121(! ") 502 ALEXANDERS ROAD ") 539 (! Lake Gellie (! 120 501 (! ") 587 #0 327 (!245 239 (! ")(! 535 VITE VITE-STREATHAM ROAD !( (! CHATSWORTH-BOLAC ROAD 87 (! !( ")242 SMITHS LANE 238 "/ BROOKS ROAD ! 534 (! ") 555 !. ") VITE VITE ROAD EAGANS LANE ( !( (!510 536 608 533 (! (! GALES LANE 326 506 511 88 (! Chinamans Swamp ") 243 118 (! (!221 WOORNDOO-STREATHAM ROAD !(!. 507 91 !((! (! 505 ") 512 ")") 515 ENNERDALE LANE 508 *# (! (! (! (! (! *# (! VITE VITE-MOUNTBUTE ROAD LYONS LANE 90 (! 117 509 ") 538 *#(! (!514 DARCYS LANE 246 92 #0 (! 513 310 VITE VITE ROAD 537 585 (! FASHAMS LANE Lake Terrinallum 606 607 309 (! !( CARLONS ROAD") COWLINGS ROAD *# !( (! (! WILSONS LANE 111 *# !( !. (! BOLAC PLAINS ROAD POST OFFICE LANE (! 244 WOORNDOO-CHATSWORTH ROAD !.(! 112 324 !( DAWES LANE (! GEDDES LANE VEALS LANE ") 307308 (! 609 584*# (!325 ")(!*# 311 *# "/(! 113 (! 602 254 WOORNDOO-DUNDONNELL ROAD 114 (! ") (! #DARLINGTON-NERRIN"/ ROAD (!* (! (!(! Tiverton "/")(! MERCERS ROAD 141 ") 313! (! 97 137 ") #*0( (! 605 138 (! VITE VITE ROAD ") !((! TERRINALLUM ROAD ! Lake Sheepwash ( (!306 139 ")"/(! ") (! !.(110 236 ") 301 (! ") CHATSWORTH ROAD STREATHAM ROAD ") 247 328 !.!(")(! ") !.(! 235 145 ?

318 101 144 WOORNDOO-DARLINGTON ROAD !( (! 305 9 MOUNT FYANS LANE !.(") STREATHAM ROAD ") ")(! 252 !(304 "/ NINE MILE LANE 109 (! 586 !("/ WORDLEYS LANE (! ") 11 MANSE ROAD (! (! 248 "/)(! (! !( 103 "/")(! 303 !(!(!( (!")Lake Barnie Bolac ")253 ") !(")(!") DARLINGTON-SKIPTON ROAD !( 302 !. HEARDS ROAD LLOYDS ROAD

MT ELEPHANT ROAD

CREARERS LANE ") ")!( 5 603 ")!( 604") WOODCUTTERS LANE DUNDONNELL-MORTLAKE ROAD ") Long Dam NORTH STATION (!

(! BASS ROAD ")"/")(! ")") CENTRE ROAD 251 ") GEELENGLA ROAD CASTLE CAREY ROAD DARLINGTON ROAD JELLALABAD ROAD

DARLINGTON ROAD

KURWEETON ROAD DERRINALLUM-LARA ROAD MORTLAKE-ARARAT ROAD

CASTLE CAREY ROAD DERRINALLUM LARRA ROAD NORTH STATION

CASTLE CAREY ROAD PRICES LANE

KURWEETON ROAD

DARLINGTON ROAD

HAMILTON HIGHWAY

BURNS ROAD STEELES LANE SIX MILE LANE

HARDYS LANE DARLINGTON-TERANG ROAD

Legend Kilometers ANDERSONS ROAD

Wind Farm Boundary VBA FIVE MILE LANE Landowner 0 2.5 5 10

KURWEETONROAD EIGHT MILE LANE KINGS LANE Wind Farm Area of Concern *# One-off Flocking #0 One-off Flocking CAIRNLEA(10km Buffer)LANE Figure 19: Historical and recent Brolga records (! .! BOONERAH ESTATE ROAD Traditional Flocking Traditional Flocking Roads WAGGS LANE CEMETERY LANE )" /" Breeding Project: Dundonnell Wind Farm MILL STREET Breeding Wetland Quality (2011) )" Breeding- wetland drained, DARCEYS LANE !( Post1980 Sightings High not suitable as breeding site anymore MOORNONG ROAD Client: Dundonnell Wind Farm Pty Ltd KURWEETON LARRA ROAD HOOD AVENUE PIT ROAD PRENTICES LANE DSE Unpublished 2010 BL&A 2009-2013 InvestigationsDARLINGTON-TERANG ROAD Low STOCKDALES LANE

POUND LANE

PURCELLS LANE KINGS LANE *# One-off Flocking Project No.: 9184 Date: 12/09/2014 Created By: M. Ghasemi Moderate *# One-off Flocking " (! Traditional Flocking Not accessible ) Breeding )" Breeding Birdlife Sheldon ¯ !( *# One-off Flocking Sightings )" Breeding (! Traditional Flocking BL&A 2007-2008 Investigations !( Sighting )" Breeding Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

5. LEVEL THREE ASSESSMENT The Level Three Assessment involves four steps to identify suitable mitigation measures for the proposed development to produce a zero net impact on the Victorian Brolga population (DSE 2012a). The triggers and steps for Level 3 Assessments are noted in Table 16. Table 16: Level 3 Assessments: triggers and steps

Level Step Assessment triggers*

Qualitative risk assessment (AusWEA 2005) of project Trigger for Level 3 following site design is greater than "low"

Avoid or mitigate all potential impacts to Brolga breeding 1 and flocking home ranges within the radius of investigation with turbine-free buffer areas

Develop a site-specific collision risk model for Brolgas 2 utilising or moving through the radius of investigation 3 Use PVA to estimate the impact of the proposed 3 development

Identify appropriate compensation strategies to ensure a 4 zero net impact on the Victorian brolga population

Key features of the four steps are outlined below. . Step One: Avoid or mitigate potential impacts Turbine-free buffers should be designed to remove any significant impact on Brolga breeding and flocking home ranges within the RoI. These buffers should control the layout of turbines in a proposed wind farm. . Step Two: Collision risk model (CRM) The objective of CRM is to estimate the residual number of Brolga flights which have the potential to interact with wind turbines on the proposed site and from this estimate the annual collision risk. . Step Three: Population Viability Assessment (PVA) model The site-specific collision risk output is then used in the PVA to model the potential impact of the proposed wind farm on the Victorian Brolga population. . Step Four: Compensation and offset options Improving Brolga breeding habitat to enhance breeding success is considered an appropriate offset. These steps are dealt with in more detail in the sections below.

Page | 84 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

5.1. Step One: Avoiding and mitigating potential impacts

5.1.1. Background Turbine-free buffers should be designed to remove any significant impact on Brolgas within their breeding and non-breeding habitats (DSE 2012a). These buffers should control the layout of turbines in a proposed wind farm. The turbine-free buffer means that no part of the turbine, including the tips of turbine blades, will be placed within the buffer. Avoiding significant impact on breeding Brolgas is discussed in the Brolga Guidelines as follows (p. 8): “In the case of breeding habitat ‘turbine siting would be used to exclude any significant reduction in breeding success caused by turbines” (Brolga Scientific Panel 2008). This will be achieved by establishing turbine-free buffers around all potential Brolga nesting sites sufficient to have no significant impact on the likelihood of successful reproduction.” For the current project, a suitable buffer around a breeding site has been set in a way that avoids both direct impacts on breeding birds, such as turbine collision, and indirect impacts such as disturbance. Avoidance of both these key potential disruptions to breeding activity will ensure minimal impact on the likely outcome of a breeding attempt at the buffered site. Avoiding significant impact at non-breeding or flocking sites is discussed in the Brolga Guidelines as follows (p. 8): “At non-breeding or flocking sites, turbine-free buffers should be designed to exclude any significant impact on the survivorship of Brolgas whilst occupying that flocking site (Brolga Scientific Panel 2008).” For the current project, a suitable buffer around a flocking site has been set in a way that avoids both direct impacts on birds moving from flocking sites to surrounding foraging habitats, such as turbine collision, and indirect effects on them, such as disturbance. The principal means of avoiding and mitigating impacts of wind farms on Brolgas is the establishment of turbine-free buffers. Turbine free buffers have been identified based on a range of Brolga investigations on and around the Dundonnell Wind Farm site. Avoiding and mitigating Brolga impacts has been a key input to defining these buffers. The method for establishing turbine-free buffers is described below and shown in Figure 20. Buffers related to breeding sites are considered first, followed by traditional flocking site buffers.

5.1.2. Breeding sites

Methods For the current project, a suitable breeding home range buffer11 around a breeding site has been set in a way that avoids both direct impacts on breeding

11 The Brolga Guidelines refer to ‘breeding’ and ‘non-breeding’ home ranges. Strictly speaking, they refer to them outside the narrow technical definition of the term ‘home range’ in ecology. Therefore, the terms ‘breeding home range buffer’ for breeding sites and ‘flocking

Page | 85 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) birds, such as turbine collision, and indirect impacts such as disturbance. Avoidance of both these key potential disruptions to breeding activity will ensure minimal impact on the likely outcome of a breeding attempt at the buffered site as it would avoid injury to or mortality of breeding adult birds through turbine collision and avoid disruption to foraging in key habitat around the breeding wetland. Turbines and associated infrastructure have been excluded from a breeding site home range buffer defined based on predicted habitat use around the breeding site, with the added 300 m disturbance buffer required in the Brolga Guidelines (p. 11). Two methods have been used in recent years to predict habitat use by Brolgas around their breeding sites: . A habitat use model that outlines areas most likely to be used based on empirical observations of habitat choice across a range of nest sites (e.g. Stockyard Hill Wind Farm, BL&A 2010); and . Home range mapping, using kernel analysis of observations of Brolga locations around a breeding wetland (e.g. Biosis 2011). The first method has been applied to the Dundonnell Wind Farm project due to the lack of regular or continued breeding activity within 3.2 km of the proposed wind farm. A comparison of results between the two methods (see Appendix 3) indicated that the first method, based on habitat modelling, encompassed a significantly larger breeding home range buffer than the latter. It is therefore considered to be conservative and may well exclude more area from turbines than is actually required. However, given that the latter method involved observations over a single breeding event, it is possible that in different years a pair using the same site may behave differently if habitat conditions around the breeding site change. The former, habitat modelling approach allows for possible between-year differences in conditions and behaviour, and therefore is more likely to encompass the area used by a breeding pair over the life of the proposed wind farm project. This method is described in more detail below.

Results The breeding home range buffer from which turbines are excluded is shown notionally in Figure 20. A Brolga breeding site home range buffer was defined based on the following: . A 400 m radius around a wetland (this is the area in which breeding Brolgas showed no preference for habitat type and were deemed to be equally likely to occur anywhere); . Up to 3.2 km from a breeding site all wetlands were identified and included in the mapped home range buffer (this is because when breeding Brolgas moved further than 400 m from their nesting wetland, they showed a clear and

home range buffer’ for flocking sites are used to avoid erroneous understanding of what is being described while maintaining a link to the meaning of the terms in the Guidelines. Turbine-free buffers include these areas plus a 300 m disturbance buffer.

Page | 86 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

statistically significant preference for wetlands, based on area of available habitat and number of flights (see Appendix 3); . To avoid disturbance effects from operating turbines and associated human activities, a further 300 m disturbance buffer is placed around this home range. Final Brolga breeding site home ranges are indicated as a component of the turbine-free buffers shown in Figure 21. It is noteworthy that observed flights by Brolgas over the wind farm site during breeding activities at wetlands 602 and 254 occurred within the designated home range buffer for that breeding site.

5.1.3. Flocking sites Outside the breeding season, Brolgas gather in larger flocks at traditional flocking sites. Observations during the current investigation and at other times (BL&A unpublished data, 2007 – 2010) show that when using traditional flocking sites, birds generally behave as described below: . Birds roost at the flocking site overnight from last light until shortly before or after sunrise; . In the early morning, they fly out in groups to forage in terrestrial and wetland habitats for the morning; . When birds have finished feeding (digestive bottleneck) they either return to the flocking site to rest and digest their food during the middle of the day, or remain in a flock at the foraging site, where they cease foraging; . When they are ready, flocks of Brolgas return to foraging sites and continue feeding until dusk; At dusk, and often just on last light, flocks return to roost for the night at the flocking site. Therefore, up to four flights of flocks per day (two out and two back) can occur within the Brolga flocking home range buffer. Based on observations of the flocking events observed near the Dundonnell Wind Farm site, this routine becomes well-established and fairly consistent while a flocking site is being used. For the current project, a suitable flocking home range buffer has been set in a way that avoids both direct impacts on birds moving from flocking sites to surrounding foraging habitats, such as turbine collision, and indirect effects on them, such as disturbance. Three traditional flocking sites occur near the wind farm where application of the default five km buffer is recommended by the Brolga Guidelines unless detailed investigations reveal otherwise.

Page | 87 Figure 20: Method for establishing Turbine-free buffers at breeding sites

Project: Dundonnell Wind Farm

Client: Dundonnell Wind Farm Pty Ltd

Project No.: 9184 Date: 15/01/2014 Created By: B. Lane / M. Ghasemi Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

The Brolga Guidelines observe that home ranges are likely to vary with local habitat quality and extent, and seasonally. It is noteworthy that two separate observations of flocking site home ranges at Blackwood Lake (Biosis 2011, King 2008) showed that the foraging range of Brolgas as they moved out into the landscape from their flocking sites was significantly smaller than the default five km radius buffer area nominated in the Brolga Guidelines. Given this, this aspect of Brolga activity in the proposed Dundonnell Wind Farm RoI has been a key focus of the current Brolga investigation. As described in Section 4, this has involved searches for flocking birds, as well as flight behaviour studies during two flocking events within five km of the proposed wind farm boundary: . From June to October 2011, when between 15 and 32 Brolgas gathered at a wetland south of Chinaman’s Swamp (wetland 585); and . From April to mid-June 2013, when between 20 and 46 Brolgas gathered at Lake Sheepwash (wetland 110) then from June to August when between 20 and 34 Brolgas flocked at a lake approximately seven km east-south- east of Lake Sheepwash (wetland 304). . From February to March 2014, a flock of 17 Brolgas was observed roosting at wetland 139 (located along Woorndoo-Dundonnell Road). This flock comprised seven resident birds. The remaining ten birds moved around the landscape together and joined the resident birds to feed on grain trails 800 m north of wetland 139, and three km north-east near wetland 602. The flock also roosted at Lake Sheepwash (110). Brolga use of Lake Gellie (wetland 239) was not consistent enough during the investigation period to enable continuous observations of flight behaviour around this site during a flocking event, although observations in 2011 and 2012 showed north-south flights between wetland habitats on a number of occasions. The flocking site flight behaviour studies during these events (Section 4.3) enable a more detailed understanding of how Brolgas utilise the landscape around these particular flocking sites. This information, in combination with a range of other data, provides clear direction for defining turbine-free buffers around flocking sites. The other information of relevance includes: . Historical observations of the historical location of Brolga sightings in the RoI from the VBA and the Birdlife Australia Birdata database (see Level One Assessment – Section 3.1.2); . Information on the location and choice of land type and land use by Brolgas from landholders in the RoI (see Level One Assessment – Section 3); and . Information on the location of Brolga activity during the 2011–2013 fortnightly roaming surveys of the RoI (see Level Two Assessment – Section 4). The data gathered from all investigations represent a number of independent lines of evidence that can be used to assemble a picture of current and likely future Brolga activity around flocking sites near the proposed wind farm, namely: . historical data base records; . landholder observations on location and habitat choice; . flight behaviour studies;

Page | 89 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. utilisation surveys of the proposed wind farm site; and . roaming surveys. What this evidence has shown about how Brolgas use the landscape around flocking sites in the RoI is summarised in the points below. The detailed findings that lead to these conclusions have been presented in Sections 3 and 4. . Brolgas are most likely to forage around flocking sites on land that has previously been cultivated, cropped or subject to rock removal and/or pasture improvement (i.e. arable land); and . They are not recorded regularly on stony rise country used for grazing. Based on these conclusions it is possible to identify known and potential flocking home range buffers near the Dundonnell Wind Farm and define turbine-free buffers that protect these from direct and indirect impacts. Turbine-free buffers for flocking sites have been identified based on: . Excluding from development any area where flocking Brolgas were observed foraging, except where it is clearly and directly related to an on-site land use practice, such as grain feeding of stock, that can be avoided in the future; . Excluding any area where regular flights by flocking Brolgas were observed, except where it is clearly and directly related to an on-site land use practice, such as grain feeding of stock, that can be avoided in the future;; . Excluding any area where the land type was arable and where there had been previous sightings of Brolgas; and . Adding the 300 m disturbance buffer to the foregoing areas as required in the Brolga Guidelines. The flocking home range buffers form a component of the turbine-free buffer shown in Figure 20. Earlier versions of the proposed wind farm boundary included a number of paddocks and properties to the west of the currently proposed wind farm site. These areas have been excluded from the proposed wind farm due to their role as foraging habitat within the flocking home range buffer for the western flocking site. Early turbine layouts also include turbines in the south-eastern part of the wind farm site, now excluded as a consequence of the more accurate definition of a flocking home range buffer to Lake Sheepwash. Additional turbines have been removed in the north-east of the wind farm as a consequence of a breeding home range buffer affecting that part of the site. In each of these areas, a 300 m disturbance buffer is also included in the turbine-free buffer. Adoption of these turbine-free buffers ensures that the proposed wind farm layout complies with the mitigation requirements of the Brolga Guidelines.

Page | 90 130 128 123 517 516 124 518

129 129 122 WOORNDOO-STREATHAM ROAD 503/"

133 502 /"120 121 /"539

501 Lake Gellie

327 239 535 245

533 238 /" 555 534 326 536 506 DARLINGTON-NERRIN ROAD

Chinamans Swamp .! /"118 505 507

508 ENNERDALE LANE

515

117 513 514

DARCYS LANE 538

FASHAMS LANE Lake Terrinallum 607 DOHERTYS LANE 585.! 111

POST OFFICE LANE 112.! VEALS LANE /" 324

DAWES LANE Breeding 584 325 254

113 /"602 114

Tiverton

137/" WOORNDOO-DUNDONNELL ROAD

313

138

DUNDONNELL-DERRINALLUM ROAD .!139 Lake Sheepwash

/"236 110.!

328

247

Kilometers Legend 0 0.5 1 2 Wind Farm Boundary Brolga Records Figure 21: The constraints map (turbine free buffers) Wetlands .! Traditional Flocking " Project: Dundonnell Wind Farm Brolga Breeding Site Homeranges / Breeding Client: Dundonnell Wind Farm Pty Ltd Arable Land with Brolga Sightings Turbine Exclusion Zones Project No.: 9184 Date: 10/09/2014 Created By: M. Ghasemi Roads Brolga ¯ Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

5.2. Step Two: Collision risk modelling The Brolga Guidelines indicate that the objective of collision risk modelling is: “ …to estimate the residual number of Brolga movements which have the potential to interact with wind turbines on the proposed site and from this estimate the annual collision risk.” The way the BL&A collision risk model works and a description of how it derives the estimated collision rate is described in detail in Appendix 5 by Symbolix Pty Ltd. The techniques involved are used regularly for the same purposes and have been published and peer reviewed in relevant professional journals. An overview of how the model works and its results are provided here. The BL&A collision risk model used two inputs: . An estimate of the different rates of movement of Brolgas within the site based on a combination of the results of the utilisation surveys on the proposed wind farm site (see Section 4 for details of methods and results) converted using kernel analysis to a probability distribution of Brolga activity across the wind farm site based on the results of the roaming surveys from 2011 to 2013 (see Section 4 for methods and results); and . An estimate of the interaction of the Brolgas estimated to fly over the site with turbines using the collision estimation model of Scottish Natural Heritage (Band et al. 2007, SNH 2010), given a range of potential avoidance rates. The rate of movement of Brolgas within the wind farm was derived from two sources of information about Brolga activity on and near the proposed wind farm: . The results of the two formal on-site utilisation surveys (see Section 3) on 6th April – 25th May 2012 and 9th – 23rd May 2013, when Brolga flocking activity was occurring in the RoI at the times of the survey at Lake Sheepwash; and . A probability density map based on Brolga roaming surveys in 2011 (April – November), 2012 (August - December) and 2013 (January – early December) that incorporate both positive and negative Brolga records versus search effort. The dates for the utilisation surveys coincided with periods when Brolga flocking activity in the RoI was occurring. The use of data from this period to extrapolate annual rates of movements of Brolgas across the proposed wind farm site is therefore considered to be conservative and likely to over-estimate the number of flights. This is because extrapolation does not take into account seasonal variability in activity. It is noteworthy that the number of Brolgas in the RoI declines during the breeding season (see Section 4). To allow for this, the flocking season rate of movement was assumed to occur for half the year and a separate breeding season rate of movement was calculated for the balance of the year. The details of how this was done are presented in Appendix 5. The combination of the number of flights and the probability density map provides an indication of likely numbers of flights by Brolgas in different parts of the proposed wind farm site. This enables a range of wind turbine layouts to be compared.

