Cook PG, Simmons CT and Wallis I Martin R (2019). Bird in

Golder Associates Pty Ltd

118 Franklin Street Adelaide, South Australia 5000 Australia

T: +61 8 8213 2100 +61 8 8213 2101

A.B.N. 64 006 107 857

Golder and the G logo are trademarks of Golder Associates Corporation golder.com

11 May 2020 1897291-005-RevC

Tom Mehrtens

Senior Environment and Community Officer

Unit 7/202-208 Glen Osmond Road

Fullarton SA 5063

BIRD IN HAND GROUNDWATER ASSESSMENT SUBMISSION RESPONSES

Dear Tom,

We have reviewed the following submissions and we have provided responses to their comments in Table 1.

Cook PG, Simmons CT and Wallis I (2019a) Bird-in-Hand gold mine. Review of groundwater

characterisation and modelling. NCGRT Report to Piper Alderman and Accolade wines. 1 February

2019.

Cook PG, Wallis I and Simmons CT (2019b) Bird-in-Hand gold mine. Review of managed aquifer

recharge trials and model validation. NCGRT Report to Piper Alderman and Accolade wines. 26 June

2019.

Martin R (2019). Bird in Hand Proposed Mine Managed Aquifer Recharge Review. Martin (2019) Report

to Inverbrackie Creek Catchment Group. 30 August 2019.

Table 1: Responses to Martin (2019) and Cook et al. (2019a; 2019b) review documents.

No. Comment Response

Responses to Martin (2019)

1 The WAP for the Western Mount Lofty Ranges PWA sets out several Principles for the recharge of water to the aquifers using MAR approaches. The information presented in the Bird in Hand Gold Project Mining Lease Application (MC4473) is deficient with respect to the MAR investigations and the principles set out in the WAP.

The MLA was not written to address every principle of the WAP in detail at this stage. The work presented in the MLA is sufficient to demonstrate that the proposed MAR can be designed and operated to comply with the WAP

2 The conversion from EC to TDS in mg/L uses 0.65 compared to 0.55 recommended by the DEW which skews the results towards a higher Environmental Value

Martin (2019) points out that the use of different conversion factors to derive TDS concentrations from recorded EC levels, could lead to a different

Tom Mehrtens 1897291-005-RevC

Senior Environment and Community Officer 11 May 2020

2


category. The correct category under the EPA EP(WQ)P should be potable and therefore the injected water must meet this criterion prior to recharge as per Principle 145(a)(i).

classification of the aquifer waters under DEW’s Environmental Value criteria. We understand this and as such we did not use conversions to estimate the TDS concentrations. TDS interrogated by Martin (2019) in Table 4 of Golder (2019a) report are in fact laboratory results as indicated by the table heading. We have recorded TDS in the field with appropriate equipment and have TDS concentrations determined by NATA accredited laboratory using two methods (EC conversion and measured at 180 degrees). Comparison of laboratory results and field results are presented in the appendices of the Golder (2019a) and Golder (2019b). Based on this clarification of the method of deriving TDS, we stand by our classification of the Environmental Value for groundwater and respectfully reject Martin’s (2019) alternate classification of a higher Environmental Value.

3 A detailed risk assessment as set out under Principle 146 of the WAP has not been completed. Some risks have been identified but what has been presented is not consistent with the National Water Quality Management Strategy Australian Guidelines for Water Recycling: Managing Health and Environmental Risks (Phase 1).

The MLA is not an application for MAR under section 128 nor is it an application for drain and discharge under the WAP. We consider the risk assessment conducted is commensurate for the stage of the proposed project.

4 An appropriate operation or management plan demonstrating that operational procedures are in place to protect the integrity of the aquifer on an ongoing basis (Principle 147(g) of the WAP) has not been prepared.

We acknowledge the importance of operation and management plans, and note that these will be developed as part of the stage two approvals (PEPR). This will include mitigation measures where needed. The modelling undertaken to date will be used to inform triggers for the monitoring and management plan.

5 There has been no analysis of the water quality data collected prior to and during the trial to support the MAR option. Analysis of the water quality data shows there is a high risk of geochemical reactions - arsenic which will impact water quality and iron precipitation which will cause clogging

The data was collected, but not reported because it was not needed for the MAR trial that was conducted. The water quality data will be considered and incorporated in the design of the management and operating system for the MAR along with the data from other injection wells that will be drilled at a later stage. Similar to the point above the work done is appropriate for the stage of the project.

6 Key water quality parameters important to assessing the risks that MAR operations pose to water quality have not been collected. These include, natural organic matter, total organic carbon, dissolved oxygen, redox potential and microbiological indicators (e.g. E.coli).

As stated in our reports DO and ORP (Oxidation Reduction Potential) of native groundwater and recharge water were recorded continuously during pumping tests (stage 1 investigation) and during injection trials (stage 2 investigations) using a Horiba U52 multi probe water quality meter, which was calibrated daily. The full range of field parameters monitored using the Horiba U52 multi probe water



3


quality meter included pH, Dissolved Oxygen, Conductivity, Salinity, TDS, specific gravity, Temperature, ORP and Turbidity. We note that key hydrogeochemical processes can be derived with a range of other parameters that were tested.

7 The lack of analysis of the chemistry and the potential for geochemical reactions to occur that could compromise the existing groundwater quality is a significant knowledge gap in the investigations undertaken to date. The chemistry of the source and receiving waters are a fundamental control on the success or otherwise of a MAR system.

The groundwater studies have provided a good understanding of native groundwater and source water quality. Matters relating to possible hydrogeochemical reactions when these are mixed will be addressed once the specific MAR wells are installed and the specific water chemistry is known. The modelling and analysis are sufficient to satisfy to requirements of the MD. As a result, we respectfully disagree that there is any lack of analysis or knowledge gap. The information set out in the MLA is fit for the level of study.

8 The presence of pyrite in the rock matrix increases the risk of producing arsenic due to geochemical reactions when the oxygenated surface water, that is not in chemical equilibrium due to mixing and treatment processes, contacts the groundwater. Under Principle 151 a permit to drain or discharge water into a well must not be granted if the draining or discharging of water would have the potential to degrade underground water- dependent ecosystems or to reduce the suitability of the underground water for other purposes for which it might reasonably be used.

Martin (2019) expresses concern about arsenic release from pyrite oxidation. Martin (2019) appears to assume the aquifers are ‘anoxic’ and that contact with oxygen in the air would cause chemical changes and arsenic release. However, Martin (2019) does not explain why they consider the aquifer anoxic or otherwise, how arsenic would be released. The notable presence of nitrate and sulphate suggest the aquifer may not be anoxic and a pathway for arsenic release may not exist at all. There’s clear evidence at Bird-in-Hand that oxygenated waters have already flowed through fractures at great depth as evidenced by the oxidization of sulphides in and around fractures. The Bird-in-Hand oxidized fractures that host aquifers have been seen in drill core beneath the deepest planned stope. The deepest open fractures seen at Bird-in-Hand are in BH044 and shows at 428m that a water bearing fracture has already been oxidized. Evidence of oxidization are common in deep holes at +300m depths. RDH002 which was drilled at Ridge clearly shows oxidization down at 130m and in a photo all the way down to 159.6m EOH.

9 The aquifer system is identified to be compartmentalised therefore the claimed benefits from managed aquifer recharge at this location are not valid.

Martin (2019) states that the fact the aquifers are compartmentalized would increase the risk of MAR being sustainable. We disagree with this assertion:

At Bird in Hand we deploy a recirculation concept. Testing and modelling shows all dewatering water can physically be reinjected. The strength of this concept is that if aquifer hydraulic conductivities were low, there would be little inflow into the mine shafts, hence only small volume to be reinjected. Higher



4


conductivities would result in higher injection rates that can be achieved.

The compartmentalization limits the extent of any changes to the boundaries of the ‘compartment’. We see this as a benefit and another key strength of the concept.

10 Existing users will have no access to the recharged water.

The MAR system is designed to maintain groundwater level pressure (the pressure response spreads further than the recharge water). There is no requirement for existing licenced wells to access the recharge water.

11 Reinjection will not result in lateral spreading of the recharged water because flow will occur preferentially along the fractures. For the water to spread laterally the pore throat entry pressure of the aquifer matrix rock need to be overcome and the pressure required to do this in a metasediment, such as the Tapley Hill Formation, would be considerable and most likely exceed the safe operating pressure presented in Golder (2019).

The hydraulic response observed in monitoring wells during injection into BHRIB01 confirmed lateral spreading along the fault zone (preferential pathway) does occur, with pressure responses observed in monitoring wells outside this fault zone. The injection trial showed exceedance of safe operating pressure during mining operation will be highly unlikely, as high rates of recharge (13 L/s) equivalent to ~3 times the expected mine inflows was easily achievable in one injection well alone and the safe injection pressure was not reached. The proposal will involve the distribution of treated mine water across 8 injection wells, and therefore each well will be injecting at much lower injection rates and operating at lower pressure than those observed during the trial. Furthermore, groundwater modelling shows groundwater rise (pressure) owing to injection in the Tapley Hill (at mine year 5) was 1 m (90% grouting scenario) and 1 to 5 m (70% grouting scenario) and therefore there is no foreseeable risk of excessive aquifer pressurisation (see Figures 22 and 23 in Golder 2019).

12 The potential for MAR to support discharge to environmental receptors such as the spring and creek is limited.

Hydraulic modelling shows baseflow to Inverbrackie Creek is maintained at pre-mining rates. Baseflows are maintained by hydraulic response rather than the actual recharge water reaching the Inverbrackie Creek (see Figure 24 in Golder 2019). Maintaining base flows is a key objective of the MAR project and a fundamental principle of the design philosophy.

13 Overpressurisation of the aquifer compartments due to the MAR recharge may be the trigger that causes an unexpected inrush into the workings.

Injection into the Tarcowie Siltstone will take place outside of groundwater flow barriers, which limits recirculation of injected water back towards the mine. Over pressurisation of the well and aquifer was not observed during the trial and not predicted by modelling.



