Recharge basin. Aquifer

Water for a Healthy Country Recharge basin Water table Aquifer Artificial Recharge Potential of the Perth Region Superficial Aquifer: Lake Preston to Moore River A. J. Smith and D. W. Pollock February 2010 Water Foundation, Western Australia

Water for a Healthy Country Flagship Report series ISSN: 1835 095X Australia is founding its future on science and innovation. Its national science agency, CSIRO, is a powerhouse of ideas, technologies and skills. CSIRO initiated the National Research Flagships to address Australias major research challenges and opportunities. They apply large scale, long term, multidisciplinary science and aim for widespread adoption of solutions. The Flagship Collaboration Fund supports the best and brightest researchers to address these complex challenges through partnerships between CSIRO, universities, research agencies and industry. The Water for a Healthy Country Flagship aims to achieve a tenfold increase in the economic, social and environmental benefits from water by 2025. The work contained in this report is collaboration between CSIRO and Western Australian Department of Water, Cockburn Sound Management Council, Town of Cottesloe, Peel Harvey Catchment Council, Water Corporation, Swan Catchment Council and City of Melville. For more information about Water for a Healthy Country Flagship or the National Research Flagship Initiative visit www.csiro.au/org/healthycountry.html Citation: Smith, A. J. and Pollock, D. W., 2010. Artificial recharge potential of the Perth region superficial aquifer: Lake Preston to Moore River. CSIRO: Water for a Healthy Country National Research Flagship. Copyright and Disclaimer c 2010 CSIRO To the extent permitted by law, all rights are reserved and no part of this publication covered by copyright may be reproduced or copied in any form or by any means except with the written permission of CSIRO. Important Disclaimers: CSIRO advises that the information contained in this publication comprises general statements based on scientific research. The reader is advised and needs to be aware that such information may be incomplete or unable to be used in any specific situation. No reliance or actions must therefore be made on that information without seeking prior expert professional, scientific and technical advice. To the extent permitted by law, CSIRO (including its employees and consultants) excludes all liability to any person for any consequences, including but not limited to all losses, damages, costs, expenses and any other compensation, arising directly or indirectly from using this publication (in part or in whole) and any information or material contained in it. The views expressed are not necessarily the views of the Government of Western Australia, nor the Water Foundation. Cover Image: Basin recharge. c 2010 CSIRO

Contents Acknowledgements Executive Summary v vi 1 INTRODUCTION 1 1.1 Objectives and scope............................... 2 1.2 Study area.................................... 3 1.3 Previous work.................................. 3 1.4 Approach of this study.............................. 3 1.5 Surface infiltration and well injection...................... 5 1.6 Hydraulic load.................................. 5 1.7 Operational scale................................. 7 2 METHODS 13 2.1 Derived datasets................................. 13 2.1.1 Ground surface elevation model..................... 13 2.1.2 Base of superficial aquifer elevation model............... 13 2.1.3 Annual mean watertable surfaces.................... 14 2.1.4 Depth to watertable below ground surface............... 14 2.1.5 Aquifer saturated thickness....................... 15 2.1.6 Aquifer transmissivity.......................... 15 2.1.7 Aquifer storage coefficient........................ 15 2.1.8 Soil and basin infiltration rates..................... 16 2.2 Assessment of aquifer storage change...................... 16 2.2.1 Linear regression modelling of water level trends............ 17 2.2.2 Trend of annual mean watertable elevation............... 17 2.2.3 Trend of annual mean freshwater thickness............... 17 2.3 Assessment of artificial recharge potential.................... 18 2.3.1 Theis (1935) well model......................... 18 2.3.2 Glover (1960) basin model........................ 21 3 RESULTS 29 3.1 Derived datasets................................. 29 3.2 Aquifer storage change.............................. 30 3.3 Saltwater intrusion................................ 30 3.4 Well recharge model............................... 30 3.5 Basin recharge model.............................. 31 4 DISCUSSION 40 4.1 Hydraulic feasibility of artificial recharge.................... 40 4.2 Recharge water sources............................. 42 4.3 Management of aquifer storage and watertable elevation............ 43 4.4 Management of saltwater intrusion....................... 44 4.5 Limitations of the study............................. 45 5 CONCLUSIONS 49 Artificial Recharge Potential of Perth s Superficial Aquifer Page iii

List of Figures 1.1 Perth metropolitan water supply......................... 2 1.2 Study area and location map.......................... 4 1.3 Hydraulic loads for basin and well recharge facilities.............. 9 1.4 Infiltration rates and basin areas for example facilities............. 10 2.1 Definition sketch for spatial datasets...................... 24 2.2 Examples of the outlier analysis using a z score test.............. 24 2.3 Calculation of aquifer transmissivity....................... 25 2.4 Examples of linear regression models fitted to the annual watertable elevations 25 2.5 Potential change of freshwater thickness.................... 26 2.6 Artificial recharge through an injection well................... 26 2.7 Rising and spreading recharge cones from the Theis well equation....... 27 2.8 Artificial recharge beneath a rectangular basin................. 27 2.9 Rising and spreading recharge mounds from the Glover basin equation..... 28 3.1 Elevation surfaces................................ 32 3.2 Aquifer saturated thickness........................... 33 3.3 Aquifer hydraulic conductivity.......................... 34 3.4 Aquifer transmissivity and storage coefficient distributions........... 35 3.5 Soil infiltration rate distribution......................... 36 3.6 Ten year change of aquifer storage 1998 2007................. 37 3.7 Relative Watertable Rise at 30 days for well injection.............. 38 3.8 Relative Watertable Rise at 30 days for basin infiltration............ 39 4.1 Opportunities for artificial recharge....................... 47 4.2 Aquifer response to well injection........................ 48 List of Tables 1.1 Site suitability for managed surface infiltration................. 6 1.2 Site suitability for well injection......................... 6 1.3 Basin recharge rates worldwide......................... 8 1.4 Well recharge rates................................ 8 1.5 Tabulated data for recharge basins in Figure 1.3 and Figure 1.4........ 11 1.6 Tabulated data for recharge wells in Figure 1.3................. 12 2.1 Assumed soil properties............................. 16 2.2 Simulated well injection rates.......................... 20 2.3 Well yields for the Perth region superficial aquifer [8].............. 21 2.4 Simulated basin hydraulic loads......................... 23 4.1 Discharges from Perth s major wastewater treatment plants.......... 43 4.2 Equivalent area of watertable rise for total volume 340 ML/d and porosity 0.2 44 Artificial Recharge Potential of Perth s Superficial Aquifer Page iv

Acknowledgements This work was funded by CSIRO, the Western Australian Water Foundation and in kind contributions from the Western Australian Department of Water (DoE), Cockburn Sound Management Council (CSMC), Town of Cottesloe (ToC), Peel Harvey Catchment Council (PHCC), Western Australian Water Corporation (WC), Perth Region NRM (PRNRM) and City of Melville (CoM). Oversight and feedback on work in progress was provided by members of the project Reference Group: Chris O Boy (DoW) Dr. Tom Rose (CSMC) Geoff Trigg (ToC) Damien Postma (PHCC) Patricia Pedelty (PRNRM) Dr. Bruce Hamilton (PRNRM) James Tay (WC) The draft report was kindly reviewed by Michael Martin from the Western Australian Water Corporation, and Dr. Cahit Yesertener from Department of Water, Western Australia. Jeff Major and Karen Barlow from Department of Water provided valuable assistance with contract management and project coordination on behalf of the Western Australian Water Foundation. Artificial Recharge Potential of Perth s Superficial Aquifer Page v

EXECUTIVE SUMMARY Analytic models of watertable response to recharge wells and basins were applied spatially across the Perth region from Lake Preston to Moore River to assess the hydraulic potential for artificial recharge into the superficial aquifer. The methodology considered the aquifer properties that control groundwater recharge and flow, the likely watertable response to prescribed recharge, and the available space in the aquifer to store the recharge. Results were compared with a separate analysis of storage decline in the aquifer since 1998 to identify areas where artificial recharge is both hydraulically viable and may provide a community benefit. The quality of the source water for artificial recharge was unspecified and the analysis did not consider economic, engineering or social aspects of artificial recharge, other than in very broad terms. The coastal limestone, extensive parts of Gnangara Mound and some parts of Jandakot Mound were found to be suitable for small, medium and large scale artificial recharge operations. The presence of sandy surface soils and moderate to very large aquifer transmissivity allow large infiltration and injection rates and promote lateral spreading of recharge mounds rather than excessive vertical rise toward ground surface. Extensive inland areas south of the Swan Canning Estuary were assessed to be unsuitable for medium and large scale artificial recharge operations. Groundwater is relatively shallow in these areas and the soil and aquifer have greater clay and silt contents that restrict infiltration, injection and lateral groundwater flow. These characteristics promote excessive vertical growth of recharge mounds toward ground surface. Although hydraulic attributes of the coastal limestone make large scale artificial recharge a feasible proposition, and there is opportunity to mitigate saltwater intrusion, the potential risk to groundwater users and the environment due to preferential flow and poor attenuation of contaminants is inadequately understood. This will contribute substantial uncertainty in appraisals of artificial recharge operations that involve treated wastewater and untreated stormwater. The risk of preferential flow in the aquifer is diminished further inland within sandy parts of Gnangara Mound, and the hydraulic conditions allow better control over local watertable elevation and aquifer storage; nevertheless, opportunity for artificial recharge is limited by the lack of water sources. The land surface is elevated relative to the coastal strip and inland conveyance of water from the coastal wastewater treatment plants would require new infrastructure and ongoing pumping costs. Irrespective of the regional patterns identified in this study, artificial recharge may be able to address a specific need if the local hydraulic characteristics are suitable and a secure water source is available. For this reason, the results of this study should not be interpreted in a way that suggests artificial recharge is either possible or impossible at a particular site. Artificial Recharge Potential of Perth s Superficial Aquifer Page vi