Page | 92 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Details of the turbines to be used were provided by the proponent and factored into the application of the Band et al. (2007) model. Finally, four avoidance rates were modelled and are presented herein: 90%, 95%, 98% and 99%. As described below, the actual avoidance rate is likely to be within the 95% to 98% range. Determining an appropriate wind turbine avoidance rate for the Brolga is challenging given the lack of past interactions between Brolgas and wind turbines. Apart from the recently commissioned Macarthur Wind Farm, there are no operating wind farms in areas where Brolgas could regularly interact with turbines so that avoidance rates could be measured based on behavioural observations. Therefore, information on the behaviour of other crane species has been used. Before discussing crane avoidance behaviour, it is worth considering definitions of avoidance. Cook et al. (2012) highlight the difference between ‘macro-avoidance’ and ‘micro-avoidance’. Macro-avoidance refers to changes in flight behaviour that result in a bird avoiding a wind farm altogether. Micro-avoidance refers to the flight behaviour of a bird to avoid a turbine once within a wind farm. In practice, if birds avoid a turbine at 100 m distance, they could do so at the edge of a wind farm or several turbine ’rows’ into a wind farm, where they might come across the first turbine in front of their flight path, so the distinction is not necessarily always useful for collision risk modelling. In a mathematical sense avoidance is dealt with in the CRM as micro-avoidance and it is assumed that no macro-avoidance occurs. In practice, our only available evidence for similar species combines both forms of avoidance, with an emphasis on macro-avoidance. What these studies show is that cranes have the ability to avoid wind turbines at a range of scales. In acknowledgement of the uncertainty of estimating the exact proportion of flights that will avoid turbines, a range of avoidance rates is presented. This is an accepted way of dealing with this uncertainty in Australian wind farm impact assessments, as is discussed in Appendix 5. Brolgas belong to the crane family and bird turbine interactions have been observed for two other crane species: the European Common Crane (Grus grus) and the North American Sandhill Crane (G. canadensis). In Germany, Stübing and Korn (2006) observed Cranes near wind farms on 88 occasions over a seven year period in Rhineland. They found that in-flight Cranes never approached closer than 100 m to turbines, with distances usually between 300 and 700 m. In summer, after breeding, the Cranes approached wind farms to within 150 to 250 m but no closer. Langgemach (2013) found that avoidance action by Cranes in response to operating wind turbines was observed for individuals and small flocks up to a distance of 750 m from turbines while for larger flocks, turbines were avoided by greater distances, between 1,000 and 1,350 m. The observations of Gerjets (2006) for European Cranes, again at a wind farm in northern Germany, are summarised below: . Cranes avoided flying close to wind turbines; . Cranes have been observed flying within 200 m of operating wind turbines where turbine lines are oriented parallel with the direction of flight;

Page | 93 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. The range of distances from turbines that Cranes were observed flying in one systematic study was between 150 m and 670 m, with a median distance of 300 m, where turbines were not parallel with the direction of flight; . In another, less systematic study, crane flocks flew around operating wind turbines at distances of between 400 and 500 m where turbine lines were not parallel with the flight direction; . Flocks of Cranes have been observed flying quite close to turbines, in one case about 100 m from one and in another between two operating turbines, quite close to the rotor tips; and . Another observation involved a “V” formation flock breaking up, possibly due to downwind turbulence from a wind turbine, at a distance of 750 m from the turbine. The flock eventually flew around the turbine and regrouped after 1.5 km. The reaction of European Cranes to wind turbines therefore varies, but it is clear that they generally avoid wind turbines and a high avoidance rate above 90% is likely to be realistic in determining inputs to a collision risk model. Langgemach (2013) reviewed the impact of wind farms throughout Germany and found that up to 2013 there were seven recorded instances of Cranes colliding fatally with wind turbines, mostly during night-time autumn migration, from a population numbering in the thousands in an area with many more turbines (e.g. in 2000, Germany had around 9,300 operating wind turbines; by 2012 this figure had risen to 23,000). It is noteworthy that, unlike European Cranes, Brolgas do not undertake long-distance night migration. It is well known that night-migrating birds are more susceptible to collision with wind turbines and other structures (Drewitt and Langston 2008, Erickson et al. 2001). Observations of Sandhill Cranes in North America are also informative.12 Observations at a wind farm in South Dakota showed that out of 66 flocks that approached wind turbines, totalling more than 4,000 individuals, 92 per cent of birds showed an avoidance response (http://aweablog.org/blog/post/windpower- report-whooping-cranes-may-avoid-wind-farms-more-research-ahead, viewed 2nd December 2013). A proportion of birds that do not show an avoidance response may collide with wind turbines. Therefore, the final avoidance rate one would use for collision risk modelling would be higher than this. A recent review (Nations et al. 2012) suggested a turbine avoidance rate of between 90 and 95 per cent for collision risk modelling for the Whooping Crane. The true turbine avoidance rate for the Brolga is unlikely to be determined in the near future, given the small number of wind farms operating and being monitored specifically for Brolga impacts in south-western Victoria. An informed assumption is therefore unavoidable and has been used in previous wind farm studies in Victoria (e.g. Stockyard Hill, Mortlake, Yaloak) and been found by decision-makers to be informative. Based on the foregoing information, it was considered

12 Note that a recent study of Sandhill Cranes at four Texas wind farms by Navarrete (2011) has not been used, as a review of its statistical design found it to be flawed as it used parametric statistical tests on frequency and category data, thereby violating the assumptions of the statistical methods.

Page | 94 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

appropriate to use avoidance rates of 90%, 95%, 98% and 99% and to present the collision risk modelling results for this range of avoidance rates. Flight speed is another key input to the CRM. A literature review on the flight speed of Cranes showed the average flight speed range is between 48 to 64 km/hour (Table 17). Given that the average weight of Brolgas (6.15 kg) lies between the average weight of Whooping and Common Cranes, a conservative average speed of 60 km/hour was assumed for Brolgas for the purpose of the CRM input. Table 17: Flight speed of Cranes

av. low high av. low high av. estimated Speed/ Species weight weight Weight speed speed Speed Source speed av. kg kg kg km/h km/h km/h range weight

Journey Sandhill 40- 2.7 6.4 4.55 40 56 48 North 10.5 Crane 56km/h 2014

Melvin & Sandhill 23-83 2.7 6.4 4.55 23 83 53 Temple 11.6 Crane km/h 1982

Journey Whooping 56 - 72 6.4 7.7 7.05 56 72 64 North 9.1 Crane km/h 2014

Common LPO 40 - 80 4 6 5 40 80 60 12 Crane 2014 km/h

The collision risk model has been run on two wind farm layouts (including an estimated annual collision rate with the proposed transmission line of 0.04 birds). Given the spatial sensitivity of the model, it is possible to compare layouts and adjust them to reduce collision risk. The proponent has provided two layouts to model. The first was an initial layout not informed by the results of the collision risk modelling. The second has involved adjustments to the positions of turbines to reduce the Brolga collision risk of the project. The methods and results of the collision risk modelling are presented in detail in Appendix 5. The results are summarised below in Table 18. Note that the output of the collision risk model is technically the number of flights that results in collision, which is assumed to be equal to the number of individual Brolgas that collide. This is based on the assumption that the removal of a small number of individuals will not affect the overall flight rate, which is believed to be valid for Brolgas where a number of individuals are responsible for the flights and not only one pair producing a large number of flights, as is more likely to be the case for raptors.

Page | 95 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Table 18: Results of Brolga CRM

Layout L03.11 - 104 WTGs Layout L08 - 104WTG (original layout) (final revised layout)

Avoidance rate 90% 95% 98% 99% 90% 95% 98% 99%

Modelled average long 2.31 1.19 0.52 NA 0.91 0.46 0.18 0.09 term annual collision rate

Modelled annual 0.04 0.04 0.04 NA 0.04 0.04 0.04 0.04 powerline collision rate

TOTAL 2.35 1.23 0.56 NA 0.95 0.50 0.22 0.13

Probability of two or more 66% 31% 8% NA 23% 8% 1% <<1% collisions* in a given year

* Probability only provided for collisions with turbines The collision risk modelling predicts that the preferred, final wind farm layout will lead to a long term, annual average of between 0.09 and 0.91 Brolga collisions with wind turbines.

5.3. Step Three: PVA Model Population Viability Assessment (PVA) is a widely accepted modelling method that attempts to predict the trajectory of the population of a plant or animal and calculates the quasi-extinction probability. The PVA is required by the Brolga Guidelines to be used as part of any Level Three Assessment of a wind farm proposal. Appendix 6 presents the PVA for Brolga for the proposed Dundonnell Wind Farm. PVA relies on inputs about baseline or current population levels (numbers), population age structure, reproduction rate and survival rates. Using this information, the PVA can predict the population of a species at a range of future times and how this might alter if any parameter changes (e.g. reproduction rate increases or adult survival rates decline). Predictions are made based on an iterative sampling method that averages the findings from 10,000 runs of the model. The University of Melbourne (Associate Professor Mick McCarthy) was commissioned by the former DSE to develop a PVA for the Victorian Brolga population. The purpose of the Victorian Brolga PVA is to model population scenarios given estimates of the impacts on Brolga survival rate of wind farm developments (using collision risk modelling results) and to set targets for compensation and offset measures to ensure zero net impact on the Victorian Brolga population. Dr McCarthy was commissioned by Trustpower Australia Pty Ltd to apply the Victorian Brolga PVA to the results of the collision risk modelling for the proposed Dundonnell Wind Farm presented in the previous section of this chapter. This report is provided in Appendix 6. The results are summarised in Table 19 below,

Page | 96 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) which estimates the population size after 25 years of operation of the proposed Dundonnell Wind Farm.

Table 19: Results of Brolga PVA

No 90% 95% 98% 99% turbines avoidance avoidance avoidance avoidance and powerlines

Expected minimum 809 795 805 808 808 south-west Victorian population

Change in population N/A 13* 4 1 1

* Difference due to rounding

Based on an expected minimum population of 809 birds, the PVA predicts that after 25 years (the planned life of the wind farm project), the population size will be between 808 (99% avoidance rate) and 795 birds (90% avoidance rate), a reduction of between one and 13 birds compared with baseline conditions. The predicted result for 95% avoidance rate, which is considered the most realistic conservative collision rate, is a population size of 805 birds, representing a reduction of four birds in the population.

5.4. Step Four: Compensation to achieve zero net impact The Brolga Guidelines require that the impacts on the Victorian Brolga population predicted in Step Three (i.e. the PVA) are ‘fully offset’ (p.12). In the case of the proposed Dundonnell Wind Farm, fully offset would mean that the Victorian Brolga population should have between one and 13 more breeding adult Brolgas than would be the case if full offsetting did not occur. For example, using the mid-range 95% avoidance rate scenario for wind farm impacts, a total of at least 13 new adult Brolgas should be added to the population by the end of the 25 year life of the wind farm project to compensate for the average estimated loss of one bird every two years over 25 years. This will ensure that the expected minimum population size remains the same, allowing for the mortality of some of the 13 new birds from a combination of natural and wind farm-related causes over the 25 year compensation period. The Brolga Guidelines state that two possible compensation measures should be considered: . Reducing mortality from transmission line collisions by marking them to improve visibility and reduce the rate at which Brolgas collide with them; and . Protecting and enhancing breeding sites to improve the production of young, leading to an increase in adult numbers over time. These are discussed further below.

Page | 97 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

5.4.1. Transmission line marking Marking transmission lines to reduce collision mortality has been demonstrated to markedly reduce collision risk for other species of Cranes (Brown and Drewien 1995). The effect of collisions with powerlines on the Brolga has not been quantified across its range in Victoria. Furthermore, the only available information comes from collision risk modelling of Brolga impacts from a small number of wind farm power grid connection powerlines (such as for the current project). No empirical measurement of the extent of interaction between powerlines and Brolgas has been made. It is noteworthy that where it has been modelled, the contribution of powerline collision to Brolga mortality represents a very small proportion of the estimated impact of the project (e.g. Stockyard Hill Wind Farm; this project), suggesting that the scale of powerline marking required to compensate for wind farm impacts is unlikely to be practicable, except where there might be particular hot-spots, none of which have yet been identified. Although the Brolga Guidelines indicate that marking or undergrounding powerlines might contribute to compensating for wind farm collision impacts, the lack of a readily quantifiable project-specific compensation level for investment in powerline marking or undergrounding makes this technique of unpredictable benefit. This lack of certainty makes it difficult to define the extent of powerline marking required to give assurance a priori that the quantitative requirement of the Brolga Guidelines for zero net impact can be met. For this reason, powerline marking is not proposed to be part of the compensation package for the Dundonnell project.

5.4.2. Protection and enhancement of breeding sites Measures to protect or improve breeding sites are detailed by Herring (2005) and may include: . Restoration of the natural flooding regime of wetlands by closing drains; . Increasing inundation frequency and duration of breeding wetlands through artificial flooding; . Creating new potential breeding habitat by damming or modifying existing wetlands or dams, including blocking previously installed drains; . Management of wetland vegetation condition through controlled grazing (or stock removal) to improve suitability as a breeding site; . Addition of nesting material to potential breeding wetlands to facilitate nest building; and . Fox control at key breeding habitats. These measures have been used successfully to improve the survival of eggs and young birds and increase the chances of breeding Brolgas successfully raising young to fledging (Herring 2005). However, in the past, concern has been expressed at the effectiveness of this measure to compensate for wind farm impacts on the Victorian Brolga population (e.g. Stockyard Hill Wind Farm Planning Panel report). This is considered below.

Page | 98 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

The breaking of the long drought during the Dundonnell Wind Farm investigations had a fundamental effect on the availability and duration of inundation of Brolga breeding sites, with the result that after two years of above average or average rainfall across much of the Victorian Brolgas range, the April 2013 flocking count included 17% one- and two-year old birds (Appendix 6). This may have resulted in the 2013 Brolga flocking season count being the highest on record, although improved site coverage and searcher effort also contributed to this. That so many young were observed in flocks compared with the previous drought years (BL&A, unpubl. data) indicates how effective improving the availability of breeding habitat can be in increasing the Brolga population. For this reason, protecting and enhancing Brolga breeding sites is considered to provide the best means to ensure zero net impact. It is considered that in the context of compensating for a loss of individuals from the population due to a new, albeit minor, source of mortality (i.e. a wind farm), density-dependent population processes will be irrelevant. The objective is to have zero net impact on the population by replacing lost individuals through enhanced breeding output. It is important to estimate the number of additional Brolgas that need to be added to the Victorian Brolga population over the 25 year life of the project to meet the required zero net impact objective of the Brolga Guidelines. With an average of 0.5 extra brolga deaths per year assuming, for example, a 95% avoidance rate, this translates to an extra 13 deaths over 25 years. Therefore, mitigating these extra deaths by increasing breeding success would require an extra 13 birds being raised to adulthood over the 25 year period. The impact on the expected minimum population size estimated by the PVA after 25 years of wind farm operations using the 95% avoidance rate was a reduction by four birds (Appendix 613). Note that proportionally different figures would apply to different avoidance rates. The difference between the population reduction in this example (four birds) and the required compensation number (13 birds) is due to the loss of some of the 13 added birds from the combination of natural and wind farm related mortality during the 25 year project life. Managing Brolga breeding sites to boost the production of young, thereby adding more birds to the state’s population, is considered feasible. It is possible based on the workings of the PVA model to set a compensation target a priori. This target for the 95% avoidance rate, based on the mortality rates in the PVA, is the production of one additional fledged young every second year from a selection of pro-actively managed Brolga breeding sites. The proponent has committed to implementing, for the life of the project, a program of Brolga breeding site enhancement and management to boost the production of fledged young to the extent required. Based on this number, a Brolga compensation plan will be

13 Note that in the example of 98% avoidance rate presented in Appendix 6, the expected minimum population size reduces by one bird, but requires the replacement of the 6.4 birds lost from the population over 25 years due to wind farm operations; the difference is again related to the estimated loss of some of those 6.4 birds over the 25 year project life due to a combination of natural and wind farm related mortality.

Page | 99 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12) prepared as a condition of any project approval and subject to a secondary consent before wind farm construction is permitted to commence. Such a plan will be implemented for the 25 year duration of the project and will include the following information: . The locations of historical Brolga breeding wetlands that will be enhanced; . Evidence of landholder agreements to participate in the breeding site enhancement project for its duration; . Methods of enhancement appropriate to each site; . A program of appropriate Fox baiting leading up to each breeding season in areas subject to the plan; . Five-yearly performance targets for each site and the program as a whole, consistent with the outcomes of the PVA and the zero net impact objective (to be amended every five years depending on outcomes); . Monitoring and reporting requirements, including reporting on whether the number of sites being managed and the way management is proceeding are on target to meet the 25-year zero net impact objective; and . The plan will include details of arrangements for potential intensified adaptive management of the historical Brolga breeding wetlands, to be implemented (and monitored and reported) if the monitoring of existing enhancement efforts fails to show that the breeding enhancement program is on target to meet the 25-year objective, likely to include but not be limited to: o increasing the number of breeding sites subject to active management; and o removing anticipated threats detected during monitoring of implementation of the compensation plan. Such a plan should meet with the approval of DEPI.

5.5. Cumulative Impacts The Brolga Guidelines state that there is a requirement to avoid cumulative impacts of the wind farm industry on the Victorian Brolga population. The specific objective of the guidelines is to manage each wind farm development to achieve a zero net impact on the Brolga population, with an overall objective to avoid cumulative impacts of multiple wind farms operating independently within the Brolga range in Victoria. The cumulative effects of wind farms may impact on Brolga in three key ways, including: . Direct effects, such as collision with turbines; . Indirect effects, such as displacement or disturbance resulting in decreased habitat use; and . Barrier effects, where wind turbines may create barriers to seasonal or local flights. The following wind farms are operating, are approved or have started their public development approval process within roughly 50 km of Dundonnell Wind Farm:

Page | 100 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

. Darlington Wind Farm (EES referral submitted): 150 turbines; . Mortlake South Wind Farm (under construction): 51 turbines; . Penshurst Wind Farm (EES referral submitted): 223 turbines; . Stockyard Hill Wind Farm (approved): 157 turbines; . Oaklands Wind Farm (operating): 32 turbines; . Morton’s Lane Wind Farm (operating): 13 turbines; . Macarthur Wind Farm (operating): 140 turbines; and . Salt Creek (under construction): up to 15 turbines. Cumulatively, it is possible that the above wind farms could lead to an increased risk of collision as an individual Brolga ranges across the area that includes these projects. This may be particularly pertinent to the Brolga as during the non- breeding season they may fly up to 100 km in a short time between habitats (I. Veltheim, pers. comm.). As each of these wind farms has been assessed for its impacts on the Brolga in different ways, it is not possible to combine quantitative estimates of Brolga impacts and arrive at a definitive number of Brolgas affected. This is primarily because most of the information available on these wind farms pre-dates the Brolga Guidelines. As the Dundonnell project is the first Victorian wind farm proposal to be considered under the new Brolga Guidelines, it has been possible to apply the required approach and minimise population impacts. The application of the guidelines has resulted in the project being required to have zero net impact on the Victorian Brolga population For this reason, the proposed Dundonnell Wind Farm project is not expected to lead to a significant incremental contribution to the cumulative impact of wind farms within the species’ range in Victoria.

5.6. Summary and Conclusions The application of the Brolga Guidelines to inform the layout of the proposed Dundonnell Wind Farm has resulted in an assessment that provides assurance that the project will have zero net impact on the Victorian Brolga population. The assessment has reduced the possible impacts of the Dundonnell wind farm on Brolgas to acceptable and manageable levels as it has: . resulted in significant reductions in the extent and refinements to the layout of wind turbines in the final layout by the adoption of well-researched turbine- free buffers around all known breeding and flocking sites; . provided estimates of residual population impacts through collision risk modelling and population viability assessment; and . identified a corresponding scale of residual impact compensation, though enhancing the breeding success of the species for the life of the project that is readily achievable.

Page | 101 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Table 20: Summary of Brolga risks, responses and outcomes for the Dundonnell Wind Farm

Risk (as per Brolga Guidelines) Responses Outcome Collision with Turbines and Powerlines Collision by breeding birds with wind . Initial screening of historical information to . Significant reduction in the extent of the turbines, leading to a significant locate past breeding sites (2009-10) proposed wind farm to confine it to an reduction in breeding success area with no historical breeding records . Interviews and discussions with 76 . Further refinement of the project landholders to locate additional past boundaries to avoid areas where Brolgas breeding sites and to understand historical breed patterns of habitat use by Brolgas . Field investigations since 2009 to search for breeding brolgas within the 10 km RoI, including fortnightly roaming surveys of all suitable wetlands and liaison with landholder network to track occurrences and habitat suitability . Detailed analysis of past BL&A data on the . Development of breeding site home movements of Brolgas around their range buffers for all observed and likely breeding sites in south western Victoria (n = future breeding sites within 3.2 km of 163) to develop a habitat modelling the proposed wind farm. approach to breeding site home range . Removal of all turbines from breeding mapping, in the absence of significant site home range buffers (including the numbers of prolonged breeding attempts 300 m disturbance buffer) within 3.2 km of the proposed wind farm.