5


14 Overpressurisation of the aquifer compartments due to the MAR recharge has the potential to significantly reduce the grouting effectiveness.

The numerical modelling and MAR trial show the recirculation of injected water back to the mine is low and therefore interference with the grouting process is not expected (injection will be undertaken outside of groundwater flow barriers).

15 Based on a review of the available information including the results of the MAR injection trial at the Bird in Hand site it is concluded that MAR targeting the Tarcowie Siltstone aquifer is high-risk and not sustainable in the long-term due to compartmentalisation and the risk of the aquifer becoming artesian.

See above comments 9, 13 and 14

16 The proposed treatment process, coupled with the open ponds presents, several risks to the receiving groundwater and aquifer:

The management of the pond will be addressed as part of the PEPR which will outline operational controls and barriers to prevent potential groundwater impacts.

17 The open pond will be subject to contamination from surface runoff, mine affected runoff and waterfowl increasing the risk of pathogens. There is no treatment proposed to manage pathogen levels in the recharge water.

As mentioned previously the PEPR will outline operational controls and barriers to prevent contamination from runoff or waterfowl increasing risk of pathogens. Furthermore, aquifers behave as natural treatment barriers to pathogens entering groundwater systems through a process known as attenuation. There is significant pathogen attenuation in aquifers explained as follows. Pathogen survival in groundwater is affected by physical, chemical, and biological processes (Toze and Hanna, 2002; Gordon and Toze, 2003). The inactivation of pathogens in aquifers highlights their potential use as robust treatment barriers in the multi-barrier approach for management of risks to human health. Research has documented that residence time in the aquifer of up to 10 days is sufficient to achieve greater than 3 log10 removal of some pathogens and the national guidelines for stormwater harvesting and reuse recognise the importance of the aquifer as an effective barrier and preventative measure. Pathogen studies carried out by Toze (2005) by suspending specially modified chambers down a drill hole completed in the T2 aquifer as part of the Bolivar reclaimed water ASR trial indicated that one log10 removal (die-off) for all pathogens tested was achieved within 20 days. International studies (Lundh, 2009; Liesel, 2009) on pathogen removal have indicated that three log10 removals can be achieved within the first 4 m of aquifer matrix around the recharge bore. This equates to a residence time of 10 to 20 days within the aquifer to achieve 3 log10 removal, however this removal is dependent on the raw water pathogen concentration.



6


18 The ion exchange process acts to break the bond between targeted ions which in turn causes a pseudo reduction in the measured salinity but results in a water composition that is no longer in chemical equilibrium. This significantly increases the risk of geochemical reactions occurring in the aquifer that would impair the water quality.

The water recharged in the FRA is sourced from the same FRA, and as such the mineral composition of the injection water would be similar. Ion exchange processes would not significantly change the groundwater composition under such circumstances.

19 Ponding the water at surface will change the redox state and therefore increases the risk of geochemical reactions (such as arsenic) due to the presence of PAF minerals in the host rock.

Refer to response 8.

20 The use of chlorinated mains water will introduce disinfection by-products into the aquifer (e.g. Trihalomethane (THM) and Haloacetic Acid (HAA)) which will impact on water quality. In other MAR operations where mains water is used it is typically recovered to meet demands. In the Bird in hand operations there is no planned recovery of the injected water increasing the risk of cumulative concentration over time of the disinfection by-products.

We agree chlorine in mains water, if this were to be used to augment flows, could release disinfection by-products if organic matter is present in the aquifer. There is no indication that this is the case, but there are relatively straightforward solutions to remove chlorine before reinjection, such as a carbon filter. Martin (2019) did not suggest solutions.

21 The proposed targets for meeting suspended solids result in 3.4 kg/day of sediment being recharged into the bore. This level of suspended solids would result in clogging in the injection well within a matter of days.

It is unclear how Martin has derived the figure of 3.4kg/day. However, to address the issues of clogging and TSS quantities in reinjection water the following points are made: - The water treatment study at appendix J1 of the

MLA states that the treated water will have TSS <10 mg/l, which is in line with the Australian Water Guidelines.

- Target TSS and actual TSS limits will be stipulated in the drainage and discharge permits.

- Chapter 10 of the MLA outlines the management strategy for monitoring and managing clogging of wells.

- Turbidity (which is an indicator of TSS) removal will be undertaken through inline filtration as part of the water treatment plant as outlined in Appendix J1 of the MLA.

22 There are several references throughout the Terramin documentation that the MAR option has been peer reviewed. This statement is misleading as IGS only reviewed the updated groundwater numerical model that incorporated the MAR option. IGS did not review or comment on the results of the trials, water chemistry, treatment process, aquifer mineralogy, or

The peer review undertaken by IGS meets the requirements of section 2.6.1 of the Ministerial determination for the minimum requirements for a Mining lease application.



7


the aquifer hydraulic responses during testing.

23 Key water quality parameters e.g. redox potential, EH, pH and salinity were only measured in grab samples as part of the MAR injection trial and not continuously in-line throughout the trial. These parameters are fundamental measurements to assess the risk of geochemical reactions occurring that will impact the water quality in the aquifer. Failure to measure these key parameters in-line throughout the trial means that the opportunity to gain valuable information about the aquifer geochemical responses for a very low cost has been missed.

Martin (2019) sets out a list of possible water quality and well clogging effects that could be associated with MAR in general. We consider the following warrants our response: Martin (2019) states that DO and Eh are ‘key parameters’ that were not recorded. As stated in our reports DO and ORP (Oxidation Reduction Potential) of native groundwater and recharge water were recorded continuously (every 1 to 2 hrs) during pumping tests (stage 1 investigation) and during injection trials (stage 2 investigations) using the Horiba U52 multi probe water quality meter, which was calibrated daily (records provided in Appendix B of Golder 2019). This was also witnessed by DEM appointed inspectors. The full range of field parameters monitored using the Horiba U52 multi probe water quality meter including pH, Dissolved Oxygen, Conductivity, Salinity, TDS, specific gravity, Temperature, ORP and Turbidity (results collected during the injection trial are provided in Appendix B of Golder 2019 and results of pumping test are provided in Appendix D of Golder (2019a)). We note that key hydrogeochemical processes can be derived with a range of other parameters that were tested.

Martin (2019) does not make clear why continuous in-line recordings of redox potential and pH would have added value in addition to the conventional sampling and testing that has been undertaken at 1-2 hour intervals during the day. We note EC and other water quality parameters (outlined above) has been recorded continuously. We consider the parameters that have been tested suffice to inform an in-depth hydrogeochemical interpretation.

Martin (2019) expresses concern about arsenic release from pyrite oxidation. Martin (2019) appears to assume the aquifers are ‘anoxic’ and that contact with oxygen in the air would cause chemical changes and arsenic release. However, Martin (2019) does not explain why they consider the aquifer anoxic or otherwise, how arsenic would be released. The notable presence of nitrate and sulphate suggest the aquifer may not be anoxic and a pathway for arsenic release may not exist at all. There’s clear evidence at Bird-in-Hand that oxygenated waters have already flowed through fractures at great depth as evidenced by the oxidization of sulphides in and around fractures. The Bird-in-Hand oxidized fractures that host aquifers have been seen in drill core beneath the deepest planned stope. The deepest open fractures seen at Bird-in-Hand are in BH044 shows at 428m in BH044 that a water bearing fracture has



8


already been oxidized. Other deep holes that show this is a common occurrence at +300m. RDH002 which was drilled at Ridge that clearly shows oxidization down at 130m and all the way down to 159.6m EOH.

We agree chlorine in mains water, if this were to be used to augment flows, could release disinfection by-products if organic matter is present in the aquifer. There is no indication that this is the case, but there are relatively straightforward solution to remove chlorine before reinjection, such as a carbon filter. Martin (2019) did not suggest solutions.

24 Key water quality parameters (natural organic matter, total organic carbon, dissolved oxygen, redox potential and microbiological indicators) that can trigger geochemical or microbiological reactions have not been collected during background sampling and throughout the trial. This is basic information which should be collected on the source and receiving waters and throughout the trial to evaluate the risk of geochemical reactions occurring that would impair groundwater quality, impact ability of existing users to continue to use the water for current uses or impair the operation of the MAR option.

Martin (2019) claim that key parameters were not collected, and we direct them to response 23. There is no source of organic carbon due to the metamorphic grade of the host rock. Any carbon would be in the form of graphite.

25 To overcome the pore-throat entry pressure required to induce flow into the bulk rock matrix of the meta-sediments that comprise the fractured rock aquifer significant injection pressures would be required for this to occur. Such injection pressures would exceed the calculated safe operating pressure reported in Golder (2019).

We do not agree with this statement and it is not supported by the MAR trial results. Injection trials demonstrated large rates of recharge are possible (in a single well) and that injection into a single well could be undertaken well below the safe operating pressure. A further 6 injection wells are planned for the mining operation, and therefore recharge will be distributed across 8 wells. Each injection well will therefore operate at lower rates and lower pressures. The trial demonstrated preferential hydraulic response along strike with storage in less fractured adjacent rock as indicated by the observed hydraulic response in wells across strike (see section 3.6 / Table 5 in Golder 2019). This response which was observed under very low rates of pressure (as outlined under section 3.6.2 in Golder 2019).

26 Terramin Ch. 10 Groundwater (2019) identifies that the combined injection rate to the Tapley Hill Formation and the Tarcowie siltstone was 20 L/s. This statement is misleading as the trials were conducted independently of one another, not concurrently. It is therefore incorrect to

The injection wells (BHRIB01 and BHRIB02) target different hydro stratigraphic units (Tapley Hill Fm- footwall and Tarcowie siltstone – hanging wall) and the wells are not hydraulically connected and are located ~ 800 m apart. As such concurrent injection testing was not required. The injection trial was designed with this in mind and justification is provided



9


conclude that the two aquifer systems can be recharged simultaneously at 20 L/s as this is not supported by the trial results presented in Golder (2019).

in section 3.1 and 3.3 in Golder 2019. The benefit of our approach meant that the groundwater rise in each of the receiving aquifer formations and groundwater decline in each of the pumping formations (Tapley Hill Fm, followed by Tarcowie Siltstone) could be assessed separately.

