Dunedin Groundwater Monitoring and Spatial Observations
SC Cox MHJ Ettema SM Mager PJ Glassey S Hornblow S Yeo
GNS Science Report 2020/11 June 2020
DISCLAIMER
The Institute of Geological and Nuclear Sciences Limited (GNS Science) and its funders give no warranties of any kind concerning the accuracy, completeness, timeliness or fitness for purpose of the contents of this report. GNS Science accepts no
responsibility for any actions taken based on, or reliance placed on the contents of this report and GNS Science and its funders exclude to the full extent permitted by law liability for any loss, damage or expense, direct or indirect, and however caused, whether through negligence or otherwise, resulting from any person’s or organisation’s use of, or reliance on, the contents of this report.
BIBLIOGRAPHIC REFERENCE
Cox SC, Ettema MHJ, Mager SM, Glassey PJ, Hornblow S, Yeo S. 2020. Dunedin groundwater monitoring and spatial observations. Lower Hutt (NZ): GNS Science. 86 p. (GNS Science report; 2020/11). doi:10.21420/AVAJ-EE81.
SC Cox, GNS Science, Private Bag 1930, Dunedin 9054, New Zealand MHJ Ettema, Otago Regional Council, Private Bag 1954, Dunedin 9054, New Zealand SM Mager, University of Otago, School of Geography, PO Box 56, Dunedin 9054, New Zealand PJ Glassey, GNS Science, Private Bag 1930, Dunedin 9054, New Zealand S Hornblow, Otago Regional Council, Private Bag 1954, Dunedin 9054, New Zealand S Yeo, University of Otago, School of Geography, PO Box 56, Dunedin 9054, New Zealand
© Institute of Geological and Nuclear Sciences Limited, 2020 www.gns.cri.nz
ISSN 2350-3424 (online) ISBN 978-1-99-001031-6 (online) http://dx.doi.org/10.21420/AVAJ-EE81
CONTENTS ABSTRACT ...... V KEYWORDS ...... VI 1.0 INTRODUCTION ...... 1 1.1 Sea-Level Rise, Groundwater and Infrastructure ...... 1 1.2 Coastal Aquifer Systems and the Water Table...... 2 1.3 Dunedin ...... 4 2.0 DUNEDIN DATA AND REFERENCE DATUM ...... 7 2.1 Groundwater Monitoring Network ...... 7 2.2 Surveying, Reference Frames, Conversions and Other Data ...... 13 2.3 Longer-Term Context of 2019 Groundwater Observations ...... 14 3.0 INTERPOLATED STATISTICAL SURFACES OF GROUNDWATER ...... 19 3.1 Groundwater Elevation ...... 20 3.2 Depth to Groundwater ...... 22 3.3 Effect and Reach of Tides ...... 24 3.4 Effect and Efficiency of Rainfall ...... 29 4.0 VERTICAL HETEROGENEITY AND HYDRAULIC PROPERTIES ...... 31 4.1 Perched Groundwater ...... 31 4.2 Confined Groundwater ...... 32 4.3 Vertical Pressure Gradients ...... 33 4.4 Drawdown and Recovery ...... 35 4.5 Temperature ...... 38 5.0 GROUNDWATER AND WATER NETWORKS ...... 40 5.1 Background ...... 40 5.2 Network Position Relative to the Water Table ...... 41 6.0 GROUNDWATER AND FLOODING ...... 44 6.1 Background ...... 44 6.2 Heavy (>75 mm) Rainfall Events ...... 46 6.2.1 3–4 June 2015 ...... 46 6.2.2 21–22 July 2017 ...... 50 6.2.3 1 February 2018 ...... 51 6.3 Discussion: Spatial versus Temporal Variability ...... 53 6.4 Available Subsurface Storage ...... 55 7.0 EFFECTS OF SEA-LEVEL RISE...... 58 8.0 SUMMARY AND CONCLUSIONS ...... 64 9.0 ACKNOWLEDGMENTS ...... 68 10.0 REFERENCES ...... 69
GNS Science Report 2020/11 i
FIGURES
Figure 1.1 Changes to the water table associated with sea-level rise, according to different models and local influences on groundwater systems ...... 3 Figure 1.2 Number of homes <50 cm elevation above the spring high tide mark, for nine coastal urban areas in New Zealand ...... 6 Figure 1.3 Elevation model of Dunedin from the 2009 LiDAR survey ...... 6 Figure 2.1 Groundwater monitoring sites in Dunedin...... 8 Figure 2.2 Definition of ground level and measuring point as used in the Dunedin groundwater monitoring dataset, similar to that used by Environment Canterbury ...... 11 Figure 2.3 Plots of 2019 groundwater elevation against the numerical day of the year (Julian days as 0–365 values) at the four ORC sites with longer-term data ...... 16 Figure 2.4 Plots of 2019 daily rainfall, and groundwater elevation at ORC piezometers compared with the daily rainfall at Musselburgh station averaged over the previous 30 days ...... 17 Figure 2.5 Plots of 2019 groundwater elevation against percentiles of the observation statistical distribution at the four ORC sites, compared against the longer-term 2010–2018 data records ...... 18 Figure 3.1 Interpolated median groundwater elevation (MedianGWL_2019) ...... 21 Figure 3.2 Interpolated model of the range (GWLrange_2019 in m) of groundwater levels in Dunedin...... 21 Figure 3.3 Interpolated depth to median groundwater elevation (MedianDTW_2019)...... 23 Figure 3.4 Electrical conductance of groundwater (μS/cm) as piezometer spot values and an interpolated grid, overlain by modelled %Seawater contours ...... 25 Figure 3.5 Tidal efficiency at each of the monitoring sites ...... 28 Figure 3.6 Tidal phase lag at each of the monitoring sites relative to the nearest (ocean or harbour) tidal source, plotted over the tidal efficiency grid...... 28 Figure 3.7 Graphs of (logarithmic) tidal efficiency (as lnTE) and phase lag (in minutes) against distance from the nearby tidal source ...... 29 Figure 3.8 The RRI ratio as a series of piezometer point observations and an interpolated grid ...... 30 Figure 4.1 Interpreted map-limit of perched groundwater sitting at around 1–1.5 m elevation within dune sands ...... 32 Figure 4.2 Plot of the difference in groundwater level for sites where deep and shallow piezometers are situated near each other ...... 34 Figure 4.3 Council St (I44/1124) groundwater level (black in mNZVD2016) and specific conductance during 2019 ...... 37 Figure 4.4 Examples of temperature data for the period 6 Mar 2019 – 12 Feb 2020 in shallow piezometers ≤6 m deep ...... 39 Figure 5.1 Position of stormwater network relative to the median water table in the flat-lying area of Dunedin ...... 42 Figure 5.2 Position of wastewater network relative to the median water table in the flat-lying area of Dunedin ...... 42 Figure 5.3 Stormwater network position relative to the water table at MHWS...... 43 Figure 5.4 Wastewater network position relative to the water table at MHWS ...... 43 Figure 6.1 Exemplars of heavy rainfall events experienced in Dunedin, plotted against time (in hours) since the start of rain ...... 45 Figure 6.2 Gridded and contoured elevation of June 2015 floodwater ...... 47 Figure 6.3 Model of June 2015 floodwater inundation ...... 47 Figure 6.4 Model of RAINstorage_fromGWLmedian, being a grid measure of the available subsurface storage above the median water table ...... 56
ii GNS Science Report 2020/11
Figure 6.5 Contours of floodwater inundation from June 2015 (see Figure 6.3) plotted over the RAINstorage_fromGWLmedian grid ...... 57 Figure 7.1 Geometric model of the elevation (in NZVD2016) of the water table at MHWS once there has been 50 cm of sea-level rise ...... 62 Figure 7.2 Geometric model of the DTW at MHWS once there has been 50 cm of sea-level rise...... 62 Figure 7.3 Comparative figure showing geometric models of the elevation (in NZVD2016) of the water table elevation at MHWS ...... 63
TABLES
Table 2.1 Summary of groundwater monitoring collated for this study...... 9 Table 2.2 Dunedin groundwater monitoring sites operational in 2019...... 12 Table 2.3 Comparison of longer- and short-term statistics of raw data from Otago Harbour ...... 14 Table 2.4 Dunedin Port tidal level averages predicted for the period 1 July 2019 – 30 June 2020 ...... 14 Table 2.5 Comparison of statistics for longer-term and 2019 statistics of groundwater levels at four ORC monitoring sites ...... 15 Table 3.1 Summary of statistical surfaces and other datasets developed to represent Dunedin groundwater elevation during 2019 ...... 20 Table 3.2 Summary of datasets developed to represent depth to Dunedin groundwater during 2019 ...... 23 Table 3.3 Summary of tidal and rainfall-recharge grid datasets developed for Dunedin groundwater ...... 26 Table 4.1 Observations of differences in measuring point elevations of nearby monitoring piezometers... 35 Table 4.2 Vertical hydraulic gradients in nearby monitoring piezometers...... 35 Table 6.1 Groundwater levels and calculated response at four ORC monitoring sites during the 3–4 June 2015 flood ...... 49 Table 6.2 Groundwater levels at four ORC monitoring sites before and during the 21–22 July 2017 rain . 51 Table 6.3 Groundwater levels at four ORC monitoring sites before and during the 1 February 2018 rain . 52 Table 6.4 Summary of RRI calculated for the ORC piezometers...... 54 Table 7.1 Approximate years when specific sea-level rise increments (metres above the 1986–2005 baseline) may be reached for the various projection scenarios of sea-level rise for the wider New Zealand region ...... 59 Table 7.2 Summary of grids developed to examine potential effects of sea-level rise ...... 61
GNS Science Report 2020/11 iii
APPENDICES APPENDIX 1 MAP OF PLACE NAMES ...... 75 APPENDIX 2 GLOSSARY ...... 76 APPENDIX 3 METADATA ...... 79
APPENDIX FIGURES
Figure A1.1 Map with key place names referred to in this study...... 75
APPENDIX TABLES
Table A3.1 Detailed description of groundwater level GIS datasets generated from the 2019 groundwater monitoring and used in this report ...... 79 Table A3.2 Detailed description of depth to groundwater GIS datasets generated from the 2019 groundwater monitoring and used in this report ...... 81 Table A3.3 Detailed description of tidal, rainfall recharge and chemistry GIS datasets generated from the 2019 groundwater monitoring and used in this report ...... 83 Table A3.4 Detailed description of GIS datasets generated to examine the potential effects of sea level in this report ...... 85
iv GNS Science Report 2020/11
ABSTRACT
Dunedin City has a large number of assets and critical infrastructure sitting on a low-lying coastal plain that is underlain by a largely unseen and relatively poorly understood hazard. Shallow groundwater in this area limits the unsaturated ground available to store rain and runoff, promotes flooding and creates opportunities for infiltration into stormwater and wastewater networks. Groundwater levels are expected to rise as sea level rises, causing greater frequency of flooding and/or direct inundation once it nears the ground surface. Dunedin’s future flooding concerns can be addressed through expensive and complex hard infrastructure solutions or by adaptive solutions that take into account the natural underlying conditions of low-lying Dunedin. Adaptive solutions are no less complex but may save time and money in the city’s effort to manage flooding and groundwater inundation. This report outlines a basis for future planning and mitigation of natural hazards in Dunedin.
Monitoring network developments in 2019 have significantly improved information on Dunedin’s groundwater. Groundwater level, temperature and specific conductance observations at 15 minute intervals have been collated into a time-series database. A wide variety of statistics have been generated for each site, including median, maximum, minimum, 95th and 5th percentiles, mean, standard deviation and range of groundwater levels. Other collated site data include: tidal amplitude, efficiency and phase lag; distance from harbour or sea; groundwater sample pH, electrical conductivity; modelled seawater percentage; and a rainfall response index reflecting the local efficiency of rainfall recharge.
This report presents derivative ArcGIS data and spatial analysis of these groundwater observations, particularly factors that influence the water table position and geometry. A series of statistical surfaces have been generated to represent the present-day (2019) water table elevation and depth to groundwater, the response to rainfall recharge and tidal forcing, the available subsurface storage of rain and the position of stormwater and wastewater networks relative to the water table. The level of groundwater is influenced by subsurface flow and runoff from the hills in the west and north but will be further encroached from the south and east by sea-level rise in the harbour and ocean in the future.
Simple geometric models have been developed using the present shape and position of the water table, combined with tidal fluctuations, to depict scenarios of groundwater levels with 30, 50 and 80 cm sea-level rise. These geometric models are strongly empirical, with many implicit assumptions and caveats – particularly, that they do not account for groundwater flow and possible changes in water-budget mass balance. Although many variables and controlling processes are simplified into a single parameter, the projected groundwater levels highlight how local variations in the water table shape and slope interact locally with the ground elevation or infrastructure networks. As statistical surfaces representing a period of about one year, they may not necessarily represent short-term flow but are suited to engineering projects, such as foundation design and probabilistic assessment of liquefaction vulnerability.
Dunedin faces a particularly demanding challenge with sea-level rise due to the shape of its topography and coastline, with encroachment from both the harbour and the ocean onto its narrow coastal plain. A valuable lesson from this case study is that the subtle slope and shape of the water table, which has variability at kilometre-scales even in a seemingly flat low-lying area, interacts with land-surface topography to generate notable differences in the depth of groundwater. There are suburb-scale variations in the elevation of the water table and related available subsurface storage that are important for engineering solutions to maintain habitable land. The groundwater spatial datasets, such as water table elevation and rainfall
GNS Science Report 2020/11 v
recharge, provide tools from which inundation or flood-vulnerable areas can be identified and other hazards, such as liquefaction susceptibility, modelled. They will also enable monitoring to be targeted on key areas of groundwater-network exchange so that hazard thresholds and any critical tipping points that might lead to infrastructure system collapse can be identified.
KEYWORDS
Groundwater, Water Table, Tides, Rain, Flooding, Sea-Level Rise, Climate Change, Stormwater, Wastewater, Infrastructure
vi GNS Science Report 2020/11
1.0 INTRODUCTION
1.1 Sea-Level Rise, Groundwater and Infrastructure
The long-term viability of many New Zealand communities is increasingly under question as climate change begins to take effect. Initial work to quantify vulnerability to sea-level rise recently suggested that impacts to New Zealand local government infrastructure alone could reach $8 billion by the time sea level reaches 1.5 m above Mean High Water Springs (perhaps in the next 50 to 75 years) (Simonson and Hall 2019). Given that 65% of New Zealand’s population live within 5 km of the coast, costs to communities go far beyond tangible infrastructure exposure and include potential impact to economic development and growth, health and safety and social support systems. However, assessments likely underestimate the hazard associated with groundwater perched slightly above sea level but hidden in the subsurface.
Where the sea is hydraulically linked to groundwater levels in low-lying coastal areas, a rise of groundwater is expected to be a major problem. It compounds hazards and affects multiple domains and sectors in a cascading and cumulative manner (Bell et al. 2017). By decreasing the unsaturated ground available to store water, rising groundwater will increase the frequency of surface ponding, river flooding and coastal-storm inundation. More directly, it causes instability in building foundations and roads, can infiltrate into and reduce capacity of stormwater and wastewater systems, leads to dampness and mould issues in housing, increases liquefaction potential and can emerge above ground to cause flooding, pollution, salinity stress or other environmental issues. Subsurface stormwater and wastewater networks are vulnerable and potentially prone to system collapse, especially where they are old. Groundwater infiltration reduces network capacity and can result in stormwater and wastewater overflows, with associated public health and environmental costs.
A significant problem faced by society is the cost-benefit return of continued investment in infrastructure in low-lying areas as groundwater levels rise. To fully quantify the risk and exposure of assets to this hazard, there needs to be some understanding of inundation magnitude, frequency and duration and how these characteristics may change in the future. With large infrastructure redesign/renewal imminent under the Three Waters Programme (Department of Internal Affairs 2019), research is urgently needed to inform and optimise investment decisions New-Zealand-wide. At what stage do we switch from defense to managed retreat? Can we predict critical groundwater levels at which infrastructure networks will experience major step changes in efficiency or even catastrophic collapse in performance?
Investment in science to characterise groundwater and drivers of the natural fluctuations and groundwater behaviour in coastal areas is a logical first step. This can provide maps of probability and duration of inundation exceedance against local geology, distance from the coast and waterways. But there is commonly a significant knowledge gap present around the direct and indirect impacts of groundwater levels on stormwater and wastewater networks (and vice-versa). Work is also needed to elucidate local interplay between dynamic natural processes and the effects of human development and infrastructure.
GNS Science Report 2020/11 1
1.2 Coastal Aquifer Systems and the Water Table
Recent work has used LiDAR topographical surveys and GIS analysis to provide indicative calculations of community and asset exposure to effects of sea-level rise (e.g. storm inundation; Paulik et al. 2019). But analysis of vulnerability to groundwater-related hazards is hampered by a widespread lack of water table information within coastal land. Such areas are rarely monitored or studied in detail, principally because the shallow coastal groundwater is usually unsuitable for domestic or agricultural supply. It is commonly assumed that groundwater levels will equilibrate locally with sea level, and simplistic ‘bathtub’ models propagate groundwater through digital terrain models using GIS techniques (e.g. Beca 2014; PCE 2015). Such first- order approximations are useful for non-specific, generalised desktop assessments of regional asset exposure but reflect little of the principles of groundwater systems and the dynamics of subsurface flow (Figure 1.1).
Unconfined groundwater and the natural position of the water is locally dependent on subsurface storativity (void space); permeability; potentiometric gradients from surrounding topography; capillary action; the presence of large water bodies, such as lakes or the sea; and recharge/loss from rainwater and surface waterways. Groundwater surfaces are not flat but instead gently sloped, with humps, ridges and hollows caused by the natural flow system and interplay with urban development, drainage, stormwater systems and other engineering features (e.g. Rotzoll and Fletcher 2013, Hoover et al. 2017, Habel et al. 2017). Humps reflect local inability to flow quickly from rainfall or hillslope recharge or upwards flow, whereas hollows develop through more porous layers or flow to natural or human-made drainage networks.
Shallow groundwater in New Zealand coastal areas typically sits above mean sea level, perched in a marine-sediment aquitard over denser saline water (e.g. van Ballegooy et al. 2014). As a consequence, the low-lying areas potentially affected by rising groundwater can extend considerably further inland than marine inundation (Figure 1.1) and storm surge. There are large variations in facies and related permeability between relatively porous coastal sand dunes, silt-dominated estuarine and marine sediments and gravel- and sand-rich alluvial sediments.
The coastal systems are also dynamic. Groundwater levels fluctuate continuously with inter-annual and seasonal shifts as well as higher-frequency variations caused by storm-events, local river floods or tidal cycles. The groundwater system can be perturbed by changes in precipitation that may result from climate change, as well as the persistent stressor of rising sea level. Tidal fluctuations and influence of sea level on groundwater usually diminish rapidly with increasing distance inland, and the effect of sea-level rise on groundwater systems depends on ground permeability and the strength of hydraulic links to the coast (Glover 1959). Such dynamic fluctuations can be used for groundwater model calibration, and the attenuation in phase and amplitude in high-frequency variations can be tracked laterally and provide information on local ground permeability (Hsieh et al. 1987; Rotzoll and El-Kadi 2008; Rotzoll and Fletcher 2013).
2 GNS Science Report 2020/11
Figure 1.1 Changes to the water table associated with sea-level rise, according to different models and local influences on groundwater systems. Adapted after Rotzoll and Fletcher (2013).
GNS Science Report 2020/11 3
1.3 Dunedin
The city of Dunedin is one of many coastal urban centres in New Zealand facing a massive challenge with sea-level rise. Dunedin is regularly highlighted as a place where there are many assets and critical infrastructure sitting near sea level that are vulnerable (e.g. Bell et al. 2015; PCE 2015; Simonson and Hall 2019; Paulik et al. 2019). There are around 2700 Dunedin homes <0.5 m above the spring high tide mark, which is more than double that in any of the other coastal urban centres in New Zealand (PCE 2015; Figure 1.2), and over 70% are lower than half this elevation. There are also 12 schools, at least six early childhood centres, a stadium, parks and sports grounds, electricity substations, the city’s main wastewater treatment plant, a stormwater pumping station, fuel storage depots and many large commercial and public buildings all situated on land <3 m above mean sea level. Main arterial roads, railway, highways and critical trunk sewerage and stormwater pipes add further to the assets at risk. This low-lying coastal land is crucial to the city’s present functional operation.
Dunedin hills comprise Caversham Sandstone and/or volcanic rocks of the Dunedin Volcanic Group (Benson 1968; Bishop and Turnbull 1996; Barrell et al. in prep). They form local bedrock and are overlain by unconsolidated sediments. Beneath South Dunedin and Harbourside, Quaternary sediments comprising sands and silts deposited under marine to estuarine conditions, and sandy and gravelly sediments laid down in fans and river valleys, form a veneer that is typically <40 m thick, but locally seems to reach as much as 70 m (Barrell et al. in prep). A sand dune barrier and embankment along the southern coast forms a 20-m-high barrier between the Pacific Ocean and South Dunedin (Figure 1.3).
The coastal plain has formed, in geological terms, very recently. During the last glacial period, which ended about 18,000 years ago, sea level was at least 120 m lower than it is now and around 35 km further offshore. At that time, the Water of Leith catchment likely drained to the Pacific Ocean, through the area now occupied by South Dunedin, via a broad (now buried) valley flanked by the hills of the peninsula to the east and the hills to the west (Barrell et al. 2014). Following the peak of the last glacial period, sea level began to rise, reaching its present level about 7000 years ago. The Otago Harbour, as well as most of the other indentations along the Otago coastline, is a former stream valley that was ‘drowned’ by rising sea level (Barrell et al. 2014). The dune barrier subsequently formed between St Clair and Lawyers Head, and fine sediments accumulated in the sheltered water at the head of the harbour. Sediment eventually built up to just above sea level, and a salt marsh and lagoon developed in what is now South Dunedin. Much of the low-lying land was reclaimed from coastal marshes and intertidal deposits following European settlement in the late 1800s, levelled with a thin (~1.0 m) veneer and developed. The harbour margin was reclaimed in the 1960s, extending the 1850 shoreline (Figure 1.3) to the present wharf and rock wall shoreline, which raised the land to around 2 m, at a level above the earlier reclamations to the west and north.
Dunedin’s flat-lying land is susceptible to a variety of natural hazards (Glassey et al. 2003, Goldsmith and Hornblow 2016). Subsurface sedimentary materials are young and weak, amplify during seismic shaking and are predicted to be locally liquefaction susceptible (Cook et al. 1993, Barrell et al. 2014). Nearby hill suburbs locally shed rockfall and landslides, and their catchments deliver runoff onto the coastal plain. The low elevation of the coastal plain, along with its high water table, make it prone to surface ponding after moderate to heavy rain. There is a tendency for surface water to accumulate along the old harbour shoreline (URS and Opus 2011), as adjacent reclaimed land is at a slightly higher elevation.
4 GNS Science Report 2020/11
The ocean has also flooded in through breaches in sand dunes during the late nineteenth century, and a seawall protecting St Clair from coastal erosion has required continued maintenance and reinforcement in the wake of heavy seas.