Page | 102 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Risk (as per Brolga Guidelines) Responses Outcome . Targeted flight monitoring of breeding . Confirmation of the range of movements Brolgas at one active breeding site close of the birds around their breeding site to the wind farm compared with the habitat modelled breeding site home range buffer, demonstrating the validity of that home range buffer . Comparison of breeding site home range . Validation of the appropriateness of the model with actual home range mapping habitat modelled breeding site home at Penshurst (Biosis 2011). range buffers where additional behavioural studies of breeding Brolgas have defined an actual (cf. modelled) breeding site home range Collision by breeding birds with powerlines . Compilation of historical and recent data . Mapping of breeding sites as a basis for leading to a significant reduction in breeding on the occurrence of breeding Brolgas modelling the likely numbers of breeding success within 10 km of the proposed powerline Brolgas that cross the powerline. south westwards from the proposed . Calculation of the likely collision risk to wind farm. Brolgas from the powerline (see under ‘flocking birds’ for total collision risk calculation) Conclusion: Based on a combination of knowing the locations of all likely future breeding sites for the Brolga within 3.2 km of the proposed wind farm, all likely breeding site home range buffers (including the 300m disturbance buffer) have been mapped and turbines excluded from these. This will reduce the risk of breeding birds colliding with operating wind turbines to a negligible level. In combination with other risk reduction strategies (see later in this table), and given the low powerline collision risk, significant reductions in the breeding success of Brolgas in the area are not therefore anticipated as a consequence of the proposed development.

Page | 103 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Risk (as per Brolga Guidelines) Responses Outcome Collision by flocking birds with wind . Initial screening of historical information . Significant reduction in the extent of the turbines, leading to a significant impact on to locate past flocking sites (2009-10) proposed wind farm to confine it to an the survivorship of Brolgas while using the area with no historical flocking records flocking site . Landholder interviews and discussions . Further refinement of the project to locate additional past flocking sites boundaries to avoid areas where Brolgas flock . Field investigations since 2009 to search for flocking brolgas within the 10 km RoI, including fortnightly roaming surveys of all suitable wetlands and liaison with landholder network to track occurrences and habitat suitability . Targeted flight monitoring of flocking . A detailed understanding of the Brolgas at active flocking sites within 5 geographical extent and habitat km of the proposed wind farm characteristics of Brolga flocking site home range buffers in the region as a . Targeted survey of 76 landholders in the basis for defining turbine free buffers for district to determine the choice of known and likely future flocking sites habitat by flocking Brolgas . All turbines excluded from defined flocking site home range buffers (including the 300m disturbance buffer) Collision by flocking birds with powerlines . Compilation of historical and recent data . Mapping of flocking sites as a basis for leading to a significant impact on the on the occurrence of flocking Brolgas modelling the likely numbers of breeding survivorship of Brolgas while using the within 10 km of the proposed powerline Brolgas that cross the powerline. flocking site south westwards from the proposed . Calculation of the likely collision risk to

Page | 104 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Risk (as per Brolga Guidelines) Responses Outcome wind farm. Brolgas from the powerline . Collision risk to Brolgas from the proposed powerline during both the breeding and flocking seasons is estimated at 0.04 birds per year or one every 25 years. Conclusion: Based on a combination of knowing the locations of all likely future flocking sites for the Brolga within 5 km of the proposed wind farm, all likely flocking site home range buffers (including the 300m disturbance buffer) have been mapped and turbines excluded from these. This will reduce the risk of flocking birds colliding with operating wind turbines to a negligible level. This, together with the low powerline collision risk, mean that significant reductions in the survivorship of flocking Brolgas in the area are not anticipated as a consequence of the proposed development. The total collision risk is estimated at between 0.13 and 0.95 birds per year, or, most likely, between 0.22 and 0.50 birds per year. Indirect disturbance Disturbance to breeding and flocking birds . Initial screening of historical information . Significant reduction in the extent of the from construction and operation of the to locate past breeding and flocking proposed wind farm to confine it to an proposed wind farm sites (2009-10) area with no historical breeding records . Landholder interviews and discussions . Further refinement of the project to locate additional past breeding and boundaries to avoid areas where Brolgas flocking sites breed and flock . Field investigations since 2009 to search for breeding and flocking Brolgas within the 10 km RoI, including fortnightly roaming surveys of all suitable wetlands and liaison with landholder network to track occurrences and

Page | 105 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Risk (as per Brolga Guidelines) Responses Outcome habitat suitability . Adoption of the 300 m disturbance . The operation of turbines is not buffer around all breeding and flocking considered likely to lead to indirect site home range buffers from which impacts including habitat exclusion of turbines have been excluded. Brolgas near the proposed wind farm Conclusion: Inclusion of an outer 300 metre disturbance buffer in all breeding and flocking site home range buffers and definition of turbine free buffers based on these have reduced the risk of significant indirect impact on Brolgas from the project to negligible levels. Barrier effects Long turbine arrays may prevent . Initial screening of historical information . Significant reduction in the extent of the movements of Brolgas between key to locate past breeding and flocking proposed wind farm to confine it to an habitats, such as between breeding and sites (2009-10) area with no historical breeding records flocking sites . Landholder interviews and discussions . Further refinement of the project to locate additional past breeding and boundaries to avoid areas where Brolgas flocking sites breed and flock . Field investigations since 2009 to . This confined the wind farm to a single search for breeding and flocking Brolgas approximately 8 by 5 km array rather within the 10 km RoI, including than an elongated series of turbine lines fortnightly roaming surveys of all suitable that would create lengthy barrier. wetlands and liaison with landholder network to track occurrences and habitat suitability Conclusion: Adoption of a single turbine cluster ad the wind farm layout has reduced the risk of significant barrier effects on Brolgas from the project to negligible levels.

Page | 106 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

6. REFERENCES Arnol, JS, White, DM and Hastings, I. 1984. Management of the Brolga (Grus rubicundus) in Victoria, Technical report series, Department of Conservation, Forests and Lands, Victoria. AusWEA (Australian Wind Energy Association). 2005. Wind Farms and Birds: Interim Standards for Risk Assessment. Report prepared by Brett Lane and Associates and AIRA Professional Services; Report No. 2003.35(2.2), July 2005. Band, W, Madders, M, and Whitfield, DP. 2007. Developing field and analytical methods to assess avian collision risk at wind farms. In: de Lucas, M., Janss, G.F.E. & Ferrer, M. (eds.) Birds and Wind Farms: Risk Assessment and Mitigation, pp. 259-275. Quercus Books: Madrid. Biosis Research. 2009. Modelled risk of Brolga collisions with turbines at the proposed Stockyard Hill Wind Farm. Biosis. 2011. Penshurst Wind Farm – Targeted Fauna Assessment Report, Port Melbourne, Victoria. BirdLife Australia. Undated. Birdata: The Atlas of Australian Birds and Birdata. Online database. At https://birdata.com.au/about_atlas.vm. BOM (Bureau of Meteorology). 2014. http://www.bom.gov.au/climate/data/index.shtml Brown WM and Drewien RC. 1995. Evaluation of two power line markers to reduce crane and waterfowl collision mortality. Wildlife Society Bulletin, 23, 217-227 Cook, ASCP, Johnston, A, Wright, LJ, and Burton, NHK. 2012. A Review of Flight Heights and Avoidance Rates of Birds in Relation to Offshore Wind Farms. Crown Estate Strategic Ornithological Support Services. Project SOSS-02. http://www.bto.org/science/wetland-and- marine/soss/projects.Department of Planning and Community Development 2011. Policy and planning guidelines for development of wind energy facilities in Victoria. DPCD, Melbourne. DEPI. Undated. Victorian Biodiversity Atlas. At http://www.depi.vic.gov.au/environment-and- wildlife/biodiversity/victorian-biodiversity-atlas. DPCD (Department of Planning and Community Development). 2012. Policy and planning guidelines for development of wind energy facilities in Victoria. Available at http://www.dpcd.vic.gov.au/planning/planningapplications/moreinformation/windenergy. Drewitt, AL, and Langston, RHW. 2008. Collision effects of wind-power generators and other obstacles on birds. Annals of the New York Academy of Sciences 1134: 233–266. DSE (Department of Sustainability and Environment). 2009. Draft guidelines for the assessment of potential impacts of wind farms on Brolgas. DSE, 26 June 2009. DSE (Department of Sustainability and Environment). 2012a. Interim Guidelines for the Assessment, Avoidance, Mitigating and Offsetting of Potential Wind Farm Impacts on the Victorian Brolga Population, Victorian Government (DSE), Melbourne, Australia. DSE (Department of Sustainability and Environment). 2012b. Atlas of Victorian Wildlife. Viridians Pty. Ltd. And DSE, Melbourne, Australia. DSE (Department of Sustainability and Environment). 2003. Action Statement No 119. Brolga Grus rubicunda. Melbourne, Australia. Erickson, WP, Johnson, GD, Strickland, MD, Young, DP, Sernka, KJ, and Good, RE. 2001. Avian Collisions with Wind Turbines: A Summary of Existing Studies and Comparisons to Other Sources of Avian Collision Mortality in the United States. National Wind Coordinating Committee (NWCC) Resource Document. Western Ecosystems Technology Inc. Journey North 2014. Whooping Crane: Flight Formation The Vs have it., http://www.learner.org/jnorth/tm/crane/FlightFormation.html, viewed 4th March 2014. Gerjets, D. 2006. Studie zur Verträglichkeit der Windkraftplanungen Schweringhausen/Wietinghausen. mit den Erhaltungszielen des EU-Vogelschutzgebietes Diepholzer Moorniederung und desFFH-Gebietes Wietingsmoor.

Page | 107 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Herring, MW. 2005. Threatened species and farming. Brolga: management of breeding wetlands in northern Victoria. ESAI sub-project 05118, Ecologically Sustainable Agriculture Initiative, Department of Sustainability and Environment, Victoria. King, KA. 2008. Behaviour patterns and habitat use of the Brolga (Grus rubicundus) at two flocking sites in south-west Victoria, Honours Thesis, School of Life and Environmental Science, Deakin University. Kingsford, RT, Halse, SA and Porter, JL. 2008. Aerial surveys of waterbirds – assessing wetland condition, Final report to the National Land and Water Resources Audit. 60pp, University of New South Wales, Sydney. Langgemach, T. 2013 Informationen uebver Einfluesse der Windenerienutzung auf Voegel, Staatliche Vogelschutzwarte, Stand 9.10.2013. LPO (Ligue pour la Protection des Oiseaux). 2014. ‘The Common Crane’, http://champagne- ardenne.lpo.fr/English/e_grue_cendree.htm, viewed 5th March 2014. Melvin, SM and Temple, SA. 1982. ‘Migration ecology of Sandhill Cranes: a review’, Proc. 1981 Crane Workshop: 73-87. Marchant, S and Higgins, PJ (eds). 1993. Handbook of Australian, New Zealand and Antarctic Birds, Volume 2: Raptors to Lapwings. Oxford University Press, Melbourne. Nations, CS, Howlen, S and Young, DP. 2012. ‘Collaborative landscape conservation approach: modelling potential impacts to migratory Whooping Cranes from wind power development.’ Proceedings of the Wind Wildlife Meeting IX. American Wind Wildlife Institute. Navarette, L, 2011. ‘Behavioral effects of wind farms on wintering sandhill cranes (Grus canadensis) on the Texas high plains.’ A thesis in wildlife science submitted to the graduate faculty of Texas Tech University. Price, ML.2008. The effect of human disturbance on birds: a selective review. In Lunney, D, Munn, A and Meikle, D (eds) Too Close for Comfort: Contentious Issues in Human-Wildlife Encounters, Royal Zoological Society of New South Wales, Sydney. Sheldon, RA. 2004. The characterisation and modelling of Brolga Grus rubicunda flocking habitat in south-western Victoria: Relationships between habitat characteristics, Brolga abundance and flocking duration, PhD Thesis, University of Ballarat, Victoria. SNH, 2010. Use of avoidance rates in the SNH wind farm collision risk model. SNH avoidance rate information and guidance note. Scottish Natural Heritage, Inverness. Stübing, S and Korn, M. 2006. Fachgutachterliche Stellungnahme zum Konfliktfeld Kranich – Windenergie. Unpublished report commissioned by JUWI GmbH, Germany.

Page | 108 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Appendix 1: Habitat assessment during the 2011 flocking season investigation

Wetland Survey date No. Brolga % Water Habitat Quality Comments number 26/04/2011 5 - 80% Moderate Partly drained lake in approx 2007. Freshwater waterbody likely to hold water in flocking season. 18/04/2011 9 Full Moderate Permanent shallow waterbody. Some drainage and overflow creating marsh habitat.

22/04/2011 11 - Full Moderate Permanent freshwater manmade lake. 20/04/2011 75 - Full Moderate Shallow freshwater wetland with some vegetation and some signs of disturbance. 26/04/2011 82 - 90% High Large freshwater swamp with abundant aquatic vegetation. 26/04/2011 85 - 90% High Largely permanent, shallow freshwater wetland with no signs of disturbance. Large birdlife population. 26/04/2011 86 - 10% Low Drained swamp. Drains into wetland 85. 26/04/2011 87 - 10% Low Drained tussock swamps. 26/04/2011 88 - 10% Low Drained and cultivated. 23/04/2011 90 - 20% Moderate Swamp with some aquatic vegetation. Grazed by stock. 26/04/2011 91 - 75% High Large shallow freshwater swamp with aquatic emergent vegetation. Abundance of waterbirds. 26/04/2011 92 - Full Moderate Large permanent shallow freshwater lake. Lack of vegetation. No access. Viewed by roadside. 26/04/2011 97 - Full High Man made large permanent lake with emergent and aquatic vegetation creating shallow freshwater marsh. 22/04/2011 101 - Dry Low Cultivated swamp. 21/04/2011 109 - 10% Low Drained swamp. No vegetation. 19/04/2011 110 80% High Permanent lake with surrounding vegetation. Shallow water.

26/04/2011 111 - Full Moderate Small shallow freshwater marsh with abundant emergent aquatic vegetation. Surrounded by River Red Gums. 21/04/2011 112 - Full High Large freshwater lake with fringing vegetation. Surrounded by stubble. 21/04/2011 113 - Dry Low Cultivated and sown to crop. Drained. 21/04/2011 114 - Dry Low Cultivated and sown to crop. Drained. 20/04/2011 117 - 20% Low Swampy marshland with little water and likely to be ephemeral. Tussock grass vegetation. 20/04/2011 118 - Full High Large permanent shallow freshwater swamp with emergent vegetation. 23/04/2011 120 - 20% Moderate Shallow freshwater swamp with some aquatic vegetation. 25/04/2011 121 - Dry Low Ephemeral swamp with some aquatic emergent vegetation. 25/04/2011 122 - 5% Low Small wet depression. No aquatic vegetation. 23/04/2011 123 - 10% Low Likely to be drained/ephemeral. No vegetation. 25/04/2011 124 - Full High Large shallow freshwater wetland with emergent and fringing aquatic vegetation. 26/04/2011 125 - 95% High Large shallow freshwater marsh with waterbirds and swans nesting. Abundant emergent vegetation. 26/04/2011 126 - Dry Low Likely to be drained. Lack of aquatic vegetation, pasture grasses present. 26/04/2011 127 - Full High Large shallow freshwater wetland with emergent and fringing aquatic vegetation. 26/04/2011 128 - 70% Moderate Ephemeral shallow freshwater swamp with some aquatic vegetation. 26/04/2011 129 - Dry Low Cropped. 26/04/2011 129 - 10% Low Cultivated. Dam next to swamp is full. 26/04/2011 130 - Full Moderate Shallow freshwater swamp with abundant emergent vegetation. 26/04/2011 133 - Dry Low Dam only. Swamp permanently drained and sown to crop. 21/04/2011 137 3 (2A, 1J) Full High High quality vegetated wetland. 21/04/2011 138 - Full Moderate Salt Lake. Brolgas seen previously but not often. Minimal vegetation. Bare edges, rocky. 21/04/2011 139 - Full Moderate Salt lake. Minimal vegetation.

Page | 109 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Survey date No. Brolga % Water Habitat Quality Comments number 21/04/2011 141 - 80% High Permanent shallow swamp with vegetation. 26/04/2011 144 - Full Low Drain line prone to disturbance - close proximity to buildings, heavily grazed, lost habitat. 26/04/2011 145 - Full Low Drain line prone to disturbance - close proximity to buildings, heavily grazed, lost habitat. 26/04/2011 149 - Full Moderate Large permanent lake with some fringing vegetation. Located along roadside. 26/04/2011 152 - Dry Low Large ephemeral swamp with some disturbance such as weeds. 26/04/2011 153 - 90% High Large shallow freshwater swamp with aquatic emergent and fringing vegetation. 25/04/2011 209 - Full High Large shallow freshwater swamp with some emergent vegetation. Likely to be ephemeral. 25/04/2011 221 - Dry Low Wet depression in paddock. Likely to be ephemeral or drained. 25/04/2011 222 - Full Moderate Permanent lake with some aquatic vegetation. Shallow at less than 0.5m deep. 21/04/2011 223 - Full Moderate Large permanent lake with little emergent vegetation. Stony raised banks. 21/04/2011 224 2A Full High Shallow freshwater wetland with some aquatic vegetation. 21/04/2011 225 2A 20% Moderate Dam nearby swamp with ankle deep water and some aquatic vegetation. 20/04/2011 235 Full Moderate Lake with permanent water. Man made.

20/04/2011 236 4 (2A, 2J) 20% High Tussock swamp surrounded by volcanic barriers. 21/04/2011 238 - 80% High Large permanent shallow lake with fringing and emergent aquatic vegetation. Fed by Bon Nerrin swamp. 19/04/2011 239 - Full Moderate Permanent salt/brackish lake with some aquatic vegetation. 21/04/2011 240 - 10% Moderate Small dam in swamp. Native grasses present. 20/04/2011 241 - Full Moderate Freshwater lake with some fringing vegetation surrounded by rocky outcrops. 27/04/2011 242 - 20% Moderate Swamp fenced off from stock. 19/04/2011 243 3 (2A, 1J) Full Moderate Large permanent freshwater lake with some fringing vegetation. Surrounded by crop. 20/04/2011 244 - 90% High Permanent spring fed Brackish-salt lake. 19/04/2011 245 Full High Permanent freshwater swamp with vegetation and lots of birdlife.

20/04/2011 246 - Full Moderate Salty semi-permanent lake. Lack of vegetation. Shallow lake. 29/04/2011 247 - Dry Low Drained and sown to crop. 19/04/2011 248 90% High Lake often seen with Brolgas.

19/04/2011 252 Full High Permanent water. Banked spring. Vegetated.

26/04/2011 253 - 90% Moderate Swamp with shallow water and tussock vegetation. 29/04/2011 254 - 70% Moderate Ephemeral shallow wetland with emergent vegetation. 18/04/2011 301 3 (2A, 1J) Full Moderate Large grassy swamp with shallow water. 19/04/2011 302 Dry Moderate Grassy swamp with some vegetation. Grazed.

19/04/2011 303 2A Dry Low Permanently drained swamp. Disturbed and no aquatic vegetation. 20/04/2011 304 4A Full High Man made. Brackish semi-permanent lake. 20/04/2011 305 3 (2A, 1J) Full High Shallow freshwater swamp with tussocks and native vegetation. 20/04/2011 306 - - High Shallow freshwater wetland. No signs of disturbance. 20/04/2011 307 - - Moderate Salty lake. Brolgas seen mostly in pairs. 20/04/2011 308 - - Moderate Salty lake. Brolgas seen mostly in pairs. 20/04/2011 309 - Full Low Man made irrigated dam. Brolgas seen on surrounding areas. No vegetation. 20/04/2011 310 - Dry Low Cultivated swamp. 20/04/2011 311 3A <30% Moderate Small semi-dry swamp with vegetation. 20/04/2011 312 - Full Moderate Semi permanent manmade lake. Some vegetation.

Page | 110 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Survey date No. Brolga % Water Habitat Quality Comments number 21/04/2011 313 - Full Moderate Dam spring fresh water. Aquatic and fringing vegetation. 21/04/2011 314 - Dry Low Dry cultivated swamp. Grazed. 21/04/2011 315 - Full High Permanent spring fed lake with vegetation such as Rush. 21/04/2011 316 - Full Moderate Man made dam. Permanent water with some vegetation. 26/04/2011 320 - 90% High Permanent wetland with aquatic fringing and emergent vegetation. 26/04/2011 321 - 70% Moderate Ephemeral shallow freshwater swamp with vegetation. 26/04/2011 322 - Full Moderate Man made permanent lake with some fringing vegetation. 28/04/2011 323 - Full High Permanent freshwater wetland. Vegetated. 29/04/2011 324 - Full High Large freshwater lake, some fringing vegetation. 29/04/2011 325 - 90% Moderate Ephemeral swamp with emergent and fringing aquatic vegetation. Abundance of waterbirds. 3/05/2011 326 - 70% Moderate Ephemeral wetland with aquatic fringing and emergent vegetation. 3/05/2011 327 - Full Moderate Large shallow lake surrounded by volcanic rock. Some fringing and emergent vegetation. 4/05/2011 328 - Full High Shallow freshwater wetland with abundant vegetation. 19/04/2011 501 2A Dry Low Dry swamp bed with vegetation. Likely to be ephemeral and provide nesting habitat when full. 19/04/2011 502 Full High Permanent spring. Significant bird life. Vegetated lake.