27 During the inject trial hydraulic responses in adjacent monitoring wells resulted in rises in the groundwater level by 9 m showing a high degree of hydraulic connectivity (e.g. Davis well 6628- 23182). In a confined aquifer the pressure response (rising water level) can be observed up to several kilometres from the point of injection but the physical movement of water is constrained. In a fractured rock aquifer, the physical movement of water is preferentially along the fractures and can therefore move considerable distances (kilometres) away from the recharge well.

The objective of the injection was to generate hydraulic response in outer monitoring wells in order to map the hydraulic extent and strengthen model calibration. We agree that the fault acts as a preferential flow path which promotes the lateral spreading of water, however hydraulic responses were also observed outside of this zone which shows storage in adjacent less fractured rock (see comment 25).

28 There has been insufficient analysis or modelling carried out on the potential movement of the injected water preferentially along the fractures during MAR at this site. This is a significant gap in the assessment of potential risks to existing users and impacts on water quality.

The analysis and modelling carried out in relation to potential preferential flows along fractures is sufficient for the level of study required for the MLA as specified in the MD. In addition, existing users of groundwater are protected from any impacts as the recharge water quality will need to comply with Principal 145a of the Water Allocation Plan for the Western Mounty Lofty Ranges.

29 The aquifer discharge tests and injection trial identify structural controls and anisotropy within the aquifer (AGT, 2017 and Golder 2019). There is no rational presented as to why injection was then subsequently modelled to occur radially around the mine void. This is a fractured rock aquifer with evidence showing clear direction for fracture orientation, strike and dip (AGT, 2017). It would have been more appropriate to site any additional bores along fracture orientation as per two of the wells that target the Tapley Hill Formation.

The proposed locations of injection wells were informed based on the conceptual groundwater model. In the model, some wells were positioned outside of assessed groundwater flow barriers in the Tarcowie siltstone (to minimise recirculation back towards the mine) whilst other wells are located within fracture zones in the Tapley hill formation to promote the hydraulic influence.

30 Solute transport modelling (AGT 2017) was completed to assess the risk of mine inflows drawing in the surrounding saline groundwater, however, no solute transport modelling was undertaken to identify the extent of the injection envelope during MAR operations. Flow moves preferentially through the fractures and in the event of a contaminant entering the aquifer it is critical to understand what risk this may present to

See comment 28 and note groundwater at the mine area is 911 mg/L to 985 mg/L (TDS measured @180 deg) and therefore the simulation proposed by Martin (2019) where 1400 mg/L is applied to recharge is an unrealistic scenario and would not reflect the real groundwater system.



10


the nearest users. The injection trial showed hydraulic influences on nearby users wells of up to 9 metres (e.g. Davis bore 6628-23182 reported in Golder, 2019). This clearly identifies a high degree of hydraulic connection between the injection well and users’ wells. It therefore represents a significant risk to the water quality available to existing users in the event a contaminant enters the aquifer or adverse geochemical reactions occur resulting in elevated concentrations of arsenic. Best practice when evaluating the impacts of a MAR system on existing groundwater users typically include modelling simulations that predict the movement of the injection envelop. It is uncertain why this level of analysis was not undertaken but, in this case, where no recovery is planned, understanding how and where the water will move is critical. A predictive simulation could have been undertaken using the groundwater salinity from the pumped well (~1400 mg/L) against the regional background salinity of ~600 mg/L. Whilst Terramin identify that they propose to recharge water of similar salinity into the aquifer following treatment the model could have been used to test the movement of the injected water. Testing such a scenario would instil greater confidence in the reliability of the model to predict potential impacts on existing users changes to water quality.

31 The injection heads in the groundwater model underpredict the actual heads due to the influences of the model grid cell size. No attempt has been made to cross check what the actual heads in the injection wells will be using analytical approaches.

Appendix E in Golder (2019) demonstrates the model under-predicted the heads (by about 2 m) in only 3 out of the 31 monitoring wells used in the calibration and this is not related to grid cell size. Injection wells were excluded from calibration owing to well losses which cannot be accounted for in the numerical model. It is standard practice to only utilise monitoring wells for the calibration process.



11


The groundwater level change in 31 wells, including private wells (see Figure 7 in Golder 2019), was recorded during pumping and injection. These wells target a range of hydro-stratigraphic units, with a range of depths spanning approximately 19 m to 470 m bgl. This data was used for comparison with simulated groundwater response for calibration purposes. Appendix E presents the comparison between simulated and observed drawdowns in response to pumping and injection. Inspection of hydrographs in Appendix E demonstrates that the model generally reproduces the pumping and injection induced fluctuations well in the majority of monitoring wells. In particular the responsiveness to pumping and injection and the magnitudes of fluctuation are reasonably well represented. BHMB02A showed small mismatch between simulated and observed heads, however this is considered acceptable quality of fit to the nearest observation well (BHMB02) and farthest observation well (Day Windmill) were prioritised, in order to ensure that the overall extent of the zone of influence and the magnitude of groundwater level change in close proximity to the injection well were adequately represented within the model.

Responses to Cook et al. (2019a; 2019b)

1 The high injection rates which were achieved, also suggest, however, that aquifer permeabilities can be high and the question arises whether this indicates that higher than anticipated inflow rates into the mine are a genuine possibility and would have to be considered. In fact, BHRIB01, penetrating the Tapley Hill Formation resulted in airlift yields of about 50 L/sec, while BHRIB02, penetrating the Tarcowie Siltstone yielded around 30 – 40 L/sec (Figure 5 of AGT, 2017). This is of the same order of magnitude as a “significant fracture zone in the hanging wall, in the Tarcowie Siltstone. When intercepted by investigation wells, the fracture yielded up to 40 L/s. The fracture zone has transmissivity of 67 m2/day (geometric mean) and is the most significant mode of groundwater flow into the mine if intercepted by mine workings. The layout of the proposed mine has been designed to avoid the main water bearing structures to minimise groundwater inflows and groundwater impacts. Where development occurs near the water bearing structures, probe drilling and grouting ahead of development will be undertaken to

Airlift yields observed in wells during drilling cannot be linked to the mine inflows predicted by the groundwater model. Unlike the pumping tests, airlift yields observed during drilling were assessed over short durations of hours rather than days. For example, site well BHRIB02 exhibited an airlift yield of up to 30 L/s during drilling but sustained 6 L/s during the 3-day pumping test. Furthermore, it is not possible to draw a direct comparison between airlift yields from different drilling campaigns, without considering variations in submergence (pumping depth), well diameter (6 inch hole versus 8 inch hole), and compressor delivery (see Groundwater and Wells, third edition). Mine inflows are calculated by the 3D numerical model based on aquifer permeability, boundary conditions, hydraulic head and storage, incorporating inflows from a wider area over an extended period of time. Aquifer parameters representing a wider area are best evaluated by an appropriate duration pumping test with observations made from monitoring wells. Aquifer parameters that represent an area-wide groundwater system cannot be derived from drilling airlift yield of a single well. Therefore, it is not recommended to base mine inflow estimations solely based on airlifting test results. Five pumping tests



12


control groundwater inflow.” (AGT, 2017, p.iv).

were conducted around the proposed mine (footwall, lode, hanging wall) and 31 wells were monitored which yielded a range of hydraulic conductivities for faulted (such as the hanging wall fault) and non-faulted areas (e.g., in the vicinity of the decline). The airlift yield of 30-40L/s referred to by Cook et al. (2019b) was associated with the hanging wall fault above the mine. The mine plan has been designed to avoid this high permeability fault. In contrast the Tapley Hill Formation, which will host the mine decline, is characterised by lower yields and permeability.

2 The question therefore arises, whether the intercepted fracture zone is not an isolated occurrence, as suggested in AGT (2017) but rather an expression of the generally high permeability of parts of the Tapley Hill Formation and Tarcowie Siltstone at the site. It is unclear, if the injection bores, which achieved high airlift yields penetrated distinct fractures, or if they are representative of the overall geological make-up of the formations. On page 8 of Report 2, it mentions that “the injection wells …. targeted geological structures within the Tapley Hill formation and Tarcowie Siltstone, as these are commonly characterised by a higher degree of secondary porosity and therefore higher hydraulic conductivities than the surrounding rock”. There is no further detail provided on the nature of these “geological features” and if these are distinct or wide-spread or where they are located. Overall, the injection tests (Report 2, Table 6, page 30) suggest generally high transmissivities for the Tapley Hill Formation and moderate transmissivities for the Tarcowie Siltstone (Report 2, Table 9). How this translates into the model is unclear, as Table 10 suggests model hydraulic conductivity values to be <0.2 m/d for the Tarcowie Siltstone and the Tapley Hill formation except for a “faulted area” with a hydraulic conductivity of 5m/d

The hanging wall fracture is an isolated occurrence and has been mapped using RQD from ~ 50 exploration holes (Figure 11 in AGT, 2017 shows the mapped hanging wall fault and traces of exploration holes together with RQD). There are other localised faults, such as the one intercepted by BHRIB01 (extent shown on Figure 9 in Golder 2019) which also exhibit localised areas of enhanced permeability. Table 6 (Golder 2019) shows hydraulic conductivities in the range of 4.2 m/d to 4.6 m/d for the FRA solutions and Table 10 compares the K adopted by the model with the K's derived from pumping/injection tests. Hydraulic conductivity zoning is displayed in Figure 18 for both the Tarcowie and Tapley Hill formations. The injection wells targeted zones of fracturing, which were previously identified in exploration holes. Testing shows high hydraulic conductivity for these wells which reflects the zone of high permeability associated with these fractures and are not representative of the overall formations.