1 INTRODUCTION Significant reduction of rainfall across the south west of Western Australia, rapid population growth, and increasing urbanisation of metropolitan Perth has added considerable stress on the region s water resource. The impacts of less rainfall and increasing water demand have been evident as declining stores of surface water and groundwater within the Perth region for the past several decades. Current water planning is proceeding on the basis that reduced water yields from traditional sources may be permanent. Strategies for meeting future demand include the development of new groundwater supplies from un tapped sources, seawater desalinisation, reducing per capita consumption, and improved water use efficiency and water recycling. A principal aim of water recycling is to retain more water for longer in the local water supply cycle. In theory, this provides more opportunity for water reuse and improves the security of the water resource. Recycling also helps to maintain greater storage levels within the water supply system, which can lessen potential impacts on dependent ecosystems that are sensitive to water level change. Nevertheless, the amount of water that is practical to recycle can be small compared to the total resource use. The Perth region is critically reliant on groundwater for both drinking water and non potable uses Figure 1.1. Enhancing groundwater storage in the aquifer system through artificial recharge is both a practical and necessary strategy for achieving water recycling goals. Groundwater currently supplies around sixty percent of potable water use in metropolitan Perth [47] and roughly eighty five percent of the total water demand for all uses. It is estimated that the aquifer system supplied in the order of 400 500 gigalitre (GL) of groundwater during water year 2008 09. 1 This compares to around 56 GL supplied from surface water and 43 GL supplied from desalinated seawater. The major uses of groundwater in the Perth region include licensed withdrawals by industry, growers and local government authorities; extractions by Water Corporation for public drinking water supply; and unlicensed pumping from private garden bores. The main sources of water that are available for artificial recharge of aquifers in Perth include stormwater, treated wastewater and groundwater drainage. The Government of Western Australia [10] has stated that it encourages proposals for suitable artificial recharge activities but it cautions that artificial recharge will not be feasible at all sites due to hydrogeological, environmental, or cost limitations. The government has adopted a risk management approach for assessing artificial recharge proposals, which seeks to balance benefit and risk. Specific reasons for establishing artificial recharge activities at particular sites can be varied; however, one of the following hydrological objectives must be achieved to attain the desired economic, social or environmental benefits: (1) controlled manipulation of the aquifer storage, or (2) controlled manipulation of watertable elevation. For instance, to enhance groundwater supply by artificial recharge requires an aquifer in which local storage can be controlled in the desired way and by practical means. Similarly, to manage saltwater intrusion through artificial recharge, or impacts of watertable decline on wetland ecosystems requires an aquifer in which groundwater level can be manipulated within the required elevation range, over the required extent of aquifer, and at the locations where this need exists. 1 One gigalitre is one thousand million litres, which is equivalent to a cube of water measuring 100 m long by 100 m wide by 100 m high. Artificial Recharge Potential of Perth s Superficial Aquifer Page 1

The degree of hydrological manipulation that can be achieved through artificial recharge activities, and the associated benefits and risks, depends fundamentally on the aquifer hydraulic properties, hydrological conditions such as depth to watertable below ground surface, and the rate at which water can be injected or infiltrated. Spatial datasets describing how these factors vary across the Perth region have been developed during the past few years for several large scale modelling studies. An opportunity now exists to assess artificial recharge quantitatively across the Perth region making use of this information. 120 GL Unlicensed private groundwater (21%) 217 GL Licensed private groundwater (37%) 43 GL Desalinated sea water (7%) 56 GL Surface water (10%) 146 GL Public groundwater (25%) Figure 1.1: Perth metropolitan water supply 1.1 Objectives and scope The purpose of this study was to assess the hydraulic feasibility of artificial recharge of the superficial aquifer across the Perth region by developing a quantitative method to estimate aquifer response to various recharge rates. Specific tasks that were undertaken included: 1. development of a quantitative procedure for simulating local aquifer responses to artificial recharge activities, including well injection and basin infiltration, which could be applied spatially across the study area, 2. preparation of suitable spatial datasets of aquifer properties and hydrological conditions across the Perth region for input into the modelling of aquifer response, 3. analysis of change of aquifer storage during the past ten years to identify areas where artificial recharge could mitigate impacts of watertable decline, and 4. application of the procedure and analysis of the results to provide spatial maps of artificial recharge feasibility. Water quality aspects of artificial recharge were not assessed in this study. Determining potential effects of water quality on physical, chemical and biological processes that can influence infiltration, injection and groundwater flow across the entire Perth region was beyond the scope of this work. Artificial Recharge Potential of Perth s Superficial Aquifer Page 2

1.2 Study area This study was conducted in the area extending from the coast to Darling Scarp, and from approximately 17 kilometres (km) north of Guilderton at the mouth of Moore River to around 25 km south of Harvey Estuary, which forms an area of approximately 5,334 km 2. A location map is presented as Figure 1.2. For the purposes of this report, the study area is referred to as the Perth region; Lake Preston to Moore River. The physiography, geology and hydrogeology of the study area are described in [8] and [9]. 1.3 Previous work The review by Scatena and Williamson [37] presents a broad assessment of the potential role for artificial recharge across the Perth region from Mandurah to Moore River. The report provides comprehensive information about the principles and objectives of artificial recharge, the available technology options, physical and chemical factors controlling the performance of artificial recharge operations, regulatory and statutory requirements that prevailed at that time, and the physiography of the Perth region. Particular emphasis was placed on water quality aspects of artificial recharge, which included potential chemical and physical effects of recharge water quality on infiltration and groundwater flow, as well as the degree of water quality improvement achievable in the aquifer. A simple hydraulic evaluation of the suitability of the superficial aquifer for artificial recharge was conducted based on depth to groundwater and aquifer transmissivity. Areas with depth to groundwater greater than 6 metres (m) and aquifer transmissivity greater than 600 m 2 /day were considered to be potentially suitable for artificial recharge operations. Six potential artificial recharge schemes were identified for further feasibility assessment, which led to the Mosman reclaimed water prefeasibility study [34]. 1.4 Approach of this study The present study applies a quantitative approach to complement and extend the qualitative assessment by Scatena and Williamson. Simple analytic models of well injection and basin infiltration are applied spatially across the study area to predict the vertical and lateral growth of recharge mounds in response to variable hydraulic loads. The models involve a number of simplifying assumptions that are reasonable for a regional scale feasibility assessment. Improved spatial datasets and updated estimates of the aquifer hydraulic properties distributions were derived from recent regional groundwater modelling datasets, e.g., [9] and [45]. Scatena and Williamson [37] used a minimum depth to groundwater of 6 m to identify areas potentially suitable for artificial recharge on the basis that this depth would permit greater time for percolation to the watertable and attenuation of contaminants. A specific constraint on depth to groundwater was not applied in the present study. Spatial maps of annual mean depth to groundwater were prepared and used to assess the limit on the vertical growth of the modelled recharge mounds. The effect of spatial variation of aquifer transmissivity on artificial recharge potential was incorporated into the analytic models, which used the aquifer hydraulic properties and groundwater recharge rate as inputs. Artificial Recharge Potential of Perth s Superficial Aquifer Page 3

GUILDERTON! GINGIN! Moore-Hill Rivers QUINNS ROCKS! BULLSBROOK! TRIGG! Swan Coast PERTH! ROTTNEST ISLAND! FREMANTLE!! ARMADALE GARDEN ISLAND! F SAFETY BAY! 0 20 SERPENTINE! km! MANDURAH Murray River! PINJARRA! PRESTON BEACH WAROONA! Harvey River Legend! Locality Road Surface drainage Wetland Urban area Surface water management area Regolith terrestrial sediments very highly weathered bedrock Figure 1.2: Study area and location map Artificial Recharge Potential of Perth s Superficial Aquifer Page 4

1.5 Surface infiltration and well injection Artificial recharge of groundwater is accomplished using two basic techniques: 1. surface water infiltration from basins, ponds, trenches and other water holding structures to unconfined aquifers, and 2. water injection through wells into unconfined, semi confined and confined aquifers using gravity or pressure systems. Generalised criteria for assessing the suitability of a particular site for surface infiltration and well injection are presented in Table 1.1 and Table 1.2. Though some rule of thumb design principles are recognised, it is generally acknowledged that detailed field investigation and testing is required to properly assess the feasibility of artificial recharge at particular sites. The question Are there any blueprints for artificial recharge? was addressed in 1996 at an international conference in Finland [33]. It was concluded that, even within small countries, similarities between artificial recharge facilities are small and contrasts large; leading to limited value in standard designs. Surface infiltration is generally a more cost effective method of artificial recharge, provided that adequate land is available and the vadose zone hydrogeology is favourable for vertical percolation. Most large scale artificial recharge facilities use surface infiltration. The infiltration rate will depend on the depth of recharge water, permeability of the vadose zone, and clogging of the filtration surface due to physical sedimentation, binding by gas bubbles and geochemical and biological activity. Hatva [20] suggested that soil permeability in the range 0.001 0.01 m/s (86 864 m/d) is necessary for successful surface infiltration, which is typical of coarse sands and sand gravel mixes [2]. Most infiltration facilities are operated based on empirical judgements whereby basins and galleries are dried out and rejuvenated when the infiltration rate declines. This can involve removal of built up sediments and re grading of the filtration surface. Well injection is the only option for artificial recharge of deep confined aquifers, and it can be used to recharge shallower semi confined and unconfined aquifers. Pre treatment of the source water to a high standard is required to minimise clogging problems. Well efficiencies can decline rapidly when the source water contains suspended materials that clog the injection well screen, well pack or the aquifer formation. Experience around the world has shown that aquifers can be successfully recharged over long periods without significant clogging using high quality source water that is injected into secondary solution features and fractures in rock aquifer systems, e.g., [5]. As a general guide, existing large yield production wells are a good indication of suitable conditions for well injection. Regardless of site and operational conditions, periodical cleaning and redevelopment of wells is normally required to maintain the injection system efficiency. 1.6 Hydraulic load Artificial recharge capacity is normally measured as the amount of water that can be added to an aquifer during a specified period. This quantity is often referred to as the hydraulic load or loading rate. Care is required when comparing reported values of hydraulic load for different artificial recharge facilities. A particular value might be calculated based on a relatively short operational period, or it might represent the average long term performance of the facility, inclusive of non operational periods required for maintenance and rejuvenation Artificial Recharge Potential of Perth s Superficial Aquifer Page 5

Table 1.1: Site suitability for managed surface infiltration Favourable conditions Unconfined aquifer with sufficiently thick vadose zone. Vadose zone with large permeability and absence of flow restricting layers. Adequate areal extent of permeable soils. Sufficient aquifer transmissivity to prevent excessive watertable mounding, e.g., deep, wide conductive layers. Source water that is chemically compatible with the receiving groundwater. Source water and vadose zone that are free of contaminants. Unfavourable conditions Aquifers in which the basin bottom cannot be excavated into permeable strata. Existence of layers with small permeability within the vadose zone. Small aquifer transmissivity leading to excessive watertable mounding. Contaminants in the source water or vadose zone. Shortage or inhibitive expense of land area. Table 1.2: Site suitability for well injection Favourable conditions Confined or semi confined aquifer. Unconfined aquifer with confining layers in the vadose zone. Existing large yield production wells. Sufficient transmissivity to prevent excessive pressure at injection bores. Secondary permeability features and fractures to target injection bores. Good quality source water with low clogging potential (chemical, physical and biological). Source water that is chemically compatible with the receiving groundwate. Unfavourable conditions Shallow, unconfined aquifer. Existing low yield production wells. Small aquifer transmissivity leading to excessive well pressure. Poor quality source water that may cause clogging of the well or aquifer formation. Source water that might cause groundwater contamination. Artificial Recharge Potential of Perth s Superficial Aquifer Page 6