19/04/2011 503 Dry Low Dry marsh land. Likely to be ephemeral or drained.

approx 19/04/2011 504 3 (2A, 1J) Moderate Swamp with shallow water and emergent vegetation. 50% 19/04/2011 505 - Full Low Small ephemeral swamp. Heavily grazed. No aquatic vegetation. 19/04/2011 506 - Dry Low Drained swamp. Sometime forms wet depressions in wet years. Lacks aquatic vegetation. 19/04/2011 507 - 50% Low Shallow freshwater marsh. Tussock grasses, grazed. Usually dry. 19/04/2011 508 - 10% Low Overgrown with pasture grasses and tussocks. Small ephemeral swamp. No waterbirds. 19/04/2011 509 - Dry Low Drained and cropped. Seasonally wet meadow. 19/04/2011 510 - Full Moderate Permanent natural lake with some waterbirds present. Minimal vegetation present. 19/04/2011 511 - Full Low Permanent lake surrounded by crop and no aquatic vegetation. Some ducks present. 19/04/2011 512 - Dry Low Drained swamp, weedy. 20/04/2011 513 4 (2A, 2J) Full Moderate Freshwater lake with some fringing vegetation surrounded by stony knolls and barley crop. 20/04/2011 514 - 20% Low Small ephemeral swamp. Foxes present. 20/04/2011 515 - Full Moderate Natural permanent dam. Waterbirds present. Some fringing vegetation such as Tussocks. 20/04/2011 516 - 10% Low Some fringing vegetation and tussock grasses. Mostly dry and ephemeral. 20/04/2011 517 - 20% Moderate Small swamp with fringing and emergent vegetation. Good variety of bird life. 20/04/2011 518 - Dry Low Small permanently drained swamp. 20/04/2011 519 - 10% Low Small ephemeral swamp containing tussock grasses. 20/04/2011 520 - Full High Large permanent lake with fringing and emergent aquatic vegetation. 20/04/2011 521 - Dry Low Dry swamp with some vegetation. Likely to be ephemeral or drained. 21/04/2011 522 - 10% Low Dam present in swamp. Native grasses present. Some surface water. Not permanent. 21/04/2011 523 - 10% Low Small ephemeral wetland with some aquatic vegetation. 21/04/2011 524 - 5% Low Small ephemeral wetland with some aquatic vegetation. Small dam present. 21/04/2011 525 - 10% Low Large drained swamp. Heavily grazed. 21/04/2011 526 - Dry Moderate Ephemeral swamp with emergent and fringing vegetation.

Page | 111 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Wetland Survey date No. Brolga % Water Habitat Quality Comments number 21/04/2011 527 - 50% High Shallow freshwater swamp with emergent and fringing aquatic vegetation. 21/04/2011 528 - 10% Moderate Freshwater swamp with some aquatic vegetation. Grazed. 21/04/2011 529 - Dry Low Drained and grazed. No water or aquatic vegetation. 21/04/2011 530 - Dry Low Cropped and drained. Cropped to pasture. 21/04/2011 531 - 80% Moderate Overflow of dam creates marsh habitat with tussocks and emergent vegetation. 21/04/2011 532 - Full High Large shallow permanent freshwater marshland with emergent and fringing aquatic vegetation. 21/04/2011 533 - Dry Low Large swamp which is now permanently drained. 23/04/2011 534 - Full Moderate Ephemeral lake that is well drained into #118. Surrounded by trees. Some vegetation. 23/04/2011 535 - Dry Low Cropped to pasture. Permanently drained. 23/04/2011 536 - Full Moderate Large ephemeral swamp with shallow water and emergent vegetation. Usually dry most years. 23/04/2011 537 - 20% Moderate Ephemeral swamp with some aquatic emergent vegetation. 23/04/2011 538 - Dry Low Swamp has been permanently drained for crop. 25/04/2011 539 - Dry Low Drained swamp. Sometime forms wet depressions in wet years. Lacks aquatic vegetation. 25/04/2011 542 - Dry Low Improved quality north of railway line. South of railway line, swamp is cultivated. 25/04/2011 544 - Full High Permanent large lake lined with vegetation and trees. Brolgas sighted in past. 25/04/2011 546 - Full High Large shallow freshwater swamp with emergent and fringing aquatic vegetation. 25/04/2011 549 - 50% Moderate Shallow freshwater marsh with aquatic vegetation. Likely to be ephemeral. 25/04/2011 550 3 (2A, 1J) 80% Moderate Swamp net to busy road. Shallow water with high emergent vegetation. 25/04/2011 552 3 (2A, 1J) 80% Moderate Freshwater ephemeral wetland with some aquatic vegetation. 25/04/2011 555 - 90% Low May become significant as #533 drains into this area. Water is considerably fresher than Lake Gellie immediate east. 26/04/2011 556 - Full High Permanent dam that overflows into swamp creating shallow marsh area with emergent vegetation. 26/04/2011 557 2A 20% Moderate Wetland with some aquatic vegetation. Likely to be ephemeral or drained. 26/04/2011 558 - 20% Moderate Small dam that overflows into swamp creating shallow marsh area with emergent vegetation. 26/04/2011 559 - 10% Moderate Small wetland with emergent tussock vegetation likely to be ephemeral or drained. Foxes present. 26/04/2011 560 - Full High Large shallow freshwater swamp with abundance of aquatic emergent and fringing vegetation. 26/04/2011 561 - 10% Low Swamp drained into #560. Currently grazed by cattle. 26/04/2011 562 - 10% Moderate Wetland likely to hold water in wet years. Some aquatic vegetation. Ephemeral with some disturbance. 26/04/2011 564 - Dry Low Grazed, weedy and limited aquatic vegetation. Likely to be ephemeral or drained. Dam attached. 26/04/2011 575 - Dry Low Swamp dry and shows signs of disturbance. However, dam attached is full and supports aquatic fringing vegetation. 26/04/2011 576 - Full Moderate Shallow freshwater wetland with tussock vegetation. Grazed. 26/04/2011 582 - 10% Moderate Small shallow freshwater marsh. Some emergent aquatic vegetation. 26/04/2011 583 - Dry Low Likely to be ephemeral or drained. Some tussock grasses present. 26/04/2011 584 - Full Moderate Freshwater swamp. May hold water during flocking season, usually a lot drier. Emergent vegetation. 26/04/2011 585 - Dry Low Drained and cropped. 27/04/2011 586 - Dry Low Small swamp permanently drained and cultivated. 27/04/2011 587 - 90% High Permanent shallow freshwater lake with aquatic vegetation present. 12/05/2011 590 - Full High Large wetland with aquatic fringing and emergent vegetation. Access to stock. Fed by drain from Lake Kornong Wildlife Reserve. 28/06/2011 600 - Full High Permanent lake and wetland with emergent aquatic vegetation. Shallow marsh areas. Dam and drain at one end. 28/06/2011 601 2A Full Moderate Freshwater swamp with high emergent vegetation. May be ephemeral.

Page | 112 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Appendix 2: Flight paths recorded during 2012 flocking survey. Initial Initial distance Brolga Number Obs. Point Flight Habitat at direction Date* Flying/Walking Flight Direction Flight behaviour Habitat at origin from and Age (wetland) height (m) destination from observer observer (m) 17/08/2012 2 602 flying 50m circling circling unknown barriers 800 east 12/09/2012 2 514 flying 10 W flapping wet paddock wet paddock 300 SE 21/09/2012 2 west of 325 grazing-flying 40 - 60 south to Tiverton lake shallow swamp lake 200 south 22/09/2012 2 602 fly to roost landing from east to lake landing unknown lake 150 east 24/09/2012 2 602 fly to roost landing from east to lake landing unknown lake 150 east 25/09/2012 2 602 fly to roost landing from east to lake landing unknown lake 150 east 26/09/2012 2 602 fly to roost landing from east to lake landing unknown lake 150 east 27/09/2012 3 Unknown Flying 80 West flapping unknown unknown 200 North 27/09/2012 2 602 fly to roost landing from east to lake landing unknown lake 150 east 522 (for 8, 2/11/2012 21 522 flying 15 N/E gentle flapping unknown other 13 keep 1800 west flying north) 2/11/2012 8 522 flying 5 east landing unknown 522 800 S/W 2/11/2012 13 522 flying 15 N/E gentle flapping unknown unknown 800 west 2/11/2012 4 522 flying 10 N strong flapping dry swamp unknown 500 S/W 2/11/2012 2 522 flying 10 N strong flapping dry swamp unknown 500 S/W 2/11/2012 6 522 flying 10 N strong flapping dry swamp unknown 500 S/W 2/11/2012 12 522 flying 10 N strong flapping dry swamp unknown 500 S/W 2/11/2012 4 522 flying 10 N/W strong flapping dry swamp unknown 500 S/W 2/11/2012 3 522 flying 10 N/W strong flapping dry swamp unknown 500 S/W N/W, they curded over a 4/11/2012 8 500m N/E of 540 flying 8 group of Brolgas who were flapping 540 dry swamp 540 dry swamp 500 S/W unseen in canola crop Brolgas in crop field lifted 4/11/2012 10 500m N/E of 540 flying 10 off. All brolgas returned to unknown flapping edge of canola crop dry swamp 1200 west 540 4/11/2012 19 300m E of 240 flying 10 N circling erratic dry swamps unknown 500 S/W 5 broke off from 19 and 4/11/2012 5 300m E of 240 flying 10 returned to 540. Other 14 gentle flapping unknown dry swamp, 540 flew off initially landed on high spot in crop 100m N of gentle flapping and 4/11/2012 7 300m E of 240 flying/ walking 10 100m N of 240 unknown 300 west 240 and then another gliding 600m further, then gone landed about 500m N of dam in barley 4/11/2012 21 300m E of 240 flying/ walking 10 gliding unknown 800 N/W 240 crop 4/11/2012 21 300m E of 240 flying 10 N strong flapping dry swamps unknown 600 N/W 4/11/2012 14 300m E of 240 flying 10 N strong flapping 500m N of 240 unknown 500 S/W

Page | 113 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Initial Initial distance Brolga Number Obs. Point Flight Habitat at direction Date* Flying/Walking Flight Direction Flight behaviour Habitat at origin from and Age (wetland) height (m) destination from observer observer (m) 4/11/2012 2 300m E of 240 flying 10 N strong flapping dry swamp unknown 500 S/W 4/11/2012 10 300m E of 240 flying 10 N strong flapping dry swamp unknown 500 S/W 6/11/2012 3 522 flying 40 S gliding unknown dry swamp 522 1800 east 6/11/2012 3 522 flying 10 S gliding unknown 500 400 west U-shaped south, then east, crop, east of 6/11/2012 2 532 flying 12 gentle slow flapping swamp edge 250 west then north 529 canola crop edge pasture near 30/11/2012 2 515 flying 10 S/W gliding and flapping 100 N near water (515) 513 10/12/2012 2 N 139 grazing 20-60 Nth West erratic dry stony rises unknown 100 West 11/12/2012 2 N 139 grazing >60 Sth East straight dry stony rises unknown 100 West 12/12/2012 2 N 139 grazing >60 Sth East straight dry stony rises unknown 100 West 17/12/2012 3 602 flying 40 West, East flapping 602 spring 150 East 21/12/2012 2 245 flying 5 NE flapping swamp unknown 100 East 7/01/2013 2 325 flying and walking 60 NE flapping stony rises Lake 325 100 East 10/01/2013 2 325 flying 70 East straight unknown 325 100 Overhead 11/01/2013 2 602 flying 50 East straight and landing unknown spring 100 Overhead 13/01/2013 2 Spring flying 60-80 circling circling spring 602 200 North 15/01/2013 2 245 flying 4 South slow strong beating stubble (barley crop) 533, dry swamp 200 South

* Flights of Brolga flocks (≥ 10 birds) are highlighted.

Page | 114 Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Appendix 3: Home Range mapping methodology comparison report

Page | 115 Appendix 3: Home Range Mapping Comparison

Appendix 3: Home range mapping methodology comparison report HOME RANGE BUFFER MAPPING FOR BROLGA BREEDING SITES

The Victorian Brolga Guidelines require that turbine free buffers be identified for wind farm developments. For Brolga breeding sites, a default buffer of 3.2 kilometres from any breeding site is proposed, except where “site specific investigations can show with a high level of confidence the size and shape of home ranges for a project…” Home range buffers around breeding sites are mapped to encompass the range of movements expected to occur from incubation to fledging around current and historical Brolga breeding sites (i.e. wetlands in which Brolgas construct their nest).

Two methods have been used to date to define Brolga breeding site home range buffers: . A habitat modelling approach informed by empirical observations about flight distances and habitat choice of breeding brolgas developed by BL&A for the proposed Stockyard Hill Wind Farm); and . Spatial sampling of brolga activity at and around a breeding site to generate data that can be analysed using Kernel Analysis, a long-standing, widely used spatial statistical method that determines the probability of occurrence of the breeding Brolgas around its breeding site (a probability density map). These two methods are compared and contrasted in this Appendix. The current Brolga assessment has used the first method as there were very few completed breeding attempts by Brolgas within 3.2 km of the proposed wind farm on which to base breeding site home range buffers to inform the location of turbine free buffers.

1. BL&A habitat modelling method Based on previous Brolga breeding surveys for nearby wind farms (data pooled from Mortlake, Darlington and Stockyard Hill), a Brolga ‘breeding home range buffer’ has been defined as: • An inner zone of 400 metres has been mapped around each known historical and current (useable)1 breeding site. The majority of Brolga movements from breeding sites were observed within this zone and it is the most intensively used within the buffer. • Around historical and current breeding sites, all usable wetlands (i.e. not converted to farming uses) within 3.2 kilometres of each breeding site (the maximum distance breeding Brolgas were observed to move from a breeding site during breeding bird investigations) have been identified and included in the home range and any country between them and the 400 m inner zone, together with the above exclusion zone in a full ‘breeding home range’. • A further disturbance exclusion zone of an additional 300 metres around the entire ‘breeding home range’ has been established for each historical (useable) and current known breeding site. The Brolga Guidelines nominate this as the

1 The Brolga Guidelines state that an historical breeding site, once permanently drained and unlikely ever again to hold water need not be included in an assessment.

Page | 1 Appendix 3: Home Range Mapping Comparison

buffer distance that will prevent excessive disturbance of Brolgas thereby not disrupting their normal behaviour. The observations that were used to develop this habitat model are presented below. BL&A analysed data obtained during previous breeding surveys undertaken during 2007 to 2009 at three wind farms in south-west Victoria. The observations are combined for this analysis and represent up to 24 hours of observations per site of Brolga behaviour during daylight hours at a total of 24 breeding sites. Some 99 flights were observed by breeding Brolgas from their breeding wetland.

Flight distance The majority of Brolga movements observed were within 400 metres of breeding sites and many (86% of all flights recorded) occurred within 1600m (Figure 1). The greatest distance flown from a breeding site was 3.2 kilometres.

30 sample size = 99 flights 25

20

15

10 Percentage of of Observations Percentage 5

0 0 - 99 100 - 199 200 - 399 400 - 799 800 - 1599 1600 - 3200 Distance moved (m)

Figure 1: Observed Brolga movement distance from breeding sites

Habitat used Areas of pasture within 200 metres of the breeding site and the wetland habitat of the breeding site itself provided the bulk of the foraging resources during breeding as demonstrated by the considerable time spent in this part of the home range. Flights from breeding sites and the time spent at the location flown to averaged 45 minutes out of the approximately 12 hour period from sunrise to sunset at the time of the observations. On average, one such flight was made from the breeding site per day. The rest of the time was spent close to (within 200 metres) and in the breeding site wetland. Brolgas showed no preference for a particular habitat when flying up to 400 metres from their breeding site (Figure 2). However, once the limited extent of wetlands compared with pasture is considered, the habitat choice data from this inner zone indicated a statistically significant preference for wetland habitats (Binomial Test, p<0.001).

Page | 2 Appendix 3: Home Range Mapping Comparison

Observations indicated that the largest proportion of observed flights over 400 metres from the nesting site occurred towards wetlands, the species’ preferred foraging habitat (Figure 3). Given that wetlands represented a significant minority by area of the habitat available in this outer zone, the data indicated a statistically significant preference by brolgas for flying towards wetlands when foraging further than 400 metres from breeding sites (Binomial Test, p <0.001). This is not unexpected as the species is dependent on wetlands.

60

50

40

30

20 Percentage of of Flights Percentage 10

0 Wetland Pasture Crop Lake Grassland Unknown Habitats flown to less than 400m from nesting site

Figure 2: Summary of habitats to which Brolgas flew from their breeding sites (n = 99)

50 45

40 35 30 25 20 15 Percentage of of Flights Percentage 10 5 0 Wetland Pasture Crop Lake Grassland Unknown Habitat flown to more than 400m from nesting site

Figure 3: Summary of habitat to which Brolgas flew from their breeding sites (n = 99)

In conclusion, the data demonstrates that during the breeding season, Brolgas spend the majority of time foraging within the nesting site wetland and the immediately adjacent pasture. The majority of flights occurred within 400m of the site. The longer distance flights observed (400 metres to 3.2 kilometres) showed a preference for wetland habitat.

Page | 3 Appendix 3: Home Range Mapping Comparison

In summary, based on these findings, the BL&A home range mapping method includes an area within 400 metres of the breeding wetland, all wetlands within a 3.2 kilometre range of the breeding wetland plus a 300 metre additional disturbance prevention zone around all areas.

2. Kernel Analysis Method

Brolga behavioural observations have been undertaken for the proposed Penshurst wind farm (Biosis 2011). For this type of home range definition, systematically collected Brolga location data around a breeding site were used in kernel analysis to describe the breeding site home range and to inform the turbine free buffers for that project (Biosis 2011).

For breeding sites that failed before sufficient data could be collected to assign a stable home range, a conservative average radius of containment area was devised to provide a suitable buffer around these sites (Biosis 2010, 2011). This value is determined as the distance that the birds are contained within for a given percentage of time (Symbolix 2010). The radius of containment is based on data collected from pairs with stable home ranges at Penshurst and Mortlake in 2009. This average radius of 687.8m also has a 300m buffer around it as agreed with DSE. As a conservative measure the 99.9% home range was used, rather than the traditionally reported 95% home range.

The home range analysis and radius of containment principle was independently reviewed by a DSE biometrics expert and a DSE Brolga expert (Biosis 2011).

3. Comparison A comparison of three nesting sites surveyed during the 2009 breeding season at the proposed Penshurst Wind Farm is presented in Figure 4. The Kernel Analysis breeding site home ranges (solid outline) capture the 99.9% probability of brolga occurrence combined with an additional 300m buffer. The BL&A habitat-modelled breeding site home ranges (broken lines) comprise the inner 400 metres, all useable wetlands within 3.2 kilometres of the breeding site and the intervening country, and a 300 metre disturbance exclusion zone. Note that home ranges corresponding to the same breeding site are the same colour. The comparison indicates that the BL&A method is more conservative, encompassing a wider area of potential habitat. The Kernel Analysis method represents what an individual pair did during a single period of observations, whereas the BL&A habitat modelling method takes account of likely future potential movements about a breeding site by including all possible habitats to which birds may move. In this respect, the latter method provides greater assurance that the identification of turbine free buffers will buffer habitat likely to be used over a number of years, rather than those used during the particular year of observations.

Page | 4 GAZETTE LANE

KOLOR ROAD ")

KEARNEYS ROAD

") MINHAMITE-CARAMUT ROAD

STONEFIELD LANE

") PENSHURST-WARRNAMBOO ROAD JUBBS ROAD

GERRIGERRUP-MINHAMIT ROAD

JELBARTS ROAD

HAUSLERS ROAD

Legend Metres ") Brolga Breeding Records Breeding site homerange buffer 2 0 750 1,500 3,000 Roads Kernel Analysis Figure 4: Analysis of Penshurst Breeding Homerange

Wetlands BL&A habitat-modelled Project: Dundonnell Wind Farm Brolga Breeding Records 700m Buffer Breeding site homerange buffer 3 Client: Dundonnell Windfarm Pty Ltd Brolga Breeding Records 3km Buffer Kernel Analysis Project No.: 9184 Date: 15/10/2014 Created By: M. Ghasemi / B. Lane Breeding site homerange buffer 1 BL&A habitat-modelled Kernel Analysis ¯ BL&A habitat-modelled Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Appendix 4: Brolga activity spatial analysis – Symbolix

Page | 116 Brolga activity spatial analysis

Dundonnell wind farm site and surrounds 2011-12

Issue Version 2.0

17 Mar 2014

Submitted to

Brett Lane and Associates

www.symbolix.com.au

1/14 Akuna Drive

Williamstown North

VIC 3016

Brolga activity spatial analysis

Acknowledgements

Symbolix would like to acknowledge and thank the following people for their valuable insights, contributions and feedback in the preparation of this report and its contained analyses.