Large areas of South Dunedin were flooded in June 2015 by a rainfall that was both heavy (144 mm) and intense (9–12 mm/hr) (Goldsmith et al. 2015). The stormwater system was unable to cope with the combination of intense rainfall, hillslope runoff, little fall to the sea and a shallow-depth water table. Drains were overwhelmed, and the extent of surface water ponding and flooding is now widely thought to have been exacerbated by poor maintenance of the drainage infrastructure (Macfie 2016). Damage claims to insurers totalled over NZ$28 million, but the true costs are suggested to be nearer $138 million once the economic and social impacts are also considered (Elder 2016).
Problems of surface flooding and groundwater inundation in low-lying parts of Dunedin are expected to increase with climate change (Rekker 2012; Fordyce 2014; Glassey 2017). The water table is high beneath Dunedin’s coastal plain. It rises and falls in response to both rainfall and tidal flow, and there is little available storage for rainwater, or space before tidal highs begin to cause flooding. In addition, there is some evidence that the land surface may be subsiding, so local rates of relative sea-level rise may exacerbate those expected from global trends (Beavan and Litchfield 2012; Morris 2016). It has been suggested that the combination of subsidence and sea-level rise could result in the 1% Annual Exceedance Probability (AEP) (‘100 year’) coastal flooding event occurring several times a year by the mid-21st century (PCE 2015). Groundwater levels are expected to increase in coastal areas throughout New Zealand in response to sea-level rise, but groundwater beneath South Dunedin and Harbourside’s narrow coastal strip is faced with encroachment from all directions: the harbour in the east, the ocean in the south and changing runoff- and recharge- related flows from hill catchments in the north and west. For simple geometrical reasons of its geography, Dunedin’s challenge is likely to be particularly demanding compared with New Zealand’s other coastal centres.
The Dunedin City Council (DCC) continues to invest and investigate a wide variety of measures to protect its community and their assets, while also considering at what decision points non-protection measures – that is, forms of managed retreat – may eventually be necessary. How to go about managed retreat remains far from clear and was one of the research priorities identified by Local Government New Zealand (Willis 2014). These are fundamentally social and economic issues, requiring community education and consultation regarding mitigation options, risk and likely costs. However, to formulate effective solutions, there is a need for rigorous technical investigations that are informed by quality environmental monitoring and modelling. Dunedin’s future flooding concerns can be addressed through expensive and complex hard infrastructure solutions or by adaptive solutions that account for the natural underlying conditions of low-lying Dunedin. Adaptive solutions are no less complex but may save Dunedin time and money in its effort to manage flooding and groundwater inundation. Here, we present observations and initial findings from the latest groundwater monitoring. This report provides information to underpin future planning and mitigation of natural hazards in Dunedin.
GNS Science Report 2020/11 5
3000
2500
2000
1500
1000 Number of Homes
500
0
Figure 1.2 Number of homes <50 cm elevation above the spring high tide mark, for nine coastal urban areas in New Zealand (using data from PCE 2015).
Figure 1.3 Elevation model of Dunedin from the 2009 LiDAR survey. The position of the shoreline in 1850, prior to reclamation of land, is depicted with a blue line.
6 GNS Science Report 2020/11
2.0 DUNEDIN DATA AND REFERENCE DATUM
2.1 Groundwater Monitoring Network
Knowledge of groundwater in Dunedin took a major step forward with installation of a network of additional piezometers during 2019. Groundwater monitoring in the lower-elevation land of Dunedin now includes a wide variety of piezometer depths, diameters, screen lengths and types of pressure transducer. Installation of piezometers involved a series of both individual and multi-agency efforts. This report is based primarily on observations from 23 sites in the South Dunedin to Harbourside area, where groundwater is monitored by a combined Otago Regional Council (ORC) and GNS Science effort.1
Data have been managed and archived in the ORC Hilltop database, referenced using the ORC well numbers. Other information that has been utilised includes: (i) a groundwater observation campaign by University of Otago Geography student Emma Fordyce (Fordyce 2014), who monitored sites in South Dunedin between July 2012 and January 2013, and (ii) observations at Kings (I44/1105) and Bayfield (I44/1106) High Schools from piezometers installed during a Curious Minds science project awarded to the New Zealand International Science Festival in 2017. A summary of the groundwater monitoring is provided in Table 2.1, a list of piezometer numbers and names in Table 2.2 and their location in Figure 2.1.2
The groundwater levels observed in the piezometer network have been calibrated by manual observations every 6–8 weeks using an electronic tape measure, reading a depth to groundwater measurement (DTW; negative = above ground, positive = below ground) in metres from the measuring point. Measuring point elevations of the monitoring piezometer reference levels are all set at the top of the piezometer lining or casing (ToC), with a local offset for the distance of this level above or below ground (GL) in metres, following a convention where above ground measuring point (MP) produces a negative GL value from GL–MP (Figure 2.2).
Measuring point elevations were surveyed on either 21 April 2018, 13 February 2019 or 12 June 2019 using a University of Otago Survey School RTK GPS to a local reference (e.g. the top of the toby cap or top of the concrete pad) with an offset measured using a tape measure, then referenced back to the DUND base station at the University of Otago Survey School. Vertical precision varied from site to site, but was on average ± 0.04 m and at maximum ± 0.07 m. The older (prior to 2019) ORC sites have published ground levels in the earlier datum DVD58, which were resurveyed, checked and converted to NZVD2016. Piezometers installed by Fordyce (2014) had elevations surveyed relative to local survey benchmarks, using LINZ published values. These were recalculated to NZVD2016 using the 2020 published elevations for the benchmarks, but appear to have a consistent ~0.3 m value above the absolute elevation and expected ground level from the LiDAR DEM. The quality of each elevation estimate is recorded as QAR = 1 for GPS survey and QAR = 2 for Fordyce data (also following the Environment Canterbury convention).
1 Geotechnical investigations of the new hospital site in the Cumberland–Castle Street (Harbourside) area were undertaken during 2019, including subsurface drilling. Groundwater monitoring, initiated by Tonkin & Taylor from September 2019 onwards, has yet to be publicly released and has not been incorporated into this report. 2 Hilltop is a time-series software widely used by regional councils in New Zealand to manage their environmental monitoring data. See http://www.hilltop.co.nz/.
GNS Science Report 2020/11 7
Figure 2.1 Groundwater monitoring sites in Dunedin. Piezometers used in this study (coloured by campaign) are shown together with an outline of the flat-lying land and modelling extent, harbour and coastal control points included in the models. Data from ORC and Curious Minds piezometers are telemetered, whereas others are downloaded periodically. The figure also shows an area in Forbury that is problematic for modelling (yellow ellipse) and the interpreted extent of a perched aquifer in the sand dunes at St Kilda (blue). A map showing place names used in this study is included in the appendices.
Barometric corrections have been applied for groundwater monitoring sites with unvented pressure transducers where there is a need to remove atmospheric pressure from the total pressure in order to obtain a measure of groundwater pressure. The same (standard) air pressure observations have been applied for all data collected during 2019 using a barometer at the ORC office (Stafford Street). All measurements assume groundwater has equal, and constant, 1000 kgm-3 density. A brief six-week experiment showed that air pressure differences between the office barometer and an air-pressure transducer at Tonga Park were insignificant, given the levels of barometer uncertainty and 10 minute sampling intervals. Fordyce (2014) corrected unvented observations in 2012 using local readings of air pressure from a transducer in Richardson Street.
8 GNS Science Report 2020/11
Table 2.1 Summary of groundwater monitoring collated for this study.
Piezometer or Campaign Overview Pressure Transducer Well Information ORC Four long-term groundwater level monitoring Cased drill holes with 80 mm diameter Vented transducers. Telemetered 15 minute sites at Kennedy St, Bathgate, Tonga and piezometers, 4.2–6 m deep. Upstands were data. Manual dips and site inspections at Culling Parks. Long-term data ongoing from installed at Tonga and Kennedy St following monthly intervals. 2010, presently recording at 15 minute intervals. 2015 floods. Else toby box flush with ground. Fordyce Nine temporary sites recording water levels PVC pipes 50 mm diameter, c. 3 m deep, Unvented HOBO U20 level loggers at 15 minute 2012–2013 at 15 minute intervals from 13/07/2012– with holes drilled in lower 2 m covered by sampling P, air-pressure-corrected to logger at 17/09/2012, with some data collected 12/2012– 200 µm mesh sock. Hand-augured (70 mm) Richardson St. Data downloads and calibration 01/2013. Raw data was not archived, but some and placed in a sand-gravel pack with pipe fortnightly. data (pressure-corrected water levels relative to upstand of ~200 mm above ground. ground) was supplied by the author in 2018. Absolute elevations of measuring points appear to have ~0.3 m absolute uncertainties, but relative data has 2–3 mm precision. Curious Minds Two piezometers at Kings and Bayfield High 32-mm-diameter (ID 25 mm) HDPE pipes Seametrics CT2X sensor recoding P, T and Schools. Water level and temperature monitoring installed to 3.7 m using a lost cone specific conductance. 12 V system with solar at 15 minute intervals ongoing from 13/06/2017, penetrometer. Slotted over lower 3 m with filter panel. Van Walt Ultima Mark2 GSM logger but with data gaps and calibration issues. sock, cased in washed sand and capped with telemetering data to www.vanwaltconnect.com. Data contains periods where there is drift bentonite clay to seal blind pipe. Toby box Irregular manual calibration. seemingly related to battery voltage, which were flush to ground. corrected manually for this study. Consortium Eight shallow piezometers installed to depths Installed using CPT drill rig by Golder Unvented Seametrics Level Scout loggers. (ORC, DCC, between 2.8 and 6 m. Water level and Associates. 42-mm-diameter PVC pipes. Air-pressure-corrected to ORC office barometer GNS Science, EQC, temperature monitoring at 15 minute intervals, Lower 3 m slotted, cased in K2 sand (96 mm (Stafford St). Manual dips every 4–6 weeks for Oceana Gold, ongoing since 6/03/2019. Datasets contain borehole diameter) and capped with bentonite calibration. Data downloads every three months. Canterbury University) drawdown/recovery from purging for regular seal. Toby box flush to ground. Some observations with cableless Seametrics sampling. Data over these periods needs to be CT2X electrical conductivity logger (roving removed to understand natural fluctuations and between sites). statistics.
GNS Science Report 2020/11 9
Piezometer or Campaign Overview Pressure Transducer Well Information NZ SeaRise Three deep (21, 22 and 45 m) and one shallow 125 mm boreholes installed by Roto Sonic 3” Unvented Seametrics LevelSCOUT loggers. (GNS Science and ORC, (6 m) drill hole installed May–June 2019. drilling (McNeil Drilling Group) with 80-mm- Air-pressure-corrected to ORC office barometer with VUW) Extra geological work done on core material diameter casing. 3 m screens near base of (Stafford St). Manual calibration dips every for NZ SeaRise project. Water level and holes, cased in sand and gravel, capped with 4–6 weeks. Data downloads every 3 months. temperature monitoring at 15 minute intervals, bentonite backfill. Two with lockable steel ongoing since 13/06/2019. Datasets contain upstands, two with toby box flush to ground. drawdown/recovery from purging for regular sampling. Harbourside Five drill holes (6.5–16 m deep) in the 125 mm boreholes installed by Roto Sonic 3” Unvented Seametrics LevelSCOUT loggers. (ORC and GNS Science) harbourside installed May–June 2019. drilling (McNeil Drilling Group) with 80-mm- Air-pressure-corrected to ORC office barometer Water level and temperature monitoring ongoing diameter casing. 3 m screens near base of (Stafford St). Manual calibration dips every since 13/06/2019. Datasets contain holes, cased in sand and gravel, capped with 4–6 weeks. Data downloads every 3 months. drawdown/recovery from purging for regular bentonite backfill. Three with lockable steel sampling. upstands, two with toby box flush to ground.
10 GNS Science Report 2020/11
Figure 2.2 Definition of ground level and measuring point as used in the Dunedin groundwater monitoring dataset, similar to that used by Environment Canterbury. Figure from van Ballegooy et al. (2014).
GNS Science Report 2020/11 11
Table 2.2 Dunedin groundwater monitoring sites operational in 2019.
Total Depth and GPS MP GL–MP Easting Northing Lowermost Height. Number Name Elevation Offset (m) (m) Depth of Precision NZVD 2016 (m) Screen, if (95%) Different (m) I44/0005 Bathgate Park 1405334.7 4914364.8 6.0 1.113 ± 0.042 0.050
Tonga Park I44/0006 1405348.5 4913652.2 6.0 1.574 ± 0.036 -0.855 (shallow)
I44/0007 Kennedy St 1405662.3 4912961.9 6.0 2.032 ± 0.050 -0.855
I44/1094 Culling Park 1407253.0 4913766.7 4.2 0.464 ± 0.034 0.065
I44/1105 Kings 1405673.4 4913894.6 3.6 0.735 ± 0.026 0.035
I44/1106 Bayfield 1407847.5 4914693.2 3.7 1.767 ± 0.036 0.025
I44/1113 Marlow St 1407672.7 4914098.9 5.9 0.693 ± 0.024 0.110
I44/1114 Timaru St 1406879.1 4914385.2 5.6 1.232 ± 0.026 0.100
I44/1120 Pretoria Ave 1405339.1 4913409.3 5.0 0.739 ± 0.034 0.080
I44/1121 Fitzroy St 1404905.8 4914105.3 2.8 4.517 ± 0.045 0.095
Andersons I44/1122 1406198.0 4915115.0 6.0 0.996 ± 0.026 0.100 Bay Rd
I44/1123 King Edward St 1406244.3 4914166.8 6.1 0.683 ± 0.028 0.150
I44/1124 Council St 1406458.7 4913576.9 6.1 0.777 ± 0.027 0.115
I44/1125 Alma St 1405825.4 4913390.2 6.0 0.824 ± 0.034 0.070
CE17/0100 Oval 1405869.7 4915713.6 14.9 (6.5) 2.746 ± 0.020 -0.555
Queens CE17/0101 1406325.5 4916556.1 15.9 2.937 ± 0.041 -0.575 Gardens
CE17/0102 Tewsley St 1406860.7 4916498.5 14.7 1.865 ± 0.026 0.045
CE17/0103 Harrow St 1406959.1 4917159.3 14.7 (13.5) 2.169 ± 0.027 0.055
CE17/0104 Logan Park 1407775.4 4917644.8 14.5 (14.3) 2.753 ± 0.062 -0.455
Tonga Park CE17/0105 1405365.4 4913645.6 44.5 (19.95) 1.381 ± 0.066 -0.650 (deep)
CE17/0106 De Carle Park 1406382.7 4913559.0 20.7 (19.9) 1.491 ± 0.034 -0.660
Moana Rua CE17/0107 1406276.7 4913174.4 21.8 3.679 ± 0.037 0.055 (deep)
Moana Rua CE17/0108 1406281.7 4913175.7 6.3 3.744 ± 0.038 0.045 (shallow)
12 GNS Science Report 2020/11
2.2 Surveying, Reference Frames, Conversions and Other Data
Detailed information on surveying coordinates, the latest datum, heights and conversions are provided by Land Information New Zealand (LINZ; see https://www.linz.govt.nz/data/geo detic-system/datums-projections-and-heights). All coordinates in this report are given in metres relative to New Zealand Geodetic Datum 2000 (NZGD2000) and Vertical Datum 2016 (NZVD2016). Maps and diagrams are projected in New Zealand Transverse Mercator (NZTM2000). Data have been converted to NZVD2016 from the local vertical datum Dunedin 1958 (DVD58) by subtracting 0.377 m using a local relationship grid (as opposed to a national geodetic adjustment) and from local chart datum by subtracting a 1.368 m offset. A local Dunedin drainage datum has a zero reference 100 m below DVD58.
As there is a tidal amplification effect due to the geometry and bathymetry of Otago Harbour, this study draws on sea-level observations from two sites. Levels in the open Pacific Ocean are monitored every 1 minute at Green Island (1397490 mE, 4907950 mN), 9 km southwest of the city, by ORC. Levels in the harbour are monitored at a 10-minute frequency at Fryatt St wharf (1406517 mE, 4916154 mN) in Dunedin by Port Otago Ltd.
The harbour and ocean statistics are distinguished in this report by the terms Mean Level of Sea (MLOS) and Mean Harbour Level (MHL), where necessary. Mean Sea Level (MSL) is used in a more general context, for example, when referring to global models of sea level. The differences can be seen through simple comparison of the Green Island and Fryatt Street dataset statistics (Table 2.3), which are raw values without corrections for storm surge, wave set up, waves, etc.
The predicted values for Dunedin Harbour 2019–2020 are given in Table 2.4 (after the LINZ website and Bell et al. 2017) and are comparable to statistics for observed harbour levels. Grid model interpolations generated in this report to represent the 2019 period of groundwater monitoring use coastal control points along the harbour and ocean margins of the models, which have been assigned ‘2019’ statistical values based on harbour or ocean ‘2019’ values from Table 2.3. Mean High Water Springs (MHWS) have not been calculated separately, as there are many ways of doing this and the method is open to interpretation (Bell et al. 2017). Instead, MHWS in the Otago Harbour was taken as 1.11 m above MHL based on the Table 2.4 predictions, and MHWS in the ocean was taken as 0.85 m above MLOS (Wild et al. 2005, based on the New Zealand tidal model of Walters et al. 2001).
The 2009 Dunedin LiDAR survey and derived digital elevation model (DEM) of the land is an important reference dataset for this study (Farrier 2009). Data initially obtained in DVD58 have been converted to NZVD2016 using the grid transformation published by LINZ. Equipment manufacturer’s specification suggests accuracy should be <10 cm (1 sigma) at the chosen sampling frequency and flight elevation. The grid was checked against a local survey network, with elevation differences returning ± 0.12 m at 95% confidence level (Farrier 2009).
Rainfall observations (daily and hourly) presented in this report are based on New Zealand MetService data collected at Musselburgh since 1918 (Automatic Weather Station 93892), which is taken as the longest, most representative, site for rainfall on Dunedin’s coastal plain. Models of the frequency of high-intensity rainfall events, in depth-duration-frequency tables and future climate change projects, are provided by the National Institute of Water & Atmospheric Research (NIWA) and have been determined for the Musselburgh site (Carey-Smith et al. 2018).
GNS Science Report 2020/11 13
Table 2.3 Comparison of longer- (shaded light grey) and short-term statistics of raw data from Otago Harbour (Fryatt St, Port Otago Ltd) and the open ocean (Green Island, ORC). Elevation data are in metres NZVD2016.
Harbour Harbour Ocean Ocean ‘2010–2018’ ‘2019’ ‘2015–2018’ ‘2019’ Start date/time 6/03/2010 00:00 6/03/2019 00:00 1/01/2015 00:00 6/03/2019 00:00
End date/time 5/03/2019 11:50 10/01/2020 15:40 5/03/2019 11:59 12/02/2020 00:00
Mean level -0.235 -0.217 -0.103 -0.073
Standard deviation 0.625 0.603 0.571 0.559
No. of observations 470,892 44,694 1,954,553 493,874
Maximum level 1.359 1.182 1.367 1.281
95th percentile level 0.732 0.702 0.772 0.769
Median level -0.249 -0.224 -0.109 -0.072
15th percentile level -1.168 -1.123 -0.963 -0.917
Minimum level -1.701 -1.546 -1.677 -1.411
Range 3.060 2.727
Table 2.4 Dunedin Port tidal level averages predicted for the period 1 July 2019 – 30 June 2020 (www.linz.govt.nz/sea/tides/tide-predictions/standard-port-tidal-levels), with conversion to NZVD2016, and the published average value that was observed for the period 1986–2005 (Bell et al. 2017) and AFEQ elevation (LINZ).
Predicted Elevation Observed Abbreviation Tidal Term 2019–2020 2019–2020 NZVD2016 Chart Datum NZVD2016 HAT Highest Astronomical Tide 2.43 1.07 -
MHWS Mean High Water Springs 2.22 0.86 -
MHWN Mean High Water Neaps 1.81 0.45 -
MHL Mean Harbour Level 1.11 -0.25 -0.301
MLWN Mean High Water Neaps 0.44 -0.92 -
MLWS Mean Low Water Springs 0.07 -1.29 -
LAT Lowest Astronomical Tide -0.14 -1.50 -
AFEQ Local Survey Marker AFEQ 3.73 - 2.37
2.3 Longer-Term Context of 2019 Groundwater Observations
This report is focused principally on presenting new observations from 2019 (and associated statistical data) that have been collated for the period 6/03/2019–12/02/2020. To draw conclusions from these data, it is important to examine their context within longer-term variability. In comparison to observations of rainfall and sea level, which are readily available over many decades, Dunedin groundwater data are only available over a relatively short period at four shallow <6 m piezometers installed by ORC: since 2010 at Bathgate Park (I44/0005), Tonga Park (I44/0006) and Kennedy St (I44/0007), and since 2014 at Culling Park (I44/1094).
14 GNS Science Report 2020/11
Statistics from 2019 were plotted against those from 2010–2018 (Table 2.5) to evaluate the new observations against recent interannual variability. Median and mean groundwater levels in 2019 are within 0.04 m of the 2010–2018 values, but there was no systematic climatic shift of groundwater to lower or higher levels during 2019. Depending on the site in question, the median (or mean) values for 2019 may be either higher or lower than 2010–2018 equivalent. There is nothing to suggest from basic statistics that the 2019 monitoring period reflected any anomalously wet, or dry, condition. Instead, it seems the 2019 groundwater levels are a reasonable proxy (± 0.04 m) for average conditions during the past decade.
Plots of groundwater level against the numerical day of the year (Julian day) show some tendency towards lower levels during late summer to early autumn (April) but are not clearly sinusoidal, suggesting well-defined or regular seasonal behaviour (Figure 2.3). Kennedy Street (I44/0007) has strong tidal behaviour that is rarely disturbed from expected solar and lunar cycles. Elsewhere, fluctuations are dominated by peaks and recessions over daily to weekly intervals that are coincident with rainfall events. There is a hint of some cyclicity at a period of 90–100 days that may reflect cumulative rainfall and recharge affected by the frequency of storms. The cyclicity shows good correlation with daily rainfall averaged over the previous 30 days (Figure 2.4).
Table 2.5 Comparison of statistics for longer-term (shaded light grey) and 2019 statistics of groundwater levels at four ORC monitoring sites. Levels are given in metres NZVD2016.
ORC No. I44/0005 I44/0006 I44/0007 I44/1094 Name Bathgate Park Tonga Park Kennedy St Culling Park
Start date 6/03/2010 6/03/2019 6/03/2010 6/03/2019 6/03/2010 6/03/2019 1/06/2014 6/03/2019
End date 5/03/2019 12/02/2020 5/03/2019 12/02/2020 5/03/2019 12/02/2020 5/03/2019 12/02/2020
Sample rate 5–15 min 15 min 5–15 min 15 min 5–15 min 15 min 5–15 min 15 min
Mean level 0.352 0.388 0.159 0.134 0.211 0.197 -0.233 -0.208
Standard 0.079 0.079 0.104 0.100 0.173 0.163 0.234 0.215 deviation
Observation 465851 32834 451801 32907 465547 32903 428727 30104 No.