of filtration surfaces. Large differences in reported hydraulic loads can also reflect differences in the scale of operation, the available water supply, aquifer characteristics, technology options and adoption, and the objectives of the artificial recharge operations. Examples of indicative infiltration and injection rates achieved at operation facilities around the world are listed in Table 1.3 and Table 1.4. The hydraulic load that can be achieved by surface infiltration is controlled by the infiltration rate, the area of infiltration, and the hydraulic properties of the underlying aquifer. A number of factors can influence the infiltration rate and how it changes with time, including soil permeability, water quality, basin water depth, hydraulic gradient and bed clogging, e.g., [21]. The capacity of injection wells is largely determined by the aquifer transmissivity, well construction, water quality and well and formation clogging. These factors control the rate at which water can be injected and transmitted through the aquifer, and the height and extent of the resulting recharge cone. The hydraulic load of an injection well field depends on the total number of injection wells and their respective capacities, and the overall watertable response reflects the superposition of the individual well responses. 1.7 Operational scale Figure 1.3 and Figure 1.4 present examples of hydraulic loads achieved at artificial recharge facilities in Perth, Australia and internationally. 2 The scale of these operations extends across five orders of magnitude from less than 0.1 megalitres per day (ML/d) to greater than 100 ML/d. One possible classification of the operational scale is indicated in Figure 1.3, as follows: small scale: hydraulic load < 1 ML/d medium scale: hydraulic load 1 10 ML/d large scale: hydraulic load > 10 ML/d Figure 1.4 depicts the relationship between average infiltration rate, area of infiltration and hydraulic load. A large hydraulic load can be achieved in a relatively small area only if the soil permeability is large enough to allow rapid infiltration and the underlying aquifer can conduct the recharge water laterally to prevent excessive watertable mounding, e.g., San Jacinto. Large recharge rates are also achieved using large infiltration areas, e.g., Sand Hollow, Leaky Acres and Yavne. The three largest injection loads in Figure 1.3 were achieved using different strategies, including: thirteen wells at Enfield Haringey injecting 48 ML/d into a fissured chalk aquifer with transmissivity up to 1,300 m 2 /d; two wells in the Las Vegas Valley injecting 22 ML/d into a cemented sand, gravel and calcrete aquifer with transmissivity of approximately 2,000 m 2 /d; and three hundred stormwater gravity drainage wells injecting 7.7 ML/d into a karst limestone aquifer with estimated transmissivity 20,000 m 2 /d. In comparison to these scales of artificial recharge operation, the daily discharge rates from Perth s three major wastewater treatment plants are: Subiaco 60 ML/d, Beenyup 41 ML/d and Cape Peron 48 ML/d. 2 Tabulated data are in Table 1.5. The Mosman injection wells were proposed in 2004 but were not implemented; there are no operational well injection schemes in the superficial aquifer in Perth. Artificial Recharge Potential of Perth s Superficial Aquifer Page 7

Table 1.3: Basin recharge rates worldwide Country or region Infiltration rate, m/d Reference World 0.08 0.8 [3] Finland 0.7 2.9 [20] Sweden 0.9 4.6 [16] Sweden 1 3 [23] Sweden (13 cities) 1.5 16 [43] California 0.6 [38] Ntherlands 0.15 0.46 [16] Netherlands 0.05 0.5 [32] USA (14 regions) 0.12 2.9 [43] USA 0.03 61 [16] Table 1.4: Well recharge rates Country or region Infiltration rate, m/d Reference USA (12 regions, pre 1960) 2.8 65 [43] USA (Florida, pre 1960) 5.7 595 [43] Australia 0.3 3.1 [12] Artificial Recharge Potential of Perth s Superficial Aquifer Page 8

200 Recharge basins Perth Australia International Recharge wells Perth Australia International Yavne (Israel) 100 80 60 40 20 Leaky Acres (USA) Sand Hollow (USA) Soreq (Israel) Kariboe Weir (QLD) San Jacinto (USA) Enfield-Haringley (UK) Las Vegas Valley (USA) Large scale Opth. river (WA) Hydraulic load, ML/d 10 8 6 4 2 1 0.8 0.6 0.4 0.2 Opth. ponds (WA) Gordon Rd. Redbank Ck. (QLD) O Reilly Weir (QLD) Halls Head Flagstone Ck. (QLD) Three Moon Ck. (QLD) Arrenaes (Denmark) Caddadup Yanchep East Meadow (USA) Sedgwick (USA) Boundary Ck. (VIC) Ma Ma Ck. (QLD) Mt. Gambia (SA) *Mosman proposed (WA) Angus-Bremer (SA) Cape May (USA) East Meadow (USA) West St. Paul (USA) Greenfields (SA) The Paddocks (SA) Clayton (SA) Andrews Farm (SA) Medium scale Small scale 0.1 0.08 0.06 Two Rocks Northfield (SA) Torrens Park (SA) Figure 1.3: Hydraulic loads for basin and well recharge facilities Artificial Recharge Potential of Perth s Superficial Aquifer Page 9

1000 100 0.01 0.1 1 10 Average infiltration rate, m/d 1000 100 10 1 0.1 0.01 1E-3 Sand Hollow Yavne Leaky Acres Soreq Opth. river basins Opth. ponds Caddadup Gordon Rd. Halls Head Arrenaes San Jacinto HYDRAULIC LOADING ML/d East Meadow Sedgwick 100 Yanchep Perth Australia International 0.1 5 2 0.5 Floreat gallery 50 20 10 1 10 1 0.1 Total area, ha 0.01 1E-3 Figure 1.4: Infiltration rates and basin areas for example facilities Artificial Recharge Potential of Perth s Superficial Aquifer Page 10

Table 1.5: Tabulated data for recharge basins in Figure 1.3 and Figure 1.4 Facility name and location Total area of infiltration basins, ha Hydraulic load, ML/d Average infiltration rate, m/d Arrenæs, Denmark 0.6 0.7 0.11 [31] Reference Caddadup, Perth 1.0 0.53 0.05 Pers. Com. 2005 East Meadow, USA 0.3 6.0 1.8 [22] Gordon Road, Perth 4.0 6.0 0.15 Pers. Com. 2005 Halls Head, Perth 1.6 2.3 0.14 Pers. Com. 2005 Leaky Acres, USA 47 83 0.17 [28] Opthalmia ponds, WA 11 6.7 0.06 [6] San Jacinto, USA 1.6 31 1.9 [26] Sand Hollow, USA 530 64 0.012 [21] Sedgwick, USA 0.18 4.7 2.6 [36] Soreq, Israel 24 39 0.16 [24] Yanchep 0.02 0.19 1.0 Pers. Com. 2005 Yavne, Israel 42 139 0.33 [24] Artificial Recharge Potential of Perth s Superficial Aquifer Page 11

Table 1.6: Tabulated data for recharge wells in Figure 1.3 Facility name and location Number of wells Hydraulic load, ML/d Average injection rate, L/s Andrews Farm, SA 1 0.16 1.9 [12] Angus Bremer, SA 30 2.7 1.1 [12] Cape May County, USA 4 1.58 4.6 [25] Clayton, SA 1 0.19 2.2 [12] East Meadow, USA 5 0.7 1.7 [22] Enfield Haringley, UK 13 48 42.7 [30] Greenfields, SA 1 0.27 3.1 [12] Las Vegas Valley, USA 2 22 126 [4] Mosman, Perth 4 4 12 [34] Mt. Gambia, SA 300 7.7 0.3 [12] Northfield, SA 1 0.11 1.3 [12] The Paddocks, SA 1 0.22 2.5 [12] Torrens Park, SA 1 0.11 1.3 [12] West St. Paul, USA 1 0.54 6.3 [35] Reference Artificial Recharge Potential of Perth s Superficial Aquifer Page 12

2 METHODS In this study, analytic models of watertable response to recharge wells and basins were applied spatially across the Perth region to assess the potential for artificial recharge into the superficial aquifer. The analytic models were applied in each 25 hectare (ha) grid cell across the study area, corresponding to 21,335 locations. The ensemble results for a particular recharge method and hydraulic load were then combined to produce maps of aquifer response and suitability. A complimentary assessment of change of aquifer storage during the past ten years was conducted by estimating the change and trend in freshwater thickness in the superficial aquifer during the period 1998 to 2007. For the purpose of applying the analytic models, the study area was discretised into 500 500 m (25 ha) contiguous cells, wherein average homogeneous aquifer and soil properties were defined. Datasets were then assembled in spatially concurrent raster formats and incorporated into a project geodatabase, which is accessible using Microsoft Access and ArcGIS software applications. A definition sketch for the spatial datasets is presented in Figure 2.1, and greater detail of the analysis methods that were employed in the study is presented in the following sub sections. 2.1 Derived datasets 2.1.1 Ground surface elevation model A digital elevation model of ground surface elevation across the study area was constructed from the following pre existing information sources: Digital topographic data 1:2,000 cultural and relief (June 2004) 3 Digital topographic data 1:25,000 relief (June 2004) 4 Australian Bathymetry and Topography Grid (June 2005) 5 A Triangular Irregular Network (TIN) for the Perth metropolitan area was constructed using the 1 m elevation contours from the 1:2,000 digital topographic data. A separate TIN covering the full extent of the study area was constructed using the 5 m elevation contours and spot heights from the 1:25,000 digital topographic data. The Australian Bathymetry and Topography Grid was constrained to be less than or equal to zero m elevation within the study area. These three surfaces were then combined by giving preference to the highest resolution surface within overlapping areas. Thus, first preference was given to the 1:2,000 topographic data, second preference to the 1:25,000 topographic data, and third preference to the Australian Bathymetry and Topography Grid. 2.1.2 Base of superficial aquifer elevation model A digital elevation model of the bottom surface of the superficial aquifer was constructed from the following data sources: PRAMS Perth Regional Aquifer Modelling System [9] PHRAMS Peel Harvey Regional Aquifer Modelling System [45] 3 https://www2.landgate.wa.gov.au/interragatorplus/metadata/f529aac9-dae7-a5cb-427d-95c6bc1ecce3.html 4 https://www2.landgate.wa.gov.au/interragatorplus/metadata/b4e6c25f-3f10-77bd-9170-08720614235d.html 5 http://www.ga.gov.au/meta/anzcw0703008022.html Artificial Recharge Potential of Perth s Superficial Aquifer Page 13