Teisha Sloane, Inga Kulik, Brett Lane & Mahsa Ghasemi – Brett Lane & Associates Version Control

Doc ID: BLABROL20120302

Main Author: Elizabeth Stark

Version Status Date Approved Issued to Copies Comments for release

0.9 For 5/3/12 S.Muir I Kulik, E For client review review T Sloane, B Lane

0.91 For 6/3/12 S.Muir As above E Added chi squared statistic review information

0.92 For 6/3/12 As above As above E Incorporated additional landowner review data, updated maps and analysis.

1.0 Issue 6/7/12 S.Muir As above E Incorporated comments on recent BUS surveys, reviewed maps.

1.1 Issue 15/4/13 S.Muir I Kulik E Amended error in table 1.

1.2 Issue 14/1/14 S.Muir I Kulik, E Added explanatory note in Summary B Lane

2.0 Issue 17/3/14 S.Muir I Kulik, E Updated maps and additional B Lane explanations as per peer reviewer comments

Approved for Release: 17/3/2014

Signed Date

Limitation: This report has been prepared in accordance with the scope of services described in communications between Symbolix Pty Ltd and Brett Lane and Associates. Any findings, conclusions or recommendations only apply to the aforementioned circumstances and no greater reliance should be assumed or drawn by Brett Lane and Associates. Furthermore, the report has been prepared solely for use by Brett Lane and Associates and Symbolix Pty Ltd.

Copyright: The concepts and information contained in this document are the property of Symbolix Pty Ltd. Use or copying of this document in whole or in part without the written permission of Symbolix Pty Ltd constitutes an infringement of copyright.

The Symbolix logo is a registered trade mark of Symbolix Pty Ltd. - i - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Table of Contents

1! Executive Summary ...... 1! 2! Methodology ...... 3! 2.1! Wetland habitat scores ...... 3! 2.2! 2011 brolga observations ...... 3! 2.2.1! Accounting for survey effort ...... 3! 2.2.2! Determining location of brolgas ...... 3! 2.2.3! Calculating relative brolga utilisation across the landscape ...... 4! 2.3! Historical data ...... 4! 2.4! Landholder survey ...... 5! 2.4.1! Assessing correlations between brolga presence and land use/type ...... 6! 2.5! Combining and using the data ...... 6! 3! Results ...... 7! 3.1! Wetland habitat scores ...... 7! 3.2! 2011 brolga observations (initial) ...... 7! 3.3! Historical data ...... 8! 3.4! 2012 Brolga utilisation survey (initial) ...... 8! 3.5! Landholder survey ...... 9! 4! Discussion ...... 11! Appendix A! Maps ...... A-1!

- ii - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

List of Figures

Figure 1 Number of times each wetland recorded as visited for observations in 2011...... A-1!

Figure 2: Wetlands surrounding the proposed wind farm, ranked according to brolga habitat quality...... A-2!

Figure 4: Relative likelihood of brolga observation at a given location, weighted by the number of brolgas in an observation (larger groups weighted higher) and by survey effort. A-4!

Figure 5: Relative likelihood of brolga observation at a given location, based on historical data only...... A-5!

List of Tables

Table 1: Number of property areas with different levels of brolga activity, against land type...... 10! Table 2 Number of property areas with different levels of brolga activity, against land use. . 10!

- iii - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

1 Executive Summary

Note: This report was originally prepared in 2012 as part of the Level 2 risk assessment under the ‘Interim guidelines for the assessment, avoidance, mitigation and offsetting of potential wind farm impacts on the Victorian brolga population’ (Victorian Government, 2011). It provided a spatial analysis of patterns in historical brolga data and assessed the relationship between land use/land type and brolga activity.

In addition, it provided an initial spatial analysis (based on data collected in 2011) and assessment of the 2012 brolga utilisation survey. This initial analysis was used to inform the final site layout and as input to the studies for the Level 3 risk assessment. As such, all maps in this report refer to the original proposed site layout. For final results pertaining to the final (2011-2013) spatial utilisation and activity rates for the wind farm site, please refer to the report “Dundonnell Wind Facility residual risk analysis: Turbine and powerline collision risk models”, Symbolix (2014), as it supersedes the results in this document.

The proposed Dundonnell Wind Farm is located in an area with potential brolga (Grus rubicunda) use. This requires an assessment of the potential for the wind farm to impact on the population. This report provides analytical support for assessing the likelihood of interaction between local brolgas and the wind farm and deciding on wind turbine placement that may reduce that likelihood.

The rarity of brolgas makes collecting enough direct data for a quantitative assessment of likelihood difficult. Where limited direct data exists it may be necessary to use more qualitative information (e.g. from historical sightings).

In this case, care must be taken to ensure that patterns in the data are not due to the particular data collection methodology. Assessing data from a range of different sources can reduce this risk, which is the approach taken in this report.

All data and data collection protocols (for direct observations and land holder surveys) were provided by Brett Lane & Associates (BL&A). The data sources used in this assessment include:

• Historical brolga records. These include approximate location and various metadata.

• Direct brolga observations throughout 2011. Two local landholders collected the data via roaming surveys and brolga location near local wetlands was noted.

• Landholder surveys. Local landowners were surveyed to determine if they had observed any brolga activity (small groups, flocks and/or nesting) on their land, and if so, where. Land use and land type was also recorded for areas within the properties.

• Wetland quality assessment. Each wetland in the area was assessed and rated according to habitat suitability by BL&A ecologists.

Some flight tracks were also recorded by the 2011 observers, but these were recorded too infrequently to be of any use in mapping a reliable flight path utilisation.

A formal brolga utilisation survey was also carried out in Autumn 2012, which collected observations from randomly placed sites across the proposed wind farm site. This dataset acts to validate the patterns detected during the roaming surveys in a randomly sampled, controlled survey design.

- 1 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Overall, the available data sets combine to provide a reasonable spatial description of areas of relatively higher and lower recorded activity.

When considered together, the data suggests the following key findings regarding brolga usage on the proposed wind farm, relative to the rest of the study area:

• Activity adjacent to the proposed wind farm site itself is low (relative to the rest of the study area) in the south and eastern portions of the site. The area adjacent to the south-eastern boundary is also relatively low usage.

• The north-west portion of the proposed site (and directly adjacent to this outside the wind farm boundary) consistently represents an area of reasonable wetland quality, high relative observed usage and high relative reported usage.

• In developing risk assessments and risk avoidance strategies, it is reasonable to consider the north-western area of the site as having a higher level of utilisation than the rest of the wind farm into the future. This may translate into a higher relative likelihood of interaction, depending on the proposed turbine layout and other wind farm specifications.

- 2 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

2 Methodology Each data set provided was processed and assessed separately. They are detailed below, along with the philosophy behind combining and comparing them for an overall assessment of relative brolga landscape use.

2.1 Wetland habitat scores Wetland habitat suitability was provided to Symbolix as a 3-point ranking (low, medium, high). No additional processing was applied to this data.

2.2 2011 brolga observations Data was collected by two landowners in approximately 80 separate observation sessions from April to November 2011. Observations included flocking and nesting behaviours. The observations were roaming within a 10km radius in and around the wind farm. On each survey a set of counts were made at individual or groups of wetlands. In total 807 of these individual brolga count observations were made. 647 (80%) of these observations recorded no brolga activity.

Because the surveys were roaming, information is not available on the exact length of time spent at each wetland group for each observation. This is why we are unable to generate a rate of flight for input into a traditional collision risk model.

2.2.1 Accounting for survey effort The survey effort covers a large portion of the year and has resulted in many observations, but the survey effort was not constant in space, which must be accounted for in the analysis. Figure 1 highlights the variation in survey effort for each wetland visited in 2011.

Of the wetlands surveyed, 50% were visited 4 or fewer times. However some wetlands were visited up to 48 times, including many of those adjacent to the proposed wind farm. If the brolga count were not scaled for this variation in survey effort, it would be biased towards those sites that were visited the most.

To account for this variation, we report brolga use as the average number of brolga per observation session per wetland.

2.2.2 Determining location of brolgas For each observation the observers recorded the number of brolgas and the wetland(s) under observation at the time. GPS location information was not available for this analysis.

To map the brolgas’ approximate locations a weighted brolga count was ascribed to the midpoint of each wetland for each observation. The weighted brolga count was the number of brolgas observed, divided by the number of wetlands in that observation. For example, if three brolgas were recorded whilst observing at wetland 1, 2 and 3, a weighted brolga count of one brolga is attributed to each of the three wetlands.

This has the effect of ‘spreading out’ the brolga observation over the recorded wetland area. By aligning the observations with the wetland locations, we are also able to scale for survey effort per wetland (see previous section).

- 3 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

If (as is usually assumed) brolga activity is centred on wetlands, this approach will not be biased. If, however, the brolgas in this area are commonly active away from wetlands, then this method may overestimate activity at wetlands relative to non-wetland areas.

To mitigate this risk we also assess historical brolga records, and a recent utilisation survey, as a comparison. This data set has specific latitude/longitude location information and so can be used to confirm the patterns of brolga activity near wetlands.

2.2.3 Calculating relative brolga utilisation across the landscape To better understand the likelihood of brolga interactions with the proposed wind farm, and to inform risk avoidance measures, we must answer the question: “Given that brolgas were recorded in these locations, where do we expect them to occur across the broader landscape?”

To answer this question, the individual observations are transformed into a probability density map, which gives the relative probability of a brolga observation occurring at a given location.

This was done by ‘smoothing’ out each of the individual observations using a kernel approximation (see for example, Silverman, 19861). Observations were weighted by survey effort (see previous sections). All calculations were carried out in ESRI ArcGIS.

A smoothing length of 2500m was chosen, based on rule of thumb methods from Silverman. Note that this loosely corresponds to a 95% probability that the observation’s true location was within 5 kilometres of the recorded location, with a maximum likelihood of being at the recorded point. The selection of a smoothing length amounts to a balance between bias and variance. Too large an estimate results in too flat a distribution, with little variation. Too small an estimate results in spurious detail and differences. As additional brolga observations are included (for Level 3 assessment) a smaller smoothing length should be applicable, which would allow finer detail to resolve.

2.3 Historical data A set of over 400 historical brolga observations were also provided for analysis. These were collected by land owners, professional and amateur observers over an unspecified period of time. This kind of data cannot be scaled for survey effort and no absence data is available.

For this reason it is not reliable as a single assessment of brolga location, as it is probable that areas with high brolga count are simply areas where observers have been more active.

The historical data is useful for the overall assessment, however, as it provides another stream of data collected according to a different methodology to that used for the formal roaming surveys. It contains location information which can be used as an additional background check of areas of brolga utilisation. This includes a check on whether brolga activity was regularly historically recorded away from wetlands. This will provide a measure of support for the 2011 observations, which were biased toward wetlands. It is also over a much longer temporal baseline. However, in recognition of the fact that historical records

1 Silverman, B.W., Density Estimation for Statistics and Data Analysis, Monographs on Statistics and Applied Probability, London,: Chapman & Hall, 1986.

- 4 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis may also be biased towards wetlands, we have also included a formal point count survey which is randomised with respect to habitat type (section 3.4).

A kernel analysis (also with smoothing length of 2500m, to be consistent with the 2011 observations) was carried out using the historical data.

2.4 Landholder survey A fourth stream of data was collected by surveying local landholders (within a 10km radius of the proposed wind farm).

Landholders were questioned about historical brolga use on areas of their properties, and the land use(s) and land type(s) were also recorded. Although the quality of this information will vary according to the landholders’ interest and length of time in the area, this data set provides a set of independent information.

Importantly, it is also information gathered across the entire study area, notwithstanding some properties where access was not granted or unavailable (e.g. Mt Fyans). As such it provides additional information in the areas away from wetlands or other features that might not be chosen for field surveys because they were not considered good brolga habitat beforehand.

This data was collected and provided as free form notes for each property, and maps indicating property areas with different attributes.

This information was scored according to the following criteria:

Brolga Frequency: Small/familial group sightings: 0: none recorded, 1: very infrequent,

2: infrequent, 3: sightings (no frequency level was recorded), 4: frequent, 5: very frequent, N/A: not available

Historical frequency of large groups/flocks: Ranked as above

Historical presence of breeding (frequency of 0: No breeding/nesting recorded nesting was not generally recorded, so a simple 1: Sightings of breeding/nesting recorded presence/absence code was used): N/A: not available

Land type: Categorical: • Arable • Stony • Aquatic (includes wetlands, dams and other water sources) • Mixed • Cleared/Semi cleared

- 5 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

• N/A

Land use: Categorical: • Cropping • Grazing • Mixed

2.4.1 Assessing correlations between brolga presence and land use/type If there exists a significant correlation between brolga use and the underlying land use or land type, this can provide additional guidance to help assess risk in areas where observational data is less available.

A standard Chi-squared test was applied to test for correlations between each of the brolga behaviour sets and the land use/type data.

2.5 Combining and using the data Each of the data streams provided has unique strengths and limitations. As such, no attempt has been made to combine the streams formally into a single assessment of brolga landscape use. This is to avoid introducing unknown biases or variation into this single assessment.

Instead, each data stream is used to map areas of higher or lower expected brolga use. As each set has different potential biases, patterns that persist across all three data streams can be considered as more reliable sources of evidence for decision making. Conversely, patterns that appear in only one set, and can be reasonably explained by the limitations of that set, should be considered less reliable evidence sources.

- 6 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

3 Results

Note: This report was originally prepared in 2012 as part of the Level 2 risk assessment under the ‘Interim guidelines for the assessment, avoidance, mitigation and offsetting of potential wind farm impacts on the Victorian brolga population’ (Victorian Government, 2011). As such it provides an initial assessment of the 2011 roaming brolga observations (section 3.2) and 2012 brolga utilisation survey (section 3.4). This initial analysis was used to inform the final site layout and as input to the studies for the Level 3 risk assessment. For final results pertaining to the activity rates for the wind farm site, please refer to the report “Dundonnell Wind Facility residual risk analysis: Turbine and powerline collision risk models”, Symbolix (2014), as it supersedes the results in this document.

Each of the individual data streams was evaluated and a spatial map of the expected brolga distribution was produced (see Appendix A). This section outlines the individual results and the combined implications are discussed in the following section.

3.1 Wetland habitat scores Figure 2 highlights that wetlands rated as high brolga habitat are fairly evenly spaced around the area adjacent to the proposed wind farm. This includes wetland IDs 118 and 112, directly adjacent to the north-west boundary of the wind farm site.

No strong spatial pattern is evident in this data alone.

3.2 2011 brolga observations (initial) Figure 3 codes each wetland by the average weighted brolgas count ascribed to that wetland per observation. There are a couple of small wetlands in the south of the study area where flocking events caused high number of brolga to be recorded over only one or two visits.

The second highest rate of activity is recorded at the wetlands in and adjacent to the north- west portion of the wind farm site. Wetland IDs 118, 513 and 585 recorded between two and four brolgas per observation, on average.

There are also chains of activity to the south and south east of the study area. In general, wetlands to the north recorded a lower weighted activity rate. It should be noted that brolga counts were more often ascribed to multiple wetlands (resulting in lower weighted count) when those wetlands were small and closely situated, as happens throughout the northern region of the study area.

This is partially accounted for by smoothing the observations using a kernel approximation to convert the raw data into a relative probability of observation.

Figure 4 maps the probability density, and also shows the location of the raw observations, for reference. The probability density is weighted by the number of brolga observed (larger groups are more highly weighted) and scaled by survey effort (so that areas with higher survey effort do not necessarily record higher probability).

This map also highlights that brolga activity in the region is more likely to be recorded in the area in an adjacent to the north-west boundary of the wind farm site. Higher activity is also recorded in the south and south east of the study area.

- 7 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Lower relative usage is recorded in the south and east of the wind farm. As we would expect from the survey design, recorded activity is centred on wetlands. To assert whether this is a reasonable pattern or is simply an artefact of a survey design that centred on wetlands, we apply the same kernel analysis to the historical data.

3.3 Historical data Figure 5 suggests a strong peak in activity in the south-east of the study area, more strongly than that suggested by the 2011 observations, or the landholder survey data below. We note that the majority of the observations in historical data set were contributed by a single observer, and that these were largely located in the south-east region. This region of high activity is likely biased as a result of this contributor.

The area to the north-west of the proposed wind farm has a higher relative likelihood historically, but the peaks in the historical data are situated further outside the wind farm boundary than in the 2011 data. Without further information regarding the nature of the historical data, including property access, the activity pattern in the 2011 data should be considered a more reliable source of evidence in this region.

We note, however, that the general pattern throughout study region suggests areas of high activity are centred on and between wetlands, as does the 2011 data. The area to the south and east of the wind farm, with few wetlands, has relatively lower observation likelihood. This suggests that, although the 2011 survey exhibits a tendency bias towards wetlands, the patterns of brolga activity observed within it align broadly with the patterns seen in the independent, historical data.

In recognition of the fact that we have limited knowledge of the survey methodology (and inherent biases) in the historical data, we have also sought additional streams of data to provide insight into brolga habitat preferences.

3.4 2012 Brolga utilisation survey (initial) As the observer information (such as proportion of observations that recorded no brolga, and proportion of observations near wetlands) is unavailable for the historical data, it provides some validation of the 2011 roaming surveys but is not ideal.

A round of brolga utilisation surveys were carried throughout April and May 2012, in part to further validate patterns of usage previously recorded. Brolga records were collected during 15-minute surveys from 20 sites across the proposed wind farm. The sites were randomly distributed, and included locations on a mix of land types.

Only two observations occurred on stony/grazing land, with the remainder on arable land. The majority of observations occurred in the north-west of the site, with two observations near wetlands in the south-west. No observations were recorded in the east and south-east (no major wetlands and dominated by stony rises).

This initial data set provides a descriptive confirmation that patterns recorded during the roaming surveys of 2011 are supported when land is surveyed with random site selection.

- 8 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis 3.5 Landholder survey Figure 6 and Figure 7 provide an overview of the basic land use and land/soil types in the study region. We can see that, broadly, the region covering the south and east of the wind farm site is stony ground used for grazing. Cropping on arable and mixed soils is more common in the north-west region of the wind farm site and throughout the north of the study area.

The landowner survey also collected qualitative information on the frequency of brolga use on the properties.

Table 1 summarises the number of counts of each frequency level against land type for small and large group sightings. Some frequency levels have been combined to obtain sufficient counts to allow a reliable correlation analysis. Mixed and cleared land types have been combined due to low count.

Testing for correlations (Chi squared) between brolga groups and land type found evidence of a non-random pattern for both small and large groups at the 99% confidence level.2 Both small and large groups are, on average, noted by landholders more frequently in aquatic or arable areas than on stony ground.

Table 2 displays the counts for frequency level against land use. Frequent use was significantly less reported on grazing land than cropping or mixed use (p <0.001 for small groups and p<0.05 for large groups)3.

These trends indicate a pattern of brolga land use that echoes the patterns seen in the 2011 and historical observations.

The reported frequency of historical brolga groups is mapped in Figure 8 - Figure 10. Again, this supports the previous data’s suggestion that brolga occurrence is highest in the north- west of the area, with geographically smaller areas of high usage in the south-east and a few pockets in the north-east.

In the area with and directly adjacent to the wind farm, higher relative frequency is recorded by landowners in the north-west region in and directly adjacent to, the site. The south and east of the wind farm is an area of lower relative recorded use.

Mixed/semi- Aquatic Arable cleared Stony SUM Small Groups Frequent/V. frequent 14 17 8 2 41 Sightings 10 5 3 2 20 Infrequent/V infrequent 21 17 14 18 70 None recorded 9 15 3 31 58 SUM 54 54 28 53 189

2 Small groups: χ2 = 39; df = 9; p< 0.001. Large groups: χ2 = 18.8; df=6; p<0.005 3 Small groups: χ2 = 34.5; df = 6; p<0.001. Large groups: χ2 = 13; df=4; p<0.05

- 9 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Large Groups/Flocks Frequent/V. frequent/ Sightings 11 9 5 2 27 Infrequent/V. infrequent 5 15 4 4 28 None recorded 38 30 19 46 133 SUM 54 54 28 52 188

Cropping Grazing Mixed SUM Small Groups Frequent/V frequent 17 5 15 37 Sightings 5 2 7 14 Infrequent/V infrequent 24 15 23 62 None recorded 17 40 9 66 SUM 63 62 54 179

Table 1: Number of property areas with different levels of brolga activity, against land type.

Large Groups/Flocks Frequent/sightings 10 2 7 19 Infrequent/V. infrequent 10 5 13 28 None recorded 42 55 34 131 SUM 62 62 54 178

Table 2 Number of property areas with different levels of brolga activity, against land use.

- 10 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

4 Discussion Each set of data provided has unique limitations, relating to the survey design used or the inherent uncertainties in historical, open datasets or landholder’s memories. As a combined meta-set, however, the limitations of each individual data set can be mitigated somewhat by comparison with another data set unhindered by that particular limitation.

Also, the individual data streams provide a reasonably consistent picture of relative brolga use across the study area. Activity is (perhaps unsurprisingly) centred on wetland and aquatic areas.

Activity adjacent to the proposed wind farm site itself is low (relative to the rest of the study area) in the south and eastern portions of the site. The area adjacent to the south-eastern boundary is also relatively low usage.