Maximum 0.611 0.529 0.989 0.349 1.294 0.730 0.528 0.468 level
95th percentile 0.459 0.487 0.313 0.266 0.493 0.465 0.162 0.160 level
Median level 0.364 0.399 0.172 0.150 0.208 0.193 -0.218 -0.190
5th percentile 0.205 0.221 -0.02 -0.076 -0.058 -0.053 -0.63 -0.602 level
Minimum 0.052 0.178 -0.121 -0.118 -0.349 -0.195 -0.793 -0.645 level
Range 0.559 0.351 1.366 0.467 1.643 0.925 1.321 1.113
Diff. medians 0.035 -0.022 -0.015 0.028
GNS Science Report 2020/11 15
Figure 2.3 Plots of 2019 groundwater elevation (orange) against the numerical day of the year (Julian days as 0–365 values) at the four ORC sites with longer-term data (blue).
16 GNS Science Report 2020/11
Figure 2.4 (Top) Plots of 2019 daily rainfall, and (bottom) groundwater elevation at ORC piezometers compared with the daily rainfall at Musselburgh station averaged over the previous 30 days (blue).
The highest recorded groundwater levels were those associated with flooding on 3–4 June 2015 (144 mm rain, 9–12 mm/hr) and high levels also occurred after an intense rainfall 21–22 July 2017 (93 mm, <7 mm/hr). Graphs of percentile distributions (Figure 2.5) highlight that the groundwater level extremes experienced at Tonga Park and Kennedy Street during 2010–2018 were significantly higher than maximum levels recorded in 2019. Tonga Park (I44/0006) reached 0.64 m higher during the 2015 flooding than the maximum experienced in 2019, and Kennedy St reached (I44/0007) 0.56 m higher than its 2019 maximum. On this basis, it seems that groundwater levels in future can certainly be expected to exceed the maximum levels measured in 2019, at least locally, by 0.5 m or more.
GNS Science Report 2020/11 17
Figure 2.5 Plots of 2019 groundwater elevation (red 2019/03–2020/02) against percentiles of the observation statistical distribution at the four ORC sites, compared against the longer-term 2010–2018 data records (blue 2010/03–2019/02 and green 2014/06–2019/02). Ground level at Bathgate Park (not shown) is 1.15 m (NZVD2016).
Although event rainfall and seasonal variability can be assessed directly from the decadal records, the groundwater level records are of insufficient length to directly assess if there are longer-period fluctuations in Dunedin. Variations in El Niño–Southern Oscillation (ENSO) cycles occur at interannual two- to four-year cycles, while Inter-decadal Pacific Oscillation (IPO) cycles occur at longer (15- to 30-year) intervals. These have the potential to affect local precipitation and groundwater levels. It is possible that models could be developed by extrapolating 2010–2019 groundwater-rainfall relationships backward in time, but this was deemed beyond scope of this initial exploration of the 2019 data.
18 GNS Science Report 2020/11
3.0 INTERPOLATED STATISTICAL SURFACES OF GROUNDWATER
Statistical values have been derived for each groundwater monitoring site, then interpolated into a series of raster grids with 20 m cells. While kriging is a geostatistical method commonly adopted for interpolating potentiometric surfaces, the small number of monitoring sites in Dunedin and their spatial distribution does not lend itself to global statistical approaches such as this. Instead, a Topo to Raster function in ArcGIS was adopted, which is a discretised thin plate technique, based on the ANUDEM program (Hutchinson 1989; Hutchinson and Dowling 1991), that was designed for the creation of hydrologically correct digital elevation models. It recognises that topographic landscapes have many ridges (local maximums) and few sinks (local minimums) and uses this information to impose constraints on the interpolation. Similarly, shallow groundwater surfaces can be expected to have numerous maxima (recharge points) and fewer or smoothed minima. The method provides hydrologically informed surfaces that are fitted through the observation points and can be extrapolated beyond data points.
‘Statistical surfaces’ are quite distinct from ‘static surfaces’ or ‘event-related surfaces’, in that the latter represent the state of groundwater at a particular point in time, or are associated with a short change during an event, whereas the statistics represent a range of states during a longer period – in this case, of at least one year or more. The maximum level of groundwater, for example, can occur at different times at various sites. Therefore, an interpolated grid of maximum values may not necessarily represent the highest level of groundwater level reached during a single storm or rainfall event. An important caveat is that, while statistical interpolations may approximate the potentiometric pressure of groundwater over extended periods, such as long-term hydraulic gradients, caution is required when they are used to interpret short-term behaviour such as directions of flow. Interpolation of discrete data points on a surface also does not necessarily mean there is hydraulic connection to other points nearby. Statistical surfaces are particularly useful for engineering models and solutions, such as incorporating maximum to minimum ranges in foundation design or distributions in the probabilistic assessment of liquefaction vulnerability (e.g. van Ballegooy et al. 2014, 2015).
The Otago Harbour and Pacific Ocean are water bodies that both have some influence on the behaviour of groundwater beneath the city and have important potential to impinge on the groundwater in future. Observations of groundwater chemistry and specific electrical conductance can be used as a proxy for salt water / freshwater mixing and show the degree of saline incursion into the groundwater system. The maximum recorded is ~80% at Kennedy St (I44/0007), but there are also some strongly tidal sites (e.g. Tewsley St, CE17/0102) that have relatively fresh groundwater that is loaded/unloaded elastically without flow of saline water in and out of the sediment (see Section 3.3 below).
Interpolated surfaces in this study have been constrained at the harbour and coast using a series of boundary ‘control points’, where values have been assigned to help rationalise the physics and improve extrapolation beyond the monitoring sites. Essentially, it has been assumed that there is some degree of hydraulic connection, or unconnected loading effect, between groundwater and the harbour or sea. Harbour and coastal points are assigned statistics, such as MHL and MLOS derived from Fryatt Street or Green Island tide gauges, then included in the interpolation. Control points have also been included along Andersons Bay inlet, which equilibrates with the upper harbour at high tide but is constrained to a minimum level around 0.19 m, due to a causeway constructed for Portobello Road.
GNS Science Report 2020/11 19
3.1 Groundwater Elevation
A series of statistical surfaces representing the elevation of the water table and overall state of groundwater during 2019 have been developed from monitoring observations between 6/03/2019 and 12/02/2020. A summary of these data, and how they were derived, is provided in Table 3.1.
The median level of groundwater (Figure 3.1) is everywhere elevated above the median levels of the sea in the harbour and open ocean during the same period (MHL= -0.25 or MWOS = -0.11 m [NZVD2016] – see Table 2.2). A prominent ridge, between 0 and 0.5 m, extends across South Dunedin from Forbury racecourse to King Edward Street. Highest median values are those adjacent to the hill suburbs, with slopes suggestive of flow out through the flat-lying South Dunedin and Harbourside land. The interpolation shows steep gradients in the Caversham to Forbury area that are largely a function of observations at Fitzroy St (I44/1121) but with values ~0.005 m/m that are not unreasonable for hydraulic gradients expected in young, unconsolidated, fine-grained sediments recharged from nearby hills (e.g. see Freeze and Cherry 1979). A model of the 2019 range of groundwater levels is dominated by tidal changes in the harbour and ocean, with little variation in the centre of South Dunedin (Figure 3.2).
Table 3.1 Summary of statistical surfaces and other datasets developed to represent Dunedin groundwater elevation (GWL) during 2019. For a more detailed description of these data, see Table A3.1.
File Name Summary GWLmax_2019 Maximum level of groundwater (metres NZVD2016) interpolated from measurements in 2019.
GWLmedian_2019 Median level of groundwater (metres NZVD2016) interpolated from measurements in 2019.
GWLmedian_2019_25cm Contours Contours (at 25 cm intervals) of Median level of groundwater (metres NZVD2016) interpolated for 2019.
GWLmin_2019 Minimum level of groundwater (metres NZVD2016) interpolated from measurements in 2019.
GWLp05_2019 5th percentile groundwater level (metres NZVD2016) interpolated from measurements in 2019.
GWLp95_2019 95th percentile groundwater level (metres NZVD2016) interpolated from measurements in 2019.
GWLrange_2019 Range (maximum to minimum) of groundwater levels (m) interpolated from measurements in 2019.
GWLmean_2019 Mean level of groundwater (metres NZVD2016) interpolated from measurements in 2019.
GWLstdev_2019 Standard deviation of groundwater level (m) interpolated from measurements in 2019.
GWLmhws_mean2019 A modelled MHWS level of groundwater above the mean value from GWLmean_2019.
GWLmhws_median2019 A modelled MHWS level of groundwater above the median value from GWLmedian_2019.
GWLdiff_medianmean A grid of difference between interpolated median and mean groundwater levels in Dunedin 2019. Values in metres.
20 GNS Science Report 2020/11
Figure 3.1 Interpolated median groundwater elevation (MedianGWL_2019), with elevation ranges and gradients highlighted by contours at 25 cm and 1 m intervals. Monitoring sites used in the interpolation, annotated with their median values, are indicated by white dots.
Figure 3.2 Interpolated model of the range (GWLrange_2019 in m) of groundwater levels in Dunedin.
GNS Science Report 2020/11 21
3.2 Depth to Groundwater
A series of depth to groundwater (DTW) surfaces (Table 3.2) were derived by subtracting the groundwater elevation grids from the digital elevation model generated from the 2009 LiDAR survey. The grid surfaces (20 m cells) map the depth (m relative to ground level) based on the 2019 grids interpolated from monitoring during 6/03/2019–12/02/2020.
The median DTW is shown, together with the median level contours and piezometer locations, in Figure 3.3. The figure highlights that DTW does not necessarily reflect topographic elevation, in that the lowest-lying suburbs do not necessarily coincide with shallowest groundwater (as suggested by ‘bathtub models’). Instead, there is also shape and slope to the water table, which are important at the kilometre-scale from suburb to suburb. However, much of South Dunedin groundwater is at depths less than 1 m, on average, and in many places it is less than 0.5 m, which leaves little subsurface storage available for rainfall or seawater incursion (further discussed in Section 6).
Negative DTW values potentially indicate places where groundwater is actually above ground and causes inundation. Alternatively, they may simply occur where the model may be at fault because of differences in the nature and spatial precision of topographic and groundwater datasets. For example, in the Forbury to Caversham area, interpolations are strongly influenced from the Fitzroy St (I44/1121) piezometer. Gradients projected west from this site into Caversham valley appear sensible and in line with groundwater drainage from a hillslope catchment. However, interpolation to the east produces a smooth surface (all the way to Bathgate I44/0005 and Tonga I44/0006 parks) with DTW<0 because the groundwater interpolation is above a local step in the ground surface. Although this area was hard hit in the June 2015 flood and is vulnerable to flooding, this may have been largely due to surface runoff from the nearby hills, rather than groundwater inundation. Here the negative values may simply be an artefact of the piezometer distribution and high-precision LiDAR surveying. It is not conclusively established that Forbury groundwater has positive artesian pressures or produces springs in this area, and there could be distinct layers and/or compartmentalisation between Fitzroy St to Caversham. The area is problematic for modelling, given the distribution of piezometers (see Figure 1.1). The DTW datasets (e.g. Figure 3.3) highlight an important place to invest in a greater density of monitoring sites.
The magnitude of any negative values, while potentially indicative of groundwater at the surface, are very unlikely to reflect the actual magnitude of any artesian pressure, nor indicate potential ponding depth – if only because these are statistical rather than static surfaces or event-related surfaces. To remove potential confusion, some DTW grids with the suffix ‘ag0’ have been generated where grid cells with negative ‘above ground’ values were reset manually to zero.
22 GNS Science Report 2020/11
Figure 3.3 Interpolated depth to median groundwater elevation (MedianDTW_2019) with median level contours at 25 cm and 1 m intervals. Monitoring sites used in the interpolations are indicated by white dots.
Table 3.2 Summary of datasets developed to represent depth to Dunedin groundwater during 2019. For a more detailed description of these data, see Table A3.2.
File Name Summary DTWmax_2019 Depth (below ground) of the maximum level of groundwater interpolated for 2019. Values in m below ground level.
DTWmean_2019ag0 Depth (below ground) of the average level of groundwater interpolated for 2019, with negative (above ground) 'depths' reclassified to value=0.
DTWmedian_2019 Depth (below ground) of the median level of groundwater interpolated for 2019. Values in m below ground level.
DTWmedian_2019ag0 Depth (below ground) of the median level of groundwater interpolated for 2019. Values in m below ground level.
DTWmin_2019 Depth (below ground) of the minimum level of groundwater interpolated for 2019. Values in m below ground level.
DTWp05GW_2019 Depth (below ground) of the 5th percentile of groundwater level interpolated for 2019. Values in m below ground level.
DTWp95GW_2019 Depth (below ground) of the 95th percentile of groundwater level interpolated for 2019. Values in m below ground level.
DTWmean_2019 Depth (below ground) of the average level of groundwater interpolated for 2019. Values in m below ground level.
DTWmhws_mean2019 A modelled depth (below ground) MHWS level of groundwater based on the mean value from GWLmedian_2019.
DTWmhws_median2019 A modelled depth (below ground) MHWS level of groundwater based on the median value from GWLmedian_2019.
GNS Science Report 2020/11 23
3.3 Effect and Reach of Tides
Ocean tides often cause water levels in coastal aquifers to fluctuate. These fluctuations result from at least two different causes. In confined aquifers, the additional weight of the water on the surface at high tide increases the pressure on the water at depth, which causes an increase in water levels in piezometers and/or observation wells located near the coast. In this case, no movement of water between the ocean and the aquifer need have occurred. In unconfined aquifers, however, the loading effect does not occur. Instead, saltwater that has moved directly through the pores within the aquifer causes a groundwater level increase.
Chemical analyses provide an independent way to examine saline incursion, seawater to freshwater mixing and how it varies in the South Dunedin and Harbourside groundwater. During 2019, groundwater from the piezometers was sampled at regular intervals. Samples were initially collected using bailer to obtain the water standing in the piezometer, but later, samples were taken using a peristaltic pump to purge the standing water first and then collect samples after water levels had recovered. Values of pH and electrical conductivity (EC) were measured at all sites using either a Hanna instruments HI 98130 pH/EC meter, calibrated using solutions pH = 4.01 and 7.01 and EC = 84 µS/cm, or a YSI Pro Plus3 that has an upper limit of 200,000 µS/cm. All values of conductivity were converted to specific conductance at 25°C in units of µS/cm using the most commonly adopted 0.019 (19%) temperature correction coefficient (being the value derived for a pure potassium chloride solution).
After electrical conductivity meters had been installed, in some piezometers it was discovered that purging had generated a lasting effect on chemistry of the groundwater in the piezometer (see Section 4.4) and that some samples were likely to be mixtures of the water and surface water. All analyses were re-examined, then a set of representative conductance values were selected for each piezometer – where possible, based on the field measurement of static water, but otherwise by correcting post-purge samples to a pre-purge expected background value. The conductivity values reported in Fordyce (2014) have also been used, although their units appear to have been incorrectly labelled in parts of the original work, with values reinterpreted as being reported in S/m. All electrical conductivity values in this study and its database are in units of µS/cm.
To model the incursion of seawater, a grid model of selected conductance values was generated using control points assigned with values from water samples collected in the harbour and sea (Figure 3.4), then a grid and contours of %Seawater were derived from a simple freshwater to seawater mixing model (where the freshwater end member was assumed to have specific conductance of 100 μS/cm and seawater 50860 μS/cm).
3 https://www.ysi.com/proplus
24 GNS Science Report 2020/11
Figure 3.4 Electrical conductance of groundwater (μS/cm) as piezometer spot values and an interpolated grid, overlain by modelled %Seawater contours (at 10% contour intervals). These data are not applicable to the overlying perched aquifer in the St Kilda sand dunes (blue diagonal hatch areal), which is relatively fresh and appears not to mix directly with the ocean.
Comparison of tidal responses with groundwater chemistry (and electrical conductivity in particular) indicates that there are examples of both mixing and loading processes generating tidal responses in South Dunedin and Harbourside groundwater: Tewsley St (CE17/0102) has relatively fresh groundwater that is loaded/unloaded elastically by harbour tides (tidal efficiency ~34%) without flow of saline water in and out of the piezometer; whereas Kennedy St (I44/0007) has a similarly strong tidal response (tidal efficiency ~30%) but instead is brine-rich (~80% sea water), reflecting direct mixing of groundwater with inland flow from the ocean.
Regardless of the mechanism, the tidal fluctuations can be described by measuring phase lag (in minutes), or delay, between the tidal peak and the groundwater level response peak, and by the tidal efficiency (TE) amplitude ratio between the magnitudes of the tidal range of groundwater (ΔW) and the tidal range (ΔT).
The dimensionlessTE = ΔW/ΔT tidal efficiency ratio is presented as a percentage in this report. We found it useful to distinguish the full tidal range (which occurs between high and low tide over 12.42 hours) from the tidal amplitude (between mid- and high tide over 12.42 hours). The tidal amplitude is half the tidal range. A series of interpolated grids (20 m cell size) representing the tidal efficiency, range and amplitude of groundwater tidal changes were developed from monitoring data (Table 3.3).
GNS Science Report 2020/11 25
The upper harbour at Dunedin lags 80 minutes behind tides at Green Island, with an amplified tidal range (~120% tidal efficiency) due to the harbour geometry (and bathymetry in particular). Tidal efficiency and phase lag were determined at each site relative to the tidal range of the spatially nearest (harbour or ocean) tidal source. Maps of tidal efficiency and phase lag (Figures 3.5 and 3.6) show general consistency regarding distance from the tidal source and/or values of nearby data points.
Strong tidal signals are recorded at three sites within 250 m of the harbour or ocean: I44/1106 Bayfield, separated from the Andersons Bay Inlet by reclaimed land, has an efficiency of 51% but only experiences the upper half of the tide because of a constant base level in the inlet; CE17/0102 Tewsley St, a deeper 14.7 m drill hole screened in silty sands below reclaimed land where groundwater pressure is loaded to 34% of the harbour tide; and I44/0007 Kennedy St, separated from St Clair beach by coastal sand dunes and experiences 30% of ocean tides. The amplitude of tidal shifts in these strongly tidal piezometers is generally much larger than those produced by rain, so the tides locally obscure all but the largest (>30 mm) rainfall events.
Weak tidal signals are present at about 16 sites, where tidal efficiencies of between 0.1 and 1% are recorded up to 1.6 km from the ocean. Sites with weak tidal signal appear to have tidal fluctuations obscured by shifts in groundwater elevation associated with rainfall recharge. For many sites, the tidal signals can only be observed at periods of particularly stable groundwater levels between rainfall events. Sites positioned along the old 1850s harbour shoreline, now separated from the harbour by around 0.5 km of reclaimed land, have tidal efficiencies of 0.6–0.8% and phase lag of 130–160 minutes (I44/1122 Andersons Bay Rd, CE17/0101 Queens Gardens and CE17/0103 Harrow St) relative to the harbour.
At 12 sites, any tidal influence is below the detection limits (~2–3 mm) of the pressure transducers, for which a 745 minute phase lag was assigned (representing 12.42 hours = one full tidal phase shift). These sites are scattered amongst the tidal monitoring sites.
Table 3.3 Summary of tidal and rainfall-recharge grid datasets developed for Dunedin groundwater. For a more detailed description of these data, see Table A3.3.
File Name Summary Piezo_data_202003 Point dataset containing observations of tidal behaviour at each monitoring site.
TE_percent Tidal efficiency (in %) ratio interpolated from site observations of groundwater tidal changes in response to the nearest tidal influence (either the harbour or ocean).
TIDE_amplitude_m Amplitude (in m) of 6.2 hour tidal change in groundwater, interpolated from piezometers, sea and harbour. Tidal amplitude = 0.5*Tidal Range measurements.
TIDE_range_m Amplitude (in m) of 12.42 hour tidal change in groundwater, interpolated from piezometers, sea and harbour.
RRI_selected Grid model of the dimensionless rainfall recharge index (RRI), based on selected site values from event observations of change in groundwater level divided by the rainfall amount (dGWL/TotalRain).
RRI_contours Contours of the dimensionless RRI in intervals of 1.
RAINstorage_fromGWLmedian Grid model providing an estimate of the rainfall limit (in mm) and/or available storage before groundwater can be expected to reach the ground surface, based on the median position of groundwater (GWLmedian_2019) and the RRI grid of response site values observed to c.30 mm rainfall events in 2019 and 2012 (using Fordyce data). For further details, see Section 6.4.
26 GNS Science Report 2020/11
Groundwater response to tidal forcing can be used to derive aquifer hydraulic properties (Ferris 1952, Hsieh et al. 1987). A 22 m/d hydraulic conductivity was derived by Rekker (2012) using the strong tidal signal at I44/0007 Kennedy St together with an interpreted depth of sand derived indirectly from a geophysical survey of resistivity, then used to help constrain a hydraulic model of South Dunedin groundwater. Collection of monitoring data from other sites during 2019 suggests that the strongly tidal Kennedy St piezometer is locally anomalous compared with the wider network.
Where multiple measurements are available, tidal efficiency and phase lag commonly show exponential and linear relationships, respectively, with distance from the coast (e.g. Rotzoll and Fletcher 2013). Slopes of tidal efficiency or phase lag versus distance have been used to estimate the bulk diffusivity and storage of aquifers (e.g. Erskine 1991). Plots for South Dunedin and Harbourside data are provided in Figure 3.7, with a caveat reminder that 12 of the 32 monitoring sites had no discernible tidal behaviour. The plots show piezometers subdivided according to the strength of their tidal response (strong = red; weak = blue), with those showing weak tidal behaviour further classified by depth (<5 m, 5–10 m and >10–24 m). There is much scatter in these projections of data.
A tendency for deeper piezometers to have smaller lags and larger tidal efficiencies has been observed in other studies (e.g. Erskine 1991) but is not replicated in the Dunedin data. Vertical lines connect shallow and deep piezometers at the same location on Figure 3.7: being Moana Rua (CE17/0106-CE17/0107) 247 m from the coast and Tonga Park (CE17/0105-I44/0006) 910 m from the coast. The deeper piezometer at Moana Rua has a longer lag and lower efficiency, whereas the deeper piezometer at Tonga Park has a longer lag and larger efficiency. Similarly, there are systematic position of <5 m, 5–10 m and >10 m piezometer values relative to each other on these plots.
According to theory, best-fit lines of both tidal efficiency (logarithmic lnTE) and phase lag versus distance should pass through a time lag of zero and tidal efficiency of 1 at the shoreline. But regression lines through the weakly tidal Dunedin sites cannot be fitted with any degree of confidence (R2 = 0.024, 0.015 – see Figure 3.7). Rather than piezometer depth, diameter or style of construction/completion, the scatter (and local absence of responses) may reflect variations in subsurface geology of the local site and/or between the piezometer and the sea. Alternatively, they may reflect local influence of and interaction with stormwater and wastewater networks.
Alternatively, Erskine (1991) provided examples of when theories based on one-dimensional flow in confined aquifers may not always be applicable to unconfined groundwater. The tidal responses in Dunedin do not necessarily reflect simple velocity flows of saltwater in and out of land mixing with the unconfined groundwater. Locally, at least, they involve processes such as poroelastic loading and potentially could reflect the passage of pressure waves (celerity compared to McDonnell and Beven 2014) and/or up-gradient restrictions to groundwater discharge (as per ‘tidal freshwater zone’ processes in rivers – see Hoitink and Jay 2016). Detailed or careful examination and dissection of the tidal behaviour, and lack of tidal behaviour, will be needed before these data can be confidently used to derive quantitative estimates of hydraulic properties.
