Interconnection of surface and groundwater systems – · PDF fileLeading policy and reform in sustainable water management Interconnection of surface and . groundwater systems –

Leading policy and reform in sustainable water management

Interconnection of surface and groundwater systems – river losses from losing-disconnected streams

Publisher

NSW Office of Water

Level 18, 227 Elizabeth Street GPO Box 3889 Sydney NSW 2001

T 02 8281 7777 F 02 8281 7799

[email protected]

www.water.nsw.gov.au


August 2011

ISBN 978 1 74263 018 2

This report may be cited as:

Brownbill R.J., Lamontagne S., Williams R.M., Cook P.G., Simmons C.T.,

Merrick N., 2011 Interconnection of surface and groundwater systems

– river losses from losing-disconnected streams. Technical final report,

June 2011, NSW Office of Water, Sydney

NOW 11_187.1

The NSW Office of Water manages the policy and regulatory frameworks

for the state’s surface water and groundwater resources to provide a

secure and sustainable water supply for all users. The Office of Water also

supports water utilities in the provision of water and sewerage services

throughout New South Wales.

© State of New South Wales through the Department of Trade, Industry, Regional Infrastructure and Services, 2011

This material may be reproduced in whole or in part for educational and non-commercial use, providing the meaning is unchanged and its source, publisher and authorship are clearly and correctly acknowledged.

Disclaimer: While every reasonable effort has been made to ensure that this document is correct at the time of publication, the State of New South Wales, its agents and employees, disclaim any and all liability to any person in respect of anything or the consequences of anything done or omitted to be done in reliance upon the whole or any part of this document.

Acknowledgements

This project was funded by the NSW Office of Water, and the National Water Commission through the Raising National Water Standards Program. This Australian Government program supports the implementation of the National Water Initiative by funding projects that are improving Australia’s national capacity to measure, monitor and manage its water resources.

This project was a collaboration involving CSIRO Land and Water, Flinders University South Australia, the National Centre for Groundwater Research and Training, Heritage Computing and Groundwater Imaging Pty Ltd. The project team wishes to acknowledge the contribution of the Groundwater Management and Hydrometric Services Units of the NSW Office of Water.

.

Authors

Brownbill R.J.1, Lamontagne S.2, Williams R.M.1, Cook P.G.2,3,

Simmons C.T.3, Merrick N.4

1NSW Office of Water, 2CSIRO Land and Water, 3Flinders University and National Centre for Groundwater Research and Training, 4Heritage Computing.

Note: There have been changes made to Table 4 and 5 in this report. The previous version of the report numbered NOW 11_187 published in July 2011, is replaced by this August update.


Contents

Executive summary ................................................................................................................................v Key results......................................................................................................................................v

Theoretical and applied modelling studies...........................................................................v Site studies..........................................................................................................................vi

Key recommendations.................................................................................................................. vii Numerical modelling .......................................................................................................... vii Water monitoring................................................................................................................ vii Water management............................................................................................................ vii Further investigation .......................................................................................................... vii

1. Introduction.................................................................................................................................... 1

2. Conceptual and numerical modelling ............................................................................................ 3 2.1 Background and purpose.................................................................................................... 3 2.2 Disconnection: terminology, physics and implications........................................................ 3

2.2.1 Terminology ............................................................................................................ 3 2.2.2 Physics.................................................................................................................... 4 2.2.3 Implications for modelling losing-disconnected systems........................................ 6

2.3 Limitations of numerical models.......................................................................................... 7 2.4 Model calibration................................................................................................................. 8 2.5 Field data ............................................................................................................................ 9 2.6 Recommendations ............................................................................................................ 10

3. Establishment of field study sites ................................................................................................ 12 3.1 Regional setting and site selection ................................................................................... 12

3.1.1 Reach selection .................................................................................................... 12 3.1.2 Piezometer transect location ................................................................................ 13