The north-west portion of the proposed site (and directly adjacent to this outside the wind farm boundary) consistently represents an area of reportedly reasonable wetland quality, high relative observed usage and high relative reported usage. The combination of historical and recent field data suggests that these patterns are resilient and can be expected to persist in the future.

In developing risk assessments and risk avoidance strategies, it is reasonable to consider the north-western area of the site as having a higher level of utilisation than the rest of the wind farm into the future. This may translate into a higher relative likelihood of interaction, depending on the proposed turbine layout and other wind farm specifications.

The data presented here provides information that can be used in constraints mapping to help reduce overall risk. If further modelling of residual risk is required, the patterns reported here may help inform utilisation scenarios for input into a collision risk modelling platform.

- 11 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Appendix A Maps

Figure 1 Number of times each wetland recorded as visited for observations in 2011.

- A-1 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 2: Wetlands surrounding the proposed wind farm, ranked according to brolga habitat quality.

- A-2 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 3: Average number of brolgas, per wetland, per observation.

- A-3 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 4: Relative likelihood of brolga observation at a given location, weighted by the number of brolgas in an observation (larger groups weighted higher) and by survey effort.

- A-4 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 5: Relative likelihood of brolga observation at a given location, based on historical data only.

- A-5 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 6: Land use categories as reported by landholders.

- A-6 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 7: Land type categories as reported by landholders.

- A-7 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 8: Historical brolga frequency - small groups of brolga, as reported by landowners.

- A-8 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 9: Historical brolga frequency – large groups and flocks, as reported by landowners.

- A-9 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

Figure 10: Historical brolga frequency – breeding occurrence, as reported by landowners.

- A-10 - BLABROL20120302, Version 2.0

Brolga activity spatial analysis

About Symbolix

Symbolix specialise in industrial process modelling, forecasting and support. We use your existing knowledge to audit, improve and expose your current business systems, or to design new approaches. Symbolix consultants will work alongside you and your team, providing you with analysis in conjunction with communication and project management services, Symbolix can provide a complete solution package, or expert consultants to assist at any stage of your project development.

Find out more at www.symbolix.com.au.

1/14 Akuna Dr Williamstown Nth VIC 3016 Telephone: +61 3 9397 2520 [email protected]

Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Appendix 5: CRM Report for proposed Dundonnell Wind Farm and transmission line

Page | 117

Dundonnell Wind Facility residual risk analysis

Turbine and powerline collision risk models

Issue

Version 2.3, 23 June, 2014

Submitted to

Brett Lane & Associates

www.symbolix.com.au

1/14 Akuna Drive

Williamstown North

VIC 3016

Dundonnell Wind Facility residual risk analysis

Version Control

Doc ID: BLABROL20131021

Main Author: Elizabeth Stark

Version Status Date Approved for Issued to Copies Comments release

0.8 Draft 21/10/13 internal Internal e For comment

0.9 Review 30/10/13 S. Muir I.Kulik; e For review – references B.Lane incomplete

0.91 Review 30/10/13 S.Muir I.Kulik; e Complete references and B.Lane final internal check

1.0 Issue 14/01/14 S.Muir I.Kulik; e Updated with final turbine B.Lane; and boundary layout A.Stewart

2.0 Issue 15/03/14 S.Muir I.Kulik, e Major revision based on B.Lane peer review comments

2.1 Issue 7/4/14 S.Muir I.Kulik, e Revision based on review B.Lane subsequent peer review comments

2.2 Issue 18/6/14 S.Muir I.Kulik, e Updated powerline model B.Lane results based on additional breeding sites

2.3 Issue 23/6/14 S.Muir I.Kulik, e Added clarifying B.Lane statements

Approved for Release: 23/6/14

Signed Date

Limitation: This report has been prepared in accordance with the scope of services described in communications between Symbolix Pty Ltd and Brett Lane & Associates. Any findings, conclusions or recommendations only apply to the aforementioned circumstances and no greater reliance should be assumed or drawn by Brett Lane & Associates. Furthermore, the report has been prepared solely for use by Brett Lane & Associates and Symbolix Pty Ltd.

Copyright: The concepts and information contained in this document are the property of Symbolix Pty Ltd. Use or copying of this document in whole or in part without the written permission of Symbolix Pty Ltd constitutes an infringement of copyright.

The Symbolix logo is a registered trade mark of Symbolix Pty Ltd.

Release at client discretion - i - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Table of Contents

1 Executive summary ...... 1 1.1 Final results ...... 1 1.1.1 Turbine collision risk model ...... 1 1.1.2 Powerline collision risk model ...... 2 2 Data overview ...... 3 3 Model details – turbine collision risk ...... 5 3.1 Methodology overview ...... 5 3.2 Key model inputs - summary ...... 5 3.3 Key model inputs – technical details ...... 7 3.3.1 Breeding/flocking season length ...... 7 3.3.2 Estimated number of flights (n) – flocking season ...... 8 3.3.3 Estimated number of flights (n) – breeding season ...... 10 3.3.4 Probability of a flight interacting with a turbine (Pr(I)) ...... 11 3.3.5 Probability of collision, following a turbine interaction (Pr(C|I)) ...... 12 3.4 Results ...... 13 3.4.1 Estimated number of flights onsite (n - flocking) ...... 13 3.4.2 Spatial probability scenarios (Pr(I)) ...... 15 3.4.3 Turbine layouts ...... 16 4 Model details – powerline collision risk ...... 18 4.1 Methodology ...... 18 4.1.1 Key model assumptions ...... 19 4.2 Data inputs ...... 20 4.2.1 Summary ...... 20 4.3 Results ...... 21 4.3.1 Flight distance and height ...... 21 4.3.2 Estimated crossings ...... 24 5 Discussion & interpretation ...... 26 6 References & further reading ...... 28

Release at client discretion - ii - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

List of Figures

Figure 1: Roaming surveys per wetland, from 2011 - 2013 inclusive...... 4

Figure 2: Brolga point count survey locations (and proposed turbine locations, for reference) used to determine overall brolga flight numbers on-site...... 9 Figure 3: Detection function and observed distance - brolga utilisation point count surveys. 14

Figure 4: The conditional probability density of flights (i.e. if a flight occurs, what is the probability that it will be in this area?) for the proposed Dundonnell site and surrounds, weighted to account for varying survey effort...... 16

Figure 5: Fitted flight distribution showing the cumulative distribution (black solid line) with individual data points (crosses), fitted gamma distribution (red line) and estimated 95% confidence interval on data (dashed line using Turnbull algorithm). Distribution fitting accounted for partial records (see section 4.2...... 22

Figure 6: Regional Data for the Breeding period, showing empirical distribution (solid line), estimated 95% confidence window (Turnbull algorithm) and fitted gamma distribution (red)...... 22

Figure 7: Empirical data, only records with known distance, showing the relationship between flight and height, and the resulting model. Top plot shows log-log relationship (as modelled), bottom shows model relationship with raw data...... 23

Figure 8: Range of possible brolga collisions per year for a combined (turbine + power line) average rate of 0.90. The green band is within the 95% confidence window, yellow are within the 99% confidence window and hashed are outside the 99% confidence bound...... 27

List of Tables

Table 1: Results from Collision Risk model with number of brolga estimated from BUS study and spatial utilisation density estimated from roaming surveys...... 2

Table 2: Results from powerline model with brolga height and flight distances collected during breeding ad flocking surveys at Dundonnell and throughout south-west Victoria. Estimated collisions based on empirical assumption of 1 collision per 10,000 crossings as used in subsequent population viability analysis (M. McCarthy, pers. comm.) ...... 2 Table 3: Key inputs, data sources and stages of the turbine collision risk model...... 7 Table 4: Engineering and biological model inputs ...... 12 Table 5: Flight height distribution from on-site brolga utilisation surveys...... 13 Table 6: Distance corrected, site-wide movement rate (flights above 30m)...... 15

Table 7: Results from Collision Risk model with number of brolga estimated from BUS study and spatial utilisation density estimated from roaming surveys...... 17

Release at client discretion - iii - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Table 8: Key inputs, data sources and stages of the powerline crossing model...... 18 Table 9: Powerline collision model input values...... 21

Table 10: Results from powerline model with brolga height and flight distances collected during breeding ad flocking surveys at Dundonnell and throughout south-west Victoria. Estimated collisions based on empirical assumption of 1 collision per 10,000 crossings as used in subsequent population viability analysis (M. McCarthy, pers. comm.) ...... 24

Table 11: Per wetland results from powerline model with brolga height and flight distances collected during breeding ad flocking surveys at Dundonnell and throughout south-west Victoria ...... 25

Table 12: Plausible (to 95% confidence) manifestations of the predicted annual collision rate. The proposed layout, with 95% avoidance is highlighted...... 26

Release at client discretion - iv - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

1 Executive summary The proposed Dundonnell Wind Farm is located in an area used by brolga (Grus rubicunda). Symbolix was contracted by Brett Lane & Associates (BL&A) to provide data advice and analytical support for assessing the likelihood of interaction between brolga and wind farm infrastructure.

Previous analysis (Symbolix 2014a) assessed a series of data sets collected during 2011/12, along with historical data. This provided information on the spatial distribution of brolga in the region and the relative activity levels across different land types. This information was used to inform initial turbine layouts and brolga activity buffers to minimise the interaction between brolga and wind farm infrastructure.

This subsequent report analyses the residual collision risk by estimating the potential for brolga collision with the final proposed turbine/powerline layout. To do this, we input data on the activity rate, flight height and the relative spatial flight distribution into a probabilistic model to predict potential for collision.

This report documents two specific collision risk models:

1. Turbine collision risk was estimated using an avian collision risk model, based directly on Band (2001,2007). The Brett Lane & Associates (BLA-BAND) application updates the model to accept spatial data inputs. This model has previously been applied to assess collision risk for Moorabool Wind Farm (Victoria). For full details see Section 3.

2. The number of powerline crossings was estimated using a geometric model and data on flight heights and distance from Dundonnell and other south-west Victorian sites. This model has been previously applied for Stockyard Hill Wind Farm. For full details see Section 4.

The results of these models have been provided specifically to input into a population viability analysis to estimate the potential population level effects.

These models provide an estimate of the residual risk to the brolga population.

1.1 Final results

1.1.1 Turbine collision risk model The following table summarises the output of the turbine collision risk model for the final layout (L08.0). Scenarios were built using avoidance rates of 90%-99% but we suggest that the 95% or 98% avoidance scenarios best balance a precautionary approach with what is recommended from other studies (see section 3.3.5 for a full discussion).

Two turbine layouts were considered, both with 104 wind turbine generators (WTGs). Both layouts were provided as GIS points from Trustpower. Following the initial collision risk modelling results, a discussion was held between Symbolix analysts, BL&A ecologists and Trustpower wind modellers. This considered the spatial density of flights and resulted in a final proposed layout, which moved and re-optimised the layout based on further buffering in the north-west area.

Release at client discretion - 1 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Layout L08.0 - 104WTG (final revised layout)

Avoidance rate 90% 95% 98% 99%

Modelled average, long-term, annual 0.91 0.46 0.18 0.09 number of collisions.

Probability of 2+ collisions in a given year 23% 8% 1% <1%

Table 1: Results from Collision Risk model with number of brolga estimated from BUS study (and breeding season scenario) and spatial utilisation density estimated from roaming surveys.

1.1.2 Powerline collision risk model Annual powerline crossings were estimated for the breeding and non-breeding (flocking) seasons separately. Following the current Victorian Guidelines (Victorian Government 2011), it was assumed that each season lasts for up to six months. The modelling results are summarised below.

Number Number of Total number Estimated of Sites individuals of crossings average at height per annual year collisions

Breeding Season (6 34 12.6 83 0.0083 months) (6 pair + juveniles)

Non-breeding/flocking 10 30 353 0.0353

TOTAL 436 0.044

Table 2: Results from powerline model with brolga height and flight distances collected during breeding ad flocking surveys at Dundonnell and throughout south-west Victoria. Estimated collisions based on empirical assumption of 1 collision per 10,000 crossings as used in subsequent population viability analysis (M. McCarthy, pers. comm.)

Release at client discretion - 2 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

2 Data overview All data used for this analysis were provided and validated by Brett Lane & Associates (BL&A). As such, details on field data collection and design are beyond the scope of this report. Some of these data streams have been previously analysed in Symbolix (2014a) and this report updates and supplements the data in that report. The data sources used in the collision risk assessment include:

• Direct brolga observation throughout 2011 (April – November), 2012 (August - December) and 2013 (January – early July). Two local landowners (trained by BL&A staff) collected the data via roaming surveys, noting brolga presence/absence at local wetlands and landmarks. In total 1459 individual wetland/landmark observations were recorded (Figure 1).

This data set is not the statistical ‘gold standard’ of a fully designed, random survey. However, it provides good spatial and seasonal coverage, and has allowed for a large dataset to be collected. Because care was taken to note brolga absence (i.e. when a wetland was surveyed and no brolga were present), it was possible to generate an activity rate (brolga per wetland per survey) from the data.

• Brolga utilisation survey from 6th April – 25th May 2012 and 9th – 23rd May 2013. Data were collected from randomly placed sites across the proposed wind farm site. This data was used (in part) to validate the patterns detected during the roaming surveys in a randomly sampled, controlled survey design (Symbolix 2014a).

This data does not account for any seasonal variation. However, it was timed to be during a time of expected high activity in the non-breeding season (B. Lane, pers. comms.). Concurrent roaming surveys confirmed that flocking was active in the local area during the survey period.

• Flocking surveys (2013): When flocking behaviour was detected during the roaming surveys, BL&A instigated flocking surveys, focussed on collecting flight track and height information at the flocking site.

• Flocking and breeding season data from other sites: BL&A provided flight height and distance data from other SW Victorian sites to supplement the Dundonnell data for powerline modelling.

Release at client discretion - 3 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Figure 1: Roaming surveys per wetland, from 2011 - 2013 inclusive.

Release at client discretion - 4 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

3 Model details – turbine collision risk

3.1 Methodology overview Applied to the development of wind energy facilities, collision risk models (CRMs) use the avian activity at a site to determine the statistical likelihood of turbine – bird interactions.

Although generally acknowledged that there must be other risk factors involved in determining the likelihood of strike, the current state of the art uses only the likelihood of proximity to the turbine. While this approach cannot provide insight into why an individual might collide, it is a common and accepted approach to inform risk assessments for windfarms.

Collision risk modelling (CRM) requires a step-wise risk model (Reason 1997), where the total risk is the probabilistic combination of the risk of each step in the process. The process can be summarised by the equation:

Ncollisions = n * Pr(I) * Pr(I|C) (1) where:

• Ncollisions is the estimated number of flights ending in collision

• n is the estimated number of flights in the region (expressed here as flights per hectare per hour)

• Pr(I) is the probability of a flight interacting with a turbine (given a flight in the region)

• Pr(C|I) is the probability of collision, given an interaction occurs.

Note that the output of the model is technically the number of flights that result in collision, which is implicitly assumed to be equal to the number of individual brolgas that collide. This requires that the removal of a small number of individuals will not affect the overall flight rate. This assumption may not be valid for all species (e.g. for resident raptors, a single pair of birds might be responsible for a large number of flights) causing this model to overestimate potential loss (as the loss of half the local population could substantively reduce the potential for flight/turbine interactions). In the case of brolga, where there are a number of individuals in the region (either multiple breeding pairs or a flock), this assumption is believed to be valid.

3.2 Key model inputs - summary Equation (1) shows that there are three key model inputs to the CRM, as summarised below.

Release at client discretion - 5 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Variable Data required Data Source Model input produced

Length of Rainfall data to establish the Annual rainfall data All year flocking season likelihood of a year as wet as from Penshurst (1906 estimated as 2010 occurring, under the – 2013) was used (see occurring in 5% assumption that the wet year section 3.3.1). of years. precipitated breeding season flocking. n Scenario for brolga breeding Regional population 1.93x10-05 flights (breeding) flights, based on estimate of 6 and activity estimates per hectare per breeding pairs (12.6 individuals) provided by BL&A hour in the region, with an average of based on breeding two flights daily. We have some season surveys at available (historical) data on Dundonnell and other flight heights, but to ensure SW Victorian sites (see transparency, simplicity, we section 3.3.2). assumed all flights could potentially be higher than 30m (this is likely to be a conservative assumption) n Distance corrected point count 2012/13 brolga point 3.8235x10-04 (flocking) data from on site (from 2012/13 counts (See section flights per brolga point counts. Refer to 3.3.2 and 3.4.1) hectare per hour table 6 for results and

confidence interval etc.)

Pr(I) The probability density (or 2011-13 local area Various values – utilisation map) of flights was observations (see taken from value derived for the region from section 3.3.4) of the 2D survey data. The spatial flight probability These data captured density was then directly density function both breeding and calculated at the turbine at each turbine. flocking events, so the locations (which allows the combined distribution model to compare different is used for both turbine layouts). seasons.

Pr(C|I) – Mechanical turbine All data from These inputs all turbine specifications. Trustpower feed into a parameters calculation based directly on Band (2001,2007) “Stage II)

Blade number 3

Rotor Diameter 126 m (minimum height ~30m)

Release at client discretion - 6 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Maximum Chord 4

Pitch 30o

Operational Frequency 0.2 rotations per second

Pr(C|I) – Brolga size and indicative Data provided by brolga flight speed (refer table 4). Brett Lane & parameters Associates

Based on data for whooping and common crane, which Flight Speed 60 kph are similar mass to brolga (e.g. Journey North 2014, LPO 2014)

Wing span 2.00m

Total Length 1.65m

Typical Flight Flapping

Pr(C|I) – 90%, 95%, avoidance Avoidance Rate 98%, 99% rate

Table 3: Key inputs, data sources and stages of the turbine collision risk model.

The BLA-BAND model’s use of the spatial activity levels (for Pr(I)) allowed us to allocate potential risk on a per-turbine basis. This information was used (in collaboration with BL&A ecologists and TrustPower wind engineers) to adjust the proposed turbine layout, with the intent of reducing the potential collision risk.

3.3 Key model inputs – technical details All data used for this analysis were provided and validated by Brett Lane & Associates (BL&A). Some of these data streams have been previously analysed in Symbolix (2014a) and this report updates and supplements the data in that report.

3.3.1 Breeding/flocking season length Data source: Bureau of Meteorology annual rainfall data (Penshurst, 1904-2013)

In building the model, differences in local activity between breeding and flocking (non- breeding) seasons were taken into account. In particular, for the flocking season, n is derived from onsite Brolga utilisation surveys (distance corrected). For the breeding season the smaller number of flights make it difficult to derive n directly, so a scenario model was used (see table below).

This leads to a separate estimate of collision for each season, which is summed to provide an annual expected rate, i.e.

(annual) breeding flocking Ncollisions = N + N

Release at client discretion - 7 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

According to the brolga guidelines (Victorian Government 2011) each season is nominally six months (182.6 days) long, however, it is possible (under the right conditions) to have flocking occurring during the breeding season. During the 2011-13 roaming surveys breeding season flocking (mainly juveniles) was observed in 2011. This followed a particularly wet year (2010), which was thought to precipitate the flocking event. It is important to account for this likelihood, as flocking poses a larger potential threat of collision than breeding.

To estimate the likelihood of a year this wet occurring, annual rainfall data (1904 – 2013) from nearby Penshurst was sourced from the Bureau of Meteorology. Analysis of the distribution of the rainfall showed that 2010 annual rainfall was in the 96th percentile. By comparison 2011 and 2012 were in the 66-68th percentiles.

To account for this the model allows for flocking all year for 5% of years (on average).

This leads to adjusted season lengths of 191.75 days (flocking/non breeding) and 173.5 days (breeding).

3.3.2 Estimated number of flights (n) – flocking season Data source: On site brolga utilisation survey (Apr – May, 2012/13)

The brolga utilisation survey data ranged from 6th April – 25th May 2012 and 9th – 23rd May 2013. This data was used (in part) to validate the patterns detected during the roaming surveys in a randomly sampled, controlled survey design (Symbolix 2014a). Brolga activity data were collected from randomly placed sites across the proposed wind farm site (Figure 2). Four survey locations were excised from the wind farm site following a change in site layout after 2012. These were excluded from analysis, leaving a total of 22 sites with 4705 observation minutes.

This data does not allow for any seasonal variation to be taken into account. However, it was chosen to be during high activity in the non-breeding season (B. Lane, pers. comms.). Flocking activity was recorded in the local area during the survey’s duration.

Release at client discretion - 8 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Figure 2: Brolga point count survey locations (and proposed turbine locations, for reference) used to determine overall brolga flight numbers on-site.

Release at client discretion - 9 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Data processing – distance correction

An observer’s efficiency can reasonably be expected to diminish as the observed object’s distance from the observer increases. To combat this effect from artificially biasing our risk assessment downward, we applied a distance correction (Buckland et al. 2001, Buckland et al. 2007, Thomas et al. 2009), which effectively accounted for the propensity to miss activity at a distance.

This resulted in an ‘effective detection radius1’ (EDR) that was used to scale the observed brolga flights up to an expected number of flights per hectare. Assuming brolgas are active for up to 12 hours a day, and 365 days per year, provided an estimated number of flights for the site.