The data were combined by giving preference to the PHRAMS layers in a small area of overlap at the southern boundary of PRAMS and northern boundary of PHRAMS. 2.1.3 Annual mean watertable surfaces Datasets representing the estimated average elevation of the watertable for a particular year were constructed for each year between 1998 and 2007. Water level elevation records for monitoring wells and piezometers held within the WIN (Water INformation) database were supplied to CSIRO by the Western Australia Department of Water. Time series from a total of 933 sites were analysed using the following procedure. Outlier analysis was performed on each time series record to remove incongruent data that could introduce bias in the mean and linear regression analyses. Outliers were identified using a z score test [27]. A third order polynomial was fitted to the raw water level data and a population of residuals was calculated as the differences between the data points and fitted polynomial. All data points with a residual greater than three standard deviations of the residuals population were classified as potential outliers and were excluded from further analyses. Examples are illustrated in Figure 2.2. The raw water level records in each time series were aggregated to annual average values to minimise seasonal effects, and to minimise temporal bias that is introduced by irregular sampling. Water level records from monitoring wells were sequentially aggregated and averaged at daily, monthly, half yearly and yearly time periods [41]. Daily average groundwater levels were first calculated for all days with measured water levels; the daily means were then averaged for all months containing days with daily means; and so on. The final annual averages were considered to provide reasonable point estimates of the annual mean watertable elevation for that year. An initial representation of the watertable surface was generated for each year as an TIN using the derived annual mean point data, and by constraining groundwater level to mean sea level of -0.059 m Australian Height Datum (AHD) along the coastal and estuarine margins of the superficial aquifer. 6 The result was then intersected with the ground surface elevation model to identify locations were the watertable was above the base of river and stream beds. Additional constraints were then added along those sections of the surface drainage by setting watertable elevation equal to stream bed elevation. The final mean annual watertable surface was interpolated as a TIN using the annual mean point data, the coastal and estuarine sea level constraints, and the stream bed elevation constraints. The TIN was then converted to raster format for application of the well and basin recharge analytic models. 2.1.4 Depth to watertable below ground surface A discretised spatial dataset, representing the annual mean watertable depth below ground surface in 2008, was constructed by subtracting the mean annual watertable elevation model for 2008 from the ground surface elevation model. 6 Mean sea level determined as local mean sea level of 0.704 m at Hillarys for period 1998 2008 (http://www.bom.gov.au/ntc/ido71062/ido71062sld.txt) minus the Hillarys elevation datum of 0.763 m AHD (http://www.bom.gov.au/oceanography/projects/abslmp/data/data.shtml). Artificial Recharge Potential of Perth s Superficial Aquifer Page 14

2.1.5 Aquifer saturated thickness A discretised spatial dataset, representing the annual mean total saturated thickness of the superficial aquifer in 2008, was constructed by subtracting the base of superficial aquifer elevation model from the annual mean watertable elevation model for 2008. 2.1.6 Aquifer transmissivity A dataset representing the spatial distribution of aquifer transmissivity was constructed from layer thicknesses and calibrated hydraulic conductivity distributions in PRAMS [9] and PHRAMS [45]. Because the superficial aquifer was represented in PRAMS and PHRAMS by two layers with different hydraulic property distributions, the following relationships were used to calculate single, vertically integrated values of aquifer transmissivity and hydraulic conductivity within each raster cell (Figure 2.3) T = K 1 B 1 +K 2 B 2 where: K = T (B 1 +B 2 ) T is aquifer transmissivity of the superficial aquifer, K is the vertically integrated value of hydraulic conductivity, K 1 is hydraulic conductivity of the upper layer representing the superficial aquifer, K 2 is hydraulic conductivity of the lower layer representing the superficial aquifer, B 1 is saturated thickness of the upper layer representing the superficial aquifer and B 2 is saturated thickness of the lower layer representing the superficial aquifer. This relationship assumes that piezometric head in the lower layer is identical to watertable elevation in the upper layer (i.e., the vertical hydraulic gradient between the layers is negligible). In that situation, applying Darcy s Law to calculate total lateral groundwater flow per unit hydraulic gradient would yield the same flow value using either the integrated transmissivity value T or the individual layer transmissivity values T 1 = K 1 B 1 and T 2 = K 2 B 2. If vertical head gradients between the layers are relatively small, then using T also provides a reasonable approximation of total flow through the aquifer. The analytic models used in this study to predict watertable change in response to well injection and basin infiltration required a single value of aquifer transmissivity as input; therefore, T was calculated for each cell in the study area and that value was used to apply the models. 2.1.7 Aquifer storage coefficient A spatial dataset representing the superficial aquifer storage properties was constructed from the calibrated storage coefficient distributions in PRAMS [9] and PHRAMS [45]. The analytic models of basin and well recharge required a single value of the aquifer storage coefficient to represent the aquifer fillable porosity above the watertable. Values of specific yield from the upper most layers representing the superficial aquifer in PRAMS and PHRAMS were adopted for this purpose (i.e., S 1 in Figure 2.3). Artificial Recharge Potential of Perth s Superficial Aquifer Page 15

Table 2.1: Assumed soil properties Soil texture class Hydraulic conductivity, m/d sand 15 1 limestone and sand 15 1 sand with clay, silt and peat 5 0.3 Infiltration rate, m/d clay and silt 0.05 0.003 2.1.8 Soil and basin infiltration rates The recharge basin analytic model assumes a constant value of the infiltration rate through the bottom of the basin, which can be considered to represent the time averaged mean. Surface geology of the Perth region is variable and there is limited data to assess the magnitude of sustained infiltration that is achievable using recharge basins across the study area. Scatena and Williamson [37] (p. 33) reported results from a recharge study in Canning Vale during the 1980s, which achieved infiltration of secondary treated wastewater at a rate of approximately 0.5 m/d into Bassendean Sand. Operational facilities in sandy areas of Perth also can achieve infiltration rates of around 1 m/d (Figure 1.4). Physical and empirical based infiltration equations have been developed for the purpose of estimating surface percolation rates, e.g., [2] (p. 252) but to be reliable they require detailed field measurements of local soil properties, which are typically undetermined at regional scales. In this study, the following simplified approach was used to assign indicative infiltration rates based on soil texture and associated saturated hydraulic conductivity values. Four soil texture class distributions (Table 2.1) were developed from 1:50,000 environmental geology data (http://mapserver.doir.wa.gov.au/datacentre/). The generalised soil texture classes were: (1) sand, (2) limestone and sand, (3) sand with various amounts of clay silt and peat, and (4) predominantly clay and silt. 7 Representative hydraulic conductivity values were assigned to each soil texture class based on currently applied values for the Swan Coastal Plain sediments, e.g., [9]. Sandy areas were assumed to be capable of achieving an average continuous infiltration rate of 1 m/d. Infiltration rates for the other soil texture classes were then derived based on their hydraulic conductivity values relative to sand. For example, if the hydraulic conductivity was one third the value of sand then the infiltration rate was similarly set equal to one third the value of sand. 2.2 Assessment of aquifer storage change Spatial datasets representing the linear trend in aquifer storage during the past ten years (1998 2007) were constructed from the annual mean watertable elevations derived in this study. The estimated change of storage was represented as both the linear trend of the annual mean watertable elevation, and the linear trend of the annual mean freshwater thickness in the aquifer. The two datasets differed only at the coast where saltwater intrusion is possible. A simple hydrostatic approximation was used to estimate the potential change of elevation of the saltwater freshwater interface in response to change of watertable elevation. 7 Typical literature values are: medium sand 5 20 m/d; loam soils 0.1 1 m/d; clay soils 0.01 0.2 m/d [2]. Artificial Recharge Potential of Perth s Superficial Aquifer Page 16

2.2.1 Linear regression modelling of water level trends For each water level time series record, linear regression models were fitted to the annual mean watertable elevations derived from the raw water level records. The slopes of the linear models were considered to provide reasonable estimates of the trend in average groundwater level during 1998 to 2007. Examples of the raw water level records from the WIN database, the calculated annual mean watertable elevations, and the fitted linear regression models are presented in Figure 2.4. The significance of each fitted trend was assessed using the F test p value according to the following criteria: F test p value 0.05 the fitted slope is significant and provides a reasonable estimate of average watertable change from 1998 to 2007 F test p value > 0.05 the fitted slope is non significant, which implies that the linear model does not represent the trend any better than a line with zero slope 2.2.2 Trend of annual mean watertable elevation Annual mean watertable surfaces were interpolated for the first and last years of the trend analysis period (i.e., 1998 and 2007) using the values from the linear regression models. A spatial dataset representing the average change of watertable elevation from 1998 to 2007 was then constructed by subtracting the two surfaces. For both years, an initial watertable surface was interpolated using the annual mean point values from the linear models, and by constraining groundwater level to mean sea level of -0.059 m AHD along the coast and estuaries. The resulting surfaces were intersected with the ground surface to identify stream bed elevation constraints. The final watertable surfaces were interpolated using the annual mean point data, the coastal and estuarine sea level constraints and the stream bed elevation constraints. 2.2.3 Trend of annual mean freshwater thickness The Ghyben Herzberg approximation was applied in cells along coastal and estuarine margins of the aquifer to account for potential change of aquifer storage caused by movement of the saltwater freshwater interface. A definition sketch is presented in Figure 2.5. In most situations, this hydrostatic approximation over estimates the inland position of the interface because it neglects mixing of saltwater and freshwater. In this respect, it can be considered to provide an estimate of the maximum likely inland position of the interface for the particular circumstance. It does not consider time and therefore represents the maximum potential movement of the interface that could occur under transient conditions. A mathematical statement of the Ghyben Herzberg approximation can be written [1] where: h s = h f ρ s ρ f 40h f h s is the depth of freshwater below mean sea level, h f is the depth of freshwater above mean sea level, ρ s = 1.025 kg/m 3 is the mean density of seawater and ρ f = 1.0 kg/m 3 is the mean density of fresh groundwater. Artificial Recharge Potential of Perth s Superficial Aquifer Page 17

This calculation was applied to the annual mean watertable surfaces that were constructed for the first and last years of the trend analysis period, which provided corresponding datasets representing the average freshwater thicknesses for those years. The two datasets were then differenced to obtain an estimate of the change of freshwater thickness from 1998 to 2007. Several other methods for simulating the saltwater freshwater interface in a coastal aquifer are available. The simplest steady state method, Ghyben Herzberg approximation, was applied in this study because it only requires knowledge of the density difference between seawater and coastal groundwater and the base elevation of the aquifer. Glover [17] presented a steady state, 2D sharp interface solution that also includes the rate of groundwater discharge to the ocean but assumes the saltwater is static. Application of this model requires an estimate of the groundwater discharge rate to the ocean at the particular locations of interest. A more general quasi 3D finite difference model called SHARP [15] was developed by the Unites States Geological Survey to simulate freshwater and saltwater flow separated by a sharp interface in multi layered aquifers. At the most realistic end of the modelling spectrum, full density coupled numerical models can be developed in modelling systems such as FEFLOW [11] which also incorporate saltwater convection and hydrodynamic dispersion of salt across the interface. Nevertheless, density coupled models require very high mesh resolution to give accurate solutions and are often not practical to apply to large regional aquifers. 2.3 Assessment of artificial recharge potential 2.3.1 Theis (1935) well model Watertable response to a single, fully penetrating injection well placed at the centre of each 25 ha grid cell was simulated using the Theis [42] well equation. A definition sketch is presented in Figure 2.6. The model assumes homogeneous aquifer properties and predicts that the recharge cone will grow according to the following mathematical relation [2] H w (r,t) = Q w 4πT W(u) u = r2 S 4Tt where: W(u) = u e u u du H w is the height of the recharge cone above the static water level at radial distance r and time t, Q w is the well injection rate, S is the aquifer storage coefficient, Artificial Recharge Potential of Perth s Superficial Aquifer Page 18