3 There are also other questions related to the interpretation of MAR injection tests. Report 2 notes that only observation wells that were located along-strike were used to analyse the injection test for BHRIB01 (Report 2, p.30). However, there is no clear explanation of why other wells were not also considered. It is possible that errors in derived hydraulic properties of formations

See paragraph 3 on pg 30 (Golder 2019) Due to high well losses commonly associate with an injection well (resulting in higher injection heads), type curves were matched to groundwater rise in monitoring wells only. The monitoring wells selected for pumping test analysis included BHMB01 (located 19 m distant from BHRIB01), BH52_VW57m (located 60 m distant from BHRIB01) and 6628-23182 (592 m distant from BHRIB01). BH52_VWP57m and 6628-23182 were



13


will be created if only a subset of wells is used for the analysis. Based on interpretation of the tests, the report concludes that bore BHRIB02 is not in direct connection with the mine area (p.35). However, due to the relatively short duration of the injection testing, this cannot be conclusively determined. Interpretation of water quality changes during the MAR trials has not been undertaken.

selected in favour of other wells as they are positioned along strike of the fault intercepted by BHRIB01 (Figure 9). Furthermore, it is unwise to select monitoring wells outside of the fault zone for determination of hydraulic conductivity and storativity. This is because monitoring wells located outside of the pumped fault have lower connection to the injection well and will drawdown less during a pumping test than those targeting the same fault as the pumping well. This will result in artificially high hydraulic conductivity values being derived from test results.

4 See paragraph [ph 1 on pg 35 (Golder 2019) Overall the above results indicate there is some structural control (groundwater flow barriers) that exist between IB4 at the proposed mine area and BHRIB02 and therefore BHRIB02 is not in direct connection with the mine area (i.e. well IB4). It should be acknowledged that the groundwater response induced by the injection test reflect the duration and rate at which the test was undertaken. We have already acknowledged that connection may not be determined on its own over the duration of injection test, however there are also other lines of evidence (discussed under section 3.6 of the Golder 2019) which indicated the presence of structural control and poor connection, such as the pumping tests conducted at the mine area (well IB4).

5 Ultimately, the groundwater model is the tool that will be used to predict the impacts of groundwater extraction and MAR. However, it is unclear what changes were made to the groundwater model as a result of the injection trials. Changes were made to horizontal hydraulic conductivity, but Figure 18 and Table 10 are not detailed enough to understand what has been changed and where. The documents describe a clay aquitard near the mine decline, and notes that the properties of this aquitard were “refined during the transient calibration process” (p.44). However, the nature of this aquitard, how it has been mapped, and how its properties were refined has not been clearly described. Also, have vertical hydraulic conductivity values changed as a result of the differences in horizontal hydraulic conductivity values? Anisotropy needs to be more carefully considered, as it can be crucial to the success of the planned injection. Yet, it is not clear that there were

We acknowledge the extent of the clay aquitard (which is shown on a cross-section in Figure 2 and mapped extent on Figure 3 in Golder 2019) has not been discussed in the Golder reports, but further information on this can be found in the MLA. A summary of changes performed on the model can be found in section 4.2 and zones that have been added (zones 1 to 3) are shown on Figure 18. 31 monitoring wells were monitored during the pumping test and injection test to measure the anisotropy.



14


sufficient observation wells to fully characterize it.

6 We observe that the model refinement produced worse a steady state model calibration (i.e., the calibration to heads prior to mine operations), but this is not mentioned in the reports. Furthermore, the newly calibrated model greatly under-predicts water level changes (relative to measured changes) at some bores and over-predicts water level changes at other bores. In particular, simulated drawdown at BH36 is less than one third of that observed (approximately 2 m versus 8 m after 10 days). Groundwater rise is under-predicted at 6628-9156, a bore close to the boundary of the mine site. No map is presented showing areas of over-prediction and underprediction, and the implications of the poor model fits for the reliability of predictions is not discussed. No quantitative measure is presented of the goodness-of-fit of the model to the MAR tests (e.g., RMSE)

It appears Cook et al. (2019b) reviewed an earlier version of the report which was released to the WCCC and not the final version which was submitted as part of the MLA. Several of these comments were addressed in the latest report submitted with the MLA. The steady state results are presented in a scatter plot of observed versus calculated groundwater elevation (Figure 19 of Golder 2019) which reported a scaled root mean squared (SRMS) error of 6.6 %. This compared well to previous SRMS of 6 % (AGT, 2017). The Australian groundwater modelling guidelines (Barnett et al., 2012) highlight that goodness-of-fit is subjective. However, an SRMS of <10% is referred to as an example target within these guidelines, which is satisfied by the current steady-state calibration process. No further refinement of the steady state calibration was warranted. Appendix E (Golder 2019) presents the comparison between simulated and observed drawdowns in response to pumping and injection. Inspection of hydrographs in Appendix E demonstrates that the model generally reproduces the pumping and injection induced fluctuations well in the majority of monitoring wells. In particular the responsiveness to pumping and injection and the magnitudes of fluctuation are reasonably well represented. BHMB02A showed small mismatch between simulated and observed heads, however this is considered acceptable quality of fit to the nearest observation well (BHMB02) and farthest observation well (Day Windmill) were prioritised, in order to ensure that the overall extent of the zone of influence and the magnitude of groundwater level change in close proximity to the injection well were adequately represented within the model. The miss match observed at BH35 (a historic exploration hole, converted to a monitoring well by maximus) was corrected in the subsequent version.

8 Overall, given the moderate to high transmissivities obtained during the MAR trials, it appears questionable, if “worst case” inflow scenarios into the mine are being explored with the current set of model parameters.

See response 1



15


9 Our earlier report identified a large number of limitations with the groundwater model. These included uncertainties concerning the possible influence of anisotropy in hydraulic properties on pumping test results, lack of anisotropy in the predictive model, and inadequate model calibration. While the new work shows that MAR may be a feasible option (due to the high injection rate at one of the two sites), it has raised additional concerns about the heterogeneity of the site, and the extent to which this heterogeneity is understood. Due to the geological complexity of the site, we believe that greater site characterization is required than has been carried out to date. This should include the use of surface geophysical techniques to better characterize geological structures. The poor fits between the model simulations of the MAR trial and the observation data apparent at a number of sites are most likely a result of incorrect description of aquifer properties in the model. This raises concerns about the confidence of model predictions. Currently, the assumed worst-case is the 70% grouting scenario - which may be optimistic, given the large air lift yields achieved during drilling. Given the uncertainties in the model and the underlying difficulties in characterising the fractured rock aquifer, it would be advisable to evaluate a more conservative worst-case scenario to cover potential larger mine inflows and therefore MAR injection volumes.

In response to uncertainties concerning the possible influence of anisotropy in hydraulic properties on pumping test results see comment 12. In response to lack of anisotropy in the predictive model, and inadequate model calibration see comment 18 and comment 22. In response to Due to the geological complexity of the site, we believe that greater site characterization is required than has been carried out to date. This should include the use of surface geophysical techniques to better characterize geological structures. Based on the volume of data collected we don’t feel additional geophysical surveys are warranted. The site has been adequately characterised by 14 pumping tests and two injection tests to assess the key structural features which influence groundwater flow. This is in addition to the 9 investigations wells (4 of which were subject to downhole geophysical surveys), 31 site monitoring wells were monitored during pumping to inform anisotropy and boundary conditions and ~50 exploration holes which were used to inform the site geology which in turn was used to inform the development of the groundwater model In response to the poor fits between the model simulations of the MAR trial and the observation data apparent at a number of sites are most likely a result of incorrect description of aquifer properties in the model. Transient calibration was revisited resulting in improved matches and this is also discussed under comment 11 and 22. Both iterations were reviewed by IGS and considered acceptable. Furthermore, we consider the calibration to be acceptable and to be based on correct description of aquifer properties. In all several lines of evidence were used for parametrisation / model calibration which included 1) pumping tests conducted on 14 wells and injection tests conducted on 2 wells (shown on comment 14) 2) a further 31 wells have been monitored during these pumping / injection tests, 3) the pumping effects from neighbouring pumping wells (6628-9154, 6628-23182, 6628-10248, 6628-8936 and 6628-23182) also contributed to parameterisation / calibration, 4) the model was validated with anecdotal pumping observations from mine dewatering (at rates of 40-60 L/s) over 120 days of pumping and observations at the Ridge and Two in the Bush mine shafts, 5) seasonal pumping from irrigation wells and the monitoring of 42 observations wells throughout the sub catchment have been captured over 3 years. 6) Pressure transducers were



16


placed in key monitoring wells to measure drawdown from summer irrigation pumping to investigate/detect the influence of third party wells at the project site. In response to Currently, the assumed worst-case is the 70% grouting scenario - which may be optimistic, given the large air lift yields achieved during drilling. This statement is incorrect. Mine inflows are calculated based on aquifer hydraulic properties and you cannot relate airlift yields of a well to mine inflows see comment 1.

11 Aquifer (pumping) tests were conducted on 8 wells: IB1, IB2, IB3, IB4, IB5 and three private wells. Some information was also obtained from examining response of private irrigation wells to irrigation cycles. However, most of these wells are located within a relatively small area (see Figure 6), and pumping tests on the three private wells were of very short duration, and so the extent of the aquifer sampled by these would have been very small and unlikely representative of larger, regional scale hydraulic conductivity values. This is particularly problematic in an aquifer that is likely to be highly heterogeneous

Since this review (and the AGT 2017 report) a further 3 pumping tests (conducted on IB4, BHRIB01, BHRIB02) and 2 injection trials (conducted on BHRIB01 and BHRIB02) spanning between 6 and 15 days were undertaken which involved the monitoring of 31 monitoring wells. This included the monitoring the hydraulic influence and pumping rates of pumping wells surrounding the site (6628-9154, 6628-23182, 6628-10248, 6628-8936 and 6628-23182). See figure 7 in Golder 2019 for locations of pumping/injection wells, monitoring wells and private wells monitored during the injection trials. The additional pumping tests brings the total pumping tests to 14 in addition to the two injection trials and series of airlift pumping test analysis which occurred during drilling. The distribution of wells subject to pumping tests is outlined below (circled wells) and the spread of 31 monitoring wells are not focused in one area.