3.2 Drilling location and piezometer design ............................................................................ 14 3.3 Hydrometric monitoring..................................................................................................... 14

4. Determination of connection status and infiltration rates ............................................................ 16 4.1 Background and purpose.................................................................................................. 16 4.2 Site reports........................................................................................................................ 16 4.3 Connection status ............................................................................................................. 16 4.4 Streambed hydraulic conductivity ..................................................................................... 18 4.5 Infiltration through the streambed ..................................................................................... 19 4.6 Temporal hydraulic head response................................................................................... 21 4.7 Environmental tracers ....................................................................................................... 22

4.7.1 Background.............................................................................................................. 22 4.7.2 Recharge source...................................................................................................... 23 4.7.3 Groundwater dating ................................................................................................. 24

4.8 River gauging .................................................................................................................... 25 4.8.1 Background.............................................................................................................. 25 4.8.2 Results ..................................................................................................................... 25

5. Geophysics.................................................................................................................................. 28 5.1 Reach-scale geophysics: waterborne resistivity imaging ................................................. 28 5.2 Site-scale geophysics ....................................................................................................... 29

5.2.1 Down-hole electromagnetic induction and gamma logging.................................. 29 5.2.2 Cross-sectional resistivity imaging ....................................................................... 29

i | NSW Office of Water, August 2011


6. Applied modelling research......................................................................................................... 32 6.1 Background and Purpose ................................................................................................. 32 6.2 Stream–aquifer interaction sensitivities by modelling scale ............................................. 32

6.2.1 Spatial scale............................................................................................................. 32 6.2.2 Temporal scale ........................................................................................................ 33

6.3 River site models............................................................................................................... 34 6.3.1 Regional models—river loss calculations and connection status ............................ 34

6.4 Local (reach-scale) models—river loss calculations and connection status .................... 35 6.4.1 Model conversion..................................................................................................... 35 6.4.2 Results and comparison with regional models ........................................................ 38

6.5 Findings............................................................................................................................. 39

7. Discussion ................................................................................................................................... 40 7.1 Disconnection status......................................................................................................... 40 7.2 Disconnection potential..................................................................................................... 41 7.3 Infiltration rates.................................................................................................................. 41

8. Conclusions................................................................................................................................. 43 8.1 Recommendations ............................................................................................................ 43 8.2 Future studies ................................................................................................................... 43

9. References .................................................................................................................................. 45

Appendix A: Compendium of project publications............................................................................... 48

Tables

Table 1 Assessment of connection status at the 12 riparian transects in 2009 using bank tests, streambed fluid pressure measurements, and depth of the water table in the riparian zone relative to streambed elevation. Bank tests (holes dug along

the banks) were first made to preliminarily classify each site (wet holes = connected, dry holes = transitional or disconnected) and guide the choice of streambed fluid pressure (ψ) measurement technique. The detailed procedures

are described in Lamontagne et al. (2010, 2011a, c). .................................................... 18

Table 2 Mean and 95 per cent confidence intervals for the streambed infiltration rate measurements at losing-connected sites, expressed as the specific infiltration

rate (q, m day ) and the lineal infiltration rate (Q , ML km day ). A negative infiltration means groundwater discharge to the river (because q = –K i). A conservative approach was used to estimate the confidence intervals

(Lamontagne et al. 2011c, e, f). ...................................................................................... 21

–1L

–1 –1

v

Table 3 Differential gauging results from site surveys. The results for each river are shown in an upstream to downstream sequence. .......................................................... 26

Table 4 Summary of river losses (kL day km ) and connection status (% of flow duration) — regional models........................................................................................... 35

–1 –1

Table 5 Summary and comparison of river losses (kL day km )—local models...................... 38 –1 –1