GNS Science Report 2020/11 27
Figure 3.5 Tidal efficiency at each of the monitoring sites, including those from Fordyce (2014), and a grid model of the tidal efficiency as a percentage of the groundwater tidal range to nearby sea or harbour tidal ranges.
Figure 3.6 Tidal phase lag at each of the monitoring sites relative to the nearest (ocean or harbour) tidal source, plotted over the tidal efficiency grid. Annotated values are the phase lags in minutes. Those sites with no observable tidal signal (white) are assigned a 12.42 hour (745 minute) phase lag.
28 GNS Science Report 2020/11
Figure 3.7 Graphs of (logarithmic) tidal efficiency (as lnTE) and phase lag (in minutes) against distance from the nearby tidal source. Regression lines through sites with weak tidal signal are poorly fitted and are shown for discussion purposes only. Vertical lines connect piezometers at different depths at the same location: Moana Rua (CE17/0106-CE17/0107) 247 m from the coast; and Tonga Park (CE17/0105-I44/0006) 910 m from the coast.
3.4 Effect and Efficiency of Rainfall
Groundwater response to rainfall recharge can be characterised at each site using a dimensionless Rainfall Recharge Index (RRI), representing the ratio of the groundwater level change to a measured amount of rainfall:
where ΔW isRRI the = ΔW/ERtotal change in piezometer groundwater level (converted to mm) divided by ER, the nearby measured total event rainfall (in mm).
Sites can be expected to have variable responses depending on local stormwater drainage, extent of 'hard' urban surfaces (i.e. ground imperviousness), piezometer construction method (or ‘well-completion’, including piezometer protection to surface water runoff and screen size and type) and subsurface sediment storativity and permeability.
RRI values were initially calculated for each monitoring site using the largest rainfall event during monitoring of 2019, which involved a prolonged 33.6 mm of rain from 27–30 October (that also coincided with a highest perigean spring ‘red tide alert’ tide). The values were then checked and verified against other rainfall events that exceeded 30 mm in 24 hours. Changes in groundwater level were normalised against rainfall measured at the Musselburgh weather station, which effectively assumes that any spatial differences in total rainfall across the low-lying parts of the city were negligible.
GNS Science Report 2020/11 29
Figure 3.8 The RRI ratio as a series of piezometer point observations and an interpolated grid, which provides a model of the relative local rise of groundwater for rain events of around 10–50 mm over a 24–48 hour period.
A slightly different approach was adopted by Fordyce (2014), who plotted ΔW versus ER graphs for around thirty 48-hour rainfall events recorded during the 2012 monitoring campaign. RRI was derived for each site from the fitted slope of the graphs. Differences in RRI, and the threshold needed to initiate a groundwater response, both appeared to show some relationship with site geology, with higher responses, small lag times and rapid recession occurring where sediment was sand-dominated. Results derived from 2012–2013 appear reasonably comparable with those from 2019 at nearby sites, despite differences in the scale of the rain events (ranging from <5 to 38 mm), and associated differences in evapotranspiration and/or stormwater runoff and network performance.
A grid of RRI (Figure 3.8) was interpolated by selecting response site values from both 2019 and 2012 observations. At the harbour and the coast, around the margins of the model area, an RRI value of 1 was chosen under the assumption that, on average, the response rise of the harbour and sea will equal the total event rainfall. The expectation is that the RRI grid should be a reasonable representation of recharge during rainfall of 10–50 mm over periods of 24–48 hours but are not necessarily applicable for extreme events. Heavy rainfall events of >75 mm are discussed further in Section 6.
30 GNS Science Report 2020/11
4.0 VERTICAL HETEROGENEITY AND HYDRAULIC PROPERTIES
Compared with locations in New Zealand where aquifers are used for domestic or agricultural water supply, there has been a paucity of information on sediments and groundwater in the flat-lying land of Dunedin, as the groundwater is not used. There have been no aquifer tests and, until 2019, the groundwater monitoring has been entirely restricted to shallow piezometers installed at depths of less than 6 m. While these provide a measure of the water table surface and its fluctuations, they provide little information on the vertical flow or compartmentalisation of groundwater. With surrounding hills locally exceeding 100 m elevation, there is potential for topographically driven groundwater flow and substantial vertical pressure gradients.
A series of deeper drill holes were completed during 2019 to provide more information on the nature and thickness of the sediment and vertical hydraulic gradients. Details of the drilling and geological units encountered are provided elsewhere (e.g. Glassey et al. submitted; Barrell et al. in prep), but here we present some preliminary observations relating to groundwater. Three relatively deep drill holes in South Dunedin were drilled at 20.7, 21.8 and 44.5 m depths, then screened over 3 m lengths with bases at 19.9, 14.3 and 19.9 m (Table 2.2), thereby providing connection to groundwater at the sediment-bedrock interface. Four drill holes in the Harbourside area were completed at depths of around 15 m and screened at their base within the sediment pile. No artesian pressures or substantial vertical flows were encountered during drilling, and post-drilling water levels have equilibrated with below-ground pressure.
A key discovery from the drilling programme was that Quaternary sediments form only a very thin veneer beneath the city, substantially less than the >100 m target depths expected based on previous geophysical surveys (e.g. Cook et al. 1993; Cournane 1992; Fletcher 2016; Lutter 2018; Rees 2018; Sangster 2019). Bedrock is commonly deeply weathered volcanic rock and/or sandstone, with depths constrained to be: 11 m at Queens Gardens (CE17/0101), 18 m at Tonga Park (CE17/0105), 13.5 m at De Carle Park (CE17/0106) and 16.2 m at Moana Rua (CE17/0107). Other drilling investigations recorded sediments >45 m thick at the Edgar Centre (Rees 2018). 3D models are presently being developed by GNS Science to define the depth of Quaternary sediment and thickness of the upper Holocene-aged part of the sequence, as well as detailed drilling logs and descriptions of sediment type and their variations (Glassey et al. submitted). This section summarises some of the recent observations pertinent to understanding vertical heterogeneity of groundwater.
4.1 Perched Groundwater
A zone of perched groundwater has been found within the sand dune area at St Kilda (Fordyce 2014). A shallow piezometer at Moana Rua (CE17/0108) and, previously, the F5 Holiday Park and F6 OMES sites installed by Fordyce (2014; Figure 2.1) record its behaviour. Groundwater sits above sea level and has a relatively low electrical conductivity and fresh composition. It is recharged by rainfall, contains little tidal signal and is more subdued in its fluctuations compared with groundwater across the rest of South Dunedin.
The extent of the perched aquifer has been interpreted and mapped from local geomorphology (using the LiDAR survey), land elevation and the distribution of clean sand-rich sediment encountered in boreholes. Locally, there is old landfill material also present within the sand dunes. which may complicate the behaviour and chemistry of perched groundwater.
GNS Science Report 2020/11 31
Figure 4.1 Interpreted map-limit of perched groundwater sitting at around 1–1.5 m elevation within dune sands (white outline), with the Perched_Aquifer_GWL (groundwater elevation grid) overlain on the GWLmedian_2019.
Grid surfaces of the perched aquifer have been interpolated and treated separately using a series of modelling points around the mapped outline to provide the minimum possible elevation of the freshwater lens. Files are distinguished from those in Tables 3.1 and 3.2 by the prefix ‘perched’ (e.g. perched_GWLmean_2019, perched_DTWmedian_2019) and can be overlain on top of the other groundwater surfaces (e.g. Figure 4.1).
4.2 Confined Groundwater
The shallow sediment sequence in the Harbourside area north of the harbour basin is composed of alluvium (volcanic rock gravels with clay-rich sands), buried beneath approximately 5 m of marginal marine-estuarine sediment (dark-grey silty clay), and an overlying, variable thickness (several metres) cap of anthropogenic fill (gravel and sand). Harrow St (CE17/0103) piezometer appears to sit entirely in alluvial material, with the marginal marine-estuarine layer missing.
The Harbourside piezometers Queens Gardens (CE17/0101), Tewsley St (CE17/0102), Harrow St (CE17/0103) and Logan Park (CE17/0104) all show a tidal signal – Tewsley St is particularly notable, as it has strong tidal fluctuations at 34% of the harbour tide. Yet the electrical conductivity of the groundwater in these piezometers is moderate (Figure 3.4). Of these piezometer sites, Logan Park has the highest specific conductance at ~3300 μS/cm (i.e.~ 5 % seawater), whereas the strongly tidal Tewsley St is relatively fresh with 400–900 μS/cm conductance (<2% seawater), even though it is only 110 m from the harbour. It appears the groundwater in alluvium is not hydraulically connected with the harbour and the marginal marine-estuarine sediment locally provides a confining layer. Only non-flowing pressure conditions have been experienced to date, with the maximum groundwater levels remaining >0.5 m below ground during 2019 at Queens Gardens, >1.0 m at Tewsley St and >2 m below ground at Harrow St and Logan Park.
32 GNS Science Report 2020/11
As the harbourside piezometers are all screened at depths around 12–15 m in alluvium, there is no information yet on the extent of unconfined shallow groundwater and tidal flow in and out of the overlying shallow marine sediment and anthropogenic fill. Given the importance of this industrial land and its infrastructure, installation of two shallow, 6 m deep, piezometers in this area (e.g. at either end of Sturdee St) would be a valuable addition to the groundwater monitoring network and improve models of the effects of sea-level rise.
Small differences in deep and shallow groundwater levels occur at Tonga Park and in the De Carle Park to Council Street area. The vertical gradients are only on the order of 1% (see Section 4.3 below) but indicate that the silty sediments beneath South Dunedin do provide a slight degree of semi-confinement of deep groundwater.
4.3 Vertical Pressure Gradients
Indications of vertical flow can be obtained at places where nearby piezometers have different screen depth positions. There are three places where there are both shallow and deep monitoring piezometers within 80 m of each other: Tonga Park (CE17/0105 and I44/0006), De Carle Park to Council St (CE17/1016-I44/1124) and Moana Rua (CE17/0107 and CE17/0108). Vertical pressure-gradient data were available from 14/06/2019 once the deeper NZ SeaRise drilling had been completed. A plot of the difference of deep minus shallow groundwater levels, with data shown as numerical (Julian) day of the year, shows both positive and negative values (Figure 4.2).
Measuring point elevations were surveyed for all Dunedin monitoring piezometers by RTK GPS (Section 2.2). In all instances, the measuring points are the top of the casing, which is offset from ground level by some amount. The vertical precision of each measurement using RTK GPS varies, but is between 0.020 and 0.066 m at 95% confidence levels (Table 2.2). However, the uncertainties compound when we want to determine differences in groundwater elevation between two piezometers, such that differences in groundwater elevation have near-double uncertainties of between 0.040 and 0.132 m.
Differences in deep minus shallow groundwater level (Figure 4.2) are significant beyond these compounded uncertainties. Groundwater levels at Tonga Park are consistently higher in the deep piezometer (CE17/0105), indicating a positive (upwards flowing) pressure regime that likely reflects topographic head associated with proximity to the hills. Small steps and departures towards lower values reflect periods of rain when rise of groundwater in the shallow piezometer (I44/0006) leads that in the deep piezometer.
At Moana Rua, the deep minus shallow difference is consistently negative (downwards flowing), which reflects the presence of a shallow perched aquifer. A climbing trend towards less negative values from winter to summer appears to be mostly due to lowering of groundwater in the shallow piezometer/perched aquifer (CE17/0108). Deep groundwater at De Carle Park (CE17/0106) is also less variable than the shallow groundwater at Council St (I44/1124), but differences in groundwater level between De Carle Park to Council Street switch between positive and negative. The negative values are higher levels of shallow groundwater in response to rain, which recess over time back to weakly positive, upwards flowing background gradients.
GNS Science Report 2020/11 33
Figure 4.2 Plot of the difference in groundwater level for sites where deep and shallow piezometers are situated near each other. Difference values (in m), derived from deep minus the shallow groundwater, are plotted against the numerical day of the year (Julian days as 0–365 values). Data shown here are for the period 14/06/2019–12/02/2020.
To improve precision and quantification of vertical hydraulic gradients, we undertook a dumpy level survey to measure local differences in elevation of the measuring points and water levels. An uncertainty of ± 0.005 m is assumed for each dumpy level reading (based on the spacing of graduations on the measuring staff), making a compounded uncertainty for relative height differences of ± 0.010 m. By using the GPS elevation for the piezometer in Table 4.1 (leftmost column) as a reference, the differences in the height of the groundwater using the dumpy level values and dipped DTW on 12/09/2019 are calculated in Table 4.2.
A pressure gradient based on the difference in screen depth of the piezometers, in m/m and percentage, was determined using the difference in depth between the top of the screen in the deep piezometer and the bottom of the shallow piezometer screen (Table 4.2). Note that these vertical gradients assume hydraulic connection throughout the entire 3-m-screened sections within each piezometer and do not take account of any differences in the depth of any particularly transmissive horizons. Based on these calculations, there was an upwards positive pressure gradient of 1.1% (and potentially flow) at Tonga Park and a slight positive pressure gradient of 0.6% at Council Street to De Carle Park on 12/09/2019, at a time when the groundwater level differences appear reasonably representative (Figure 4.2). By way of contrast, the perched groundwater in the sand dunes in the St Kilda area, monitored in the shallow Moana Rua piezometer, had a downwards vertical gradient of 5%.
34 GNS Science Report 2020/11
Table 4.1 Observations of differences in measuring point (MP) elevations of nearby monitoring piezometers.
Piezometer with MP Piezometer with MP MP Difference (m) GPS MP Elevation Elevation (m) (from Elevation (m) (from Measured with Difference (m) RTK GPS) NZVD2016 RTK GPS) NZVD2016 Dumpy Level CE17/0105 I44/0006 Tonga (deep) Tonga Park (ORC) -0.193 ± 0.102 -0.210 ± 0.010 1.381 ± 0.066 1.574 ± 0.036
CE17/0106 I44/1124 De Carle Park Council St 0.714 ± 0.061 0.708 ± 0.010 1.491 ± 0.034 0.777 ± 0.027
CE17/0107 CE17/0108 Moana Rua (deep) Moana Rua (shallow) -0.065 ± 0.075 -0.050 ± 0.010 3.679 ± 0.038 3.744 ± 0.037
Table 4.2 Vertical hydraulic gradients in nearby monitoring piezometers.
Shallow Difference in Deep Piezometer, Piezometer, Depth MP Difference Local Depth to Top of to Base of Screen, (m) Measured Groundwater Vertical Screen, and DTW and DTW (m) with Dumpy Elevation (m) Gradient (m/m) (m) Measured by Measured by Level 12/09/2020 Manual Dip Manual Dip Deep–Shallow CE17/0105, 16.95 m I44/0006, 6.0 m Upwards +0.12 ± 0.02 Tonga (deep) Tonga Park (ORC) -0.210 ± 0.010 0.011 over 10.95 m 1.025 ± 0.005 1.355 ± 0.005 (1.1%)
CE17/0106, 16.9 m I44/1124, 6.07 m Upwards +0.068 ± 0.02 De Carle Park Council St 0.708 ± 0.010 0.006 over 10.83 m 1.185 ± 0.005 0.545 ± 0.005 (0.6%)
CE17/0107, 18.8 m CE17/0108, 6.25 m Downwards -0.673 ± 0.02 Moana Rua (deep) Moana Rua (shallow) -0.050 ± 0.010 -0.053 over 12.55 m 3.370 ± 0.005 2.747 ± 0.005 (-5.3%)
4.4 Drawdown and Recovery
Piezometers were purged by peristaltic pump during regular sampling runs throughout 2019– 2020 by the University of Otago (Sarah Mager / Sarah Yeo) and GNS Science (Simon Cox). Purging is a commonly adopted procedure when sampling wells for groundwater quality. The rationale is to remove any contamination within the well – for example, from static water and/or interacting with the casing or from pollutants that may have leaked into the well from the surface. Periods of anomalously low groundwater level associated with purging are immediately obvious in monitoring data from Dunedin. ‘Drawdown’ and ‘recovery’ effects are present throughout the 2019 dataset. Although data from these periods were removed for statistical analysis of groundwater levels and modelling of water table surfaces, the drawdown and recovery periods serve as useful experiments from which observations can be made on groundwater flow behaviour and the hydraulic properties of the sediment.
Drawdown and recovery responses are commonly utilised in simple models of hydraulic conductivity, where the Thiem equation of radial flow is applied in falling or rising head tests (Hvorslev 1951, Bouwer and Rice 1976, Bouwer 1989, Sanders 1998). While some Dunedin piezometers are unable to be drawn down with a small peristaltic pump working at 3 l/min or had fast recovery within the 15 minute sample interval of transducers, other sites show
GNS Science Report 2020/11 35
consistent exponential-type recoveries lasting between several hours or several days. At some sites, the electrical conductance has also been measured continuously, together with pressure and temperature, to provide a record of chemical changes associated with drawdown.
For example, a series of purges at the Council St (I44/1124) piezometer during 2019 are illustrated in Figure 4.3. The specific conductance (red line) was at about 10300 μS/cm before a purge on 14/04/2019, dropped to 50 μS/cm and then took approximately six days to recover after the purge, but only to a new level at 9800 μS/cm. After the next purge on 12/06/2019, it took nearly three weeks to recover to this level of conductance, and the chemical recovery was equally long after a purge on 8/08/2019. In comparison, there was a quick recovery of the groundwater levels, almost inevitably within two or three days (black line). It appears that pumping drew less-conductive water (presumably near-surface rainwater) downwards into the piezometer, and it took as much as three to four weeks for it to chemically equilibrate, presumably by diffusion. Any sample collected after the purge would have recorded a low electrical conductivity for weeks after the purge.
Similarly, induced influx of freshwater (presumably down the piezometer casing sides, the bentonite-sediment margin and/or through its construction sand pack) and delayed chemical recoveries have now been observed in response to purging at King Edward St (I44/1123), Alma St (I44/1125), Kings (I44/1105) and Bayfield (I44/1106). In some instances, the regular three-month purging has generated a lasting perturbation on the electrical conductivity. Elsewhere, by way of contrast, electrical conductivity is inversely related to groundwater level and both respond quickly in response to rainfall-recharge (e.g. Pretoria Ave I44/1120).
Delayed chemical compared to water-level recovery provides clear indication of low volumes of groundwater exchange and/or throughflow of rainfall recharge and places where sediment has low hydraulic conductivity. It also seems the rising-head response after the purging, or any falling response from slug tests (e.g. Fordyce 2014), most likely record the piezometer installation method and/or sand-pack permeability. Aquifer tests are required to derive larger-scale hydraulic properties of groundwater in the sediment, but in the meantime the continuous electrical conductance data, when combined with temperature and water level data, are a fertile source of complementary information for future work.
36 GNS Science Report 2020/11
Figure 4.3 Council St (I44/1124) groundwater level (black in mNZVD2016) and specific conductance (red in μS/cm) during 2019, illustrating the effects of a series of regular purges to sample water from the piezometer. Water levels recovered within a few days of each purge, but the electrical conductivity did not recover for weeks after each purge. Rainfall (blue – hourly values from Musselburgh) generates small peaks in groundwater levels, followed by recessions, but rainfall appears insufficient to effect electrical conductivity of groundwater through dilution.
GNS Science Report 2020/11 37
4.5 Temperature
Most of the transducers in Dunedin have been set to record temperature as well as pressure, so an extensive dataset has been collected on thermal behaviour across the monitoring network. Most of the transducers are Seametrics LevelSCOUT loggers with absolute accuracy of ± 0.1°C and relative precision of ± 0.01°C. Although the temperature data are yet to be studied in any detail, some preliminary observations from the Hilltop database are presented here.
The style of groundwater thermal behaviour and variability of temperature is largely a function of the depth of the piezometer below ground. Shallower sites (of less than 6 m depth) tend to have prominent sinusoidal behaviour reflecting seasonal variation but no obvious diurnal changes. Transient departures from regular sinusoidal behaviour are caused by: (i) some larger rainfalls, which warm groundwater during summer or cool it during winter (rare); (ii) condensation, which forms noisy fluctuations of ± 0.2°C in winter as water forms inside the piezometer casing and trickles down into the water column; and (iii) recordings of air temperature when the water level has fallen below a logger (rare) or loggers have been removed for data retrieval (spikes, but on known dates).
Deeper piezometers, screened at depths of 10 m or more, generally have average values of about 12–13°C and small (± 0.5°C) temperature ranges. In the Harbourside area, any seasonal sinusoid is weak and/or disturbed by other thermal influences such as hillslope runoff or the harbour itself (e.g. Oval CE17/0100, Tewsley St CE17/0102 and Harrow St CE17/0103). Queens Gardens (CE17/0101) seems a notable deep piezometer exception, where a ± 3.5°C seasonal fluctuation is disturbed ~0.5°C on weekly (or more) intervals by cumulative rainfall events. Weak sinusoidal behaviour at 12–13°C ± 0.5°C is seen in the deep Tonga (CE17/0105) and Moana Rua (CE17/0107) piezometers in South Dunedin, whereas De Carle Park (CE17/0106) has a near constant 13.2 ± 0.1°C temperature.
Examples of the seasonal behaviour in shallow ≤6 m piezometers in South Dunedin is provided in Figure 4.4. Most have temperatures within a 12–15°C range, with ± 1.5°C sinusoidal amplitude, which is warmer than average air temperatures at Musselburgh weather station (where monthly means range from about 7–15°C during the year). Absolute differences in groundwater temperature from site to site seem to reflect a combination of depth of groundwater below the ground surface with a lesser effect of the depth of the transducer hanging below the water surface (seen only at some sites when hanging depths were changed during the year).
Although the annual groundwater temperatures are strongly sinusoidal in response to seasonal variation, there are also significant differences in the dates of temperature maxima and minima. Many of the shallow ≤6 m sites show 4–6-month thermal phase delay, with temperatures that are warmest in June (winter) and coldest in December (summer). Absolute thermal diffusivities can be calculated from the delay, combined with the change in temperature amplitude (Tabbagh et al. 1999; Seward and Prieto 2015), but have yet to be derived from these data. Regardless of absolute values, the thermal lags in these South Dunedin piezometers, together with the general absence of thermal effects from rainfall, indicate heat transfer is substantially a conductive rather than convective process (Tabbagh et al. 1999; van Manen and Wallin 2012, Seward and Prieto 2015). It provides indication that there are low volumes of groundwater exchange and/or throughflow of rainfall recharge, supporting the notion (above) that the sediment is locally impermeable and/or with low hydraulic conductivity. As well as providing information on the thermal resource (for example, for development of shallow ground- source heat pumps), the temperature data are a fertile resource for future work.
38 GNS Science Report 2020/11
Figure 4.4 Examples of temperature data for the period 6 Mar 2019 – 12 Feb 2020 in shallow piezometers ≤6 m deep. There is a strong sinusoidal seasonal fluctuation with site-to-site differences in amplitude and phase. Fitzroy Street (I44/1121) is notably anomalous, but commonly groundwater levels at this site are at, or below, the depth of the logger.