T is the aquifer transmissivity and W(u) is the well function, which is a form of the exponential integral function. Though the Theis solution is traditionally used to estimate watertable draw down in response to a pumping well, it can also be applied to predict watertable up coning in response to injection, e.g., [43], particularly when the height of watertable rise is relatively small compared to the aquifer saturated thickness. Other suitable analytic models for a transient recharge well are not available. The model assumes that the well has an infinitesimally small diameter, and the aquifer has infinite lateral extent. The predicted size and rate of growth of the recharge cone are dependent on the values of the aquifer transmissivity, the aquifer storage coefficient and the well injection rate. An equilibrium response to continuous injection is not predicted because the model does not incorporate groundwater discharge; therefore, the recharge cone continues to grow with time but at an ever decreasing rate as the injected water spreads further into the aquifer. The model does not account for increasing transmissivity as the aquifer thickness increases in response to the recharge. In this sense it provides a conservative estimate of the time require for the recharge cone to rise a specified distance; however, in areas of large transmissivity, where a recharge well is likely to be feasible, the effect is small and can be neglected. It will be most significant in areas of small transmissivity where recharge wells are unlikely to be viable in any case. Rising and spreading responses Figure 2.7 illustrates examples of rising and spreading types of recharge cones generated from the Theis equation. A rising type recharge cone occurs if the aquifer transmissivity is small. Lateral groundwater flow is restricted and the mound grows vertically as injected water fills the local storage. After 30 days, the centre of the recharge cone is growing relatively slowly and there is minimal watertable change at the cell edge. Dependent on local depth to groundwater, it is possible that the cone centre may rise to the ground surface, and artificial recharge operations may be problematic. In contrast, a spreading type recharge cone occurs if the aquifer transmissivity is large. In Figure 2.7, the centre of the spreading type recharge cone is growing very slowly after 30 days. In this situation it is unlikely that mounding will be a constraint on artificial recharge operations unless depth to groundwater is small. Radius of influence The Theis equation predicts radial symmetric flow of the injected water outward from the well. In this situation, it is possible to define a radius of influence within which the aquifer storage has been increased by a specified portion of the injected volume after a specified period of time. Total outward flow at radial distance r from the well is given by the expression Q r = 2πrT( H w / r) which leads to the following relationship between groundwater flow and well injection ( ) Q r = exp r2 S Q w 4Tt With minor re arranging, the following expression describing the radius of influence is obtained R ǫ (t) = 4Tt S ln(1 ǫ) Artificial Recharge Potential of Perth s Superficial Aquifer Page 19

Table 2.2: Simulated well injection rates Scale of operation Injection rate, ML/d L/s sand 0.5 6 medium 1 12 large 5 58 where ǫ = 1 Q r /Q w. For example, if ǫ = 0.95 then R 0.95 (t) represents the radius within which aquifer storage has been increased by 95 percent of the injected water volume at time t. The storage is not increased by 100 percent of the injected volume because some of the existing storage moves outward from the well in response to the injection. Considering the rising type recharge cone in Figure 2.7, we obtain R 0.95 = 63 m after one day and R 0.95 = 346 m after thirty days. For the spreading type recharge cone R 0.95 = 1,896 m after thirty days. In this case the same increase in aquifer storage is distributed in a much larger volume of the aquifer, resulting in smaller values of watertable rise over a larger area. Model application The Theis model was applied at the centre of each raster cell using the cell aquifer properties, and by solving for the height of the recharge cone at radius 0.1 m and time 30 days for the three values of the well injection rate specified in Table 2.2. The predicted heights for each injection rate were then divided by the annual mean depth to groundwater in corresponding cells to obtain the Relative Watertable Rise (RWR) RWR = H w DSW where H w is the height of recharge cone above static water level at radial distance r = 0.1 m and time t = 30 d, and DSW is depth to static watertable below ground surface. A value less than 1 indicated that the predicted recharge cone did not exceed ground surface elevation after 30 days and artificial recharge by well injection was potentially manageable at that location. A value greater than or equal to 1 indicated that the predicted recharge cone exceeded ground surface elevation before 30 days and therefore watertable mounding was likely to be a restriction on well injection at the specified rate. Evaluating the Theis model at 30 days was considered to be a reasonable choice because one month represents a practical length of time for continuous operation of an artificial recharge scheme. Furthermore, the rate of growth of the predicted recharge cone is normally relatively small after 30 days (Figure 2.7). The three injection rates listed in Table 2.2 were chosen to be indicative of small (0.5 ML/d), medium (1 ML/d) and large (5 ML/d) scales of operation. The choice of values was based on well yield data presented by Davidson [8] for typical pumping wells in the superficial aquifer; these data are summarised in Table 2.3. In comparison, Water Corporation production wells in the superficial aquifer typically produce 1 2 ML/d; Local Government irrigation wells normally yield around 0.5 1 ML/d; and domestic garden wells are designed to yield around 0.1 0.2 ML/d [8]. Artificial Recharge Potential of Perth s Superficial Aquifer Page 20

Table 2.3: Well yields for the Perth region superficial aquifer [8] Formation Well yield, ML/d L/s Tamala Limestone > 10 > 116 Bassendean Sand 0.5 2 6 23 Guildford Clay < 0.1 < 1.2 2.3.2 Glover (1960) basin model Watertable response to a single, square recharge basin placed at the centre of each raster cell was evaluated using the Glover [18] solution as depicted in Figure 2.8. In an unconfined, homogeneous aquifer of infinite lateral extent, the model predicts that the elevation of the recharge mound is given by the following equations [46] H b (x,y,t) = I b 4S P(u 1,u 2,u 3,u 4 ) P(u 1,u 2,u 3,u 4 ) = t 0 (erfcu 2 erfcu 1 )(erfcu 4 erfcu 3 )dτ u 1 = x L 2 d,u 2 = x+ L 2 d,u 3 = y W 2 d,u 4 = y + W 2 d,d = 4T S (t τ) erfcu = 2 e z2 dz π u where: H b (x,y,t) is height of the recharge mound above static water level at location x,y and time t, I b is the infiltration rate, L is basin length in the x direction, W is basin width in the y direction, S is the aquifer storage coefficient, T is aquifer transmissivity, P(u 1,u 2,u 3,u 4 ) is defined here as the basin function, with units of length, and erfc is the complimentary error function. The hydraulic load for the basin Q b is determined by the relationship between the infiltration rate and the basin area. In this study, infiltration rates were assigned based on the soil type Artificial Recharge Potential of Perth s Superficial Aquifer Page 21

properties in Table 2.1. A particular choice for Q b implied a particular basin size, which was used to define a practical limit on the achievable hydraulic load in some areas. As for the Theis model, an equilibrium response to continuous recharge is not predicted by the Glover equation; instead the recharge mound continues to grow with time but at an ever decreasing rate. Maximum mound height occurs beneath the centre of the recharge basin at the coordinate origin. The size and rate of growth of the recharge mound is dependent on the aquifer transmissivity, the aquifer storage coefficient, the hydraulic load and the basin size. Other methods for estimating watertable rise beneath rectangular and circular recharge basins are also available. Three methods for circular basins and three methods for rectangular basins were compared and evaluated by Warner et al. [46]. The two highly recommended methods for rectangular basins were Glover [18] and Hantush [19]; however, Warner et al. showed that the methods provide mathematically identical solutions, and therefore the choice of method is not critical in this study. An analytic method was chosen because it can be applied rapidly at many locations across large study areas. Rising and spreading responses Figure 2.9 illustrates examples of rising type and spreading type recharge mounds beneath a square basin, which correspond to the examples for the Theis well model in Figure 2.7. The basin infiltration rate applied in these examples was I b = 1 m/d, which translates to a basin with side length L = W = 1000/1 = 32 m. Similarities and differences between the Glover model and Theis model are evident from visual inspection of the figures. The Theis model applies the recharge at a point, whereas the Glover model distributes the same quantity of recharge uniformly over the basin area. Locally, the basin recharge mound is broader and lower than the well recharge cone; however, the solutions become progressively more similar with increasing distance from the well and basin. In the far field region the solutions are almost identical because the same amount of water is being transmitted outward from a central source through an identical aquifer. It is evident that after 30 days, the predicted vertical growth rate of the rising type mound is relatively slow, and the predicted vertical growth rate of the spreading type mound is very slow. Model application The Glover model was applied at the centre of each raster cell using the cell aquifer properties, and by solving for the height of the recharge mound directly beneath the centre of the basin at time 30 days and for the three values of hydraulic load specified in Table 2.4. The predicted heights for each hydraulic load were then divided by the annual mean depths to groundwater in the corresponding cell RWR = H b DSW where H b is the height of recharge mound above static water level at x = 0, y = 0 and time t = 30 d. A value less than 1 indicated that the recharge mound beneath the basin did not rise above ground surface after 30 days and artificial recharge by basin infiltration was potentially manageable at that location. A value greater than or equal to 1 indicated that the predicted recharge mound would exceed ground surface elevation before 30 days, and watertable rise was likely to be a restriction on successful basin infiltration at the specified Artificial Recharge Potential of Perth s Superficial Aquifer Page 22

Table 2.4: Simulated basin hydraulic loads Scale of operation Hydraulic load, ML/d sand 1 0.1 medium 5 0.5 large 10 1 Equivalent infiltration rate for a 100 100 m basin, m/d hydraulic load. Evaluating the Glover model at 30 days was considered to be a reasonable choice because one month represents a practical length of time for continuous operation of a basin, and the rate of growth of the predicted recharge mound is relatively small by that time (Figure 2.9). The three hydraulic loads in Table 2.4 are suggestive of small (1 ML/d), medium (5 ML/d) and large (10 ML/d) scales of operation. Table 2.1 lists the four soil texture classes and the associated infiltration rates assigned in this study. These vary from 1 m/d for sandy soils to 0.003 m/d for predominantly clay and silt soils. In sandy areas, the three hydraulic loads listed in Table 2.1 translate to square recharge basins with side lengths 32 m (0.1 ha), 71 m (0.5 ha) and 100 m (1 ha), respectively. In predominantly clay soil areas, the basin sizes required to achieve the same hydraulic loads would be 577 m (33 ha), 1,291 m (167 ha) and 1,826 m (333 ha). Because some of these basins are impractically large, a maximum basin area of 25 ha was enforced; this corresponds to a square basin with side length 500 m. If the combination of hydraulic load and infiltration rate implied that the basin would need to be larger than 25 ha then the analysis of mound height was not conducted in those cells. Artificial Recharge Potential of Perth s Superficial Aquifer Page 23

Definitions: 1. Ground surface elevation 2. Base of superficial aquifer elevation 3. Superficial aquifer watertable elevation 4. Depth to watertable below ground surface 5. Superficial aquifer saturated thickness 6. Superficial aquifer freshwater thickness 4 1 3 Sea level 6 5 5, 6 Seawater - freshwater interface 2 Aquifer Cross Section Figure 2.1: Definition sketch for spatial datasets a. b. 9.0 11.0 Water level, m AHD 10.5 10.0 Measurement Fitted polynomial Upper and lower outlier bounds Water level, m AHD 8.5 8.0 7.5 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Measurement Fitted polynomial Upper and lower outlier bounds 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 Date Date Figure 2.2: Examples of the outlier analysis using a z score test a. no outliers detected, b. five outliers detected Artificial Recharge Potential of Perth s Superficial Aquifer Page 24