The EDR was determined using all brolga utilisation survey observations, regardless of height. Restricting the distance correction flights at rotor-swept height would have resulted in too few records to carry out the analysis. In addition, the distance curve and EDR are calculated using only the first individual recorded in each observation (i.e. a brolga pair is treated as one observation). This is because we assumed that the detection of the second individual would be influenced by the fact we have detected the first. This simplification is only applied to calculate the EDR. All activity rates are based on the total number of individuals.

3.3.3 Estimated number of flights (n) – breeding season For the breeding season the smaller number of flights make it difficult to derive n directly from on ground surveys. Instead, a scenario model was used to estimate the total activity in the study region (that could interact with the site).

The key scenario assumptions are

• Breeding season movements take place for six months of the year (Victorian Government 2011). An adjusted season length of 173.5 days was used (see section 3.3.1) to allow for the situation where flocking might occur all year.

• Assumed six breeding pairs/year in the area, making an average 12.6 individuals/year, using the juvenile to adult ratio of 0.05 (Herring 2001 (in McCarthy 2008)).

• A average of two flights per day gives a total of 347 flights per individual per breeding season.

• Assume all flights could potentially be at risk of collision. There is some available data on typical flight heights, but this assumption was made for transparency and simplicity of the scenario. It is likely to be a conservative assumption.

1 Formally, this is the radius of the circle around each point such that as many objects are detected beyond this radius as remain undetected within the radius (Buckland et al 2001)

Release at client discretion - 10 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

3.3.4 Probability of a flight interacting with a turbine (Pr(I)) Data source: Local area brolga observations (roaming): 2011-13

Brolga observations were made throughout 2011 (April – November), 2012 (August - December) and 2013 (January – early July). Two local landowners (trained by BL&A staff) collected the data via roaming surveys, recording brolga presence/absence at local wetlands and landmarks. In total 1459 individual wetland/landmark observations were recorded.

This data set is not the statistical ‘gold standard’ of a fully designed, random survey. However, it provides good spatial and seasonal coverage, if the variation in survey effort is properly accounted for. If the brolga count were not scaled for this variation in survey effort, it would be biased towards those sites that were visited the most.

To account for this variation, we report brolga use as the average number of brolga per observation session per wetland. This is possible because care was taken to note brolga absence (i.e. when a wetland was surveyed and no brolga were present). The time spent at each wetland was recorded in 2012/13, but not in 2011, so the activity rate is presented per observation session, rather than in units of time.

Data processing - Determining location of brolgas

For each observation the observers recorded the number of brolgas and the wetland(s) under observation at the time. GPS location information was not available for this analysis.

To map the brolgas’ approximate locations a weighted brolga count was ascribed to the midpoint of each wetland for each observation. The weighted brolga count was the number of brolgas observed, divided by the number of wetlands in that observation. For example, if three brolgas were recorded whilst observing at wetland 1, 2 and 3, a weighted brolga count of one brolga is attributed to each of the three wetlands.

This has the effect of ‘spreading out’ the brolga observation over the recorded wetland area. By aligning the observations with the wetland locations, we are also able to scale for survey effort per wetland (see previous section).

If (as is usually assumed) brolga activity is centred on wetlands, this approach will not be biased. If, however, the brolgas in this area are commonly active away from wetlands, then this method may overestimate activity at wetlands relative to non-wetland areas.

To mitigate this risk we previously (Symbolix 2014a) also used data from local (within 10km of the site) land-owner surveys to establish a statistically significant trend in brolga activity related to land type. These results supported the assertion that brolgas preferentially use wetlands and arable land, with lower activity recorded in stony paddocks. This pattern was supported by the brolga utilisation surveys, historical data and roaming surveys.

Data processing - calculating relative brolga utilisation across the landscape

To better understand the likelihood of brolga interactions with the proposed wind farm, and to inform risk avoidance measures, we must answer the question: “Given that brolgas were recorded in these locations, where do we expect them to occur across the broader landscape?”

To answer this question, the individual observations were transformed into a probability density map, which gives the relative probability of a brolga observation occurring at a given location. This was done by ‘smoothing’ out each individual observation using a kernel

Release at client discretion - 11 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis interpolation (see for example, Silverman, 1986). Observations were weighted by survey effort (see previous sections).

3.3.5 Probability of collision, following a turbine interaction (Pr(C|I)) Key turbine & biological specifications:

The key specifications used in the model are provided in Table 4.

Turbine specifications Brolga biological inputs Item Specification Item Specification Blade number 3 Flight Speed 60 kph Rotor Diameter 126 m Wing span 2.00 Maximum Chord 4 Total Length 1.65 Pitch 30o Typical Flight Flapping Operational Frequency 0.2 rotations per second Table 4: Engineering and biological model inputs

Micro-Avoidance Rate:

This number is not well established (and difficult to measure) for individual species and can vary depending on the particular CRM used. In addition, the avoidance rate (as positioned here) can excapsulate more than the simple, physical propensity to diverge around an obstacle in flight (Madders & Whitfield (2006)), and avoidance rates from different studies and models may not always be immediately comparable.

This model also considers only micro-avoidance. Cook et al (2012) defines macro-avoidance as the propensity for a bird to avoid an entire wind farm site, whereas micro-avoidance is the ability to avoid individual turbines within a facility. As there is no data available on macro- avoidance (and no reasonable published argument that brolgas would avoid the wind farm site as a whole) we have assumed that the measured flight activity rates (n) and the larger scale activity density (Pr(I)) are largely unchanged following construction.

Studies of avoidance and benchmarks of the Band (2007) CRM are limited in the Australian context, but a number of European studies have used post construction monitoring, and direct measurement with radar, to benchmark the model and suggest appropriate avoidance parameters for different species classes.

SNH 2010 and Band 2012 suggest 98% avoidance as a rule of thumb, and most published papers suggest avoidance of 95% and above for similar species, with 95% seen as precautionary (SNH 2010, 2012). Recent studies suggest that generally collision risk models tend to over-estimate the risk when compared to post construction data, leading to avoidance rates of 99% and above being employed (see for example paragraph 6 in SNH 2012).

In Victoria, avoidance rates of 95% and above were used in estimating brolga collision risk for the Stockyard Hill Wind Farm (Biosis 2008). We present results for a range of avoidance values, in accordance with current Australian practice and international guidelines, e.g. Cook, et al. (2012) who recommend modelling a range from 95% - 99.5% avoidance. For precautionary purposes, we present a range from 90% - 99% avoidance. However, current international guidelines and Victorian precedent suggest rates of 95% and above are most reasonable.

Release at client discretion - 12 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis 3.4 Results

3.4.1 Estimated number of flights onsite (n - flocking) Number of individuals recorded during utilisation surveys

During 2012/13 a total of 30 individual observations of (flying and walking) brolga were seen across the 22 onsite locations. The flight heights were recorded in ranges, which were split into minimum and maximum flight heights for analysis. The distribution of minimum and maximum recorded flight height is shown in Table 5.

The 18 flights with minimum height = maximum height = 0m were brolgas observed walking only. All observations in this table were used to generate an effective detection radius for distance correction (see below). Of these, only two flights were recorded as being between 20-40m high. The set of turbines under consideration for this site all have a minimum rotor height above 30m, so these two flights are the only two observed as ‘at risk height’ for the purposes of modelling.

Height 0m2 0-5m 5-10m 10-20m 20-40m

Maximum flight height (top of recorded 18 2 4 4 2 range)

Minimum flight height (bottom of 18 4 2 6 0 recorded range)

Table 5: Flight height distribution from on-site brolga utilisation surveys. The bold numbers are those flights that could be within rotor swept height.

Distance corrected activity rate

As described in Section 3.3.2, distance correction was applied to account for limited observer visibility. This was done by calculating the Effective Detection Radius (EDR) from the reported distance from observer to brolga at first detection. This is used to correct the flight rate, which is used as an input to the collision risk model.

This process used the distance at first detection, recorded during the survey. As only two flights were recorded at rotor swept height, all heights in Table 5 were included to calculate the EDR. However, a number of observations were of pairs, where the second observation cannot be considered independent of the first. In these cases, the observation of the pair was taken as a single observation for the purpose of determining the distance distribution.

Using DISTANCE 6 (Thomas et. al. 2009) a set of detection functions were fit to the distribution of distances and the most appropriate chosen using AIC (Akaike Information Criteria). The chosen model was a hazard rate key of the form k(x) = 1 – exp(-(x/A)-B), where A = 274.0 (s.e. 127.1) and B = 2.733 (s.e. 0.53).

Table 6 summarises the results of one of the goodness of fit tests applied (the χ2 test). Overall the model represented a reasonable fit to the observed distance data at larger

2 The 18 flights with minimum height = maximum height = 0m were brolgas observed walking only.

Release at client discretion - 13 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis distances. The model appears to overestimate the detection probability at small radii (Figure 3). In the author’s experience, this is common for larger species. Although the detection probability is (by definition) equal to one at a radius of zero, in practice larger species are usually detected before they are overhead/at the observer.

Histogram Bin (m) Observed Values Expected Values 0 : 500 9 8.73 500 : 1000 4 3.54 1000 : 1500 0 1.43 1500 : 2000 1 0.79 2000 : 2500 1 0.51 Total χ2 = 2.0277; DoF = 2, p = 0.36 Table 6: χ2 test of agreement between model and observed value for detection distance.

Figure 3: Detection function and observed distance - brolga utilisation point count surveys.

The Effective Detection Radius (EDR) was 461m (95% confidence interval [229m, 925m]).

The adjusted flight activity rate is summarised in Table 7. Accounting for the size of the development site we determined an average of 1.54 flights at height, per hour on-site.

Release at client discretion - 14 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Flights per hectare per hour Coefficient of variation 95% confidence interval3

3.8235e-04 66% (1.046e-4, 1.405e-3)

Table 7: Distance corrected, site-wide movement rate (flights above 20m).

3.4.2 Spatial probability scenarios (Pr(I)) The relative spatial activity density is shown in Figure 4. This includes all data from 2011- 2013. Previously (Symbolix 2014a) we reported the activity density based on the first year of surveys. This updated figure still suggests that the highest activity density occurred to the north-west of the site. Although a number of flights were recorded during 2013 in the wetlands to the south-east, the north-west displays a higher relative flight density after accounting for survey effort.

Once the density map was built, the relative probability of a flight at the turbine locations (given that a flight occurs) was extracted from the GIS.

3 The 95% confidence range has been determined using the technique of Burnham et al 1987, with the variation that rather than Zα, a Student’s t-distribution better reflecting the true degrees of freedom was employed (Satterthwaite (1946))

Release at client discretion - 15 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Figure 4: The conditional probability density of flights (i.e. if a flight occurs, what is the probability that it will be in this area?) for the proposed Dundonnell site and surrounds, weighted to account for varying survey effort.

3.4.3 Turbine layouts Two turbine layouts were considered, both with 104 wind turbine generators (WTGs). Both layouts were provided as GIS points from Trustpower. The L03.11 layout incorporated brolga

Release at client discretion - 16 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis buffer regions from earlier stages of the risk assessment and other (non-ecological) constraints. The turbine placement was then optimised for wind energy generation.

As Table 8 shows, the site-wide estimated average collision rate for the L03.11 layout was similar to the simple ‘flat’ assumption, which simply assumes the flight activity density is equally distributed across all turbines. Given previous analyses (Symbolix 2014a) that suggest preferential use of wetlands and arable land, the flat assumption is simplistic, but useful as a validation of the model. The L03.11 layout (although optimised for wind resource) is essentially randomly distributed throughout the site relative to brolga use. Therefore, when the individual turbine risks are summed, we expect a total collision rate similar to the flat scenario (and this is what we see).

The limitation of the flat scenario is that it offers no insight into how the residual risk might be further reduced. Instead, the spatial activity density was used for this purpose.

Following the L03.11 results, a discussion was held between Symbolix analysts, BL&A ecologists and Trustpower wind modellers. This considered the spatial density of flights and resulted in a revised layout (L08.0), which moved and re-optimised the layout based on further buffering in the north-west area.

The final scenario predictions are shown below.

Flat distribution (assume 104 Layout L03.11 - 104 WTGs Layout L06.2 - 104WTG (final WTGs with equal probability (original layout) revised layout) of flight activity) Avoidance 90% 95% 98% 99% 90% 95% 98% 99% 90% 95% 98% 99% rate

Modelled average, long-term, 1.3 0.65 0.26 0.13 1.23 0.61 0.25 0.12 0.91 0.46 0.18 0.09 annual collision rate

Probability of 2+ collisions 37% 14% 3% <1% 35% 13% 3% <1% 23% 8% 1% <1% in a given year

Table 8: Results from Collision Risk model with number of brolga estimated from BUS study and spatial utilisation density estimated from roaming surveys.

Release at client discretion - 17 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

4 Model details – powerline collision risk

4.1 Methodology The model is a simple projection of activity across the landscape. This is done via geometrical assumptions about the number of flights, the overall distribution of flight lengths, and the flight height at a given flight distance. It assumes (and models) a relationship between the flight height and the distance flown. The model gives a range of possible heights for any distance.

The spatial inputs were the locations of all traditional flocking and breeding wetlands. The provided data included the radial distance to the closest point on the powerline for all wetlands.

Using flight height and distance information, we calculate the (symmetric) flight probability distribution as a function of radius from the wetland. The number of crossings was estimated from this flight probability at the powerline location and the total number of flights expected across the region.

We need not limit the total flight distance, but assume that all birds land and rest within some maximum distance (currently 10km) of the take off point.

Stage Data required Data Source Model input produced

1 Expected number of Brett Lane (pers. Potential flights per wetland site. flights in the region comm.). See Section each year, associated 4.1.1. with breeding and flocking

2 Flight distance 2013 flocking season Percentage of flights that travel observations during observations and required distance (between flocking and breeding. historical BL&A data wetland and powerline). Obtained (breeding). by fitting a distribution to the flight distance data.

3 Flight distance and 2013 flocking season Percentage of flights that are at the height observations observations (checked requisite height (after travelling the during flocking and against historical required distance). Obtained by breeding. breeding set – height- fitting a generalised linear model to distance relationship was relate flight height to distance, and equivalent). using the residuals of this model to estimate the likelihood of a flight through the powerline height range 15-22m.

Table 9: Key inputs, data sources and stages of the powerline crossing model.

Release at client discretion - 18 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

4.1.1 Key model assumptions Annual powerline crossings were estimated for the breeding and non-breeding (flocking) seasons separately. Following the current Victorian Guidelines (Victorian Government 2011), it was assumed that each season lasts for up to six months. The other input assumptions are listed below. All data (flights and wetland/powerline locations) and ecological assumptions are provided by Brett Lane & Associates, except where indicated.

Both breeding and flocking

Model parameters:

• The modelled relationship between flight height and flight distance was based on flight track data from Dundonnell WF area in 2013 (see below), for both the breeding and flocking model. This relationship was similar to that seen at other sites.

• Powerline is 15-22 metres above the ground.

• Assume all birds, irrespective of flight length, land and rest within 10 km from take- off point.

Technically, being a geometric representation there is little need for the last dot point. The outcomes display asymptotic behaviour as one allows this maximum distance to approach infinity. The maximum allowed radius is only applied to generate a better estimate of the range of flight heights when crossing the transmission line’s transect.

Breeding season

Model parameters:

• Breeding season movements take place for six months of the year (Victorian Government 2011)

• Assumed six breeding pairs/year in the area, making an average 12.6 individuals/year, using the juvenile to adult ratio of 0.05 (Herring 2001 (in McCarthy 2008))

• A average of two flights per day gives a total of 365 flights per individual per breeding season

• Assume equal probability of using any breeding site in any one year

• Assume mutually exclusive (individual) use of wetlands (i.e. each pair breeds at one site only)

• The distribution of flight distance is taken from combined BL&A breeding data from a range of sites through south-west Victoria (due to there being no suitable breeding data at Dundonnell). See section 4.2 for more information.

• For powerlines that intersect breeding sites, assume 50% of flights are in the direction of the line, and all of those flights are potentially at risk of intersection.

Non-breeding/Flocking season

• We use all sites denoted as “traditional” flocking sites (excluding “one-off” flocking sites) in the provided data.

• Assume 30 individuals in the area. This number was supported by a review of the flocking and roaming survey data from 2011-13, to confirm common flock sizes.

Release at client discretion - 19 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

• Assume six-month long flocking season.

• Birds average four flights per day per individual, resulting in 730 flights per bird per season.

• Assume mutually exclusive use of wetlands (flocks are based around one wetland at a time) and there is an equal probability of using any of the flocking wetlands

• The distribution of flight distances taken from 2013 Dundonnell flocking study. See section 4.2 for more information.

4.2 Data inputs All data used for this analysis were provided and validated by Brett Lane & Associates (BL&A).

Flocking surveys (2013):

When flocking behaviour was detected during the roaming surveys in 2013 (section 3.3.2), targeted flocking surveys were instigated, focussed on collecting flight track and height information at the flocking site. In these surveys, two observers with radio communication noted the number of brolga, flight height, direction, start and end habitats/wetland identifiers, distance and direction from observer and start/end times. They also mapped flight tracks, which were digitised to extract flight distances.

Statistically speaking, the flight data was “censored data”, which simply means that we were missing some information (either the start or end point) on some records. For flights where the end point (or initial point) was not a landmark, the recorded flight’s distance was treated as a minimum value.

Flocking and breeding season data from other sites:

BL&A provided flight height and distance data collected during planning for other Victorian wind facilities (Darlington, Mortlake, Stockyard Hill) to supplement the Dundonnell data for powerline modelling. In this data, all flight distances were estimated by the field observer directly.

4.2.1 Summary The full set of powerline model data inputs are summarised below.

Release at client discretion - 20 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Input Value

Flight distance distribution – from Dundonell and historical data Used to fit distance distribution.

Relationship between flight height and distance – from Dundonnell Used to fit regression and historical data. model/estimate height range probability

Location & number of wetlands Refer Table 10

Data for both breeding and flocking

Length of seasons Both 6 months

Height of powerline 15-22m

Max flight length before rest 10km

Breeding season

Number of breeding pairs in area 6

Juvenile to adult ratio 0.05

Average flights per day per bird 2

Flocking season

Number of individuals in area 30

Average flights per day per bird 4

Table 10: Powerline collision model input values.

4.3 Results

4.3.1 Flight distance The first stage of the model is to estimate the proportion of flights that travel the requisite distance to the powerline.

For this site, we were required to provide information for both the breeding and the flocking periods. The propensity to fly a given distance is different during these two times.

Flocking data was collected (as detailed previously) and a distribution fitted. This process accounts for the fact that many flights only have an effective minimum flight distance, as flights may disappear from site before the bird lands (i.e. statistically, the data is called “right- censored”).

The cumulative flight distance distribution is shown in Figure 5.

Release at client discretion - 21 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Figure 5: Fitted flight distribution showing the cumulative distribution (black solid line) with individual data points (crosses), fitted gamma distribution (red line) and estimated 95% confidence interval on data (dashed line using Turnbull algorithm). Distribution fitting accounted for partial records (see section 4.2.

As no data were collected during breeding events on this site, BL&A provided regional breeding season data from other Victorian sites. This demonstrates that, the propensity to fly longer distances than 4 km is much smaller during breeding than during flocking (Figure 6).

Figure 6: Regional Data for the Breeding period, showing empirical distribution (solid line), estimated 95% confidence window (Turnbull algorithm) and fitted gamma distribution (red).

Release at client discretion - 22 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

4.3.2 Probability of flight at powerline height The final input to the model was the relationship between flight height and distance. The data is transformed4 and a linear model fit (R2 = 0.98, Figure 7) 5. The residuals are modelled using a Z distribution to estimate the probability of a flight within the powerline height band, so the normality of the residuals must be maintained. In this case the model residuals show neither skew (Z= -1.06, p=0.3) nor kurtosis (Z = 0.24, p =0.8) (Omnibus test χ2 = 1.17, p = 0.55).

Figure 7: Empirical data, only records with known distance, showing the relationship between flight and height, and the resulting model. Top plot shows log-log relationship (as modelled), bottom shows model relationship with raw data.

4 The data is transformed (log(Height) ~ log(Distance)) to ensure the model’s residuals are Gaussian, independent and identically distributed. This is particularly important in this case because the residual spread is required to calculate the probability.

5Statistical note: The R2 value reflects the model fit relative to the null model. In this case the fitted model has only a gradient term, no intercept, so the null model variance reduces to the variance of the residuals. This is noted as it is a slightly different interpretation than is usual.

Release at client discretion - 23 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

The model (and the range of residuals) is then used to generate a model of the relative proportion of all flights that will fall within the heights of the transmission lines transect. In effect, we are modelling the band that contains the dots, rather than the red line that predicts the average.

4.3.3 Estimated crossings The total number of estimated, annual powerline crossings are provided in Table 11. The model results per wetland are provided in Table 12.

Number Number of Total number of crossings Estimated average of Sites individuals at height per year annual collisions

Breeding Season 34 12.6 83 0.0083 (6 months) (6 pair + juveniles) Non-breeding/ 10 30 353 0.0353 flocking

TOTAL 436 0.044

Table 11: Results from powerline model with brolga height and flight distances collected during breeding ad flocking surveys at Dundonnell and throughout south-west Victoria. Estimated collisions based on empirical assumption of 1 collision per 10,000 crossings as used in subsequent population viability analysis (M. McCarthy, pers. comm.)