GNS Science Report 2020/11 39
5.0 GROUNDWATER AND WATER NETWORKS
5.1 Background
A key aspect of the city’s management of stormwater and wastewater involves understanding the system capacities and areas for improvement. Much of the stormwater network in the flat-lying area of Dunedin only has capacity to convey small rainfall events, with shallow ponding predicted to occur in many locations (primarily in road channels) during storms as small as a 1-in-2-year ARI4 rainfall event (URS and Opus 2011). Within the 8.9 km2 Harbourside to South Dunedin area of groundwater surfaces interpolated in this report, there are 92 km of stormwater pipes, which have a volume of 49,000 m3. The static volume of stormwater pipe storage for this area is equivalent to around 5 mm of rain.
There are a number of hydraulic ‘bottlenecks’ in the system that create a lower overall level of service in the network and regular ‘nuisance flooding’ (50–300 mm deep). Blocking of intake screens leads to backwater effects and exacerbates overland flow, which can be mitigated through regular cleaning and maintenance, but deep flooding (>300 mm) risk appears to remain to some properties during events as small as a 1-in-10-year ARI rainfall (URS and Opus 2011). Infiltration and inflow of groundwater and stormwater occur in any wastewater system to some degree, but locally appears to be greater than that allowed for in the system design and so causes problems in Dunedin (Opus and URS 2011). Three main issues are caused: (i) overflows or flooding can be triggered where the system lacks capacity, (ii) diluted wastewater can cause problems for biological treatment and (ii) energy may be wasted in pumping unnecessary additional volumes.
High groundwater levels limit the available drainage space and creates opportunities for infiltration into both stormwater and wastewater networks. Potential for infiltration has been assessed on a catchment-scale basis and considered likely where wastewater dry weather discharge per capita per day exceeds water demand per capita per day by more than 20% (Opus and URS 2011). Alternatively, if wastewater dry weather discharge rate was more than 30% lower than water demand rate, then there is potential for exfiltration. High infiltration was identified in the low-lying parts of the city and attributed to wastewater pipes being below the water table such that groundwater enters through leaking joints or cracks (Opus and URS 2011). There is also a baseflow volume in the stormwater network in the absence of rainfall, which discharges around 430 m3/day at the Portobello pumping station outlet (Rekker 2012) and appears to be sourced, at least in part, from groundwater. Aging pipes are particularly vulnerable to infiltration and leakage (Opus and URS 2011; URS and Opus 2011).
The intrusion of saltwater into wastewater pipelines has been of ongoing concern for the Dunedin City Council due to its effect on pipe condition and, more importantly, the wastewater treatment plant processes. Increased salinity at the Tahuna Wastewater Treatment Plant kills the bacteria used to treat the wastewater to a secondary level (Opus and URS 2011). Some of the worst problem areas in the wastewater network (e.g. Harbourside at Fryatt St) have been found and replaced through monitoring of electrical conductivity together with flow volumes (Osborn and Sinclair 2011, Opus and URS 2011). Saltwater intrusion into the stormwater system occurs regularly via the outfall pipes, but there also seems to be some infiltration of saline groundwater along the pipelines that reduces network capacity, particularly during high tides.
4 ARI = Average Recurrence Interval. It is also known as 'return period'. It is the average number of years that are predicted to pass before an event of a given magnitude occurs.
40 GNS Science Report 2020/11
There has been relatively little specific examination of the relationship between groundwater levels and infiltration, for either stormwater or wastewater networks. Peaks in wastewater flow were shown to coincide more closely with the water table elevation than rainfall during short periods of monitoring by Fordyce (2014). Losses and gains to the stormwater network are considered a significant component in groundwater flow models (e.g. Rekker 2012). The likely implications of climate change and sea-level rise, with associated rising groundwater, is an exacerbation of infiltration. Together with increases in direct inflow from rain, reduced capacities of both stormwater and wastewater systems has been predicted (Opus and URS 2011; URS and Opus 2011), particularly if aging infrastructure is not upgraded.
5.2 Network Position Relative to the Water Table
The installation of a wider groundwater monitoring network during 2019, and the detailed mapping of groundwater levels and electrical conductivity provided herein, is a significant new knowledge step. It enables places to be identified where the stormwater and wastewater networks sit above or below the water table. A network GIS layer provided by the Dunedin City Council has elevations for a large number of network nodes, being sumps, manholes, network joins, etc. Node invert levels and pipe lowest ‘down elevations’ were supplied in Dunedin Drainage Datum and converted to NZVD2016 by subtracting 100.377 m. Each node or pipe was assigned a value for the local position of the water table from median (GWLmedian_2019) or MHWS (GWLmhws_median2019) groundwater grids.
Stormwater and wastewater node and pipe lower points have been coloured according to their elevation (in metres) above or below the water table in Figures 5.1 and 5.2, respectively. Comparison of the figures highlights that the wastewater network is deliberately constructed at a level deeper than the stormwater, so is more commonly situated below the water table. Interpolated models of network elevations were also derived by interpolation of the stormwater and wastewater node heights (background shading Figures 5.1 and 5.2). At the first order, the networks follow topography, deviating locally with design-slopes to enable gravitational flow where possible.
Grids of difference between the network elevations and groundwater levels were derived by subtracting the water table grid from the stormwater or wastewater elevation grids. The situation at MHWS (Figures 5.3 and 5.4) highlights places (blue) where the networks are substantially below the water table and vulnerable to infiltration. They can be combined with electrical conductivity data and models of %Seawater (Figure 5.3) to examine potential for saline incursion and/or models of inundation associated with sea-level rise (e.g. Section 7.0). Such models can also be used to identify places where infiltration might switch to exfiltration (leakage and losses) at times when groundwater levels fall. Stormwater and wastewater monitoring can potentially be targeted at places that may enable thresholds for groundwater- network exchanges to be defined in relation to network flows, age and construction material and methods. In turn, this may enable critical tipping points for system collapse to be foreseen. Concurrent monitoring of network flows and groundwater fluctuations at a wider number of sites is the next logical step to fully understand the groundwater infiltration and rationalise long-term investments in infrastructure.
GNS Science Report 2020/11 41
Figure 5.1 Position of stormwater network relative to the median water table in the flat-lying area of Dunedin, with nodes and pipes coloured according to their position (in metres) above or below the median groundwater model GWLmedian_2019. Red nodes and pipes are >0.5 m above the median water table and orange ± 0.5 m above, whereas blue nodes are below the water table and are potential sites for infiltration. Background transparent shading is a surface elevation model of the stormwater network in metres NZVD2016.
Figure 5.2 Position of wastewater network relative to the median water table in the flat-lying area of Dunedin, with nodes and pipes coloured according to their position (in metres) above or below the median groundwater model GWLmedian_2019. Colours follow the same class ranges as the stormwater network (Figure 5.1) to enable the two diagrams to be easily compared.
42 GNS Science Report 2020/11
Figure 5.3 Stormwater network position relative to the water table at MHWS, derived by subtracting the GWLmhws_median2019 model from an elevation model derived from the stormwater network nodes. Blue areas are significantly below the water table and are potential places for infiltration.
Figure 5.4 Wastewater network position relative to the water table at MHWS, derived by subtracting the GWLmhws_median2019 model from an elevation model derived from the sewer network nodes. Blue areas are significantly below the water table and are potential places for infiltration.
GNS Science Report 2020/11 43
6.0 GROUNDWATER AND FLOODING
6.1 Background
Much of the Dunedin coastal plain and flat-lying land is subject to some degree of flood hazard (URS and Opus 2011; Goldsmith and Hornblow 2016). As well as direct accumulation of rain, eight small catchments (totalling 14.8 km2) contribute runoff to South Dunedin, and the Water of Leith (42 km2) and Lindsay Creek (12.5 km2) catchments shed runoff to the North Dunedin area. Factors contributing to the hazard are heavy rainfall, impervious urban surfaces, rapid surface runoff and a shallow water table. Reduced capacity or failure of stormwater drains and pumping stations also exacerbates surface water ponding. As the only rainfall events experienced during 2019 were low- and medium-sized events, and none caused flooding, this section draws on exemplars of groundwater behaviour at ORC longer-term sites during heavy (>75 mm) rain events in June 2015, July 2017 and February 2018 (Figure 6.1).
The four telemetered ORC piezometers (Bathgate, Tonga, Kennedy St, Culling Park) and the Musselburgh rain gauge contribute to the ORC forecasting of flooding in South Dunedin. There are also nearby automatic rain gauges at Sullivan’s Dam and Pine Hill, but these are of more direct relevance to North Dunedin.
Flood warning for the Water of Leith and Lindsay Creek relies primarily on rainfall sites at Sullivan’s Dam and Pine Hill and from stream flow monitoring sites in each catchment. Rainfall of greater than 30 mm in less than 6 hours at either rainfall site will generally initiate an alarm. Three different types of events are recognised to have potential to cause floods (see below), with the worst damage typically occuring when type B or C conditions are followed by intense rainfall of type A: A. Short-duration, high-intensity rainfall (e.g. 6–9 mm in 15 minutes), generally resulting in small scale floods, with ≤30 minutes lag time between the max rainfall and the flow peak. B. Low-intensity, steady-state rainfall (e.g. drizzling; 1–2 mm per 15 minutes over half a day or more, resulting in medium-range flood peaks of 20–30 cumecs). Floods generally build up slowly, giving several hours of warning for downstream residents. C. Medium-intensity rainfalls (3–3.5 mm per 15 minutes), lasting for several hours or more, which can result in slightly larger floods. Lag time between the maximum intensity rainfall and flood peak can be <1 hour, depending on whether the catchment is already saturated.
44 GNS Science Report 2020/11
Figure 6.1 Exemplars of heavy rainfall events experienced in Dunedin, plotted against time (in hours) since the start of rain. Cumulative hourly rainfall (top left graph) from the Musselburgh weather station shows variations in total rainfall and intensity during the storms. The lower four graphs show groundwater levels at a constant vertical scale (but not absolute value) relative to 2010–2018 median levels, 95th and 5th percentiles, the ground surface and nearby levels of stormwater pipes and inverts.
GNS Science Report 2020/11 45
6.2 Heavy (>75 mm) Rainfall Events
6.2.1 3–4 June 2015
A rainfall event from 3–4 June 2015 caused widespread surface flooding in South Dunedin. More than 400 emergency calls were made (McNeilly and Daly 2015) when 144 mm of rain fell within 36 hours, reaching peak intensities of 9–12 mm/hr. With at least 2400 claims, the event cost insurers at least NZ$28.2 million, but a true cost of $138 million has been suggested once the economic and social impacts are also considered (Elder 2016).
The June 2015 event involved rain that was both continued and intense (Figure 6.1). Its total amount was the second highest on record at Musselburgh weather station (Goldsmith et al. 2015) 5, which has the longest continuous rainfall record within the Dunedin area (beginning in 1918, the record total was 229 mm in April 19236, and record intensity 15.8 mm/hr in February 2005). Estimates of the return period of this event can vary: 63 years using the total set of observation data (Goldsmith et al. 2015) or >100 years using a shorter data record (Carey-Smith et al. 2018).
Rain started about 04:00 on 3 June and reached peak intensities between 11:00 and 13:00. Surface flooding in South Dunedin was well-established by mid-afternoon, at a time when cumulative rainfall totals had reached around 80 mm. Floodwater was a combination of rain that fell directly on the flat land and runoff from the surrounding hill catchments.
Floodwater levels were measured shortly after the event, using the maximum height of flood-debris marks at 150 locations. The floodwater elevation, surveyed by RTK GPS to ± 0.03 m vertical accuracy (Smeaton 2015), had a clear pattern sloping from northwest (Forbury Corner) to south and east approximately following topography (Goldsmith et al. 2015; Figure 6.2). A lack of floodwater in the lowest part of South Dunedin (Tainui and St Kilda), furthest from the hills, is an important observation from the event. Anecdotal observations and insurance claims also suggest flooding in the area with lowest topography was limited.
Flooding inundation depths were not measured at each locality. Instead, these were determined for each site from the difference between the elevation of the debris mark and ground level derived from the 2019 LiDAR survey (Goldsmith et al. 2015). There is considerable localised variability in the inundation depth, in part reflecting the ± 0.15 m uncertainties compounded by using two different survey methods (± 0.03 and ± 0.12 m). It is recommended that any future flood surveys should take two measurements at each site: both the elevation of debris marks (or ground level) by GPS survey and a relative local water depth (or maximum height above ground) using a (higher precision) mm-scale tape measure. Rather than determining inundation at each site and interpolating these values with their inherent uncertainty, for this study an inundation grid was derived from the difference between grids of the LiDAR ground level grid and interpolated elevation of floodwater level (Figure 6.3).
5 Note that a former investigation engineer with ORC, Neil Johnstone, is reported as saying South Dunedin’s stormwater system was tested more severely in 1968 when 158 mm of rain fell over 24 hours at intensities of 23 mm/hr, but all except for 98 houses remained dry (Macfie 2016). As groundwater levels are not available for this event, it has not been reviewed for this report. 6 Rainfall in 1923 also caused widespread flooding in South Dunedin (Goldsmith and Hornblow 2016).
46 GNS Science Report 2020/11
Figure 6.2 Gridded and contoured elevation of June 2015 floodwater, using data from a post-flood survey within the pink box extent (Smeaton 2015). Floodwater elevation is overlain here on the median water table contours and a semi-transparent grid surface for 2019 (GWLmedian_2019) using the same colours for elevation classes.
Figure 6.3 Model of June 2015 floodwater inundation. Derived by subtracting the LiDAR digital elevation model from a grid of floodwater elevations surveyed by RTK GPS (Smeaton 2015). Note that absolute uncertainties are ± 0.15 m, but relative uncertainties are likely smaller. Contours at 0.15 m intervals are shown to highlight areas with significant inundation.
GNS Science Report 2020/11 47
There were only four ORC piezometers monitoring in June 2015 from which the behaviour of groundwater levels during the flood, or its role in flooding, can be examined. Groundwater responses at these sites (Figure 6.1, blue lines) are summarised in Table 6.1. In the month prior to the June 2015 flooding, there had also been two rainfalls of 24 and 27 mm on 12 and 25 May 2015, respectively, that lifted groundwater levels from early autumn seasonal low levels. The May rains caused step-increases of 0.2 m at Tonga Park (I44/0006) and 0.3 and 0.5 m rises at Culling Park (I44/1094). Recession following these rainfalls had been insufficient to return groundwater to pre-rain positions, so the groundwater levels had been rising towards winter levels.
However, immediately prior to the 3–4 June 2015 rainfall, the levels of groundwater were not particularly high or anomalous when compared with statistical distributions from 2010–2018 (Table 2.3, Figures 2.4 and 6.1). Only the Tonga Park piezometer had notably high groundwater levels prior to the June 2015 rain, equating to an 86th percentile level. Other sites had groundwater levels somewhere between 55th and 67th percentile levels (Table 6.1, Figure 6.1). Pre-rain depths to groundwater appear to have been at least 0.5 m throughout South Dunedin; however, groundwater levels at Tonga Park and Culling Park rose rapidly to the surface. It took <90 mm of rain for surface water to begin to pond at these sites. Calculated RRI values were between 1.6 and 6.9 but depend on whether the ponded surface water is included in the calculation (Table 6.1).
48 GNS Science Report 2020/11
Table 6.1 Groundwater levels and calculated response at four ORC monitoring sites during the 3–4 June 2015 flood. Levels are given in metres NZVD2016.
ORC Piezometer Number Parameter I44/0005 I44/0006 I44/0007 I44/1094 Name Bathgate Park Tonga Park Kennedy St Culling Park
Ground elevation (± 0.05) 1.113 0.719 1.252 0.529 as surveyed 2019
Pre-rain groundwater level groundwater elevation at 0.375 0.263 0.297* -0.179 00:00 3/06/2015
Pre-rain groundwater 55th percentile 86th percentile 67th percentile 57th percentile elevation as a percentile
Pre-rain DTW at 00:00 0.738 0.456 0.955 0.708 3/06/2015
Time at which ground 11:30 23:30 15:00 surface was reached by N/A 3/06/2015 3/06/2015 3/06/2015 groundwater
Total rain when ground N/A (>144 mm) 59 mm 135 mm 82 mm level reached
Time of peak 14:30 4/06/2015 00:40 00:15 4/06/2015 15:45 groundwater elevation after rain stopped 4/06/2015 coinciding with low tide 3/06/2015
Peak groundwater level 0.611 0.989 1.294 0.528
DTW at peak 0.502 -0.346** -0.042** 0.001
Groundwater elevation change, including surface 0.236 0.726 0.997 0.707 ponding
RRI based on 144 mm rain 1.6 5.0 6.9 4.9 and including surface water
RRI without surface water 1.6 2.6 6.6 4.9 component
* Daily mean, so tide is removed. ** Negative value reflects above ground surface water.
GNS Science Report 2020/11 49
6.2.2 21–22 July 2017
There were media reports on 20 July 2017 that Dunedin ‘had sandbags at the ready’ and was ‘bracing for flooding’. Rain started to fall at 11:00 on 21 July 2017 and continued for 24 hours until around 11:00 on 22 July. Parts of Dunedin experienced their wettest July day on record (since 1918), but a comparatively ‘small’ 93 mm fell at Musselburgh weather station with peak intensities only reaching 6 mm/hr. While hundreds of homes were evacuated, and major highways were cut off elsewhere in Canterbury and Otago (a regional state of emergency was declared), no serious surface-floodwater damage was reported in South Dunedin.
The 21–22 July event was a lower intensity heavy rainfall compared to June 2015 but significant because groundwater levels prior to rain were already very high. Groundwater at Bathgate, Tonga and Culling parks piezometers (away from the strongly tidal-influenced area) were at 83rd percentile levels or greater (compared to 2010–2018 data) (Table 6.2; Figure 6.1, red lines). The event resulted in the second highest groundwater levels recorded at Bathgate Park (I44/0005) and Kennedy St (I44/0007) piezometers – albeit delayed at the former until 04:00 on 24 July, 65 hours after rain started, and at the latter in conjunction with storm runup and large tides in the ocean nearby (Figure 6.1).
The July 2017 event provides a useful comparator to the June 2015 floods, because the rain fell at a lower intensity and at a time of relatively high groundwater. There has since been widespread discussion drawing on anecdotal observations over whether the lack of surface water and flood damage (compared to June 2015) could be due to improved maintenance and clearing of the stormwater system (following after Macfie 2016). RRI values were similar to those in June 2015 (Table 6.2 compared to 6.1), so recharge to groundwater did not appear to have greatly changed. The role of rainfall intensity and groundwater level response in flooding needs closer examination (see discussion below).
50 GNS Science Report 2020/11
Table 6.2 Groundwater levels at four ORC monitoring sites before and during the 21–22 July 2017 rain. Levels are given in metres NZVD2016.
ORC Piezometer Number Parameter I44/0005 I44/0006 I44/0007 I44/1094 Name Bathgate Park Tonga Park Kennedy St Culling Park
Ground elevation (± 0.05) as 1.113 0.719 1.252 0.529 surveyed 2019
Pre-rain groundwater level 0.478 0.249 0.282* -0.031
Pre-rain groundwater elevation 97th percentile 83th percentile 65th percentile 84th percentile as a percentile
Pre-rain DTW 0.635 0.470 0.970 0.560
Time at which ground surface N/A N/A N/A N/A was reached by groundwater
03:15 22/07/2017 Time of peak groundwater 04:00 24/07/2017 12:50 10:50 coinciding with elevation after rain stopped 22/07/2017 22/07/2017 high tide
Peak groundwater level 0.588 0.521 0.581 0.477
DTW at peak 0.525 0.198 0.671 0.052
Groundwater elevation change 0.110 0.272 0.299 0.508
RRI based on 96 mm rain 1.1 2.8 3.1 5.3
* Daily mean, so tide is removed.
6.2.3 1 February 2018
Dunedin declared a state of civil defence emergency as the remnants of tropical cyclone Fehi hit the city on 1 Feb 2018. Musselburgh weather station recorded 78 mm of rain between 00:00 and 16:00, which started slowly but reached near-record intensity peaks at around 13 mm/hr between 12:00–13:00 (Figure 6.1). Schools closed early, and people sand-bagged properties and businesses. By late afternoon, there was widespread surface flooding at ‘nuisance levels’, some of which entered the city’s wastewater system, but the rain eased. Several streets and residential properties were inundated (Surrey St, Macandrew Rd, Forbury Rd, Loyalty St, South Rd and Marne St), but for the most part there were relatively few problems with water entering living spaces. Radius Fulton rest home in Hillside Road was one of the few places significantly flooded and damaged.
Groundwater levels had been very low prior to the 1 February 2018 rain (Figure 6.1, green lines). Levels at Bathgate, Tonga and Culling Park piezometers (away from any strong influence of tides) were at their 6th percentile or lower (compared to 2010–2018 data). Despite widespread presence of surface ponding, the groundwater does not appear to have reached the ground surface at any of the piezometers – the closest being 0.1 m below ground at Culling Park (I44/1094). The 1 February 2018 rain provides a useful comparator to the June 2015 floods because it was associated with relatively low groundwater and a high intensity rainfall. RRI values were notably higher than usual at the Culling Park piezometer (Table 6.3), but otherwise similar to the recharge in June 2015 and July 2017 at other piezometers.
GNS Science Report 2020/11 51
Table 6.3 Groundwater levels at four ORC monitoring sites before and during the 1 February 2018 rain. Levels are given in metres NZVD2016.
ORC Piezometer Number Parameter I44/0005 I44/0006 I44/0007 I44/1094 Name Bathgate Park Tonga Park Kennedy St Culling Park
Ground elevation (± 0.05) 1.113 0.719 1.252 0.529 as surveyed 2019
Pre-rain groundwater level 0.217 -0.051 0.330* -0.639
Pre-rain groundwater 6th percentile 2nd percentile 75th percentile 5th percentile elevation as a percentile
Pre-rain DTW 0.896 0.770 0.922 1.168
Time at which ground surface was reached by N/A N/A N/A N/A groundwater
17:30 1/02/2018 16:30 1/02/2018 Time of peak groundwater 19:40 coinciding with 15:15 but kept rising for elevation 1/02/2018 high tide and 1/02/2018 a week some storm runup
Peak groundwater level 0.310 0.287 0.550 0.380
DTW at peak 0.303 0.432 0.526 0.149
Groundwater elevation 0.093 0.338 0.220 1.019 change
RRI based on 96 mm rain 1.2 4.4 2.8 13.1
* Daily mean, so tide is removed.
52 GNS Science Report 2020/11
6.3 Discussion: Spatial versus Temporal Variability
The groundwater level responses to the heavy (>75 mm) rainfall events at the longer-term ORC monitoring sites (Figure 6.1), as well as some additional moderate (20–35 mm) rainfalls, are summarised in Table 6.4 using the RRI ratio of groundwater level change/rainfall as a measure of the water table sensitivity to infiltration. Bathgate Park (I44/005) piezometer is relatively insensitive to rainfall (RRI is low at 0.8–1.7), with peak levels commonly reached after rain has finished. By way of contrast, Culling Park (I44/1094) piezometer is always highly responsive to rainfall (RRI is high, 4.9–22) and, at least in part, as there is no upstand protecting the piezometer from surface runoff, levels change quickly and respond immediately to rainfall. Reponses at Kennedy Street (I44/007) are complicated by tidal fluctuations, which tend to dominate over rainfall response and are themselves partly a function of storm runup in the nearby ocean. The RRI values calculated at Kennedy St tend to be moderate (3–7) for larger storm events, and low (1–2) during smaller events – the latter being when the effect of rainfall can be difficult to discriminate from fluctuation caused by tides. Kennedy St and Tonga Park (I44/0006) groundwater responses are also difficult to interpret, as upstands were installed in late 2016 to stop surface water from leaking into the piezometers (which potentially decreased the RRI thereafter).