17


The hydraulic extent from the more recent injection tests was detected at a distance of 1 km and therefore we disagree with the statement that the aquifer sampled by these would have been very small and unlikely representative of larger regional scale hydraulic conductivity values. Furthermore, additional work (outlined under comment 9) was undertaken to inform parametrisation of the broader sub catchment.

12 A number of different mathematical models are used to interpret the pumping test results (p.38 and Appendix B). However, there is no discussion of why particular models are chosen for particular situations, and insufficient discussion of the flow geometry (and assumptions) underlying these different models. There is also insufficient discussion of the extent to which these models fit the observation data, and hence what the accuracy of the estimated hydraulic parameters might be. Precise values are given in Table 5, with little discussion of their accuracy. Changes in the shape of the drawdown plot that occurred during pumping well IB-4 are interpreted as the effect of a zone of low hydraulic conductivity (p.68), but the description of the changes in drawdown do not clearly match the plots presented in Appendix B. In any case, it is unclear how the location of this boundary is determined with sufficient accuracy for it to be included in the regional groundwater model (see below). Some of the modelling simulations (described below) also suggest that significant anisotropy is present in the aquifer system. If this anisotropy is not specifically included in models used to interpret pumping tests, then the results of the pumping test analyses may be seriously in error (National Research Council, 2012).

A suite of analytical solutions were adopted to derive aquifer properties such as T, K and S from different pumping wells and monitoring wells. In both studies > 30 pumping test analysis were performed on the pumping test data which gave a range of values depending on the 1) pumping well used in the analysis, 2) the monitoring well used in the analysis 4) ratios of Ky/Kx (where applicable) and 4) whether the drawdown, recovery or both phases of the pumping test were evaluated (See Table 5 in AGT 2017 and Table 10 in Golder 2019). As AQTESOLV curve matching produced good fits for the range of solutions applied, the geometric mean of K was adopted by the numerical model and refined during the transient calibration process (within bounds of tested values). Detailed discussion is provided under section 4.3 in Appendix F. Furthermore, a sensitivity analysis was performed based on the tested range of hydraulic parameters and doubling K See Table F3 in Appendix F of AGT, 2017.

In response to in any case, it is unclear how the location of this boundary is determined with sufficient accuracy for it to be included in the regional groundwater model. The presence of the boundary condition was re-evaluated by subsequent pumping / injection tests (reported in Golder 2019). See comment 4.

In response to If this anisotropy is not specifically included in models used to interpret pumping tests, then the results of the pumping test analyses may be seriously in error. The Gringarten-Witherspoon w/vertical fracture solution was also used as this allows ratios of Ky/Kx to be varied. For example, a Ky/Kx of 1 resulted in a Kx of 1.44 m/d, where as a Ky/Kx of 0.4 resulted in a Kx of 2.35 m/d. These results fall within the range of K’s derived by other solutions shown in Table 5 of the AGT 2017 report.

13 Information on variation in hydraulic properties with depth is based largely on a plot of bore yield versus depth (Figure 40). Based on this plot AGT state that the most

The plot of bore yield versus depth was based on catchment wide airlift data obtained from drillers records in WaterConnect. This was undertaken to demonstrate the range of the fracture zone below



18


productive zone of the Tapley Hill Formation is at 50 – 150 m below ground level, which they suggest might be due to a higher fracture density or larger fracture apertures within this zone (p.111). However, it is not clear that the 50 – 150 m depth zone has, on average, higher well yields than shallower and deeper zones. To establish this, the authors would need to present mean values (and possibly standard deviations) of well yield within different depth zones. No such statistics are presented.

ground surface (50 m – 150 m) which is also supported by the fact that the average well depth in the catchment is about 73 m,. Calculating the mean and standard deviation values as suggested by Cook would be of little significance in this instance because the data used is not for a specific depth, but rather the end of the hole and represents the whole of the well. Typically, above the fracture zone is comprised of weathered rock and below this zone is the fracture extinction zone where aquifer permeability decreases. Note, this is an observation based on WaterConnect drillhole data sourced from the Inverbrackie Creek sub catchment, and it therefore applies to the broader Inverbrackie sub catchment rather than the mine area where the focused drilling investigations took place. This approach for determining the productive zone for model development is also consistent with the approach used by DEW (formally Dept for Water) in their groundwater flow model of the Cox Creek Catchment, Mount Lofty Ranges (Stewart and Green, 2010).

14 Recharge across the catchment is estimated using the chloride mass balance approach. The description of the application of this method is particularly poor, with no justification for the assumed chloride concentration in rainfall (10 mg/L) or annual runoff rate (10% of rainfall). The authors claim that this is a reliable technique in fractured rock aquifers (p.118) and reference a report by Cook (2003). However, the report by Cook (2003) describes “two major difficulties” with using the chloride mass balance approach in a fractured rock system. Probably the most significant of these problems is the requirement by the chloride mass balance method that the system be in steady state. Studies in the Clare Valley have shown that this assumption may not be reasonable and will result in underestimation of recharge (Love et al., 2002). This problem is not discussed within any of the AGT reports. Due to uncertainties associated with the chloride mass balance method for estimating recharge, it would be prudent to also apply other approaches, perhaps including the water table fluctuation method and use of geochemical groundwater age indicators. It is unclear whether or not flow

The use of the CMB method for recharge estimates in the Inverbackie Creek Sub Catchment is consistent with other recharge studies undertaken by DEW (formally DWLBC) in the neighbouring sub catchments of the WMLR (Forreston, Fox Creek and Mylor sub-catchments) see Green et al 2007 (Banks et al 2006). DWLBC applied the CMB technique and compared this method to groundwater age indicators (CFC12 and 14C). DWLBC concluded that at all sites the CMB technique compared well and proved to be a good indication of recharge. Given hydrogeological and climatic settings are similar, the use of CMB technique in the Inverbrackie sub-catchment is justified. In the AGT report, the CMB technique utilised chloride concentrations from about 50 wells to assess spatial trends.



19


in Inverbrackie Creek was measured as part of the characterisation (p.47 refers to monitoring of flow, but no data is provided in the report). If it was not monitored, then it is not clear why this is the case, as it is a relatively simple process and would have been useful data.

The water-table fluctuation (WTF) method is based on the premise that rises in groundwater levels in unconfined aquifers are due to recharge arriving at the water table. The WTF method is not suited to this setting and is most applicable in areas with shallow water tables that display sharp rises and declines following rainfall events and at sites which are not influenced by groundwater abstraction from pumping wells. In this setting, we observe large drawdowns in monitoring wells from summer pumping. For example, in the Tapley Hill Fm, large summer drawdowns of up to 15 m occurred due to nearby summer pumping making the WTF method unsuitable for several wells. Monitoring wells near the proposed mine site (revealed deeper groundwater levels of 40 m below surface) and showed muted seasonal groundwater level fluctuation of 1.5 m. These wells target deep water bearing parts of the aquifer > 125 m deep which are less responsive to rainfall. Therefore, we do not believe the use of the WTF will produce reliable recharge rates in this setting.

Baseflow to the Inverbrackie Creek was not measured. Further discussion is provided under comment 15.

15 The report argues that the model is “validated” through comparison with flows in Inverbrackie Creek. Simulated groundwater flow (baseflow) to Inverbrackie Creek (and tributaries) in pre-mining simulations is approximately 732 ML/y (23.2 L/s), and it is noted that this is “adequately comparable” with a baseflow estimate of 874 ML/y (27.7 L/s) estimated by Zulfic et al. (2002). However, the latter is extremely unreliable due to the methods used to derive it. It is unclear why no attempt was made to measure creek flow by the project proponents. Mining activities are predicted to decrease baseflow by approximately 150 ML/y by year 5 – 6 with no mitigation (Figure F35).

Flow rates to the Inverbrackie Creek have been published in two government reports. DEW (formerly DWLBC) published a baseflow value of 874 ML/y (see Barnett et al., 2002) and the Western Mount Lofty Natural Resources Management Board published a value of 867 ML/y (see Table 5.4 in the Water Allocation Plan for the Western Mount Lofty Ranges). It is acknowledged that there exists uncertainty in these values, however the similarity of the two estimates suggests a degree of reliability that is greater than extreme unreliability. Both published values for baseflow are consistent with modelled baseflow value of 732 ML/y.



20


16 Numerous faults are presented in Figure 9 (p.49), although the source of this information is not described. Their role on the regional hydrogeology also is not discussed. Faults can act as either barriers or conduits to flow (Bense et al., 2013), and some consideration of their possible role would appear to be warranted.

We agree that faults can act as either barriers or conduits to groundwater flow. An aspect of model conceptualisation and parameterisation was the characterisation of fault behaviour. Field programs (14 pumping tests) and data collection during pumping / injection (31 monitoring wells) focused on informing the behaviours of faults which were implemented in the groundwater model. Key faults assessed by pumping test include: • Faults which behave as conduits to groundwater flow such as the hanging wall fault assessed by IB4 and the NE-SW trending fault intercepted by BHRIB01 (the latter being incorporated during the model update presented in Golder (2019b)) • Faults which behave as groundwater flow barriers – Two NW-SE trending faults near the proposed mine were evaluated several times (i.e. during the pumping test IB4 and injection into BHRIB01, injection into BHRIB02 and pumping from nest egg) Despite these conceptualisation/parameterisation components being identified, modelling was used to test the influence of each as part of the uncertainty analysis (see Section 7.5 in AGT, 2017). This was particularly important in determining the likely extent of drawdown, as barrier behaviour is likely to restrict the area impacted by drawdown (see Figs 63 and Fig 64). The work of Golder (2019b) showed further field-based evidence of these structural controls.



21


17 Water quality data has been obtained from investigation bores and from a number of private wells. Although much of the groundwater within the catchment is of relatively low salinity (most bores are less than 1500 mg/L TDS), some areas of higher salinity exist, particularly to the southeast of the mine site. However, based on data contained in Appendix C1 and C2, there are also areas of saline water (> 3000 mg/L) much closer to the mine site than depicted in the regional groundwater salinity map (Figure 28). It is unclear why this data (e.g., from 6628-23530 and 6628-8301) is not discussed in the report and omitted from the map. It suggests that a more thorough understanding of the distribution of water quality in the regional aquifer (including how it changes with depth) is warranted.