Distance from % of Flights Type wetland to Projected Minimum Height that cross power line [m] power line Breeding 2808 30.43 0.19% Breeding 2839 30.58 0.18% Breeding 2706 29.95 0.22% Breeding 7091 20.96 0.00% Breeding 4614 27.47 0.03% Breeding 2362 28.26 0.36% Breeding 8639 15.08 0.00% Breeding 3758 29.34 0.07% Breeding 1458 22.96 1.52% Breeding 6720 22.09 0.00% Breeding 9218 11.87 0.00% Breeding 8117 17.35 0.00% Breeding 2752 30.17 0.20% Breeding 5873 24.43 0.01% Breeding 7440 19.83 0.00% Breeding 4385 27.99 0.04%

Release at client discretion - 24 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Breeding 2555 29.22 0.27% Breeding 8772 14.42 0.00% Breeding 7324 20.21 0.00% Breeding 8847 14.03 0.00% Breeding 8341 16.42 0.00% Breeding 9831 6.16 0.00% Breeding 8297 16.61 0.00% Breeding 4862 26.91 0.03% Breeding 9191 12.05 0.00% Breeding 4156 28.49 0.05% Breeding 5536 25.29 0.01% Breeding 5083 26.39 0.02% Breeding 9457 10.15 0.00% Breeding 1225 21.31 2.26% Breeding 248 10.73 5.47% Breeding 5956 24.21 0.01% N/A (assume all flights in Breeding 0 direction of powerline 50% potentially at risk) Breeding 6500 22.72 0.01% Flocking 8376 16.27 0.00% Flocking 8002 17.80 0.00% Flocking 9835 6.11 0.00% Flocking 5418 25.58 0.01% Flocking 8282 16.68 0.00% Flocking 9238 11.74 0.00% Flocking 4271 28.24 0.04% Flocking 826 17.99 4.50% Flocking 9116 12.51 0.00% Flocking 7979 17.89 0.00% Table 12: Per wetland results from powerline model with brolga height and flight distances collected during breeding ad flocking surveys at Dundonnell and throughout south-west Victoria

Release at client discretion - 25 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

5 Discussion & interpretation The long-term, average rates estimated above provided a quantitative estimate of the likelihood of collision (the consequence of collision was dealt with elsewhere, through population viability modelling).

However, in assessing the overall risk, and to provide practical guidance for the operational phase of the development (e.g. adaptive management triggers), it is instructive to consider how the average collision rate might manifest in a given year (or over the life of the farm).

This also provides insight into the practical difference between two predicted averages. Table 13 compares the potential collision after one, five and 10 years for the preferred layout (L08.0). The original layout (L03.11) is shown for comparison.

We also show more detailed, per year counts for the 95% avoidance scenario, with the preferred (L08.0) layout (Figure 8). The 95% scenario is chosen as a reasonable precautionary scenario, as most Band CRM research recommends this value or above (see section 3.3.5). This figure shows years of operation down the rows, and potential mortalities after that time along the columns. For example, after five years, we expect (to 95% confidence) somewhere between 0 – 6 total mortalities.

Layout L03.11 Layout L08.0 Scenario 90% 95% 98% 99% 90% 95% 98% 99% Long-term average # of 1.23 0.61 0.25 0.12 0.91 0.46 0.18 0.09 turbine collisions per year Estimated power line 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 collisions each year Combined average turbine and power line 1.27 0.65 0.29 0.16 0.95 0.5 0.22 0.13 collisions per year Potential mortality range (95% confidence interval): After 1 year 0-4 0-2 0-2 0-1 0-3 0-2 0-1 0-1 After 5 years 2-11 0-7 0-4 0-3 1-9 0-6 0-3 0-2 After 10 years 7-18 2-11 0-6 0-4 4-15 1-9 0-5 0-4 Table 13: Plausible (to 95% confidence) manifestations of the predicted annual collision rate. The proposed layout, with 95% avoidance is highlighted. Note that the avoidance rate applies only to the turbine collision risk estimate. For powerline collisions we have applied the avoidance rate of 1 collision per 10,000 crossings, as this was used in subsequent population viability modelling (M. McCarthy, pers comm.)

Release at client discretion - 26 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Figure 8: Range of possible brolga collisions per year for a combined (turbine + power line) average rate of 0.90. The green band is within the 95% confidence window, yellow are within the 99% confidence window and hashed are outside the 99% confidence bound.

Release at client discretion - 27 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

6 References & further reading

Band, W. 2001. Estimating collision risks of birds with wind turbines. SNH Research Advisory Note. Scottish Natural Heritage: Edinburgh.

Band, W., Madders, M., & Whitfield, D.P. 2007. Developing field and analytical methods to assess avian collision risk at wind farms. In: de Lucas, M., Janss, G.F.E. & Ferrer, M. (eds.) Birds and Wind Farms: Risk Assessment and Mitigation, pp. 259-275. Quercus Books: Madrid.

Band, B. 2012. Using a collision risk model to assess bird collision risks for offshore wind farms. SOSS report, The Crown Estate. Online report: http://www.bto.org/sites/default/files/u28/downloads/Projects/Final_Report_SOSS02_Band1M odelGuidance.pdf. Accessed 24/10/2013

Biosis. 2008. Collision risk modelling report, November 2008. In: Mortlake wind energy facility panel report – August 2010.

Burnham K., Anderson, D.,White, G., Brownie, C. and Pollock, K. 1987 Design and Analysis Methods for fish survival experiments based on release-recapture American Fisheries Society, Monograph No. 5 pp 212.

Cook, A., Johnston, A., Wright, L. and Burton, N. 2012. A review of flight heights and avoidance rates of birds in relation to offshore wind farms. Report of work carried out by the British Trust for Ornithology on behalf of the Crown Estate. Project SOSS-02. www.bto.org/sites/default/files/u28/downloads/Projects/Final_Report_SOSS02_BTOReview.pd f. Accessed 24/10/13.

Journey North. 2014. Whooping Crane: Flight formation, the V’s have it. http://www.learner.org/jnorth/tm/crane/FlightFormation.html. Viewed 4/3/14.

Ligue pour la Protection des Oiseaux (LPO). 2014. The common crane. http://champagne- ardenne.lpo.fr/English/e_grue_cendree.htm. Viewed 5/3/14.

Reason, J. 1997. Managing the Risks of Organisational Accidents. Ashgate Publishing Ltd: Aldershot England.

Buckland S., Anderson D., Burnham K., Laake J., Borchers D., & Thomas L. 2001. Introduction to Distance Sampling: Estimating abundance of biological populations. Oxford University Press: Oxford. Reprinted 2008.

Buckland S., Anderson D., Burnham K., Laake J., Borchers D., & Thomas L. 2007. Advanced Distance Sampling: Estimating Abundance of Biological populations. Oxford University Press: Oxford.

Madders, M. & Whitfield, D. 2006. Upland raptors and the assessment of wind farms. Ibis. 148(s1): 43-56.

Satterthwaite, F. 1946. An approximate distribution of estimates of variance components, Biometric Bulletin. 2:110-4

Silverman, B. 1986. Density Estimation for Statistics and Data Analysis. Monographs on Statistics and Applied Probability. Chapman & Hall: London.

Release at client discretion - 28 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

Scottish Natural Heritage . 2012. Guidance: Avoidance Rates for Wintering Species of Geese in Scoland at Onshore Windfarms. Online report www.snh.gov.uk/docs/A916616.pdf. Accessed 24/10/2013.

Scottish Natural Heritage. 2010. Guidance: Use of avoidance rates in the SNH wind farm collision risk model. Online report http://www.snh.gov.uk/docs/B721137.pdf. Accessed 24/10/2013.

Symbolix. 2014a. Brolga activity spatial analysis: Dundonnell wind farm and surrounds. Report to Brett Lane & Associates. Version 1.2. 14 January 2014.

Thomas, L., Laake, J.L., Rexstad, E., Strindberg, S., Marques, F.F.C., Buckland, S.T., Borchers, D.L., Anderson, D.R., Burnham, K.P., Burt, M.L., Hedley, S.L., Pollard, J.H., Bishop, J.R.B. and Marques, T.A. 2009. Distance 6.0. Release “x”1. Research Unit for Wildlife Population Assessment, University of St. Andrews, UK. http://www.ruwpa.st- and.ac.uk/distance/

Victorian Government. 2011. Interim guidelines for the assessment, avoidance, mitigation and offsetting of potential wind farm impacts on the Victorian brolga population. Published by Department of Sustainability & Environment, http://victorianbrolgastudy.com.au/brolga- wind-farm-guidelines/. Accessed July 2013.

Release at client discretion - 29 - BLABROL20131021, Version 2.3

Dundonnell Wind Facility residual risk analysis

About Symbolix

Symbolix specialise in industrial process modelling, forecasting and support. We use your existing knowledge to audit, improve and expose your current business systems, or to design new approaches. Symbolix consultants will work alongside you and your team, providing you with analysis in conjunction with communication and project management services, Symbolix can provide a complete solution package, or expert consultants to assist at any stage of your project development.

Find out more at www.symbolix.com.au.

1/14 Akuna Dr Williamstown Nth VIC 3016 Telephone: +61 3 9397 2520 [email protected]

Dundonnell Wind Farm – Brolga Assessment Report No. 9184 (4.12)

Appendix 6: PVA Model Report by Dr McCarthy

Page | 118

Results for Brolga PVA for Turbine Layout at Dundonnell

Michael McCarthy

School of Botany, The University of Melbourne

17 March 2014

This report documents a population viability analysis (PVA) of brolga to predict the impacts of a proposed wind farm at Dundonnell. The impact of the wind farm was predicted based on modelled extra mortality from collisions with wind turbines and powerlines. The extra mortality from turbine collisions for a specified turbine configuration (Layout L08 - 104WTG (final revised layout)) and powerline crossings were provided by Symbolix (Elizabeth Starks; report dated 15 March 2014).

The population viability analysis was based on an age-structured model, with individuals classified as being one year old birds (juveniles), two year-old birds (sub-adults) and mature birds (adults). Let J, S and A be the abundances in each of these age classes. The per capita fecundity rate (f) and survival rate (s) define the transitions among these age classes, which can be represented diagrammatically:

f

J s S s A

one-year-olds two-year-olds mature birds

The fecundity rate f is the product of the probability that an adult breeds, the average number of chicks produced, and the survival of any chicks to one year of age. The transitions between age classes can be defined by a matrix (M):

0 0 f    M = s 0 0 0 s s 

Estimation of the parameters s and f is problematic for brolga because mark-resighting data are not available. However, estimates can be derived from the observed ratio of immature (one-year-old and two-year-old birds) to mature birds and assumptions about the population trend with particular assumptions. If the population is stable (abundances of each of the age classes are the same from year to year):

School of Botany The University of Melbourne Victoria 3010 Australia

J = f A, S = s J, and A = s (S + A).

Solving these equations leads to:

f = (1 – s) / s2, and s = √(R + 1) / (R + 1),

where R equals the ratio of immature to mature birds ((J + S) / A). Therefore, if R = 0.05, which is approximately the case for contemporary populations in southeastern Australia (Herring 2001), s = 0.976 and f = 0.025, which are used as the standard set of parameter values. The estimate of s can be compared to predictions from an allometric model (McCarthy et al. 2008, but with additional data on cranes: Bennett and Bennett 1990, Link et al. 2003, Masatomi et al. 2007) that predicts the annual survival rate of adult birds from body mass. Based on a body mass of 6 kg, the predicted annual survival rate of cranes is 0.91 with a 95% credible interval of [0.77, 0.96]. Therefore, the estimate based on age structure is higher than might be expected for a crane of this size but not inconceivably so. Nevertheless, a survival rate less than 0.976 may be possible, and in fact may be likely.

The population growth rate based on the matrix model can be obtained by eigenanalysis of the transition matrix M, and is the (real) solution to the cubic equation:

λ3 – s λ2 – f s2 = 0.

A closed form solution can be obtained, but it is unhelpfully complicated (result not shown). However, the solution can be approximated using a first-order Taylor series expansion around the point f = 0, leading to λ ≈ s + f. The next term in the expansion is –f2/s, which is small when f is small and s is large. Therefore, λ ≈ s + f is a good approximation if f ≈ 0 and s ≈ 1, which is the case for the brolga. This means that reductions in the population growth rate due to decreased survival of brolgas can be approximately compensated by an increase in fecundity of the same magnitude.

Cranes are one of the most common bird taxa that collide with powerlines (Morkill and Anderson 1991, Bevanger 1998, Janss and Ferrer 2000 and references therein). Specific instances of brolgas being killed by powerlines are known (Goldstraw and Du Guesclin 1991), but the rate of mortality is unknown. Information on likely impacts can be obtained from other cranes in other parts of the world. Janss and Ferrer (2000) estimated 5×10–5 deaths for each crossing of a 400 kV powerline by common cranes in southwest Spain. For sandhill cranes in the Nebraska, USA, the estimated mortality rate was 1×10–4 per crossing for high-voltage (69 - 345 kV) powerlines (Morkill and Anderson 1991). Therefore, the chance of a crane dying when flying over a powerline was assumed to be one in ten thousand.

This assumption was used to convert the number of crossings to an annual number of deaths, which was added to the predicted number of brolgas colliding with wind turbines (Table below). The number of deaths was converted to an annual probability by dividing the number by the assumed size of the southwest Victoria brolga population (907; DSE report of simultaneous flock counts in April 2013).

School of Botany The University of Melbourne Victoria 3010 Australia

Initial abundances in the simulation were set at 79 1 year-olds, 75 2 year olds, and 753 adults, reflecting the 2013 simultaneous flock counts. The numerous younger birds (17% of the population) suggests that annual recruitment rate in the last two years has been approximately 10% in the last couple of years following high rainfall. Assuming annual survival of 0.976, and a reproduction rate of 2.5% on average to obtain a stable population size, this suggests large variation in reproduction. Previous versions of the model had assumed a comparatively modest coefficient of variation (CV), but these newer data suggest much greater variation – a value of 100% for the CV seems plausible. This value was chosen for the CV for fecundity, and 50% was chosen for the CV of mortality (which translates to small variation in survival).

Simulations were based on 10,000 stochastic iterations for each set of parameters within RAMAS/GIS (Akçakaya & Root 2002).

Results

In the absence of losses to wind turbines and powerlines, the probability of falling to or below 800 brolga within the next 25 years based on these parameters is 0.43, while over 50 years the probability is 0.60. Probabilities of decline to or below lower thresholds (650, 700, 750) are lower (Table 1). The corresponding expected minimum population sizes were 808.6 and 761.7. Impacts of the turbines and powerlines can be compared to these values. Results are shown for risks over 25 years, and for 50 years both when the windfarm impacts are present for 25 years and 50 years. Small apparent decreases in the risk in the presence of the wind farm (Table 1) are due to random variation arising from the stochastic simulations.

The expected minimum population size (EMP) is a useful metric of the risk of decline. It is calculated from stochastic simulations of the model. The smallest population size in each of the simulations is recorded. The EMP is the average of these, representing the average degree by which a quasi-extinction risk curve is from extinction (McCarthy 1996; McCarthy and Thompson 2001). The standard deviation of the minimum population size is not as easy to interpret as a measure of risk, but it is reported in the table below because it was requested.

The linear approximation of the population growth rate (λ = f + s) indicates the number of births that would required to offset mortality events from collisions. Because fecundity and survival have approximately additive effects on growth rate, each mortality event would need to be mitigated by an extra bird being raised to adulthood. This might be achieved by improvement to breeding habitats or reduced collision with other infrastructure such as existing powerlines or fences (Beaulaurier 1981; Alonso et al. 1994; Brown & Drewien 1995). For example, with an average of 0.25 extra brolga deaths per year with a 98% avoidance rate, this translates to 6.25 extra deaths over 25 years. Therefore, mitigating these extra deaths by increasing breeding success would require an extra approximately 6.25 birds being raised to adulthood over the 25 year period. If survival of juveniles to adulthood is j, then the required number of extra juveniles can be calculated as 6.25/j. Assuming survival from 1 year of age to adulthood of 0.976, this equates to 6.4 one year old birds.

School of Botany The University of Melbourne Victoria 3010 Australia

Table 1. Results of simulations of the south-west Victorian brolga population without turbines, and for a specified turbine layout and three different turbine avoidance rates for each.

Layout L06.2 No turbines 104WTG Avoidance rate NA 90% 95% 98% 99% Modelled annual mortality 0 0.91 0.46 0.18 0.09 events (turbines) Modelled annual mortality 0 0.04 0.04 0.04 0.04 events (powerlines) Total added annual mortality 0 0.95 0.50 0.22 0.13 events Total added annual mortality 0 0.105 0.055 0.024 0.014 rate per 100 brolgas

25 year results EMP 808.6 795.3 804.9 807.5 807.9 SDMP 77.0 80.0 76.6 77.1 76.4 QE Prob (650) 0.03 0.05 0.03 0.03 0.03 QE Prob (700) 0.10 0.13 0.10 0.10 0.10 QE Prob (750) 0.23 0.28 0.24 0.23 0.23 QE Prob (800) 0.43 0.49 0.43 0.42 0.42 Reduction in EMP - 13.3 3.8 1.1 0.7

50 year results; 25 yrs of impacts EMP 761.7 747.3 755.2 761.5 763.2 SDMP 103.5 105.2 101.2 103.2 102.2 QE Prob (650) 0.15 0.19 0.16 0.16 0.15 QE Prob (700) 0.27 0.32 0.30 0.28 0.27 QE Prob (750) 0.43 0.48 0.45 0.43 0.42 QE Prob (800) 0.60 0.66 0.63 0.59 0.59 Reduction in EMP - 14.4 6.5 0.1 -1.5

50 year results EMP 761.7 744.4 761.2 761.6 761.6 SDMP 103.5 105.5 101.7 102.8 102.0 QE Prob (650) 0.15 0.20 0.15 0.16 0.16 QE Prob (700) 0.27 0.33 0.27 0.27 0.27 QE Prob (750) 0.43 0.49 0.42 0.42 0.42 QE Prob (800) 0.60 0.66 0.61 0.60 0.60 Reduction in EMP - 17.3 0.5 0.1 0.1

School of Botany The University of Melbourne Victoria 3010 Australia

References

Akçakaya HR, Root W 2002. RAMAS Metapop: Viability analysis for stage-structured metapopulations (version 4.0). Applied Biomathematics, Setauket, New York.

Alonso JC, Alonso JA, Muñoz-Pulido R 1994. Mitigation of bird collisions with transmission lines through groundwire marking. Biological Conservation 67:129-134.

Beaulaurier DL 1981. Mitigation of bird collisions with transmission lines. Bonneville Power Administration, Portland, Oregon, USA.

Bennett AJ, Laurel A Bennett 1990. Survival Rates and Mortality Factors of Florida Sandhill Cranes in Georgia. North American Bird Bander 15: 85-88.

Bevanger K 1998. Biological and conservation aspects of bird mortality caused by electricity power lines: a review. Biological Conservation 86:67-76.

Brown WM, Drewien RC 1995 Evaluation of two power line markers to reduce crane and waterfowl collision mortality. Wildlife Society Bulletin 23:217-227.

Janss GFE, Ferrer M 2000. Common crane and great bustard collision with power lines: collision rate and risk exposure. Wildlife Society Bulletin 28: 675-680.

Herring MW 2001. The Brolga (Grus rubicunda) in the New South Wales and Victorian Riverina: Distribution, Breeding Habitat and Potential Role as an Umbrella Species. Honours Thesis. Charles Sturt University, Albury.

Link WA, Royle JA, Hatfield JS 2003. Demographic analysis from summaries of an age-structure population. Biometrics 59: 778-785.

Masatomi Y, Higashi S, Masatomi H. 2007. A simple population viability analysis of Tancho (Grus japonensis) in southeastern Hokkaido, Japan. Population Ecology 49:297-304.

McCarthy, M.A. (1996). Red kangaroo (Macropus rufus) dynamics: effects of rainfall, harvesting, density dependence and environmental stochasticity. Journal of Applied Ecology 33: 45-53.

McCarthy, M.A., and Thompson, C. (2001). Expected minimum population size as a measure of threat. Animal Conservation 4:351-355.

McCarthy, M.A., Citroen, R., and McCall, S.C. (2008). Allometric scaling and Bayesian priors for annual survival of birds and mammals. American Naturalist 172: 216-222.

Morkill AE, Anderson SH 1991. Effectiveness of marking powerlines to reduce sandhill crane collisions. Wildlife Society Bulletin 19:442-449.

School of Botany The University of Melbourne Victoria 3010 Australia

Glossary of acronyms

CV coefficient of variation, equal to the standard deviation divided by the mean.

EMP expected minimum population size; a measure of risk of decline of a population.

PVA population viability analysis; a model-based analysis of the risk of decline of populations.

QE Prob quasi-extinction probability; the probability of decline to or below a particular threshold population size within a specified time period.

SDMP standard deviation of the minimum population size; a measure of the variation in the minimum population size.

School of Botany The University of Melbourne Victoria 3010 Australia