Ground surface Upper aquifer layer (K 1, S1) Watertable Transmissivity T 1 = K1 B1 T 2 = K 2 B2 T = K B + K B 1 1 2 2 B 1 Conductivity K = T / ( B + B ) 1 2 B 2 Lower aquifer layer (K 2, S2) Aquifer Cross Section Figure 2.3: Calculation of aquifer transmissivity a. b. 67.0 26.5 66.5 26.0 Water level, m AHD 66.0 65.5 65.0 64.5 64.0 Measurement Annual mean water level Linear model (p-value = 0.0014) Water level, m AHD 25.5 25.0 24.5 24.0 23.5 Measurement Annual mean water level Linear model (p-value = 0.098) 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Date Date Figure 2.4: Examples of linear regression models fitted to the annual watertable elevations a. significant slope (p value 0.05), b. non significant slope (p value > 0.05) Artificial Recharge Potential of Perth s Superficial Aquifer Page 25

Ghyben-Herzberg approximation: h s = hf / ( - f) S Decline in freshwater thickness due decline of the watertable Monitoring well Ground surface Watertable Sea level h f Freshwater Seawater - freshwater interface f h s Seawater Base of aquifer S Decline in freshwater thickness due to rise of the seawater-freshwater interface Figure 2.5: Potential change of freshwater thickness a. Plan View r Recharge well Q w Ground surface D Water table H w Static water level B Impermeable aquifer base b. Aquifer Cross Section Figure 2.6: Artificial recharge through an injection well Artificial Recharge Potential of Perth s Superficial Aquifer Page 26

8 Height of well recharge cone above static water level ( H w ), m 7 6 5 4 3 2 1 Rising cone: Q w = 1,000 m 3 /d T = 100 m 2 /d S = 0.3 Spreading cone: Q w = 1,000 m 3 /d T = 1,000 m 2 /d S = 0.1 1 day 10 days 20 days 30 days Edge of raster cell 0 0 50 100 150 200 250 Radial distance from well ( r ), m Figure 2.7: Rising and spreading recharge cones from the Theis well equation a. Plan View y W (0,0) x Recharge basin L Ground surface Q b D Water table H b Static water level B Impermeable aquifer base b. Aquifer Cross Section Figure 2.8: Artificial recharge beneath a rectangular basin Artificial Recharge Potential of Perth s Superficial Aquifer Page 27

Height of basin recharge mound above static water level ( H b ), m 8 7 6 5 4 3 2 1 Rising mound: Q b = 1,000 m 3 /d T = 100 m 2 /d S = 0.3 1 day 10 days 20 days 30 days Spreading mound: Q b = 1,000 m 3 /d T = 1,000 m 2 /d S = 0.1 Edge of raster cell 0 0 50 100 150 200 250 Distance from basin centre ( x ), m Figure 2.9: Rising and spreading recharge mounds from the Glover basin equation Artificial Recharge Potential of Perth s Superficial Aquifer Page 28

3 RESULTS 3.1 Derived datasets Ground surface elevation The derived ground surface elevation model (Figure 3.1a) varied from sea level along the coast and estuaries to greater than 100 m AHD in the northern part of the study area. A relatively large region within the Murray and Harvey Rivers management areas (Figure 1.2) has ground elevation below 10 m AHD. Watertable elevation Annual mean watertable elevation for year 2008 (Figure 3.1b) varied between sea level along the coast and estuaries up to approximately 80 m AHD in the Gnangara Mound. South of the Swan and Canning Rivers estuary system the watertable elevation of the Jandakot mound was 20 30 m AHD. Generally, there is a correspondence between the topographic surface and watertable elevation surface, except along the coast and estuaries where the watertable is influenced by sea level and is consistently low. Depth to watertable Annual mean depth to watertable for year 2008 (Figure 3.1c) varied from less than 2 m across large parts of the southern study area, to greater than 40 m beneath the coastal topographic ridges. Aquifer thickness Annual mean saturated thickness of the superficial aquifer for year 2008 (Figure 3.2) varied from approximately 10 20 m in the Murray and Harvey Rivers management areas, up to 40 70 m within the central parts of Gnangara Mound. Saturated thickness of the central parts of Jandakot Mound was in the range 30 50 m. Aquifer properties distributions Figure 3.3(a) and (b) depict the hydraulic conductivity values from the upper and lower superficial layers of PRAMS and PHRAMS. Figure 3.3(c) depicts the integrated values calculated in this study. Variation of hydraulic conductivity across the study area reflects the variety of sediment types and textures that are present. Superficial sediments vary from clays and loams with relatively small permeability in the southeast, to porous coastal limestone with very large permeability. Large parts of the study area contain predominantly medium textured sands with hydraulic conductivity values in the range 5 20 m/d. The blocky pattern of hydraulic conductivity distribution reflects the geometric regions used to assign aquifer properties in PRAMS and PHRAMS. The distribution of aquifer transmissivity (Figure 3.4a) was derived as the product of annual mean saturated thickness for year 2008 and aquifer hydraulic conductivity. Transmissivity varied from very large values (> 10,000 m 2 /d) along the coastal strip, to relatively small values (10 100 m 2 /d) in eastern parts of the Murray and Harvey Rivers management areas. Across much of the study area, the derived aquifer transmissivity values were in the range 100 1,000 m 2 /d. The values of aquifer storage coefficient from PRAMS and PHRAMS (Figure 3.4b) were mostly in the range 0.2 0.35. Values of less than 0.1 occurred in the southeast of the study area. Infiltration rates Figure 3.5(a) depicts the generalised classification of soil texture derived from the Artificial Recharge Potential of Perth s Superficial Aquifer Page 29

environmental geology data. The associated distribution of soil infiltration rates calculated using Table 2.1 is presented in Figure 3.5(b). Most of the study area is covered by sandy soils where it was assumed that an infiltration rate of approximately 1 m/d is feasible. A much smaller infiltration rate was assigned in predominantly clay soil areas, which are present mostly within the Murray and Harvey Rivers management areas, and in upper reaches of the Swan Coast management area. 3.2 Aquifer storage change Estimated change of aquifer storage for the ten year period 1998 to 2007 is depicted in Figure 3.6(a). This result was derived from the linear regression analysis of mean annual watertable elevation. Coloured points correspond to locations of monitoring wells and piezometers used in the analysis. The continuous coloured rendering underneath the points represents the interpolation of the point data to a continuous surface. Colour differences between the points and rendered surface indicate areas where the fitted trends were not significant. The point data depict the values of the fitted trends, whereas the interpolation assumed that the trend was zero at locations where the linear fit was not significant. 3.3 Saltwater intrusion Figure 3.6(b) and (c) illustrate the results obtained by applying the Ghyben Herzberg approximation to the derived watertable trends. Part (b) depicts the potential increase or decrease in thickness of sea water in the aquifer due to the watertable change. The displayed values represent a potential change only because the approximation does not consider the amount of time required for the saltwater freshwater interface to move position. The base of superficial aquifer elevation model was used to check if the theoretical position of the interface was above or below the base of the aquifer. Part (c) depicts the potential change of freshwater thickness in the aquifer due to the combined effects of watertable change and potential movement of the saltwater freshwater interface. It differs from the map in part (a) only along the coastal strip where saltwater intrusion is possible. 3.4 Well recharge model The results obtained by applying the Theis well model within each 25 ha grid cell across the study area are presented in Figure 3.7. Parts (a), (b) and (c) were produced using the three injection rates in Table 2.2, which were chosen to represent small (0.1 ML/d), medium (1 ML/d) and large (5 ML/d) scales of operation. Each map is made up of 21,335 results that were obtained by evaluating the analytic model independently in each cell. The rendered results represent the Relative Watertable Rise at the centre of each cell, which is the ratio of predicted watertable rise and depth to static watertable (see Methods section). For example, a value 0.5 indicates that the watertable at the well rose half way between static water level and ground surface after 30 days; a value greater than one indicates the predicted watertable rose above ground surface after 30 days. As the simulated injection rate increases from small to large in Figure 3.7 the area of aquifer that is potentially suitable for artificial recharge decreases in extent. The value of Relative Watertable Rise does not indicate the height of the recharge cone or the depth of unsaturated sediment above the cone; therefore, it is possible in very transmissive areas that Artificial Recharge Potential of Perth s Superficial Aquifer Page 30

the predicted watertable may be relatively close to ground surface in areas where artificial recharge activities are indicated to be feasible. 3.5 Basin recharge model Figure 3.8 illustrates the results obtained by applying the Glover basin model within each 25 ha grid cell across the study area using the three hydraulic loads in Table 2.4, which correspond to small (1 ML/d), medium (5 ML/d) and large (10 ML/d) scales of operation. As above, the results are displayed as Relative Watertable Rise; however, the areas rendered in grey were excluded from the analysis because the basin would need to be larger than 500 500 m (25 ha) to achieve the particular hydraulic load. For all three hydraulic loads examined in this study, the maximum basin size was exceeded in the clay soil areas where the assigned infiltration rate was 0.003 m/d (Figure 3.5). Artificial Recharge Potential of Perth s Superficial Aquifer Page 31

Moore-Hill Rivers Swan Coast Gnangara Mound 20 km Jandakot Mound Murray River Harvey River a. b. c. Ground level, m AHD < 10 10-30 30-50 50-70 70-90 90-110 110-130 > 130 Surface water management area Groundwater level 2008, m AHD < 10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100 Depth to groundwater 2008, m < 2 2-5 5-10 10-15 15-20 20-25 25-30 30-35 35-40 > 40 Figure 3.1: Elevation surfaces a. Ground surface elevation, b. annual mean watertable elevation in 2008, c. annual mean depth to watertable in 2008 Artificial Recharge Potential of Perth s Superficial Aquifer Page 32

20 km Aquifer saturated thickness, m 0-10 10-20 20-30 30-40 40-50 50-60 60-70 Figure 3.2: Aquifer saturated thickness Artificial Recharge Potential of Perth s Superficial Aquifer Page 33

20 km a. b. c. Hydraulic conductivity upper superficial, m/d < 0.1 0.1-1 1-5 5-20 20-100 100-1,000 Hydraulic conductivity lower superficial, m/d < 0.1 0.1-1 1-5 5-20 20-100 100-1,000 Integrated hydraulic conductivity, m/d < 0.1 0.1-1 1-5 5-20 20-100 100-1,000 Figure 3.3: Aquifer hydraulic conductivity a. upper superficial, b. lower superficial, c. vertically integrated Artificial Recharge Potential of Perth s Superficial Aquifer Page 34

20 km a. b. Transmissivity, m 2 /d < 10 10-100 100-1,000 1,000-10,000 > 10,000 Storage coefficient 0.01-0.05 0.05-0.1 0.1-0.15 0.15-0.2 0.2-0.25 0.25-0.3 0.3-0.35 Figure 3.4: Aquifer transmissivity and storage coefficient distributions Artificial Recharge Potential of Perth s Superficial Aquifer Page 35