6628-8301 was inadvertently left from Fig 28 however the groundwater salinity (3,300 mg/L) measured at this well is consistent with the groundwater salinity measured in other wells sampled which target the same formation. 6628-23530 (BH35) is a converted exploration hole which was not drilled for hydrogeological purposes. The higher salinity measured in this hole is uncharacteristic of the aquifer. Nearby monitoring wells BH34, BH16, BH41 and investigation wells IB1 to IB5 (which have been subject to longer term pumping tests) revealed groundwater salinities that are more representative of the aquifer.



22

18 The groundwater model construction was informed by the characterisation described above. For the most part, the model uses constant and isotropic horizontal hydraulic conductivity values for each geological unit, although changes in hydraulic conductivity with depth occur for some units. The use of constant and isotropic hydraulic conductivity values across geological units is unrealistic for a fractured rock aquifer but is a common assumption for modelling purposes. However, it is critical that this assumption is assessed as part of sensitivity and uncertainty analysis. Small-scale variation in hydraulic conductivity may not significantly influence drawdown at distances of several hundred metres or more from the pumping sites, as drawdown at these distances is likely to respond to average hydraulic properties. However, if there is a preferred orientation to fractures, then hydraulic conductivity can be higher is some directions than in other directions (i.e., anisotropy). This can have a profound effect on drawdown. Some of the characterisation data suggest this is the case, but it is not included in pump test analysis or in the groundwater model. Errors in drawdown can also result if large fractures or fractures zones are present and not included in the model conceptualisation.

The majority of fracturing in the study area is understood to be small-scale and multi-directional. As the reviewers indicate, it is common practice to adopt an isotropic equivalent porous medium as a surrogate for modelling these geological units. In parallel with the response to comment 28 below, it is acknowledged that anisotropy parameters could be included in a more extensive uncertainty analysis process. However, this would not expected to have a profound influence on the likely range of model predictions. Whilst anisotopy in general has the potential to significantly influence drawdown propagation, variation in anisotropy parameters in the present model would be constrained to a large degree by the injection trial data (see Golder, 2019b). It is these data that assessed the key NNE-SSW trending fault in the study area, whilst no other large-scale anisotropy features were indicated. The key NNE-SSW trending fault was incorporated explicitly in the updated numerical model as an additional hydraulic conductivity zone, see Golder (2019b). The pumping test results presented in AGT (2017) indicated possible local-scale anisotropy. However, this was within in an area confined by identified flow barriers (see our response to comment 16). Therefore, the potential for this anisotropy to influence model impact predictions on a significant scale is limited.



23


19 Based on well yield data (see above), Layer 7 of the groundwater model is assigned K values one order of magnitude lower than Layer 6 (Appendix F p.9). The model does not appear to include a separate sensitivity analysis on the conductivity of this lower layer, and sensitivity analysis to hydraulic conductivity is mostly limited to factor-of-two variation. It is therefore unclear exactly what effect this low hydraulic conductivity layer has on model results, but it might limit the lateral spread of drawdown from the mine, and hence result in underestimation of the regional impact (see discussion below).

There is no significant lateral movement occurring within this layer, however when the aquifer properties of model layer 7 are changed to equal those of model layer 6, there is no change to mine inflows and drawdown.

20 The groundwater model also includes flow barriers that were inferred principally from the pumping test on well IB-4. The effect of these barriers on simulated drawdown with the Tapley Hill Formation is significant (Figure F47), and also in the Tarcowie Siltstone, particularly to the south of the mine site in the unmitigated and 70% effective grout simulations. Since the presence of these barriers has not been conclusively demonstrated, their inclusion in the model is questionable, particularly since their impact on model results is significant. Furthermore, numerous other faults are presented in Figure 9 (p. 49), but their role on the hydrology is not discussed, and these are not included in the model. We therefore cannot be sure that “worst case” conditions have been examined in the current set of model parameters and scenarios.

See comment 16. At the time of the AGT study these flow barriers were also removed from the model to address this uncertainty (present at this time) and allow the drawdown to spread (see section 7.5.1 of the AGT 2017 report). Since this work, additional monitoring wells had been installed around these structures and additional pumping/injection tests have been conducted (BHRIB01, BHRIB02 and IB4) which confirmed the presence of this structural control and their role of controlling the spread of drawdown (refer to section 3.6 of the Golder 2019 report).



24


21 Recharge zones in the model are loosely based on chloride mass balance data, although some recharge rates used by the model are lower than any of the chloride mass balance estimates of recharge (compare Figure F5 with Table 9 and Figure 43) and there is insufficient data available to accurately define boundaries between the zones. Defining recharge zones based on chloride mass balance estimates obtained from chloride concentrations measured in individual bores also ignores the fact that groundwater flows through the catchment, and so the recharge estimate obtained from a particular bore represents the recharge rate up-gradient of the bore location, not at that location. These issues together with the very high uncertainties associated with the chloride mass balance approach collectively mean that the recharge distribution used in the model must be considered to be highly uncertain. Although the model uses a sensitivity analysis to show that the impact of a zone of high recharge east of the mine does not greatly alter drawdown predictions, there is no more general examination of sensitivity to recharge values or the resulting uncertainty in the model predictions. Underestimation of recharge will likely result in underestimation of hydraulic conductivity when the model is calibrated. This could result in underestimation of the extent of the drawdown cone induced by mine dewatering and needs to be tested in a formal uncertainty analysis. It is also unclear whether the model uses constant recharge or whether recharge is varied seasonally. If constant recharge was used, then the calibration of the model to seasonal water level variations could result in errors in estimated hydraulic parameters (see discussion below).

The model development / calibration was supported by as many field-test-based inputs as possible, including both K (derived from substantial number of pumping tests) and refined through calibration process and R (based on the CMB). Justification for the use of the CMB technique is provided in comment 14. A substantial increase in R would require a substantial and pervasive increase in K (in order to maintain the ratio of R/K and thus maintain the current calibration statistics). The latter is not supported by field data.



25


22 Results of the transient calibration to pumping test data shows reasonable simulations of some well drawdown and recovery measurements, but poor simulation of others (Figure F13). The report claims that underestimation of drawdown at three wells may indicate local anisotropy in hydraulic conductivity with a northwest-southeast primary axis and note that this type of anisotropy is common in fracture zones. If such anisotropy exists, then it is concerning that the model did not try to simulate it. However, the poor calibration may result from poor model parameterisation in other areas. (The IGS report describes the parameterisation of the hanging wall fracture zone as a “major limitation of the model”, and we agree with this assessment.) The AGT (2017) report further claims that model-based underestimation of recovery rate indicates that the model is conservative (Appendix F p.31). But simulation of the pumping test data greatly underestimates drawdown at a number of the observation bores. This is not a conservative result, as underestimation of pumping test results is likely to mean that results of mine dewatering impacts will be similarly underestimated, when the calibrated model is ultimately used for mine scenario modelling.

The calibration to pumping test data (pumping from IB4 which targets the hanging wall fault) was revisited as part of the 2019 MAR investigation (Golder 2019) which involved a calibration to longer pumping period (11 days compared to 6 days) and observations from an additional 8 monitoring wells (which were not present during the first pumping test performed on IB4 in 2014). Improved matches were observed for a number of monitoring wells (such as IB1, IB2, IB5, BH36). These wells showed an underestimation in predicted drawdown in the 2014 calibration but showed a very slight overestimation in drawdown in the 2019 calibration. This slight change from a slight underestimation to a slight overestimation to drawdown and improved matches did not change the model predictions of inflows or drawdown (see results section 4.4, Fig 20, Fig 21, Golder 2019). Revised calibration showed slight reductions to both predicted inflows and drawdown at private wells. To clarify, an over prediction of drawdown arises from lower K which result in lower inflows, where as an underprediction of drawdown arises from higher K which results in higher inflows (and therefore the AGT results were conservative in terms of mine inflows).

23 Figure 52 (main report) and Figure F31 (Appendix F) appear to show the drawdown cone extending to the models eastern and southern boundaries. Since the drawdown cone cannot extend beyond this boundary, a large head gradient is created at the boundary which allows flow into the model. The flow into the model is much greater than would occur if the model extended further, and this will reduce the simulated drawdown in other areas. The proximity of the eastern and southern boundaries thus appears to be influencing the model results and causing drawdown to be less than would otherwise be the case. This appears to be a major model defect.

The assertion of “major model defect” is unjustified and based on a misunderstanding of the modelling approach. The drawdown cone only extends to the eastern and southern boundaries for the infeasible “no mitigation” scenario, wherein no mitigative measures are employed during mining. The only relevant quantitative output provided by this scenario is the unmitigated groundwater inflow into the mine, which provided the basis for developing the grouting scenarios in terms of relative inflow reduction. Thus, if anything, the larger head gradients created by eastern and southern model boundary influences would artificially inflate simulated mine inflows to some extent, resulting in all subsequent mitigation scenarios being relatively conservative. The boundary effect is nonetheless expected to be insignificant, however, as the drawdown cone only reaches the eastern and southern model boundaries in the latter stages of the simulation, limiting the amount of time available for the resultant hydraulic influence to propagate into the model domain. Moreover, the only boundaries encountered by the cone of drawdown during scenarios including



26


grouting are “no-flow” boundaries. This will in fact artificially increase simulated drawdown in the model for these scenarios, thus further increasing its conservative nature. Finally, during predictive simulations including MAR, the cone of drawdown does not encounter any boundaries. Given that MAR has been confirmed as a mitigation measure to be employed during mining, and thus these simulation results are the only truly relevant results, the proximity of model boundaries being described as a “major model defect” in the context of the project would appear to be an overstatement.