Table 6 Comparison of the connection status for the study reaches determined by different approaches. ...................................................................................................... 40

ii | NSW Office of Water, August 2011


iii | NSW Office of Water, August 2011

Figures

Figure 1 The potential for a system to become disconnected is a function of stream depth (d), the thickness of the streambed clogging layer (h ), and the ratio of

the hydraulic conductivity of the clogging layer (K ) to that of the aquifer (KDisconnection occurs where the factors limiting the feed of water from the river into the underlying strata (shallow river, hence low driving head) and thick or

poorly permeable clogging layer prevent the replacement of water draining away from beneath the river. An unsaturated (or vadose) zone is then created. Where the water table occurs a small distance below the clogging layer, the

system is in transition, and changes in groundwater table depth will still influence the infiltration rate. Where the water table is deeper, the system becomes disconnected, and the infiltration rate is no longer affected by further

lowering of the water table. ............................................................................................... 6

c

c a).

Figure 2 Locations of the river reaches selected for the losing streams project........................... 13

Figure 3 Conceptual representation of the different scales of investigation, using the

Lachlan River as an example. The study reach for each river (bottom left) was usually 5–30 km in length. Each reach was instrumented with two piezometer transects (Hillston Bridge and Gonowlia Weir on the Lachlan). Each piezometer

transect (bottom right) consisted of three piezometer nests installed perpendicular to the river. The section of the river in front of the piezometers was used for the streambed-based measurements........................................................ 17

Figure Summary of vertical hydraulic conductivity measurements made using falling-head tests at the study sites. Additional measurements were also made using constant-head tests at the Macquarie, Namoi and Dumaresq sites. The falling-

head test has a lower detection limit of 10 m s . At Billabong Creek, more accurate estimates of the hydraulic conductivity of the clays were also made by using evaporation experiments with intact streambed sediment cores

(Lamontagne et al. 2011a).............................................................................................. 19

–6 –1

Figure 5 Variations in vertical hydraulic gradient and hydraulic conductivity (from constant-head tests—K ) at Dumaresq River site 2 (Border Rivers). Negative

gradients indicate losing conditions. Stream flow was gradually receding during the measurement period. Horizontal hydraulic gradients between bank pits and the river also shown (negative values again indicate losing conditions). ....................... 20

CH

Figure 6 Changes in hydraulic head over time in the piezometer nest closest to the river at (A) a losing-connected site (Macquarie River at Macs Reserve) and (B) a losing-disconnected site (Gwydir River at Brageen Crossing). Green line,

shallow piezometer; blue line, deep piezometer. Note the flow reversal event in the Macquarie River........................................................................................................ 22

Figure 7 Patterns in the stable isotopes of water at the Namoi transects. (A) Change in

deuterium ( H) concentration with distance from the river and (B) H– O plot showing the evaporation line for the samples (dashed). ‘Pore water’ is groundwater from the streambed. The local meteoric water line (LMWL; solid)

represents the expected range in isotopic signature for rainfall in a region. Isotopic fractionation during evaporation tends to push isotopic signatures to the right of the LMWL. In the case of the Namoi River sites, all samples plot on the

same evaporation line suggesting that they originated from one source. Open symbols – Yarral East site; Closed symbols – Old Mollee site....................................... 24

2 2 18


Figure 8 Rn-inferred ages in streambed profiles on the Macquarie River. Specific infiltration rates (q) are estimated by assuming a porosity of 0.3. .................................. 25 222

Figure 9 River gauging sites on the Macquarie River, 23–24 September 2009. .......................... 27

Figure 10 Hydrograph of river discharge (ML day ) on the Macquarie River at Baroona (site 421127). The shaded area represents the differential gauging measurement period at the study reach. ........................................................................ 27

–1

Figure 11 Cross-sectional resistivity imaging with overlain down-hole resistivity for the Namoi River upstream section........................................................................................ 31

Figure 12 Cross-sectional resistivity imaging with overlain down-hole resistivity for Namoi

River downstream section............................................................................................... 31

Figure 13 Temporal variations in stream leakage estimated by regional (parent) model and sub-model (TMR). .................................................................................................... 33