20 km a. b. Texture classification Infiltration rate, m/d Sand 1 Limestone and sand 0.3 Sand with clay, silt, and peat 0.003 Clay and silt Figure 3.5: Soil infiltration rate distribution a. Generalised soil texture distributions, b. assigned infiltration rates Artificial Recharge Potential of Perth s Superficial Aquifer Page 36

Quinns Lexia-Bullsbrook West Mirrabooka 20 km a. b. c. Watertable change 1998-2007, m < -5-5 to -2-2 to -1-1 to -0.5-0.5-0.5 0.5-1 1-2 2-5 > 5 Saltwater thickness change 1998-2007, m < -5-5 to -2-2 to -1-1 to -0.5-0.5-0.5 0.5-1 1-2 2-5 > 5 Freshwater thickness change 1998-2007, m < -5-5 to -2-2 to -1-1 to -0.5-0.5-0.5 0.5-1 1-2 2-5 > 5 Figure 3.6: Ten year change of aquifer storage 1998 2007 a. change of watertable elevation, b. potential change of saltwater thickness, c. potential change of freshwater storage Artificial Recharge Potential of Perth s Superficial Aquifer Page 37

20 km a. b. c. Relative Watertable Rise Relative Watertable Rise Relative Watertable Rise Small injection rate, 0.5 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 Medium injection rate, 1 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 Large injection rate, 5 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 Figure 3.7: Relative Watertable Rise at 30 days for well injection a. small injection rate, b. medium injection rate, c. large injection rate Artificial Recharge Potential of Perth s Superficial Aquifer Page 38

20 km a. b. c. Basin too large Basin too large Basin too large Relative Watertable Rise Small hydraulic load, 1 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 Relative Watertable Rise Medium hydraulic load, 5 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 Relative Watertable Rise Large hydraulic load, 10 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 Figure 3.8: Relative Watertable Rise at 30 days for basin infiltration a. small hydraulic load, b. medium hydraulic load, c. large hydraulic load Artificial Recharge Potential of Perth s Superficial Aquifer Page 39

4 DISCUSSION Within the central part of the study area between Quinns Rocks and Mandurah, the observed pattern of watertable change during 1998 to 2007 was broadly consistent with the pattern of change from 1995 to 2004 found by Smith et al. [41]. Significant decline of storage in Gnangara Mound is evident in the northern part of the study area in which the watertable elevation fell by 1 5 m (0.1 0.5 m/yr) across the central part of the mound. Localised watertable decline of 1 5 m also is evident within the west Mirrabooka and Lexia Bullsbrook bore fields, where groundwater is extracted for public drinking water supply. In contrast, watertable elevation in Jandakot Mound and further south within the Murray and Harvey River management areas has been relatively stable during the past ten years. The exception is a narrow strip of land along the base of Darling Scarp between Serpentine and Armadale wherein the watertable has declined by 2 3 m. Comparison of hydrographs for observation bores screened in the superficial aquifer and in the underlying Leederville and Yarragadee Aquifers indicate that this is a response to pumping and drawdown of piezometric head in the deep aquifers. The superficial sediments are directly underlain by the Leederville and Yarragadee Aquifers in this region [9]. A downward hydraulic gradient has been present during 1998 to 2007 and drawdown of 3 6 m in the Yarragadee Aquifer has been reflected by drawdown of 2 4 m in the Leederville Aquifer and 2 3 m in the superficial aquifer. Application of the Ghyben Herzberg approximation indicates that some degree of saltwater intrusion may have occurred around Quinns Rocks in response to watertable decline. Groundwater level has fallen by around 0.5 1 m in apparent response to groundwater pumping from the Quinns bore field, which commenced in 1999. The results also imply there has been potential for saltwater intrusion around Safety Bay between the coast and Serpentine River; however, it is unclear if saltwater can flow from the ocean into this part of the aquifer. It is probable that a small watertable mound between the river and coast prevents saltwater movement into this area. Several layers of information are superimposed in Figure 4.1, which provides a basis for subsequent discussion. In parts (a) and (b), the regional drainage network and locations of existing and proposed wastewater treatment plants are overlayed on the previous results for a medium well injection rate (Figure 3.7b) and medium basin hydraulic load (Figure 3.8b). Part (c) depicts Perth s Underground Water Pollution Control Areas (UWPCA) overlayed on the ten year change of aquifer freshwater storage from 1998 to 2007, inclusive of potential saltwater intrusion. 4.1 Hydraulic feasibility of artificial recharge The main areas identified as being hydraulically suitable for artificial recharge operations include the coastal strip of the study area, extending approximately 10 km inland from the coast, the central part of Gnangara Mound, and smaller parts of Jandakot Mound. The eastern margin of the study area, and extensive inland parts south of Point Peron are potentially unsuited to artificial recharge activities due to a combination of relatively shallow depth to watertable and soil and aquifer sediments with relatively small permeability. The coastal strip is hydraulically suited to artificial recharge because the limestone has large transmissivity and large depth to groundwater beneath the coastal ridges. Gnangara Mound is also suited to artificial recharge because the aquifer has medium transmissivity due to Artificial Recharge Potential of Perth s Superficial Aquifer Page 40

increased aquifer thickness, and relatively large depth to static watertable. Overall, large scale artificial recharge operations in the superficial aquifer appear to be feasible only within the coastal limestone and central parts of Gnangara Mound. The potential for artificial recharge operations using wells or basins is qualitatively similar. This follows from the similarity of the analytic models applied in this study, which both simulate divergent flow from a central source. Based on international experience with well injection schemes, the feasibility for successful well recharge is anticipated to be greatest in the coastal limestone where there is opportunity to inject water into secondary porosity features. On the other hand, injection into fissures or karst may present a risk to groundwater users and receiving environments, dependent on the quality of the recharge water and the nature of preferential flow paths in the aquifer. Well injection into sandy sediments of Gnangara Mound is also technically feasible but there is greater potential for clogging problems to occur. Where adequate land is available, the potential for surface infiltration is similar in coastal and inland areas which have predominantly sandy soils. An injection well will result in greater local watertable rise than an infiltration basin with the same hydraulic load, and will therefore raise groundwater toward ground surface more rapidly. This may be desirable or undesirable dependent on the objectives of an artificial recharge operation. If it is undesirable because the watertable is raised too far, then it might be possible to achieve the same hydraulic load or greater using multiple lower capacity injection wells separated by a suitable distance (i.e., injection well field). This study has not explored the feasibility of multi wells systems or attempted to optimise the spacing of a well field. This would be interesting and useful work but needs to be the topic of a more specific study. The current study investigates the feasibility of operating a single injection well of specified capacity as a means of determining if artificial recharge by injection wells, single or multiple, is potentially feasible. Producing results for different injection rates provided an indication of the relative scale of operation that might be achievable at each location in the study area; however, they should not be interpreted as absolute limits on the injection capacity that could be achieved using an optimised multi well system. It is possible that sustained infiltration rates greater than the assumed rate of 1 m/d might be possible in the coastal sand and limestone, particularly with clean or highly treated water. If a larger infiltration rate is assumed then the same hydraulic load can be achieved using a smaller basin, in which case the basin recharge model would predict a narrower but higher watertable mound beneath the basin. From this viewpoint, a greater infiltration rate is not automatically an advantage in every situation. For instance, it is conceivable that under certain circumstances it might be preferable to achieve the target hydraulic load using a larger basin with smaller infiltration rate in order to realize a broader and lower watertable mound. The analysis in this study does not consider design optimisation of the recharge basin and therefore the results provide a relative assessment of the potential to operate a single infiltration basin. They should not be interpreted as either the desired or maximum achievable hydraulic load. The restriction on successful operation of a recharge basin is based on the assumption that watertable rise to the ground surface is likely to cause shallow groundwater and possibly groundwater seepage in the vicinity of the basin. It is worthwhile noting that this restriction does not apply if the basin is located within a local surface depression where surrounding watertable rise to the basin water level is unlikely to cause surface seepage or water logging Artificial Recharge Potential of Perth s Superficial Aquifer Page 41

of adjacent ground. It is evident from the above discussion that the regional assessment of artificial recharge potential in this study identifies broad areas where artificial recharge is more likely or less likely to be successful, as well as the regional hydraulic characteristics that control this capacity. It is also evident that local conditions and opportunities are not taken into account in the assessment, and that these conditions might improve or diminish the potential of the site for artificial recharge. Therefore, the results of this study should not be interpreted in a way that suggests artificial recharge is either possible or impossible at a particular location. State and Commonwealth Government guidelines on managed aquifer recharge should be accessed to obtain further information and advice about current procedures for proposing and investigating the feasibility of artificial recharge at a specific site, e.g., [10], [29] and [13]. 4.2 Recharge water sources The drainage network and wastewater treatment plants serve as collection infrastructures that result in water being available for reuse at particular locations. Artificial recharge of these water sources at other locations, for example, at the site of saltwater intrusion or a drying wetland, may require new conveyance infrastructure as part of the proposed reuse strategy. The majority of existing and proposed wastewater treatment plants (Figure 4.1) are located within the coastal strip where large scale artificial recharge operations are feasible. Large volumes of treated water could be infiltrated or injected; however, the recharge water may discharge to the ocean relatively quickly. There is a risk that nutrients and other contaminants in the treated wastewater would not be attenuated before entering the near shore marine system. Further work is required to characterised the nature of preferential flow paths in the coastal limestone that control the enhance transmissivity of the aquifer. Treated wastewater is currently infiltrated at the six treatment plants listed in Table 1.5. The largest, Gordon Road, achieves a hydraulic load of approximately 6 ML/d, which is equivalent to a medium size facility in this study. In Figure 4.1, the Gordon Road treatment plant is shown to be in a transition area where artificial recharge at that rate may or may not be feasible dependent on depth to groundwater and aquifer transmissivity. Watertable depth is reduced near Serpentine River where ground level is lower, and the PRAMS value of aquifer transmissivity also decreases in this area (Figure 3.4). In comparison with the infiltration amounts listed in Table 1.5, the total capacity of the major wastewater treatment plants that currently discharge to the ocean is around 340 ML/d (Table 4.1). Many urbanised areas of Perth overlie parts of the superficial aquifer where artificial recharge is feasible. In those areas there are good prospects for increasing rainwater capture by directing stormwater runoff from roofs and impermeable surfaces into local infiltration facilities. This practice is already common as an efficient means for stormwater disposal and aquifer recharge. For instance, assuming rainfall of approximately 0.8 m/yr and an average house block size of 500 800 m 2, with approximately half that area covered by roofs and impervious services, then each house block contributes approximately 300 480 KL/yr groundwater recharge to the superficial aquifer. This calculation assumes that all rainfall on to roofs and hard surfaces is redirected into the ground via domestic soak wells, and 50% of rainfall percolates as deep drainage beneath lawn and garden areas. On the same basis, it can be estimated that the aquifer receives roughly 1 GL/yr recharge from each 2,000 3,000 Artificial Recharge Potential of Perth s Superficial Aquifer Page 42