24 The steady state simulation appears not to include the presence of groundwater extraction (predominantly irrigation) wells. (Pumping data is not included in the water balance in Table F4.) However, in the transient calibration to seasonal regional pumping, AGT observes that “pumping for irrigation causes seasonal fluctuations, but no long-term decline”. However, if the groundwater system appears to be currently in steady state with current groundwater pumping, then it would appear to be essential to include this groundwater pumping in the simulation. If the recharge rates are correct, then to omit pumping (which is of similar magnitude to recharge) would increase the volumes of water moving through the aquifer, and the model would need to have higher hydraulic conductivities to simulate the observed water levels. The comparison of modelled versus observed heads on calibration bores is reasonable for many bores, but differences are in excess of 10 m on a number of bores (Figure F10). No map is provided showing the spatial distribution of these discrepancies, yet this is common practice and would be informative. The authors do comment that the fits are poorest in bores that are furthest from the project site (p.35 and Figure F18). Although this may be true, since this data was used as part of model calibration, it may have impacts throughout the model domain.

Based on present understanding, the groundwater system in the study area responds rapidly to seasonal (summer) extraction such that it recovers annually to a post-winter state that may be considered to be an effective representation of a “pre-development steady-state condition”, despite the fact that some development has occurred in the region in the form of domestic and agricultural groundwater abstraction. Steady-state calibration could equally have been undertaken based on an assumption of a late-summer effective steady-state condition that includes pumping stresses. However, the large degree of uncertainty in regional pumping rates would have propagated into the steady-state calibration process as additional, unnecessary uncertainty in estimated hydraulic properties. If regional seasonal pumping was well-estimated and a “post-summer” effective steady-state calibration was undertaken as an alternative, then estimates of hydraulic conductivity should be equivalent to the estimates of hydraulic conductivity obtained through the “post-winter” effective steady state, given that it is a material property and not dependent upon whether or not pumping is occurring.



27


25 A short transient calibration was also carried out using regional pumping data, in an attempt to reproduce seasonal water level variations. Figure F17 shows plots of simulated and observed heads over approximately 2 years at 28 bores within the region. For seven of these bores, seasonal variations in both simulated and observed heads appear to be less than 1 m and cannot be readily determined on the scale of the figures. However, for most of the remaining 21 bores, seasonal variations in heads are significantly underestimated by the model. In a number of cases, the underestimation approaches an order-of-magnitude. It would have been helpful if a plot showing simulated versus observed seasonal drawdown was included in the documentation, but this is not the case. The IGS review of the model also raised this issue. It would also be helpful if these data were plotted as absolute values rather than relative head deviation. It is not common practice to only show relative deviations, as relative values mask some of the model calibration issues. Nevertheless, the underestimation of seasonal drawdown would appear to indicate that either pumping rates are greatly underestimated, or that storage parameters are overestimated. If the latter, then the impacts of pumping on adjacent groundwater users might also be significantly underestimated. However, it might also result from not including seasonality of recharge in the model (if this is the case), or from recharge rates that are too high in these areas. The ratios of R/K and S/K (recharge divided by hydraulic conductivity, and storage coefficient divided by hydraulic conductivity) will control the ability of the model to simulate these transient water level changes, and it would suggest that at least one of these parameters is in error. The report states that the model “generally reproduces seasonal pumping-induced fluctuations” (Appendix F p.35), but this is not reflected in the comparison of field data and simulation results. It also notes that “offsets” in calibration are “not relevant to the transient calibration”. This is not correct. Offsets in calibration mean that the model is not correctly reflecting reality. At the field site it is probably largely due to errors in hydraulic conductivity values (or the lack of

It is stated that “relative values mask some of the model calibration issues”. However, we believe that, on the contrary, absolute values mask the transient calibration performance with steady-state calibration effects. Systematic offsets between simulated and observed heads stem from imperfections in the hydraulic conductivity field estimated during the steady-state calibration process. The role of the transient calibration process is to enable comparison of the simulated and observed aquifer dynamics and adjust storage parameters accordingly. Regarding the speculation as to whether pumping rates are greatly underestimated or storage parameters are overestimated, we are confident that the former is the case. It is known that not all groundwater extraction within the model domain area is accounted for due to private landholders not providing information. Further increasing this confidence is the fact that various phases of hydraulic testing undertaken at the site have provided field estimates of storage parameters which are consistent with those employed in the model (refer to Table F3 in AGT (2017) and Table 10 in Golder (2019)). In response to “This is not correct. Offsets in calibration mean that the model is not correctly reflecting reality”: We of course agree that offsets in calibration mean that the model is not correctly reflecting reality. And it is not broadly stated in the report that “offsets are not relevant to transient calibration”. This has been manipulated in the review document to appear as a broad statement, when the report is clear in referring specifically to initial, systematic offsets. It is stated in the report is that these systematic offsets reflect imperfections in hydraulic conductivity and recharge, these parameters being the ones that were estimated during the steady-state calibration process. Examining these particular offsets as part of the transient calibration process is effectively double accounting for the calibration errors that were already reported on via steady state calibration statistics. The focus of the transient calibration process was to examine storage parameters and aquifer dynamics, hence the examination of relative deviations independent from systematic offsets that were previously examined during the steady-state calibration process.



28


including spatial variability in hydraulic conductivity). Incorrect K values will affect predictions produced by the model. Since pumping was not included in the steady state calibration, inclusion of pumping in this transient calibration should have caused a declining groundwater trend over time. This is not readily apparent from Figure F17, although is briefly mentioned on page 42 of Appendix F. It is possible that this declining trend is not more readily apparent because of the relatively short period of this simulation.

26 The risk of saline groundwater intrusion from areas east of the mine site was assessed through solute transport modelling on the basis of the developed flow model. AGT (2017) states, that “hydraulic models for the 70% grouting and 70% grouting with MAR scenarios formed the basis for the salt transport models” (p.149). Based on the solute transport predictions, a “very minor movement of the saline interface, even where no mitigation is applied (grouting or MAR)” is predicted, “owing to the short mining life” (p.149). It remains unclear, which hydraulic conditions were assumed for this assessment, as the above statement (which refers explicitly to “no mitigation”) contradicts the explanation that the “hydraulic models for the 70% grouting and 70% grouting with MAR scenarios formed the basis for the salt transport models”.

The presented results pertain to the scenario involving no mitigation. The purpose of this was to demonstrate the risk of saline groundwater migration even for an extremely conservative scenario which was adopted to force a hydraulic gradient from the east (to induce groundwater flow from the Kanmantoo Formation). Movement of saline water from the east was negligible for the scenarios involving MAR, as the groundwater divide between the EMLR and WMLR was maintained in these simulations. The in-text reference to the 70% grouting scenarios is a remnant of the initially intended conservative approach to the transport modelling prior to adopting the extremely conservative approach.

27 The model simulates MAR using eight wells, with a simulated combined injection rate equal to the total mine inflows. Thirty percent of inflows are injected into one well, with the remainder evenly divided amongst the other seven wells. Whether or not these injection rates can be sustained has not yet been determined. Well loss coefficients used by the model are not specified. (Well loss coefficients describe the resistance to flow associated with the well itself. Their values will initially depend on well construction techniques, but the resistance can increase over time due to clogging of well screens by fine material and microbial activity.) Model simulations show that artesian conditions develop around one of the wells. Whether this causes problems

The feasibility of MAR has been proven by subsequent studies (Golder 2019). In regards to the comment Thirty percent of inflows are injected into one well, with the remainder evenly divided amongst the other seven wells, we note that this was demonstrated by the injection trial which showed high rates of recharge were possible in a single well and this was undertaken under low pressure. The injection rate achieved during the trial was equivalent to nearly 3 times the expected mine inflows. The proposal will involve the distribution of treated mine water across 8 injection wells, and therefore each well will be injecting at much lower injection rates and operating at lower pressure than those observed during the MAR trial. The vertical connectivity was assessed to be low which allowed injection wells to be pressurised.



29


will depend upon vertical connectivity of aquifer layers, which is currently not well known at the site. It is possible that a larger number of injection bores will be required.

Furthermore, groundwater modelling shows groundwater rise (pressure) owing to injection (at mine yr 5) was - 1 m (90% grouting scenario) and - 1 to -5 m (70% grouting scenario) and therefore there no foreseeable risk of excessive aquifer pressurisation (see Figures 22 & 23 in Golder 2019).

28 The Terramin groundwater model contains some limited sensitivity analysis, but no uncertainty analysis. The model sensitivity analysis has included analysis of the effect of doubling the hydraulic conductivity of the Tapley Hill Formation. This analysis shows a significant increase in mine inflows. Drawdown contours (Figures F52, F53 and F54) show some increase in size of the drawdown cone, but the expansion of the drawdown cone appears to be limited by the size of the modelling domain. Model results cannot be trusted once this occurs. It appears that the expansion of the drawdown cone would have been greater than these model simulations show had it not been artificially limited by the model extent. Moreover, due to the general uncertainty on hydraulic parameters in fractured rock environments, a doubling of hydraulic conductivity greatly under-represents the true uncertainty of this parameter. It therefore seems doubtful that the model has explored a realistic range for model parameters that reflect natural variability and hence uncertainty at appropriate representative scales.

The comment regarding the size of the model domain has been addressed in response 23 above. The assertion of “no uncertainty analysis” is considered misleading, given that, for example, key model predictions are presented for scenarios such as assuming grouting will be only 70% effective, which is extremely conservative based on the independent grouting effectiveness predictions. Model prediction sensitivity to more exhaustive and pervasive parameter variability could be explored at a future stage. However, Golder maintain that the sensitivity analysis conducted and presented is appropriate for the stage of the project. The currently presented uncertainty envelope (Figure F50) represents variability of inflows by up to approximately 25%. The 70% effective grout scenario represents a hypothetical failure of the grouting effectiveness to the extent that inflows are 300% greater than for the predicted 90% grout effectiveness case. Given the 25% variability observed during the sensitivity analysis (albeit employing parameter ranges that may be somewhat limited as pointed out), it was not considered feasible for any degree of realistic hydraulic property variability to result in a 300% increase in predicted mine inflows. If any model predictions based on the 70% effective grouting results were deemed unacceptable during regulatory assessment, a more thorough hydraulic property-based uncertainty analysis could be undertaken to better define the upper bound (worst-case-scenario) of the prediction uncertainty envelope (which, in terms of predicted impacts, would thus fall somewhere between the 90% and 70% grout effectiveness scenarios).