Figure 14 Macquarie River losses estimated from groundwater models at daily, weekly and monthly time scales. ................................................................................................ 34

Figure 15 The graph from Brunner et al. (2009a) shows that as river depth approaches 1

m (dotted line), the difference between flux estimates obtained by explicitly simulating unsaturated flow between the river bed and the water, and not, become negligible. .......................................................................................................... 36

Figure 16 MODFLOW model accuracy for vertical layer discretisation. The graph shows that as the normalised river leakage (q/K ) decreases, the differences between the calculated water table mount height beneath the river, estimated by

explicitly simulating the unsaturated flow between the river bed and the water, and not, become small.................................................................................................... 36

a

Figure 17 River height probability distribution for the Namoi River as an example. ....................... 37

Figure 18 River leakage probability distribution for the Namoi River as an example. .................... 37

Figure 19 Lineal river loss in the Macquarie River.......................................................................... 38

iv | NSW Office of Water, August 2011


Executive summary

A key challenge for the management of alluvial aquifers in the Murray–Darling Basin (MDB) is to

properly quantify surface water–groundwater exchanges. In particular, the contribution of losing-disconnected rivers to alluvial aquifer recharge is poorly understood. The water table in losing-disconnected rivers is well below the streambed, which is sometimes interpreted as meaning that no

exchange is taking place. This is not correct because infiltration will still occur through the unsaturated zone below the rivers.

The aim of this collaborative project was to re-evaluate the conceptual representation of losing rivers

and to develop practical guidelines and methodologies to apply this knowledge for the management of alluvial aquifers. Some of the specific objectives of the project were to:

• Define a new criterion to identify those rivers with a potential for disconnection.

• Develop guidelines for the use of the standard modelling software in the industry (MODFLOW) for

modelling losing streams.

• Revise and update six alluvial groundwater models used for the management of key NSW MDB alluvial aquifers located along Billabong Creek and the Lachlan, Macquarie, Namoi, Gwydir and

Dumaresq Rivers.

• Evaluate connectivity and infiltration rates along a reach of the six above-mentioned rivers using a

range of techniques, including riparian water table monitoring, geophysics, environmental tracers, and differential gauging. For this purpose, two riparian piezometer transects were installed at each

study reach.

Key results

Theoretical and applied modelling studies

The physical processes controlling infiltration in losing rivers were reviewed and the exchange dynamics were evaluated for simple conceptual representations of alluvial aquifers. Key outcomes

included the following:

• It is incorrect to equate disconnection with the absence of exchange between rivers and aquifers

because infiltration rates are maximal once disconnection occurs.

• Once disconnection occurs at a certain point in the stream, the infiltration flux will not change with

further lowering of the water table. However, the reach-scale infiltration flux could still increase, because a greater proportion of the reach could become disconnected with further lowering of the

water table.

• Not all streams can become disconnected. For disconnection to occur, streambeds must have a

clogging layer with a significantly lower hydraulic conductivity than the underlying aquifer.

• A new criterion was developed to quantify the potential for disconnection. The potential for a system to become disconnected is a function of stream depth (d), the thickness of the streambed

clogging layer (hc), and the ratio of the hydraulic conductivity of the clogging layer (Kc) to that of the aquifer (Ka).

• Guidelines were developed for the appropriate use of the MODFLOW modelling platform under losing-disconnection.

The guidelines were used in the revision of existing MODFLOW groundwater models for the six alluvial aquifers.

v | NSW Office of Water, August 2011


Site studies

Connection status

Water table monitoring in the riparian zone and fluid pressure measurements in the streambed

identified three losing-disconnected reaches (Billabong, Lachlan and Gwydir), two losing-connected reaches (Macquarie and Namoi) and one connected reach with both losing and gaining sections (Dumaresq). All disconnected river reaches had a well-defined clogging layer near or slightly below

the streambed, usually a clay unit 0.5 to 2 m thick. The resistivity-based geophysical techniques used were generally not successful in mapping the position of the water table near streams, probably because they also responded to variations in clay content and salinity in the subsurface. Nevertheless,

with further developments, geophysical techniques have a significant potential to evaluate connectivity at the reach scale.