Table 4.1: Discharges from Perth s major wastewater treatment plants Wastewater treatment plant Current capacity, ML/d Initial capacity, ML/d Woodman Point 160 160 Beenyup 120 150 Subiaco 60 60 Alkimos 20 (year 2010) 160 Future capacity, ML/d East Rockingham 40 (year 1025) 160 (year 2060) Total 340 690 developed house blocks. For an urban area of 63,000 ha [14] the above calculation yields an estimate of total groundwater recharge of approximately 380 GL/yr. Less potential exists in low lying urban areas in the east of the study area. Perth s drainage network (Figure 4.1) collects water predominantly from areas with shallow groundwater and less permeable soil and aquifer sediments where local artificial recharge of the superficial aquifer is more difficult and there is limited need to enhance the natural recharge. These flows include rainfall runoff and groundwater drainage. There may be opportunity to redirect this inland drainage water into the coastal superficial aquifer, particular where major drains traverse the coastal strip. Existing examples include the Subiaco and Herdsman Main Drains. Nevertheless, many main drains discharge to the Swan Canning and Peel Harvey Estuaries east of the coastal strip where there is limited opportunity to redirect the water into the aquifer, e.g., Peel and Birrega Main Drains. Conveying water from those drains to more coastal locations would require additional infrastructure development. 4.3 Management of aquifer storage and watertable elevation There is relatively good correspondence between areas with declining groundwater level and areas that are potentially suitable for artificial recharge based on hydraulic considerations. At a regional level of analysis, artificial recharge is a viable option for mitigating or slowing storage decline in the worse affected parts of the study area, including the coastal zone and within Gnangara Mound. There is less potential for addressing storage decline in the area along the base of Darling Scarp (Figure 4.1) where greater clay and silt contents are likely to restrict both surface infiltration and well injection rates. The amount of water that is potentially available for artificial recharge is significant relative to observed watertable decline. Table 4.2 compares the annual wastewater discharge capacity of 340 ML/d from Perth s major wastewater treatment plants (Table 4.1 to volumetrically equivalent changes of aquifer storage, assuming a fillable porosity 0.2. For example, 340 ML/d is volumetrically equivalent to an aquifer storage increase of 1 m/yr over an area of 621 km 2 (25 25 km). Figure 4.2 is an alternative presentation of the modelled response of the superficial aquifer to Artificial Recharge Potential of Perth s Superficial Aquifer Page 43

Table 4.2: Equivalent area of watertable rise for total volume 340 ML/d and porosity 0.2 Watertable rise, m/yr Area of watertable rise, km 2 0.1 6,205 79 0.5 1,241 35 1 621 25 Side length of equivalent square, km well injection based on the method of this study. Part (a) shows the estimated radius of influence within which the aquifer storage is increased by ninety percent of the injected water volume at 30 days. Part (b) depicts the injection rate required to raise the watertable at the injection well by an equal amount of 0.1 m, and part (c) depicts the injection rate required to raise the watertable to ground surface at 30 days. In combination, they provide a quantitative impression of the amount of water required to achieve similar storage or water level responses in the aquifer. Within the coastal limestone, large amounts of water are needed to significantly influence the watertable elevation. There is potential for using artificial recharge to supplement or freshen groundwater supplies but the feasibility of manipulating the groundwater level to recover wetlands or groundwater dependent ecosystems is limited. If the objective of artificial recharge in the coastal strip is to supplement groundwater supply then the facility should be located as far away as possible from the coast and inlet shores to minimise induced storage loss to the ocean and estuaries. There is limited opportunity for controlling only the local watertable response because the recharge water is rapidly transmitted away from the recharge site. Yesertener [49] modelled the amount of artificial recharge that would be required to raise and maintain declining groundwater level in seven caves within the Yanchep National Park cave system, which is located in Tamala Limestone approximately 50 km north of Perth and around 5 km from the coast. It was estimated that approximately 3.8 ML/d would maintain a 2 m watertable rise in Crystal cave until 2015. At the other six cave sites, it was estimated that artificial recharge rates of 0.8 1.4 ML/d would be required to maintain watertable rises of around 1 1.5 m. Thus, each cave would require a small to medium scale artificial recharge operation to maintain a relatively small watertable rise in the cave and surrounding limestone aquifer. Inland of the coastal strip within the Gnangara and Jandakot Mounds, there is potential to achieve medium to large injection and infiltration rates, and there is greater potential to address local groundwater issues through better control over local aquifer storage and groundwater elevation. Recharge water moves away from injection and infiltration sites more slowly and the recharge is unlikely to induce additional significant groundwater discharge to the ocean or estuaries. This may present opportunities to supplement groundwater supply where relatively large recharge rates can be achieved, and to recover watertable declines at wetland sites if sufficient control over watertable elevation is achievable. 4.4 Management of saltwater intrusion Large intra annual watertable change does not occur close to the coast throughout most of the Perth region because groundwater level in the coastal limestone is strongly controlled by Artificial Recharge Potential of Perth s Superficial Aquifer Page 44

sea level elevation, which is relatively invariant between years. Nevertheless, a small sustained change of watertable elevation in the coastal zone can result in a relatively large movement of the saltwater freshwater interface position and associated change of freshwater storage in the aquifer. A watertable decline of 0.5 m can produce a maximum theoretical upward movement of the interface of approximately 20 m. In practice the movement is less due to mixing of saltwater and freshwater across the interface, which weakens the fluid density contrast driving the interface movement. A reduction of the coastal freshwater resource due to a small sustained decline in watertable elevation can go largely unnoticed unless the position of the saltwater freshwater interface is monitored, or it is raised to the point where it intercepts well screens and the wells go salty. A corollary of the above discussion is that management of saltwater intrusion by artificial recharge requires the ability to permanently raise the watertable or mitigate watertable decline over a relatively large area. In turn, this requires a large and secure water source. Considering the overall water balance, it is clear that some amount of saltwater intrusion will occur unless the amount of groundwater withdrawal in the coastal strip is matched by an equal amount of recharge. These factors indicate that an artificial recharge scheme to counter saltwater intrusion would add a significant volume of recharge water to the aquifer between the recharge site and coast. The degree of natural attenuation of contaminants introduced with the recharge water, and the potential for impacts on groundwater users and groundwater receiving environments due to preferential flow are therefore significant issues for risk assessment. 4.5 Limitations of the study This report provides a regional scale analysis of artificial recharge potential at 21,335 locations across the Perth region within an area of 5,334 km 2. In applying the simplified analytic models for well injection and basin infiltration various assumptions were made or implied about the local conditions based on regional datasets. This approach allowed a consistent comparison of artificial recharge potential across the entire water resource region; however, it also has a number of limitations that should be considered when exploring and interpreting the results. In particular it does not consider the design and optimisation of basins and multi wells injection systems to suit local variation of conditions and different hydraulic objectives of the artificial recharge operation, such as the desired lateral extent and height of the recharge mound. The analytic models assume a laterally infinite and homogeneous aquifer, and constant infiltration and injection rates. The results provide a quantitative estimate of the average watertable response to an average recharge condition. Feasibility of successful artificial recharge at each location was assessed by considering the amount of local watertable rise at 30 days after the recharge was commenced. The presented results are dependent on these modelling assumptions and the specific applications of the models, and therefore they should not be taken as absolute limits on the potential capacity of an artificial recharge scheme or the aquifer response. Local conditions and facility specific operations will lead to different watertable responses. The methodology in this study does not consider the influence of local soil and aquifer layering on vertical percolation and flow, which can have critical influence on the design, performance and operation of injection wells and infiltration basins (Tables 1.1 and 1.2). Artificial Recharge Potential of Perth s Superficial Aquifer Page 45

This level of analysis was beyond the scope of this report but it is worthwhile mentioning that local soil layers are known to exist throughout the study area, although their distributions and hydraulic properties are not well studied. Survey and characterisation of such layers would need to be carried out as part of a site specific investigation, with the expectation they could significantly change the feasibility or design of a proposed artificial recharge facility if found to be present or absent. Less hydraulically conductive layers are thought to be common but discontinuous across the Perth region [39]. Xu et al. [48] reported that where fine sandy clay or limonite cemented sand layers (coffee rock) occur there is increased soil water retention by those layers and less natural recharge. Local water perching layers are also present in the superficial aquifer in the vicinity of wetlands [7]. In assessing change of freshwater storage in the aquifer, the change of freshwater thickness was calculated from the estimated change in saltwater thickness. This was performed by converting the ten year change in mean annual watertable elevation into an equivalent potential change in the position of the saltwater interface position by applying the Ghyben Herzberg approximation. It is worthwhile to reiterate that this estimate represents the maximum potential movement of the interface if change of watertable elevation persists. It does not consider the amount of time required for the interface position to adjust or hydrodynamic dispersion in the aquifer. If the change of inshore watertable elevation persists long enough for a new equilibrium position of the saltwater interface to establish, then the real adjustment of the interface position would be less than predicted by the Ghyben Herzberg approximation. Artificial Recharge Potential of Perth s Superficial Aquifer Page 46

Two Rocks Yanchep Two Rocks Yanchep Alkimos Bullsbrook Alkimos Bullsbrook Gnangara Beenyup Beenyup Perth Coastal 20 km Subiaco Subiaco Gwelup Woodman Point Woodman Point Point Peron Halls Head Kwinana East Rockingham Port Kennedy Gordon Rd. Point Peron Halls Head Gordon Rd. Kwinana East Rockingham Port Kennedy Jandakot Karnup-Dandalup Caddadup Tims Thicket Yunderup Pinjarra Caddadup Tims Yunderup Thicket Pinjarra a. b. Drain Wastewater treatment plant current proposed Relative Watertable Rise Medium injection rate, 1 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 Drain Wastewater treatment plant current proposed Relative Watertable Rise Medium hydraulic load, 5 ML/d 0-0.1 0.1-0.25 0.25-0.5 0.5-0.75 0.75-1 > 1 c. Underground Water Pollution Control Area Freshwater thickness change 1998-2007, m < -5-5 to -2-2 to -1-1 to -0.5-0.5-0.5 0.5-1 1-2 2-5 > 5 Figure 4.1: Opportunities for artificial recharge a. medium scale well injection at 30 days, b. medium scale basin infiltration at 30 days, c. ten year change in aquifer storage of freshwater Artificial Recharge Potential of Perth s Superficial Aquifer Page 47

20 km a. b. c. Well radius of influence, m < 250 250-500 500-1,000 1,000-2,000 2,000-4,000 > 4,000 Injection rate to raise watertable 0.1m at 30d, ML/d < 0.02 0.02-0.04 0.04-0.06 0.06-0.08 0.08-0.1 0.1-0.12 > 0.12 Injection rate to raise watertable to ground surface at 30d, ML/d < 0.5 0.5-1 1-5 > 5 Figure 4.2: Aquifer response to well injection a. radius of influence R 0.9 at 30 days, b. injection rate required to raise the watertable at the well by 0.1 m at 30 days, c. injection rate required to raise the watertable at the well to ground surface at 30 days Artificial Recharge Potential of Perth s Superficial Aquifer Page 48