Collated flood event data highlights that there is spatial variability in the rainfall recharge, with each site reasonably consistent in its sensitivity relative to others (Figure 6.1). RRI values determined for moderate rainfalls in 2019, described in Section 3.4 above, also showed similar spatial variability (see Figure 3.8). In an examination of a greater number of small to moderate rainfall events, Fordyce (2014) found that the threshold needed to initiate a response was also site-specific. As the longer-term ORC piezometers all have the same depth and screen construction, the spatial variability between sites in Table 6.4 and Figure 6.1 likely depends on factors such as subsurface sediment storativity and permeability, the extent of impervious urban surfaces nearby, local surface topographic high/low positions and/or presence of local stormwater drainage.
GNS Science Report 2020/11 53
Table 6.4 Summary of RRI calculated for the ORC piezometers.
I44/0005 I44/0006 I44/0007 I44/1094 Bathgate Park Tonga Park Kennedy St Culling Park Pre-Rain Total Rainfall Event Groundwater Slow to rise, commonly Upstand installed late Upstand installed late No upstand. Highly (and Max Intensity) Levels peaking after rain. 2016 – may have 2016. Daily tidal effect responsive. decreased infiltration and removed, but RRI may extreme peaks. include storm runup.
12/05/2015 24 mm (5 mm/hr) Low 2.3 9.0 4.2 13.0
25/05/2015 27 mm (3 mm/hr) Moderate 2.1 7.6 7.4 19.2
144 mm 3/06/2015 Moderate 1.6 5.0 6.9 4.9 (9–12 mm/hr)
3/06/2015 Excluding surface water - - 2.6 6.6 -
21/07/2017 96 mm (6 mm/hr) Very High 1.1 2.8 3.1 5.3
1/02/2018 78 mm (13 mm/hr) Very Low 1.2 4.4 2.8 13.1
30/05/2019 31 mm (4 mm/hr) Moderate 0.8 4.3 1.0 22.0
27/10/2019 34 mm (4 mm/hr) Moderate 1.7 3.0 2.0 16.6
RRI selected - - 1.7 3.0 1.0 16.6 for grid model
54 GNS Science Report 2020/11
What is less clear is any temporal effect that groundwater levels, or other factors, have on flood hazard. There are several pre-event conditions that may vary over time and could have a role on the vulnerability to flooding. Care and maintenance of the stormwater network is perhaps the most obvious and contentious example. Figure 6.1 and Tables 6.2 and 6.3 highlight the issue, showing how groundwater levels responded to rain when these levels were already very high in July 2017, compared with response in February 2018 when they were very low. Although a soil moisture deficit might be expected to decrease responsiveness at times when groundwater is low, Culling Park (I44/1094) instead appears to have been more responsive from very low levels in February 2018. Here, the lack of upstand to protect the piezometer from surface runoff may be an important factor to consider when interpreting these data. The comparison illustrated by Figure 6.1 suggests that issues with flooding, experienced in June 2015 and February 2018 but not July 2017, were associated with higher intensity rainfall, but significant response of groundwater levels is likely to be a combination of both duration and intensity.
A difficulty for extracting temporal information from event comparisons is the lack of knowledge of how the rainfall intensity and volumes may have varied across the city from one storm to the next. RRI calculations assume constant rainfall across the network area during an event, although, to some degree, rainfall will inevitably have varied at a kilometre-scale or greater. Stormwater infrastructure efficiency has also been improved by more regular maintenance from 2016 and has potential to decrease infiltration and RRI values since that time, but this may not be constant across the entire network. At this stage, the extent of temporal variability and potential factors influencing recharge to groundwater compiled in Figure 6.1 and Table 6.4 is too limited in number (and potentially scale of events) to reach clarity over the role of controlling variables. Changes in the stormwater system efficiency, wastewater mains integrity, soil moisture, rainfall duration and rainfall intensity are among factors that need to be more consistently quantified over a greater number of events.
6.4 Available Subsurface Storage
As there was a degree of spatial consistency in the site response of groundwater levels, we wanted to provide a measure of the spatial variation in available subsurface storage that relates the position of the water table to the RRI. An equivalent to the rainfall required to lift groundwater to the ground surface and cause inundation can be estimated from the DTW divided by the RRI.
Grids of the[RAINstorage_fromGWLmedian] available subsurface rainfall = [DTWmedianstorage (in_2019] mm) /provide [RRI] a proxy for the rainfall that can occur before groundwater can be expected to reach the ground surface. These were derived using the mean, median and maximum levels of groundwater. Ratios of DTW (e.g. DTWmedian_2019) divided by RRI were recalculated into units of mm (x1000) to reflect units in which rainfall is most commonly reported. The amounts required to lift groundwater to surface mostly ranged from 1 to 5000 mm and were greater around the raised margins of South Dunedin, where the groundwater is deep. Since values over 1000 mm significantly exceed daily rainfall expected for the city (Carey-Smith et al. 2018), grid points with values >1000 mm were reset to equal 1000.
It is unlikely that the RAINstorage grid will reflect the absolute totals of rainfall needed to generate flooding; instead, it is better used to indicate where the water table elevation and RRI combine in the subsurface to make some areas more vulnerable than others. The RAINstorage
GNS Science Report 2020/11 55
grid has been classified into places where light (0–20 mm), moderate (20–75 mm) or heavy (75–150 mm) rainfall have the potential to raise groundwater from median levels to the ground (Figure 6.4). Those areas highlighted purple (moderate) and pink (heavy) in Figure 6.4 include many parts of suburbs that experienced inundation during the June 2015 floods. Importantly, their definition is a function of DTW and RRI, as opposed to being places of low-lying topography alone (see discussion below).
Although the RAINstorage was calculated relative to median groundwater levels, many more places would have been shown to be vulnerable to equivalent rainfalls when groundwater is high. The grid hints at the role groundwater (and a relatively shallow water table) may play in local surface floodwater hazard, with potential for improving stormwater management and flood warning. When the June 2015 floodwater inundation contours have been plotted over the RAINstorage grid (Figure 6.5), there is coincidence of many areas. But, there is an equally important reminder of the many places in the lower-lying land towards the east (e.g. Tainui and parts of St Kilda) for which no serious flooding issues were reported. While some suburbs appear well-matched, the model appears to ‘fail’ through its inability to provide first order of explanation of why places with a high RRI in the east did not experience flooding. One potential explanation is that the water table median model is a statistical surface and does not account for pre-event variations – such pre-event groundwater level at Tonga Park (in the west) sitting at its 86th percentile, whereas the Culling Park groundwater level (in the east) was at the 57th percentile. Perhaps the efficiency of the stormwater network and variations in rainfall intensity are important parameters to consider along with sediment storativity and hydraulic conductivity. Further work is clearly needed to elucidate the relationship between shallow groundwater and flooding.
Figure 6.4 Model of RAINstorage_fromGWLmedian, being a grid measure of the available subsurface storage above the median water table converted to a rainfall equivalent required to lift groundwater to the ground surface and cause inundation. Values in mm of rain equivalent based on a 24–48 hour event, restricted to >0 and <1000 mm.
56 GNS Science Report 2020/11
Figure 6.5 Contours of floodwater inundation from June 2015 (see Figure 6.3) plotted over the RAINstorage_fromGWLmedian grid. While there are some spatial correlations, it is not immediately clear why the low-lying area in the east (Tainui) did not flood in 2015.
GNS Science Report 2020/11 57
7.0 EFFECTS OF SEA-LEVEL RISE
Simple geometric models to depict how rising sea level will affect groundwater have been developed using the present-day statistical surfaces. At their simplest, the models assume that the absolute position of South Dunedin and Harbourside groundwater levels will also rise and that the shape of the water table in the future will be, on average, the same as at present. By taking the mapped position of groundwater levels into account, the geometric models are quite distinct from ‘bathtub models’, which assume a horizontal water table that is everywhere equilibrated with MSL (Figure 1.1). Instead, the geometric models account for subsurface heterogeneity and highlight spatial variations in the water table height and ground elevation. These geometric models are strongly empirical, simplifying many variables and controlling processes into a single parameter, and do not account for groundwater flow and possible changes in water-budget mass balance. They have been provided in part to help constrain and inform more complicated and computationally difficult models presently being developed as part of the NZ SeaRise project.7
The geometric models contain a number of implicit assumptions: 1. Future positions of the water table will equilibrate with the harbour and ocean as they do with MSL along the coast at present. 2. Tides will continue to influence groundwater levels to the same extent they do now, depending on the permeability of the sediment and distance from the coast. 3. Processes controlling the present hydraulic gradients will not change sufficiently to alter the water table shape, with slopes continuing to be maintained by rainfall recharge, flow of water from the hill suburbs and with potential to be locally affected by urban infrastructure. 4. In areas away from the coast, the water table will continue to sit at an elevated position above sea level.
The digital elevation model is an important present-day reference frame, although the ± 0.2 m (2 sigma) uncertainty of LiDAR surveying is within the realm of the sea-level changes being modelled. Calculated groundwater elevations do not account for possible changes in land elevation (subsidence or uplift) and, for the most part, reflect relative sea-level rise (however, see below). It is assumed the initial experiences of groundwater inundation, as opposed to rainfall and surface water flooding, are more likely to be associated with periods when the highest tides begin to intersect this ground surface. The MHWS is adopted as a high-tide reference level. Until such a time as local sea-level rise models are published (e.g. by NZ SeaRise), an explicit assumption has been adopted that: 5. Future tidal ranges will be similar to those at present, and the present day 1.1 m difference between MHWS and MSL in Dunedin is a reasonable proxy for the difference in the future.
7 www.searise.nz The NZ SeaRise project is a five-year (2018–2023) research programme funded by the Ministry of Business, Innovation & Employment and hosted at Victoria University of Wellington that aims to improve predictions of sea-level rise in Aotearoa New Zealand to 2100 and beyond. As part of the NZ SeaRise project, GNS Science are re-developing, adapting and improving hydrological groundwater models first constructed by ORC (Rekker 2012).
58 GNS Science Report 2020/11
Table 7.1 Approximate years when specific sea-level rise increments (metres above the 1986–2005 baseline) may be reached for the various projection scenarios of sea-level rise for the wider New Zealand region. Reproduced from Bell et al. (2017).
Year Achieved Year Achieved Year Achieved Year Achieved Sea-Level Rise for RCP2.6 for RCP4.5 for RCP8.5 for RCP8.5 H+ (m) (Median) (Median) (Median) (83th Percentile) 0.3 2070 2060 2050 2045
0.5 2110 2090 2075 2060
0.8 2175 2140 2100 2085
Scenarios have been developed for 30, 50 and 80 cm sea-level rise increments, which match levels commonly provided in planning guidance documents (e.g. Bell et al. 2017). The time at which these increments will be reached is still very uncertain, being highly dependent on absolute versus local sea-level models and mitigation behaviour (Table 7.1).
The workflow is based on changes to the elevation of the mean groundwater surface derived from 2019 observations (e.g. GWL_median), with constant offsets of 0.3, 0.5 and 0.8 m added for sea-level rise, to produce a new, long-term equilibrated, statistical position of the water table (e.g. SLR30cm_GWLmedian). To incorporate the influence of tides, and the decay with distance away from the harbour and coast, a tidal component is then added. This is derived by multiplying the present-day 1.1 m difference between MHWS and MSL by the local tidal efficiency grid.
The geometric model equation, and nomenclature, follows the form:
- where [ ] indicates[SLR30cm_GWLmhws] an interpolated = statistical([GWLmedian_20 surface19]+0.3 grid (raster m) + ((MHWS dataset).MSL) DTW *[ TE_%]/100)grids have also been calculated by subtracting groundwater elevation from the 2009 LiDAR DEM.
-
A full list of [SLR30cm_DTWmhws]sea-level rise datasets = [DEM] is provided [SLR30cm_GWL in Tablemhws] 7.2 and exemplified in Figures 7.1, 7.2 and 7.3. While the geometric models are relatively quick to develop and a useful first interpretation, there are a number of important caveats to consider when using these data: 1. The models are simple empirical geometric approximations of statistical condition. They are not full hydrological flow models that incorporate water mass balance and so are unable to incorporate potential changes to a range of climate-related parameters that may also be associated with sea-level rise. In particular, rainfall duration, intensity and event frequency, or air pressure and storm surge, have at least some potential to cause short-term changes to the water table position and its shape. It may pay to test and rationalise whether dynamic behaviour in the future is reliably captured by statistics such as standard deviation or range, especially before using these data for hazard probabilities and/or recurrence timeframes.
GNS Science Report 2020/11 59
2. Land elevation is assumed to be constant, and the geometric models make no distinction between relative and absolute sea-level rise. There is some evidence for land subsidence in Dunedin (Morris 2016), with contemporary rates of up to 2 mm/year indicated in unpublished cGPS data (University of Otago Surveying School / P Denys pers. comm.) but, as yet, there is no evidence for land-motion over longer (geological) time frames (e.g. >1,000 years) (e.g. Beavan and Litchfield 2012). Unpublished satellite radar data also suggest subsidence that is significant in the context of sea-level rise and with differential rates across the city (GNS Science / I Hamlin pers. comm.). The geometric models may need an adjustment for land motion, particularly if subsidence is not uniform across the modelled area. 3. The geometric models assume the magnitude of astronomical tides and harbour amplification effect will not change. A constant, present-day MHWS-MSL 1.1 m offset has been used. This might need to be updated once local sea-level models have been published. 4. The models take no account of groundwater density. Freshwater is less dense and tends to sit above saline groundwater. Landward advance of the saltwater interface with rising sea level has the potential to locally raise the water table position. 5. The models do not account for development of springs, runoff and any effects of discharge to surface. Springs may develop, and losses of groundwater to surface runoff have the potential to locally constrain the water table where it intersects with ground level. 6. No attempt has been made to account for changing interaction with stormwater and/or wastewater networks or imperviousness of urban ground surfaces. As the water table rises, there is potential for increased infiltration into, or losses from, the networks. There is also potential that the networks could become less efficient, or even overwhelmed. The extent to which this may locally affect groundwater levels, and the groundwater-network interaction, needs further investigation. 7. Water table interpolation, and subsequent DTW calculations, are strongly dependent on the spatial distribution of groundwater monitoring sites. Interpolation of 2019-data- generated results appear locally problematic between monitoring sites in the Forbury to Caversham region. Some new monitoring sites and further work is required to fully resolve uncertainties in this area.
60 GNS Science Report 2020/11
Table 7.2 Summary of grids developed to examine potential effects of sea-level rise. For more detailed descriptions, see Table A3.4.
File Name Summary SLR30cm_GWLmedian Elevation (in metres NZVD2016) of a modelled, median level of groundwater equilibrated with 30 cm sea-level rise.
SLR30cm_GWLmhws Elevation (in metres NZVD2016) of a modelled MHWS level of groundwater equilibrated with a 30 cm sea-level rise, incorporating a tidal component.
SLR30cm_DTWmedian Depth (below ground) of a modelled, median level of groundwater equilibrated with a 30 cm sea-level rise.
SLR30cm_DTWmhws Depth (below ground) of a modelled MHWS level of groundwater equilibrated with a 30 cm sea-level rise, incorporating a tidal component.
SLR50cm_GWLmedian Elevation (in metres NZVD2016) of a modelled, median level of groundwater equilibrated with a 50 cm sea-level rise.
SLR50cm_GWLmhws Elevation (in metres NZVD2016) of a modelled MHWS level of groundwater equilibrated with a 50 cm sea-level rise, incorporating a tidal component.
SLR50cm_DTWmedian Depth (below ground) of a modelled, median level of groundwater equilibrated with a 50 cm sea-level rise.
SLR50cm_DTWmhws Depth (below ground) of a modelled MHWS level of groundwater equilibrated with a 50 cm sea-level rise.
SLR80cm_GWLmedian Elevation (in metres NZVD2016) of a modelled, median level of groundwater equilibrated with an 80 cm sea-level rise.
SLR80cm_GWLmhws Elevation (in metres NZVD2016) of a modelled MHWS level of groundwater equilibrated with an 80 cm sea-level rise.
SLR80cm_DTWmedian Depth (below ground) of a modelled, median level of groundwater equilibrated with an 80 cm sea-level rise.
SLR80cm_DTWmhws Depth (below ground) of a modelled MHWS level of groundwater equilibrated with an 80 cm sea-level rise.
GNS Science Report 2020/11 61
Figure 7.1 Geometric model of the elevation (in NZVD2016) of the water table at MHWS once there has been 50 cm of sea-level rise. The modelled position is derived by adding both a 50 cm offset and a tidal response to the 2019 median groundwater level.
Figure 7.2 Geometric model of the DTW at MHWS once there has been 50 cm of sea-level rise.
62 GNS Science Report 2020/11
Figure 7.3 Comparative figure showing geometric models of the elevation (in NZVD2016) of the water table elevation at MHWS modelled for 2019 (A), or after 30 cm (B), 50 cm (C), or 80 cm (D) sea-level rise. DTW models, relative to the LiDAR land surface elevation, are shown for 50 cm (E) and 80 cm (F) sea-level rise scenarios.
GNS Science Report 2020/11 63
8.0 SUMMARY AND CONCLUSIONS
The city of Dunedin is one of many coastal urban areas in New Zealand facing a massive challenge with climate change and sea-level rise. The city has a large number of assets and critical infrastructure that are situated at, or close to, sea level. Presently protected by a slightly elevated margin of reclaimed land and fragile sand dunes, the flat-lying coastal land in South Dunedin and Harbourside is crucial to present functional operations of the city. However, the land is underlain by shallow groundwater, which limits the amount of unsaturated ground available to store rain and runoff, limits drainage space and creates opportunities for infiltration into stormwater and wastewater networks. Groundwater levels are expected to increase in coastal areas throughout New Zealand in response to sea-level rise, but groundwater beneath Dunedin’s narrow coastal strip is faced with encroachment from all directions: the harbour in the east, the ocean in the south and changing runoff- and recharge-related flows from hill catchments in the north and west. For simple geometrical reasons of its geography, Dunedin’s challenge is likely to be particularly demanding compared with New Zealand’s other coastal centres.
Future flooding concerns can be addressed through expensive and complex hard infrastructure solutions or by adaptive solutions that take into account the natural underlying conditions of low-lying Dunedin. Adaptive solutions are no less complex but may save time and money in the city’s effort to manage flooding and groundwater inundation. This report outlines a basis for future planning and mitigation of natural hazards in Dunedin.
Science to monitor groundwater (levels and chemistry), characterise drivers of its natural fluctuations and understand future behaviour is an important step in gathering the information needed to support decisions regarding how long to continue investment to protect and maintain assets that serve the community. Knowledge of Dunedin’s groundwater took a major step forward with the installation of a network of piezometers during 2019. Water level, temperature and specific conductance were measured at 15 minute intervals and collated into a Hilltop time-series database.
While numerous two-dimensional hydrographs have been examined of individual site groundwater levels over time, this report was focused toward the spatial analysis of groundwater data, principally to survey the water table height and factors influencing the water table position and its geometry in the low-lying land. A wide variety of statistics were generated for each site, including median, maximum, minimum, 95th and 5th percentiles, mean, standard deviation and range of groundwater levels. Other collated site data include: tidal amplitude, efficiency and phase lag; distance from harbour or sea; sample pH, electrical conductance and modelled percentage seawater; and a rainfall response index reflecting rainfall recharge efficiency. Site data were reviewed relative to each other as a series of points and/or interpolated over an 8.9 km2 area in a series of 20 m cell-sized grids to represent the flat-lying South Dunedin and Harbourside land.
The groundwater monitoring network, for the most part, appears to be of sufficient density for confident spatial interpolation of the shallow water table, generating realistic and sensible groundwater depths relative to land elevation. Groundwater throughout South Dunedin is unconfined, but there is a locally perched freshwater aquifer in dune sand at St Kilda, and groundwater is locally confined in Harbourside. Care is needed when comparing the Harbourside piezometers with those in South Dunedin, as there are some differences in depth. Fitzroy St piezometer (I44/1121) is insufficiently deep to capture the lowest levels of groundwater. It is situated in the suburb of Forbury, where modelling groundwater is also problematic due to sharp changes in topography. As this area is a locus of hillslope runoff and
64 GNS Science Report 2020/11
surface flooding, installation of two new piezometers to depths of 6 m is strongly recommended. The Kings (I44/1105) piezometer could also be resituated and/or reinstalled to a greater depth (from 3 to 6 m) to ensure late-summer to early-autumn low groundwater levels are always able to be recorded in the centre of South Dunedin.
The report provides an outline of some key observations of groundwater over the first year of operation (6/03/2019–12/02/2020). The data contain some regular, anomalous periods associated with sampling drawdown and recovery that need to be removed for the analysis of natural processes. Groundwater levels were dominated by rainfall-related peaks and recessions, showed some tendency towards lower levels during late-summer to early-autumn and contain subtle tidal-cycle-related fluctuations. Groundwater level data from four longer- term monitoring sites indicated that records from 2019 were a reasonable proxy for average conditions during the previous decade but missed some of the extreme levels experienced during that decade. Although event rainfall and seasonal variability can be assessed directly from the decadal records, the groundwater level records are of insufficient length to directly assess the extent to which the 2–4 year El Niño–Southern Oscillation (ENSO) or 20- to 30-year Inter-decadal Pacific Oscillation (IPO) cycle. As there were no extreme rainfall events during 2019, indirect modelling will be needed to fully understand the temporal effects that heavy rain intensity and duration have on groundwater behaviour.
A series of surfaces representing the elevation of the water table and DTW were developed from monitoring observations during 2019. Statistics derived for each monitoring piezometer were interpolated into raster grids with 20 m cells, constrained at the harbour and coast using a series of boundary ‘control points’. These ‘statistical surfaces’ approximate the potentiometric pressure and condition of groundwater over extended periods (e.g. long-term hydraulic gradients), but ‘static surfaces’ or ‘event-related surfaces’ will be better for characterising short-term behaviour (e.g. directions and velocity of flow). The statistical surfaces are particularly suited for engineering, such as foundation design and probabilistic assessment of liquefaction vulnerability.
Depth to groundwater table grids, derived relative to a LiDAR survey of topography, highlight that the position of the water table relative to ground does not necessarily reflect topographic elevation. The lowest lying suburbs do not necessarily coincide with shallowest groundwater. Instead, the water table has changes in elevation and gradients at kilometre-scales that are important from suburb to suburb. Median groundwater levels (GWLmedian_2019) are mostly <1 m (everywhere sitting above MSL) and highest adjacent to hill suburbs, suggesting a degree of discharge out through the flat-lying land.
Groundwater adjacent (<1 km) to the coast has specific (electrical) conductance, most likely reflecting saline intrusion and sea-water mixing. Three sites have strong, regular, tidal fluctuations at 30–51% efficiency of the nearby harbour or ocean. Elsewhere, tidal influence is weak (0.1–1% efficiencies at <1.6 km) or undiscernible, with lagging as much as 360 minutes behind the harbour or ocean tide. Tidal efficiencies were mapped and interpolated for use in sea-level rise models.