Streambed hydraulic conductivity

There are few measurements of streambed hydraulic conductivity in the Murray-Darling Basin, one of the key parameters defining surface water–groundwater exchanges. A large range in streambed hydraulic conductivity was measured among and within river reaches using field permeametry.

However, as a first approximation, at the site level streambeds could be characterised either as having a low (<10–6 m s–1) or a high hydraulic conductivity (10–5 to 10–3 m s–1). Essentially, streambeds had a low hydraulic conductivity when lined with clays. This suggests that streambed hydraulic conductivity

in regional groundwater models in the MDB (which is often assumed to be constant) could be constrained by classifying river sections into low and high hydraulic conductivity categories from field surveys. However, further measurements are required to define a representative range in hydraulic

conductivity for clay-lined streambeds.

Infiltration

Infiltration rates at the study reaches were evaluated by:

• sampling groundwater and surface water at the riparian piezometer transects for a range of

environmental tracers, including major ions, stable isotopes of water, chlorofluorocarbons, sulfur hexafluoride and helium-4

• measuring the Darcy flux through the streambed by piezomanometry

• differential gauging.

Chloride concentrations and the stable isotopes of water indicated that significant recharge occurred along most reaches under both high and low flow conditions. Another tracer (helium-4) identified a regional baseflow component for one study reach (Dumaresq), possibly due to leakage from the

underlying Great Artesian Basin. Infiltration rates could not be estimated by groundwater dating with chlorofluorocarbons (CFCs) due to unsuitable geochemical conditions in the aquifers. However, sulfur hexafluroride (SF6), which can date groundwater over a similar age range as CFCs (1970s to the

present), appeared a suitable alternative. Overall, with some refinements, environmental tracers can be used to identify infiltration sources and rates in losing MDB rivers.

Infiltration rates estimated by the Darcy flux and differential gauging techniques were considered to

have an overall limited accuracy because of unsteady river levels prior to and during the measurement periods. Most MDB rivers have continuously fluctuating water levels, in part because pulses of water are regularly released from upstream reservoirs for irrigation or environmental needs downstream.

Where possible, river operation protocols should be developed to enable better field-based infiltration rates in the future.

vi | NSW Office of Water, August 2011


vii | NSW Office of Water, August 2011

Key recommendations

Numerical modelling

1) Carefully consider the choice of numerical model in every case so as to avoid the construction of models using codes that cannot simulate the relevant processes. Situations in surface

water–groundwater interactions have been identified where MODFLOW is inappropriate.

2) Develop a standardised protocol which ensures that the principles of model calibration are adhered to, and which can adequately assess and demonstrate model robustness and

reliability. Infiltration flux (river leakage) data is essential for model calibration where the surface water–groundwater exchange process is being simulated.

3) Revise the MDB’s Groundwater Flow Modelling Guidelines to provide a clear definition of the

term ‘disconnection’ (see Recommendation 5) and to facilitate the implementation of Recommendation 2.

Water monitoring

4) Increase investment in monitoring infrastructure in order to collect more field data pertinent to

groundwater interaction and connectivity, in particular the estimation of the river–groundwater exchange flux. Pair riparian piezometer transects with river gauging stations and instrument them to measure river and bore water levels continuously. Design observation bores to allow

sampling for environmental tracers for recharge estimation.

Water management

5) The water management community continue to use the term ‘disconnected’, however an effort should be made to explain the term correctly, completely and unambiguously and negate any

inference that it conceptually implies no exchange between aquifers and streams.

6) Water sharing plans (WSPs) for both surface water and groundwater should consider and, if required, include management rules which take into account the processes and size of the

estimated flux exchange between surface water and groundwater.