29 We also believe that the sensitivity analysis is severely limited by the one-at-a-time approach to variation in hydraulic parameters. What are the implications for the model results if more than one parameter is in error?

The response to this is equivalent to the response to comment 29 above – varying parameters together may have a slight effect on the presented prediction sensitivity envelope, but which is expected to be insignificant relative to the 70% grouting effectiveness results.



30


30 Uncertainties associated with the solute transport model will arise from uncertainty in the underlying flow model (due to errors in hydraulic conductivity and storage coefficient or conceptualisation errors), but will be compounded by potential errors in additional parameters related to solute (salinity) modelling. It would therefore also be advisable to add the parameters that are important for the solute transport modelling (particularly porosity and dispersivity) to the sensitivity analysis. This would provide an indication of the range of potential salinity migration scenarios under a range of realistic porosities and initial salinity distributions.

The solute transport modelling demonstrated insignificant movement of the interface for the unrealistically conservative “no mitigation” scenario. This was considered to represent an end-member for this model prediction, thus it was deemed unnecessary to explore the effect on this prediction of varying parameters within realistic ranges.

31 It should be noted, however, that the supergene zone is “just beneath the water table” (p.100) and as such relatively small declines in the water table may allow exposure of the material to oxygen. In addition, AMD may occur beyond the supergene zone due to dewatering of the Tapley Hill Formation. Pyritic material is evident from the investigation by Tonkin (2016) outside of the supergene layer within parts of the Tapley Hill Formation, albeit at low concentrations at the mine site. However, “regionally the Tapley Hill Formation can be rich in pyrite…” (Tonkin 2016) and therefore some generation of acid mine drainage should be anticipated during the operation of the mine. Treatment of the mine-derived seepage is necessary before injection into the ground as part of the proposed MAR operations. However, no information is available on the kind of treatment proposed. This should be clarified, in order to ensure that i) treated mine-derived water will not deteriorate the local groundwater and that ii) no adverse geochemical reactions ensure when mixing between treated mine water and ambient groundwater occurs during MAR.

The supergene zone is held within healed fractures or had formed as growths locked up in a clay aquitard. The clay aquitard is not expected to drain and oxidise. Since the Tonkin (2016) investigation was conducted Terramin has conducted an IP survey and analysis of this data shows that there are no sulphides present outside of the know mineralisation zones.

Important Information

Your attention is drawn to the document – “Important Information”, which is included in Attachment B of this

report. The statements presented in this document are intended to advise you of what your realistic

expectations of this report should be. The document is not intended to reduce the level of responsibility

accepted by Golder Associates, but rather to ensure that all parties who may rely on this report are aware of

the responsibilities each assumes in so doing.



31

References

Adelaide and Mount Lofty Ranges Natural Resources Management Board 2013 Water Allocation Plan

Western Mount Lofty Ranges

AGT (2017), Bird-in-Hand Gold Project Groundwater Assessment. Prepared for Terramin Australia Pty Ltd. 19

June 2019

Banks, EW, Wilson, T, Green, G & Love, AJ 2006, Groundwater recharge investigations in the Eastern Mount

Lofty Ranges, South Australia, Report DWLBC 2007/20, Government of South Australia, through Department

of Water, Land and Biodiversity Conservation, Adelaide.

Barnett, S.R., van den Akker, J. and Zulfic, D., 2002. Mount Lofty Ranges Groundwater Assessment, Upper

Onkaparinga Catchment. South Australia. Department of Water, Land and Biodiversity Conservation. Report,

DWLBC 2002/29.

Barnett et al (2012), Australian groundwater modelling guidelines, Waterlines report, National Water

Commission, Canberra

Cook PG, Simmons CT and Wallis I (2019a) Bird-in-Hand gold mine. Review of groundwater characterisation

and modelling. NCGRT Report to Piper Alderman and Accolade wines. 1 February 2019.

Cook PG, Wallis I and Simmons CT (2019b) Bird-in-Hand gold mine. Review of managed aquifer recharge

trials and model validation. NCGRT Report to Piper Alderman and Accolade wines. 26 June 2019.

Golder (2019a), Bird-in-Hand Gold Project – Managed aquifer recharge investigation – Stage 1 drilling and

pumping tests. Prepared for Terramin Australia Pty Ltd. 31 May 2019.

Golder (2019b), Bird-in-Hand Gold Project – Investigation into managed aquifer recharge. Stage 2 injection

tests and Stage 3 groundwater model validation. Prepared for Terramin Australia Pty Ltd. 18 June 2019.

Gordon, C and Toze, S, (2003). Influence of groundwater characteristics on the survival of enteric viruses.

Journal of Applied Microbiology 95: 536-544.

Liesel, JT, (2009) Bugs in the water - Microbial ecology studies in the Upper Floridan and Biscayne Aquifers

Aquifer Storage and Recovery ASR 9 Presentations Orlando Florida 2009.

Lundh, Dr MMA, (2008) Swedish Experience on Pathogen Removal in Managed Aquifer Recharge for

Possible Application in New Zealand. Water resources management conference Rotorua

Martin R (2019). Bird in Hand Proposed Mine Managed Aquifer Recharge Review. Martin (2019) Report to

Inverbrackie Creek Catchment Group. 30 August 2019.

Stewart, S and Green, G (2010). GROUNDWATER FLOW MODEL OF COX CREEK CATCHMENT, MOUNT

LOFTY RANGES, SOUTH AUSTRALIA, May 2010

Toze, S, (2005) Water reuse and health risks – Real vs. Perceived. In: Khana, S., Muston, M. and Schafer, A.,

(Eds.). Integrated concepts in Water Recycling 2005.13-16 February 2005. Wollongong, VIC,

Toze, R, and Hanna, J, (2002). The survival potential of enteric pathogens in a reclaimed water ASR project.

in: Dillon, P, (ed), Proceeding of the 4th International symposium on artificial recharge of groundwater ISAR-4

- Management of Aquifer Recharge for Sustainability, 22-26 Sept 2002, Adelaide: Balkeme Publishers

Australian: 139–142


118 Franklin Street Adelaide, South Australia 5000 Australia

T: +61 8 8213 2100 +61 8 8213 2101

A.B.N. 64 006 107 857

Golder and the G logo are trademarks of Golder Associates Corporation golder.com


Jason van den Akker Eric van Nieuwkerk

Principal Hydrogeologist Senior Hydrogeologist

JvdA/EvN/ro

https://golderassociates.sharepoint.com/sites/19943g/deliverables/1897291-005-rev1-golder bih submission repsonses.docx

GOLDER ASSOCIATES PTY LTD IMPORTANT INFORMATION RELATING TO THIS REPORT

Error! Unknown document property name. Page 1 of 1 GAP Form No. LEG04 RL2

5/2018

The document (“Report”) to which this page is attached and which this page forms a part of, has been issued

by Golder Associates Pty Ltd (“Golder”) subject to the important limitations and other qualifications set out below.

This Report constitutes or is part of services (“Services”) provided by Golder to its client (“Client”) under and subject

to a contract between Golder and its Client (“Contract”). The contents of this page are not intended to and do not

alter Golder’s obligations (including any limits on those obligations) to its Client under the Contract.

This Report is provided for use solely by Golder’s Client and persons acting on the Client’s behalf, such as its

professional advisers. Golder is responsible only to its Client for this Report. Golder has no responsibility to any other

person who relies or makes decisions based upon this Report or who makes any other use of this Report. Golder

accepts no responsibility for any loss or damage suffered by any person other than its Client as a result of any

reliance upon any part of this Report, decisions made based upon this Report or any other use of it.

This Report has been prepared in the context of the circumstances and purposes referred to in, or derived from,

the Contract and Golder accepts no responsibility for use of the Report, in whole or in part, in any other context

or circumstance or for any other purpose.

The scope of Golder’s Services and the period of time they relate to are determined by the Contract and are subject to restrictions and limitations set out in the Contract. If a service or other work is not expressly referred to in this Report, do not assume that it has been provided or performed. If a matter is not addressed in this Report, do not assume that any determination has been made by Golder in regards to it.

At any location relevant to the Services conditions may exist which were not detected by Golder, in particular due to

the specific scope of the investigation Golder has been engaged to undertake. Conditions can only be verified at the

exact location of any tests undertaken. Variations in conditions may occur between tested locations and there may

be conditions which have not been revealed by the investigation and which have not therefore been taken into account

in this Report.

Golder accepts no responsibility for and makes no representation as to the accuracy or completeness of the

information provided to it by or on behalf of the Client or sourced from any third party. Golder has assumed that such

information is correct unless otherwise stated and no responsibility is accepted by Golder for incomplete or

inaccurate data supplied by its Client or any other person for whom Golder is not responsible. Golder has not taken

account of matters that may have existed when the Report was prepared but which were only later disclosed to

Golder.

Having regard to the matters referred to in the previous paragraphs on this page in particular, carrying out the

Services has allowed Golder to form no more than an opinion as to the actual conditions at any relevant location.

That opinion is necessarily constrained by the extent of the information collected by Golder or otherwise made

available to Golder. Further, the passage of time may affect the accuracy, applicability or usefulness of the opinions,

assessments or other information in this Report. This Report is based upon the information and other circumstances

that existed and were known to Golder when the Services were performed and this Report was prepared.

Golder has not considered the effect of any possible future developments including physical changes to any

relevant location or changes to any laws or regulations relevant to such location.

Where permitted by the Contract, Golder may have retained subconsultants affiliated with Golder to provide some or all of the Services. However, it is Golder which remains solely responsible for the Services and there is no legal recourse against any of Golder’s affiliated companies or the employees, officers or directors of any of them.

By date, or revision, the Report supersedes any prior report or other document issued by Golder dealing with any

matter that is addressed in the Report.

Any uncertainty as to the extent to which this Report can be used or relied upon in any respect should be

referred to Golder for clarification

Documents

Cook PG, Simmons CT and Wallis I Martin R (2019). Bird in