Groundwater response to rainfall has been characterised using a RRI. Values of RRI are spatially variable from site to site depending on factors such as local stormwater drainage, extent of impervious urban surfaces, piezometer construction method and its protection from runoff, and subsurface sediment storativity and permeability. There are some hints at temporal variability of rainfall response and that groundwater levels could potentially have a role in whether heavy rain events do, or do not, cause flooding. Further work is needed to understand the relative effects of pre-event groundwater level, rainfall duration and intensity, and stormwater network performance.
GNS Science Report 2020/11 65
Thermal behaviour and variability of groundwater temperature is largely a function of the depth of the piezometer below ground. Shallower sites (≤6 m) tend to have 12–15°C median temperature; prominent sinusoidal behaviour reflecting seasonal variation, but no obvious diurnal changes; and local departures due to rain. Deeper piezometers (≥10 m) generally have average values of about 12–13°C, small (± 0.5°C) temperature ranges and disturbed seasonal sinusoids. There are significant differences in the dates of temperature maxima and minima, with many shallow sites showing 4–6-month thermal phase delay recording warmest groundwater in June (winter) and coldest in December (summer). This suggests that heat transfer is substantially a conductive rather than convective process and provides indication that there are low volumes of groundwater exchange and/or throughflow of thermal pulses associated with normal recharge events.
In the absence of aquifer tests, there are several other independent lines of evidence that suggest sediment with low hydraulic conductivity and storativity is reasonably widespread. Strong tidally influenced groundwater fluctuations are limited to within ~250 m of the harbour or ocean, and saline-incursion appears equally restricted in its inland and vertical extent. Tidal responses are otherwise weak and with substantial time-lags, indicative of low bulk diffusivity of the sediment. Delayed chemical response, compared with water-level recovery at sites purged for water sampling, provides clear indication of diffusion processes with low volumes of groundwater exchange and/or throughflow of rainfall recharge. It is a positive aspect of parts of Dunedin’s ground characteristics that may help and enable engineering work to mitigate effects of rising sea level and groundwater.
Large parts of Dunedin’s wastewater and stormwater networks lie at or below the water table and are vulnerable to infiltration. Saltwater intrusion into wastewater pipelines has been an ongoing concern, as the salinity affects secondary treatment efficiency. The likely implications of climate change and sea-level rise (with associated increased rainfall intensity, storm frequency and rising groundwater) is an exacerbation of infiltration and other inflow where infrastructure is not upgraded, which decreases system efficacy. GIS datasets have been generated that identify places where the network sits relative to the water table, including places where infiltration might switch to exfiltration with groundwater fluctuations. The analysis provides a geometric tool to target future monitoring aimed at identifying thresholds for groundwater-network exchanges in relation to network flows, age, pipeline material or construction methods. In turn, it should enable the degree of hazard posed by groundwater to be identified together with any critical tipping points that might be present for system collapse. Concurrent monitoring of network flows and groundwater level fluctuations at a wider number of sites is recommended to fully understand the groundwater infiltration and the extent to which it might be hazardous compared to problematic, so that long-term investments in infrastructure can be rationalised.
Groundwater monitoring during 2019 did not record any heavy (>75 mm) rainfalls, so investigations of the relationship between groundwater levels and surface flooding must rely on data from ORC’s longer-term piezometers (Bathgate Park I44/0005, Tonga Park I44/0006, Kennedy St I44/0007, Culling Park I44/1094). Rainfalls in June 2015, July 2017 and February 2018 occurred at moderate, very high and very low pre-event groundwater levels, respectively. Spatial variability in the recharge was reasonably consistent from site to site, similar to that determined for a moderate rainfall in 2019. However, anecdotal reports suggest there was little flood damage in July 2017, when maintenance of the stormwater system had been a focus of attention, yet flooding might have been expected if hazard predictions were based on the very high pre-event groundwater levels.
66 GNS Science Report 2020/11
Temporal variations in rainfall infiltration to groundwater from event to event are at times counter-intuitive, and any temporal variability the groundwater levels might have on RRI and flood hazard is not yet clear. Changes in the stormwater system efficiency, soil moisture, rainfall intensity or other controlling factors also need to be continuously recorded and a significant number of events analysed. Until such a time as high-fidelity rain radar is also available, or rainfall monitoring and event modelling is improved, extracting temporal information from event comparisons will be fraught by a lack of knowledge of how much the rainfall and its intensity vary across the city during an event.
A measure of the available subsurface storage, equivalent to the rainfall required to lift groundwater to the ground surface and cause inundation, was calculated as a grid model from the DTW divided by the RRI. The grid shows places where light (0–20 mm), moderate (20–75 mm) or heavy (75–150 mm) rainfall have the potential to raise groundwater from median levels to the ground surface. It delineates many of the suburbs that experienced inundation during the June 2015 floods, which shows promise, as local vulnerability is derived as a function of DTW and RRI rather than simply because those suburbs represent places of low-lying topography. Further work is needed to fully elucidate the relationship between shallow groundwater and flood hazard.
Simple geometric models to depict how sea-level rise will affect groundwater levels have been developed using the present-day statistical surfaces. At their simplest, the models assume the absolute position of Dunedin groundwater will also rise and that the shape of the water table in the future will be, on average, the same as at present. By taking the mapped position of the water table elevation into account, the geometric models are quite distinct from ‘bathtub models’, which assume a horizontal water table surface that is everywhere equilibrated with MSL. The geometric models are strongly empirical, with many caveats and implicit assumptions. They simplify many variables and controlling processes into a single parameter and do not account for groundwater flow and possible changes in water-budget mass balance.
Scenarios have been developed for sea-level rise increments of 30, 50 and 80 cm. To incorporate the influence of tides, and their decayed influence with distance away from the harbour and coast, a tidal component was added based on present-day observations of tidal efficiency. The geometric models show progressive changes to the water table position as it is encroached from east, south, west and north, highlighting a threat faced in Dunedin due to its geography. More complex models are presently under development to more accurately assess the effects of sea-level rise and range of uncertainty.
Specific knowledge gaps and recommendations for Dunedin groundwater monitoring and research include: 1. Install more (2–3) piezometers at 6 m depth in the Forbury to Caversham area. 2. Extend the depths of the Fitzroy (I44/1121) and Kings (I44/1105) piezometers to 6 m depth. 3. Install two shallow 6 m piezometers in reclaimed land in the Harbourside area. 4. Incorporate groundwater monitoring observations and data from the new Hospital development into the network. 5. To better understand the relationship between groundwater levels and surface flooding, undertake further examination on the relative effects of pre-event groundwater level, rainfall duration and intensity, and stormwater network performance to elucidate temporal variability of groundwater level and chemical responses. 6. Develop and improve datasets of rainfall recharge and groundwater response to a wider number of events, spanning a range of rainfall duration and intensity.
GNS Science Report 2020/11 67
9.0 ACKNOWLEDGMENTS
Monitoring of groundwater in Dunedin city took a major step forward in 2019 when a consortium of interested parties contributed to development of an improved network. The partnership involved efforts by staff from Otago Regional Council, Dunedin City Council, the Earthquake Commission, GNS Science, New Zealand Centre for Earthquake Resilience (QuakeCoRE) through the University of Canterbury, Oceana Gold and the University of Otago. Technical work was completed by Golders Associates and McNeill Drilling. Subsequent collection and management of groundwater data has been led by Otago Regional Council in conjunction with GNS Science. Tidal data has been supplied by Port Otago through Glen Rowe at the New Zealand Hydrographic Authority (Land Information New Zealand). The authors would like to acknowledge the partner agencies, but also leadership and support from Gavin Palmer, Jean Luc Payan, Frederika Mourot, Amir Levy, John Scott and Paul Gebert. Pascal Sirguey, Clare Lewis and Paul Denys (Otago University Surveying School) are thanked for their help surveying piezometer elevations and sea-level datum conversions.
We also wish to thank the wider community of Dunedin whose enthusiasm to understand the place where they work and live, and its environment and future, is seemingly insatiable. It has certainly helped drive the direction and quality of this work. While there are too many people to individually thank, notable champions supporting groundwater research have been: Hon. Clare Curran (Labour Member of Parliament for South Dunedin); Councillors Rachel Elder, Andrew Whiley and Jim O’Malley (Dunedin City Council); Maria Ioannou and Jared Oliver (Dunedin City Council); Eleanor Doig (South Dunedin Community Group); Ray Macleod (Greater South Dunedin Action Group); Dr Ivan Diaz-Rainey and Assoc. Prof. Janet Stephenson (University of Otago).
Work for this report has involved contributions from the Urban Groundwater project (GNS Science’s Strategic Science Investment Fund) and the NZ SeaRise Programme (an Endeavour programme funded by the Ministry of Business, Innovation and Employment) led by Victoria University of Wellington. The authors would particularly like to thank Richard Levy, Tim Naish, Conny Tschritter and Rogier Westerhoff for their leadership and management in these programmes. The report was significantly improved by internal reviews by Frederika Mourot and Peter Johnson.
68 GNS Science Report 2020/11
10.0 REFERENCES Barrell DJA, Glassey PJ, Cox SC, Smith Lyttle B. 2014. Assessment of liquefaction hazards in the Dunedin City district. Dunedin (NZ): GNS Science. 66 p. + 1 DVD. Consultancy Report 2014/68. Prepared for: Otago Regional Council. Barrell DJA, Glassey PJ, Smith Lyttle B. In prep. Dunedin urban geological map. Lower Hutt (NZ): GNS Science. Scale 1:25,000. (GNS Science urban map series). Beavan RJ, Litchfield NJ. 2012. Vertical land movement around the New Zealand coastline: implications for sea-level rise. Lower Hutt (NZ): GNS Science. 41 p. (GNS Science report; 2012/29). Beca. 2014. Assessment of options for protecting Harbourside and South City from direct impacts of sea level rise. [place unknown]: Beca Ltd. 19 p. + appendices. Prepared for: Dunedin City Council. Bell R, Lawrence J, Allan S, Blackett P, Stephens S. 2017. Coastal hazards and climate change: guidance for local government. Wellington (NZ): Ministry for the Environment. 279 p. Bell RG, Paulik R, Wadwha S. 2015. National and regional risk exposure in low-lying coastal areas: areal extent, population, buildings and infrastructure. Hamilton (NZ): National Institute of Water & Atmospheric Research. 270 p. Client Report HAM2015-006. Prepared for: Parliamentary Commissioner for the Environment Benson WN. 1968. The W.N. Benson geological map of Dunedin district [map]. Wellington (NZ): Department of Scientific and Industrial Research. 1 folded map + 1 booklet, scale 1:50,000. (New Zealand Geological Survey miscellaneous series map; 1). Bishop DG, Turnbull IM. 1996. Geology of the Dunedin area [map]. Lower Hutt (NZ): Institute of Geological & Nuclear Sciences. 1 folded map + 52 p., scale 1:250,000. (Institute of Geological & Nuclear Sciences 1:250,000 geological map; 21). Bouwer H. 1989. The Bouwer and Rice slug test – an update. Groundwater. 27(3):304–309. doi:10.1111/j.1745-6584.1989.tb00453.x. Bouwer H, Rice RC. 1976. A slug test for determining hydraulic conductivity of unconfined aquifers with completely or partially penetrating wells. Water Resources Research. 12(3):423–428. doi:10.1029/WR012i003p00423. Carey-Smith T, Henderson R, Singh S. 2018. High Intensity Rainfall Design System: version 4. Wellington (NZ): National Institute of Water & Atmospheric Research Ltd. 73 p. Client Report 2018022CH. Prepared for: Envirolink. Cook DRL, McCahon IF, Yetton MD. 1993. The earthquake hazard in Dunedin. Wellington (NZ): The Earthquake Commission. 129 p. (EQC Research Project; 91/56) Cournane S. 1992. Seismic and oceanographical aspects of lower Otago Harbour [MSc thesis]. Dunedin (NZ): University of Otago. 216 p. Department of Internal Affairs. 2019. Three Waters review. [accessed 2019 Nov 11]. https://www.dia.govt.nz/Three-waters-review Elder V. 2016 May 24. Flood’s ‘true’ cost $138 million. Otago Daily Times. [accessed 2020 April 17]. https://www.odt.co.nz/news/dunedin/dcc/flood’s-true-cost-138-million Erskine AD. 1991. The effect of tidal fluctuation on a coastal aquifer in the UK. Groundwater. 29(4):556–562. doi:10.1111/j.1745-6584.1991.tb00547.x. Farrier T. 2009. Metadata for Dunedin City Council LiDAR 2009. New Zealand Aerial Mapping. 4 p.
GNS Science Report 2020/11 69
Ferris JG. 1952. Cyclic fluctuations of water level as a basis for determining aquifer transmissibility. Washington (DC): US Geological Survey. 17 p. (Ground-water Notes; 1). Fletcher PT. 2016. The geological evolution of Otago Harbour: a high-resolution seismic reflection study [BSc(Hons) thesis]. Dunedin (NZ): University of Otago. 145 p. Freeze RA, Cherry JA. 1979. Groundwater. Englewood Cliffs (NJ): Prentice-Hall. 604 p. Fordyce E. 2014. Groundwater dynamics of a shallow coastal aquifer [MAppSc thesis]. Dunedin (NZ): University of Otago. 155 p. Glassey PJ. 2017. How future sea level rise will affect South Dunedin. International Journal of Environmental Impacts. 1(1):40–49. doi:10.2495/ei-v1-n1-40-49. Glassey PJ, Barrell DJA, Forsyth PJ, Macleod R. 2003. The geology of Dunedin, New Zealand, and the management of geological hazards. Quaternary International. 103(1):23–40. doi:10.1016/s1040-6182(02)00139-8. Glassey PJ, Barrell DJA, Smith Lyttle B, Hornblow S, Mackey B. Submitted. Characterising subsurface South Dunedin to better define multiple natural hazards. In: New Zealand Geotechnical Society Symposium 2020: good grounds for the future. 2020 Oct 15–17; Dunedin (NZ). 10 p. Glover RE. 1959. The pattern of fresh-water flow in a coastal aquifer. Journal of Geophysical Research (1896–1977). 64(4):457–459. doi:10.1029/JZ064i004p00457. Goldsmith M, Hornblow S. 2016. The natural hazards of South Dunedin. Dunedin (NZ): Otago Regional Council. 69 p. Goldsmith M, Payan J-L, Morris R, Valentine C, MacLean S, Xiaofeng L, Vaitupu N, Mackey B. 2015. Coastal Otago flood event 3 June 2015. Dunedin (NZ): Otago Regional Council. 56 p. Habel S, Fletcher CH, Rotzoll K, El-Kadi AI. 2017. Development of a model to simulate groundwater inundation induced by sea-level rise and high tides in Honolulu, Hawaii. Water Research. 114:122–134. doi:10.1016/j.watres.2017.02.035. Hsieh PA, Bredehoeft JD, Farr JM. 1987. Determination of aquifer transmissivity from Earth tide analysis. Water Resources Research. 23(10):1824–1832. doi:10.1029/WR023i010p01824. Hoitink AJF, Jay DA. 2016. Tidal river dynamics: implications for deltas. Reviews of Geophysics. 54(1):240–272. doi:10.1002/2015rg000507. Hoover DJ, Odigie KO, Swarzenski PW, Barnard P. 2017. Sea-level rise and coastal groundwater inundation and shoaling at select sites in California, USA. Journal of Hydrology: Regional Studies. 11:234–249. doi:10.1016/j.ejrh.2015.12.055. Hutchinson MF. 1989. A new procedure for gridding elevation and stream line data with automatic removal of spurious pits. Journal of Hydrology. 106(3–4):211–232. doi:10.1016/0022-1694(89)90073-5. Hutchinson MF, Dowling TI. 1991. A continental hydrological assessment of a new grid-based digital elevation model of Australia. Hydrological Processes. 5(1):45–58. doi:10.1002/hyp.3360050105. Hvorslev MJ. 1951. Time lag and soil permeability in ground-water observations. Vicksburg (MS): US Army Waterways Experiment Station. 50 p. Bulletin 36. Lutter TJ. 2018. 3D gravitational inversion modelling of the South Dunedin sub basin, Otago, New Zealand [MSc thesis]. Dunedin (NZ): University of Otago. 109 p. Macfie R. 2016 June 11. Flood fiasco: eye of the storm. New Zealand Listener. 254(3968):22–29.
70 GNS Science Report 2020/11
McDonnell JJ, Beven K. 2014. Debates – the future of hydrological sciences: a (common) path forward? A call to action aimed at understanding velocities, celerities and residence time distributions of the headwater hydrograph. Water Resources Research. 50(6):5342–5350. doi:10.1002/2013wr015141. McNeilly H, Daly M. 2015 June 4. Flooding wreaks havoc in Dunedin. Stuff. [accessed 2020 April 17]. https://www.stuff.co.nz/national/69063192/flooding-wreaks-havoc-in-dunedin Morris C. 2016 July 18. Fears South Dunedin also sinking. Otago Daily Times. [accessed 2020 April 17]. https://www.odt.co.nz/news/dunedin/fears-south-dunedin-also-sinking Opus International Consultants Ltd, URS New Zealand Ltd. 2011. Integrated Catchment Management Plans 2010–2060: phase 1 wastewater system – final report. 113 p. + 9 appendices. Prepared for: Dunedin City Council. Osborn T, Sinclair L. 2011. An investigation worth its salt: targeting saline intrusion. In: Water New Zealand Annual Conference. 2011 Nov 9–11; Rotorua (NZ). Wellington (NZ): Water New Zealand. 18 p. Parliamentary Commissioner for the Environment [PCE]. 2015. Preparing New Zealand for rising seas: certainty and uncertainty. Wellington (NZ): Parliamentary Commissioner for the Environment. 92 p. Paulik R, Stephens S, Wadwha S, Bell R, Popovich B, Robinson B. 2019. Coastal flooding exposure under future sea-level rise for New Zealand. Wellington (NZ): National Institute of Water & Atmospheric Research. 76 p. Client Report 2019119WN. Prepared for: The Deep South Challenge. Rees OW. 2018. Seismic characterisation of Otago Harbour sediments tied to a carbon-dated core record [BSc(Hons) dissertation]. Dunedin (NZ): University of Otago. Rekker J. 2012. The South Dunedin coastal aquifer & effect of sea level fluctuations. Dunedin (NZ): Otago Regional Council. 25 p. Rotzoll K, El-Kadi AI. 2008. Estimating hydraulic properties of coastal aquifers using wave setup. Journal of Hydrology. 353(1–2):201–213. doi:10.1016/j.jhydrol.2008.02.005. Rotzoll K, Fletcher CH. 2013. Assessment of groundwater inundation as a consequence of sea-level rise. Nature Climate Change. 3(5):477–481. doi:10.1038/nclimate1725. Sanders LL. 1998. A manual of field hydrogeology. Upper Saddle River (NJ): Prentice Hall. 381 p. Sangster CA. 2019. Dunedin rock and roll: 3D seismic wave velocity modelling for seismic hazard analysis [MSc thesis]. Dunedin (NZ): University of Otago. 317 p. Seward AM, Prieto A. 2015. New Zealand rock properties: determining thermal properties of shallow soils. In: Horne R, Boyd T, editors. World Geothermal Congress, Australia-New Zealand: views from down under – geothermal in perspective; Melbourne (AU); 2015 Apr 19–24. Bonn (DE): International Geothermal Association. 7 p. Simonson T, Hall G. 2019. Vulnerable: the quantum of local government infrastructure exposed to sea level rise. Wellington (NZ): Local Government New Zealand [LGNZ]. 43 p. + appendices. Smeaton D. 2015. Report on the South Dunedin Flood Survey – June 2015. Dunedin (NZ): TL Survey Services Ltd. 9 p. Prepared for: Otago Regional Council. Tabbagh A, Bendjoudi H, Benderitter Y. 1999. Determination of recharge in unsaturated soils using temperature monitoring. Water Resources Research. 35(8):2439–2446. doi:10.1029/1999wr900134. URS New Zealand Ltd, Opus International Consultants. 2011. Integrated Catchment Management Plans 2010–2060: summary report. 78 p. + appendices. Prepared for: Dunedin City Council.
GNS Science Report 2020/11 71
van Ballegooy S, Cox SC, Thurlow C, Rutter HK, Reynolds T, Harrington G, Fraser J, Smith T. 2014. Median water table elevation in Christchurch and surrounding area after the 4 September 2010 Darfield Earthquake. Version 2. Lower Hutt (NZ): GNS Science. 79 p. + appendices. (GNS Science report; 2014/18). van Ballegooy S, Wentz F, Boulanger RW. 2015. Evaluation of CPT-based liquefaction procedures at regional scale. Soil Dynamics and Earthquake Engineering. 79:315–334. doi:10.1016/j.soildyn.2015.09.016. van Manen SM, Wallin E. 2012. Ground temperature profiles and thermal rock properties at Wairakei, New Zealand. Renewable Energy. 43:313–321. doi:10.1016/j.renene.2011.11.032. Walters RA, Goring DG, Bell RG. 2001. Ocean tides around New Zealand. New Zealand Journal of Marine and Freshwater Research. 35(3):567–579. doi:10.1080/00288330.2001.9517023. Wild M, Bell R, Walsh J. 2005. Otago extreme sea level analysis. Christchurch (NZ): National Institute of Water & Atmospheric Research. 25 p. + appendices. Client Report CHC2005-114. Prepared for: Otago Regional Council. Willis G. 2014. Managing natural hazard risk in New Zealand – towards more resilient communities: a think piece for local and central government and others with a role in managing natural hazards. Wellington (NZ): Local Government New Zealand [LGNZ]. 61 p.
72 GNS Science Report 2020/11
APPENDICES
GNS Science Report 2020/11 73
This page left intentionally blank.
74 GNS Science Report 2020/11
APPENDIX 1 MAP OF PLACE NAMES
Figure A1.1 Map with key place names referred to in this study.
GNS Science Report 2020/11 75
APPENDIX 2 GLOSSARY
AEP Annual Exceedance Probability. The probability of a flood, or inundation event, of a certain size or greater occurring in any year. Expressed as a percentage. ArcGIS A programme for working with maps and geographic information. Used for creating and using maps, compiling geographic data, analysing mapped information and sharing and discovering geographic information. Artesian Referring to an aquifer or underground layer containing groundwater that is under positive pressure. If a well is sunk into the ground, the water will rise and equilibrate at a point that is above ground level. ARI Average Recurrence Interval. Also known as 'annual return period'. It is the average number of years that are predicted to pass before an event of a given magnitude occurs. Aquifer An underground layer of water-bearing permeable rock, rock fractures or unconsolidated materials from which groundwater can be extracted. Aquitard An impermeable layer of sediment, or barrier of rock, that acts as a barrier to the flow of groundwater. Confined Groundwater that is separated from atmospheric pressure by a layer of groundwater relatively impermeable material. DCC Dunedin City Council (Kaunihera a-rohe o Ōtepoti). Drawdown The reduction of the pressure head in an aquifer as the result of the withdrawal of free water. EQC The Earthquake Commission (Kōmihana Rūwhenua). Electrical A measure of the ability of water to conduct electricity that is directly conductivity related to the concentration of ions dissolved in the water. Usually measured in micro- or millisiemens per centimetre (μS/cm or mS/cm). Empirical Based on, concerned with, or verifiable by observation or experience rather than system theory or pure logic. GIS Geographic Information System. A framework for gathering, managing and analysing data rooted in the science of geography. GPS Global Positioning System. Hilltop Software for storing and analysing time-series environmental data. The software is used in New Zealand by regional councils, electricity companies and consulting engineers. http://www.hilltop.co.nz/. Hydraulic For an isotropic porous medium and homogenous fluid, the volume conductivity of water that moves in unit time under a unit hydraulic gradient through a unit area measured at right angles to the direction of flow. Commonly, though imprecisely, taken to be synonymous with permeability. Units: m/day.
76 GNS Science Report 2020/11
Hydraulic Slope of the water table or potentiometric surface. The change in gradient static head per unit of distance in a given direction. If not specified, the direction is generally understood to be that of the maximum rate of decrease in head. Hydraulic head The height above a datum plane (such as sea level) of the column of water that can be supported by the hydraulic pressure at a given point in a ground water system. For a piezometer, the hydraulic head is equal to the distance between the water level in the piezometer and the datum plane. IPCC The Intergovernmental Panel on Climate Change (IPCC). A United Nations body for assessing the science related to climate change. LiDAR Light Detection and Ranging. Provides high resolution topography datasets. LINZ Land Information New Zealand (Toitū te Whenua). Liquefaction A change in state that occurs when a saturated or partially saturated soil substantially loses strength and stiffness in response to an applied stress such as shaking during an earthquake or other sudden change in stress condition. Material that is ordinarily a solid behaves like a liquid. MHWS Mean High Water Springs. The highest level that spring tides reach on the average over a period of time – usually about 24 hours in each semi-lunation (approximately every 14 days) when the range of the tide is greatest (Spring Range). MHL Mean Harbour Level. MLOS Mean Level of Sea. MSL Mean Sea Level. Used in a more general context in this report when not referring specifically to the Otago Harbour or Pacific Ocean. NIWA The National Institute of Water & Atmospheric Research (Taihoro Nukurangi). NZ SeaRise The NZ SeaRise project is a $7.1 million, five-year (2018–2023) research programme funded by the Ministry of Business, Innovation and Employment. The overarching goal of the programme, hosted at Victoria University of Wellington, is to improve predictions of sea-level rise in Aotearoa New Zealand to 2100 and beyond. https://www.searise.nz/ ORC Otago Regional Council. Perched Subsurface water that forms a saturated horizon within porous media groundwater at an elevation higher than the local or regional groundwater table. Permeability Used in a general sense, it is a measure of how easily water can flow through a rock or unconsolidated sediment and how easy it will be to extract the water. A more strict definition is the measure of the relative ease with which a porous medium can transmit a fluid under a potential gradient. It is the property of the medium only and is independent of the fluid. Commonly, but imprecisely, taken to be synonymous with the term hydraulic conductivity (which implies the fluid is water).
GNS Science Report 2020/11 77
Piezometer An open well or standpipe installed into the ground with a slotted or screened casing within a zone where water pressure is being measured. Porosity The ratio of the volume of the interstices to the total volume of rock expressed as a fraction. Effective porosity includes only the interconnected pore spaces available for groundwater transmission; measurements of porosity in the laboratory usually exclude any void spaces caused by cracks or joints (secondary porosity). Potentiometric An imaginary surface representing the elevation and pressure head of surface groundwater and defined by the level to which water rises in a well or piezometer. The water table is a particular potentiometric surface. Rainfall The ratio of the change in groundwater level during a rainfall event to recharge index the total amount of rain during that event. (RRI) Specific A measurement of electrical conductivity that has been made at, Conductance or corrected to, 25°C. Storativity The volume of water an aquifer releases from or takes into storage per unit surface area of the aquifer per unit change in head. Tidal amplitude The elevation of tidal high water above MSL (equals half of the tidal range between high tide and low tide). Tidal efficiency The ratio of the amplitude (or range) of tidal-related fluctuations in (TE) groundwater to the tidal amplitude (or range) in the sea. Expressed herein as a percentage using the nearest water body (harbour or ocean). Transducer A measuring device that converts an applied pressure into an electrical signal. Usually consists of two parts, an elastic material which deforms under the application of pressure and an electrical part which detects this deformation. Can be vibrating-wire, pneumatic or strain-gauge in operation. Converts pressure in groundwater or air into an electrical recording. Unconfined Groundwater in an aquifer where upper water surface (water table) groundwater is at atmospheric pressure and thus is able to rise and fall. Unconsolidated A deposit consisting of loose grains that are not held together by cement. River terrace deposits are a typical example of an unconsolidated aquifer. Water table The surface of a body of unconfined groundwater at which the pressure is equal to that of the atmosphere. The static water level in a well in an unconfined aquifer.
78 GNS Science Report 2020/11
APPENDIX 3 METADATA
Table A3.1 Detailed description of groundwater level GIS datasets generated from the 2019 groundwater monitoring and used in this report. Data are available from GNS Science on request.
File Name Description GWLmax_2019 An interpolated grid surface (20 m cells) mapping the maximum elevation (metres NZVD2016) of groundwater monitored in the period 2019/03/06– 2020/02/12. Surface generated using boundary points for the harbour and Andersons Bay Inlet (set at 1.182 m) and coast (1.281 m). Excludes shallow piezometers in perched aquifer (CE17/0108, F5 Holiday Park and F6 OMES).
GWLmedian_2019 An interpolated grid surface (20 m cells) mapping the median elevation (metres NZVD2016) of groundwater monitored in the period 2019/03/06– 2020/02/12. Surface generated using boundary points for the harbour (-0.224 m), Andersons Bay Inlet (0.25 m) and coast (-0.072 m). Excludes shallow piezometers in perched aquifer (CE17/0108, OMES, Holiday Park). Appears to be reasonably representative of the long-term mean values (checked against 2010–2018).
GWLmedian_2019_25cm Contour lines (metres NZVD2016 at 25 cm intervals) representing the elevation of Contours the median surface height, generated from the interpolated surface GWLmedian_2019.
GWLmin_2019 An interpolated grid surface (20 m cells) mapping the minimum elevation (metres NZVD2016) of groundwater monitored in the period 2019/03/06– 2020/02/12. Surface generated using boundary points for the harbour (-1.546 m), Andersons Bay Inlet (0.19 m) and coast (-1.411 m). Excludes shallow piezometers in perched aquifer (CE17/0108, F5 Holiday Park and F6 OMES).
GWLp05_2019 An interpolated grid surface (20 m cells) mapping the 5th percentile elevation (metres NZVD2016) of groundwater monitored in the period 2019/03/06– 2020/02/12. Surface generated using boundary points for the harbour (-1.123 m), Andersons Bay Inlet (0.2 m) and coast (-0.917 m). Excludes shallow piezometers in perched aquifer (CE17/0108, F5 Holiday Park and F6 OMES).
GWLp95_2019 An interpolated grid surface (20 m cells) mapping the 95th percentile elevation (metres NZVD2016) of groundwater monitored in the period 2019/03/06– 2020/02/12. Surface generated using boundary points for the harbour (0.702 m), Andersons Bay Inlet (0.7 m) and coast (0.769 m). Excludes shallow piezometers in perched aquifer (CE17/0108, F5 Holiday Park and F6 OMES).
GWLrange_2019 An interpolated grid surface (20 m cells) mapping the range (maximum to minimum) of groundwater levels (m) monitored in the period 2019/03/06– 2020/02/12. Surface generated by subtracting the maximum and minimum elevations at each site, then interpolating that difference using boundary points for the harbour (2.73 m), Andersons Bay Inlet (0.99 m) and coast (2.69 m). Excludes shallow piezometers in perched aquifer (CE17/0108, F5 Holiday Park and F6 OMES).
GWLmean_2019 An interpolated grid surface (20 m cells) mapping the mean elevation (metres NZVD2016) of groundwater monitored in the period 2019/03/06– 2020/02/12. Surface generated using boundary points for the harbour (-0.217 m), Andersons Bay Inlet (0.25 m) and coast (-0.073 m). Excludes shallow piezometers in perched aquifer (CE17/108, F5 Holiday Park and F6 OMES). Appears to be reasonably representative of the long-term mean values (checked against 2010–2018).
GNS Science Report 2020/11 79
File Name Description GWLstdev_2019 An interpolated grid surface (20 m cells) mapping the standard deviation of groundwater levels (m) monitored in the period 2019/03/06–2020/02/12. Surface generated using boundary points for the harbour (0.603 m), Andersons Bay Inlet (0.4 m) and coast (0.559 m). Excludes shallow piezometers in perched aquifer (CE17/108, F5 Holiday Park and F6 OMES).
GWLmhws_mean2019 An interpolated grid surface (20 m cells) mapping the groundwater elevation at MHWS. A tidal offset to GWLmean_2019, based on the present-day difference between MHWS and MSL (1.1 m in Dunedin), was multiplied by the tidal efficiency grid of groundwater fluctuations (NB: need to divide tidal efficiency values in percentages to a proportion). GWLmhws_mean2019 = GWLmean_2019 + (1.1*TE).
GWLmhws_median2019 An interpolated grid surface (20 m cells) mapping the groundwater elevation at MHWS. A tidal offset to GWLmedian_2019, based on the present-day difference between MHWS and MSL (1.1 m in Dunedin), was multiplied by the tidal efficiency grid of groundwater fluctuations (NB: need to divide tidal efficiency values in percentages to a proportion). GWLmhws_median2019 = GWLmedian_2019 + (1.1*TE).
GWLdiff_medianmean A simple difference (20 m cells) subtracting the GWLmean_2019 grid from the GWLmedian_2019 to help decide the impact of using parametric versus non-parametric statistics. High (0.044 m) and low (-0.043 m) values indicate there is <5 cm of difference (which is less than uncertainties in the LiDAR DEM and measuring point elevations). Areas of Caversham and central South Dunedin tend to have a higher median, whereas near the harbour and coast the median is lower.
80 GNS Science Report 2020/11
Table A3.2 Detailed description of depth to groundwater (DTW) GIS datasets generated from the 2019 groundwater monitoring and used in this report. Data are available from GNS Science on request.
File Name Description DTWmax_2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to highest groundwater measured in the period 2019/03/06–2020/02/12. Surface generated using GWLmax_2019, with boundary points in the harbour (set at 1.182 m) and coast (1.281 m). Values in metres relative to the 2009 LiDAR land surface model.
DTWmean_2019ag0 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to average groundwater measured in the period 2019/03/06–2020/02/12. Surface generated using GWLmean_2019, with boundary points in the harbour (set at -0.217 m) and coast (-0.073 m). Values in metres relative to the 2009 LiDAR land surface model, with any negative above ground values reset to 0.
DTWmedian_2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to median groundwater level measured in the period 2019/03/06– 2020/02/12. Surface generated using GWLmedian_2019, with boundary points in the harbour (-0.224 m), Andersons Bay Inlet (0.25 m) and coast (-0.072 m). Values in metres relative to the 2009 LiDAR land surface model.
DTWmedian_2019ag0 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to median groundwater level measured in the period 2019/03/06– 2020/02/12. Surface generated using GWLmedian_2019, with boundary points in the harbour (-0.224 m), Andersons Bay Inlet (0.25 m) and coast (-0.072 m). Values in metres relative to the 2009 LiDAR land surface model, with any negative above ground values reset to 0.
DTWmin_2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to lowest groundwater measured in the period 2019/03/06–2020/02/12. Surface generated using GWLmin_2019, with boundary points in the harbour (-1.546 m), Andersons Bay Inlet (0.19 m) and coast (-1.411 m). Values in metres relative to the 2009 LiDAR land surface model.
DTWp05GW_2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to 5th percentile of groundwater measured in the period 2019/03/06– 2020/02/12. Surface generated using GWLp05_2019, with boundary points in the harbour (-1.123 m), Andersons Bay Inlet (0.2 m) and coast (-0.917 m). Values in metres relative to the 2009 LiDAR land surface model.
DTWp95GW_2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to 95th percentile of groundwater measured in the period 2019/03/06– 2020/02/12. Surface generated using GWLp95_2019, with boundary points in the harbour (0.702 m), Andersons Bay Inlet (0.7 m) and coast (0.769 m). Values in metres relative to the 2009 LiDAR land surface model.
DTWmean_2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to average groundwater measured in the period 2019/03/06–2020/02/12. Surface generated using GWLmean_2019, boundary points in the harbour (set at -0.217 m) and coast (-0.073 m). Values in metres relative to the 2009 LiDAR land surface model.
GNS Science Report 2020/11 81
File Name Description DTWmhws_mean2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) of the MHWS level of groundwater measured in the period 2019/03/06– 2020/02/12. Model is based on the present-day difference between MHWS and MSL (1.1 m in Dunedin), multiplied by the tidal efficiency grid of groundwater fluctuations (NB: need to divide tidal efficiency values in percentages to a proportion) using equation GWLmhws_mean2019 = GWLmean_2019 + (1.1*TE). Values are then subtracted from the DEM to give depth in metres relative to the 2009 LiDAR land surface model.
DTWmhws_median2019 An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to MHWS level of groundwater measured in the period 2019/03/06– 2020/02/12. Model is based on the present-day difference between MHWS and MSL (1.1 m in Dunedin), multiplied by the tidal efficiency grid of groundwater fluctuations (NB: need to divide tidal efficiency values in percentages to a proportion) using the equation GWLmhws_median2019 = GWLmedian_2019 + (1.1*TE). Values are then subtracted from the DEM, to give depth in metres relative to the 2009 LiDAR land surface model.
82 GNS Science Report 2020/11
Table A3.3 Detailed description of tidal, rainfall recharge and chemistry GIS datasets generated from the 2019 groundwater monitoring and used in this report. Data are available from GNS Science on request.
File Name Description Piezo_data_202003 Point dataset containing observations of tidal behaviour at each monitoring site. Includes attribute columns characterising tidal response at each site: TideRange_m; TideEff_%; TidePhaseLag_min (in which sites showing no tides are set to a default value of 745 minutes), Tidal_Comment (text field); Nearest_Forcing_Tides (being harbour or ocean); Dist_Sea_m (being the distance from the nearest forcing – harbour or ocean).
TE_percent An interpolated grid (20 m cells) of the dimensionless tidal efficiency, derived from the ratio of the tidal change of groundwater level in piezometers to the tidal change in the nearby ocean or harbour (as measured at Green Island and Fryatt St) driving the tidal response.
TIDE_amplitude_m An interpolated grid (20 m cells) of the groundwater 6.2-hour tidal amplitude measured in piezometers, interpolated with coastal (0.75 m) and harbour (0.9 m) boundary points. Tidal amplitude = 0.5*Tidal Range.
TIDE_range_m An interpolated grid (20 m cells) of the groundwater high to low tidal range measured in piezometers, interpolated with coastal (1.5 m) and harbour (1.8 m) boundary points.
RRI_selected An interpolated grid (20 m cells) of the dimensionless RRI. Site values are calculated from observations of rainfall events and representative values selected for interpolation using change in groundwater level divided by the rainfall amount (dGWL/TotalRain). Sites can be expected to have variable responses depending on local stormwater drainage, extent of 'hard' urban surfaces, piezometer/well completion and subsurface sediment storativity and permeability. Changes in groundwater level have been normalised against rainfall measured at Musselburgh, which assumes rainfall is constant across the city.
RRI_contours Contours (in values = 1) from an interpolated grid (20 m cells) of the dimensionless RRI. Site values are calculated from observations of rainfall events and representative values selected for interpolation. Sites can be expected to have variable responses depending on local stormwater drainage, extent of 'hard' urban surfaces, piezometer/ well completion and subsurface sediment storativity and permeability. Changes in groundwater level have been normalised against rainfall measured at Musselburgh, which assumes rainfall is constant across the city.
RAINstorage_fromGWLmean An estimate of the rainfall infiltration limit (in mm) and/or available storage before groundwater can be expected to reach the ground surface, based on the mean position of groundwater (GWLmean_2019) and the RRI grid of response site values observed to c.30 mm rainfall events in 2019 and 2012 (Fordyce data). Any grid points returning values >1000 mm, reflecting deep groundwater or very low RRI, have been set to equal 1000 (which exceeds the limit of any 24-hour rainfall expected for Dunedin).
GNS Science Report 2020/11 83
File Name Description RAINstorage_fromGWLmedian An estimate of the rainfall limit (in mm) and/or available storage before groundwater can be expected to reach the ground surface, based on the median position of groundwater (GWLmedian_2019) and the RRI grid of response site values observed to c.30 mm rainfall events in 2019 and 2012 (Fordyce data). Any grid points returning values >1000 mm, reflecting deep groundwater or very low RRI, have been set to equal 1000 (which exceeds the limit of any 24 hour rainfall expected for Dunedin). chemmodelpoints_201912 Groundwater chemistry point data used for modelling Dunedin groundwater. A series of field and laboratory pH and electrical conductivity observations at wells, plus assigned values at points along the coast and harbour for interpolation of grids. Many of the electrical conductivity and pH values were affected by purging. The attributes cond_2019EC (in μS/cm) and pH_2019pH were selected as representative for interpolation. Data from Fordyce (2014) corrected and included. conductivity_2019 An interpolated grid surface (20 m cells) mapping the specific conductance (in μS/cm) in groundwater. Generated using the chemmodelpoints_201912 attribute cond_2019EC. Data from Fordyce (2014) corrected and included. Provides a proxy for the incursion of seawater. A few small areas had extrapolation values <0 due to the interpolation method, so these have been reset to equal 1 μS/cm.
SEAWATER_percent Grid model (20 m cells) of saline incursion as percent seawater, based on the interpolated grid of specific conductivity (conductivity_2019 generated from electrical conductivity) and a freshwater to seawater mixing model where freshwater = 100 μS/cm and seawater = 50860 μS/cm.
SEAWATER_percent Contours (10% intervals) of the grid model of saline incursion as percent _contours seawater. Based on the interpolated grid of specific conductivity (conductivity_2019 generated from electrical conductivity) and a freshwater to seawater mixing model where fresh = 100 μS/cm and seawater = 50860 μS/cm.
84 GNS Science Report 2020/11
Table A3.4 Detailed description of GIS datasets generated to examine the potential effects of sea level in this report. Data are available from GNS Science on request.
File Name Description SLR30cm_GWLmedian An interpolated grid surface (20 m cells) mapping the elevation (metres NZVD2016) to median groundwater level equilibrated with a 30 cm sea-level rise. A simple 0.3 m offset was added to the GWLmedian_2019.
SLR30cm_GWLmhws An interpolated grid surface (20 m cells) mapping the elevation (metres NZVD2016) to groundwater at MHWS following 30 cm sea-level rise. A 30 cm offset was added to the GWLmedian_2019 to create SLR30cm_GWLmedian, then a tidal component added to lift the median to MHWS. The tidal offset was based on the present-day difference between MHWS and MSL (1.1 m in Dunedin), multiplied by the local tidal efficiency of groundwater fluctuations.
SLR30cm_DTWmedian An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to median groundwater level following a 30 cm sea-level rise. A 30 cm offset was added to the GWLmedian_2019 to create SLR30cm_GWLmedian, which was then subtracted from the ground elevation DEM. Values in metres relative to the 2009 LiDAR land surface model. Negative values represent inundation of groundwater.
SLR30cm_DTWmhws An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to groundwater at MHWS following 30 cm sea-level rise. A 30 cm offset was added to the GWLmedian_2019 to create SLR30cm_GWLmedian, then a tidal component added to lift the median to MHWS. The GWL was then subtracted from the ground elevation DEM, returning DTW in metres relative to the 2009 LiDAR land surface model. Negative values represent inundation of groundwater.
SLR50cm_GWLmedian An interpolated grid surface (20 m cells) mapping the elevation (metres NZVD2016) to median groundwater level equilibrated with a 50 cm sea-level rise. A simple 0.5 m offset was added to the GWLmedian_2019.
SLR50cm_GWLmhws An interpolated grid surface (20 m cells) mapping the elevation (metres NZVD2016) to groundwater at MHWS following a 30 cm sea-level rise. A 50 cm offset was added to the GWLmedian_2019 to create SLR50cm_GWLmedian, then a tidal component added to lift the median to MHWS. The tidal offset was based on the present-day difference between MHWS and MSL (1.1 m in Dunedin), multiplied by the local tidal efficiency of groundwater fluctuations.
SLR50cm_DTWmedian An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to median groundwater level on a 50 cm sea-level rise. A 50 cm offset was added to the GWLmedian_2019 to create SLR50cm_GWLmedian, which was then subtracted from the ground elevation DEM. Values in metres relative to the 2009 LiDAR land surface model. Negative values represent inundation of groundwater.
GNS Science Report 2020/11 85
File Name Description SLR50cm_DTWmhws An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to groundwater at MHWS following 50 cm sea-level rise. A 50 cm offset was added to the GWLmedian_2019 to create SLR50cm_GWLmedian, then a tidal component added to lift the median to MHWS. The GWL was then subtracted from the ground elevation DEM, returning DTW in metres relative to the 2009 LiDAR land surface model. Negative values represent inundation of groundwater.
SLR80cm_GWLmedian An interpolated grid surface (20 m cells) mapping the elevation (metres NZVD2016) to median groundwater level equilibrated with an 80 cm sea-level rise. A simple 0.8 m offset was added to the GWLmedian_2019.
SLR80cm_GWLmhws An interpolated grid surface (20 m cells) mapping the elevation (metres NZVD2016) to groundwater at MHWS following 80 cm sea-level rise. An 80 cm offset was added to the GWLmedian_2019 to create SLR80cm_GWLmedian, then a tidal component added to lift the median to MHWS. The tidal offset was based on the present-day difference between MHWS and MSL (1.1 m in Dunedin), multiplied by the local tidal efficiency of groundwater fluctuations.
SLR80cm_DTWmedian An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to median groundwater level on a 30 cm sea-level rise. An 80 cm offset was added to the GWLmedian_2019 to create SLR80cm_GWLmedian, which was then subtracted from the ground elevation DEM. Values in metres relative to the 2009 LiDAR land surface model. Negative values represent inundation of groundwater.
SLR80cm_DTWmhws An interpolated grid surface (20 m cells) mapping the depth (metres relative to ground) to groundwater at MHWS following 80 cm sea-level rise. An 80 cm offset was added to the GWLmedian_2019 to create SLR80cm_GWLmedian, then a tidal component added to lift the median to MHWS. The groundwater elevation was then subtracted from the ground elevation DEM, returning DTW in metres relative to the 2009 LiDAR land surface model. Negative values represent inundation of groundwater.
86 GNS Science Report 2020/11
Principal Location Other Locations
1 Fairway Drive, Avalon Dunedin Research Centre Wairakei Research Centre National Isotope Centre Lower Hutt 5010 764 Cumberland Street 114 Karetoto Road 30 Gracefield Road PO Box 30368 Private Bag 1930 Private Bag 2000 PO Box 30368 Lower Hutt 5040 Dunedin 9054 Taupo 3352 Lower Hutt 5040 New Zealand New Zealand New Zealand New Zealand T +64-4-570 1444 T +64-3-477 4050 T +64-7-374 8211 T +64-4-570 1444 www.gns.cri.nz F +64-4-570 4600 F +64-3-477 5232 F +64-7-374 8199 F +64-4-570 4657