Further investigation

7) Develop guidelines for the design of coupled surface water–groundwater monitoring networks at the regional scale and invest in an expansion of those networks. Extend the river and

groundwater time-series monitoring already in place to capture a greater variety of hydrological events and to provide better river loss estimates.

8) Determine the connection status and infiltration rates at repeated locations (and at repeated

times) as a means to calibrate regional-scale assessments of surface water–groundwater interaction made by numerical models.

9) Use environmental tracer and age dating techniques (e.g. CFCs, SF6, stable isotopes of

water) to improve the conceptualisation and measurement of exchanges between surface water and groundwater.

10) Develop methods to measure infiltration in the field at the point scale and reach scale and

develop a framework to scale these estimates to the regional scale.

11) Continue with the development of geophysical techniques as a means to scale up other point-scale and reach-scale measurements, such as streambed clogging layer thickness and depth

to the water table near or below streams.


1. Introduction

The Commonwealth Government, through the National Water Commission, made available a

$1.4 million grant to investigate the process of water loss from rivers into groundwater under ‘losing-disconnected’ conditions in the NSW portion of the Murray–Darling Basin (MDB). The NSW Office of Water contributed a further $840,000 in project management and field support and led a large multi-

team consortium, including chiefly the CSIRO and Flinders University, to investigate the condition.

Losing-disconnected is the condition where a zone of unsaturation exists between the bed of a river and the groundwater table immediately beneath it. It occurs where certain hydraulic conditions

combine to restrict the movement of water down from the river into the saturation zone of an aquifer.

It is an important hydraulic process to understand, as the dynamics and rate of water loss from rivers through unsaturated strata are different from those through a fully saturated column, and the transient

flux exchange is affected differently between the two situations by changes to the natural hydrological regime (by groundwater pumping, for instance).

This difference has implications for determining valley-scale water balances that are estimated by

analytical techniques and numerical models. Sound water management requires a good understanding of the river–groundwater flux exchange. Thus, learning more about what happens under losing-disconnected conditions and applying it in numerical models will help improve that

component of water balance estimation.

This technical final report presents a complete synopsis of the technical and scientific activities of this project, and draws conclusions and recommendations for water management from the findings. The

document is an abridgement of many component reports which have been produced as the project has progressed. It ties together these reports into a coherent standalone summary which explains the overall context of the work and the key overall findings. A compendium of the reports produced for the

project is listed in Appendix A.

The National Water Commission and the NSW Office of Water are particularly interested in understanding the differences in the hydrodynamics of losing-connected versus losing-disconnected

conditions now because of:

• growing demand on water managers to better quantify the water exchange between rivers and

groundwater, particularly as the reliability of surface water supplies decreases

• knowledge that the hydrodynamics differs between losing-connected and losing-disconnected

conditions and that our regional numerical models do not explicitly deal with the unsaturated zone (or have to make assumptions about it)

• knowledge that surface water models lump groundwater accessions into ‘unaccounted losses’

• the need to know whether the incomplete treatment of the unsaturated zone matters when we are

determining catchment water balances or the effect of different management options—and if so, to what extent

• the need to know what water monitoring data and information are required in order to allow the accurate estimation of river–aquifer water exchange.

This large, collaborative, multi-component project took place between May 2008 and December 2010. The following sections of this report describe each component of the project:

• Section 2: Investigate at a fundamental level the conceptual and numerical modelling of the disconnected condition. Examine in detail the effect of the assumptions MODFLOW makes about

the unsaturated zone and more generally the suitability of our regional MODFLOW models where disconnection exists. (Led by Flinders University.)

1 | NSW Office of Water, August 2011


• Section 3: Establish six field investigation sites on NSW MDB rivers where losing-disconnected

conditions are suspected (Dumaresq, Gwydir, Namoi, Macquarie and Lachlan rivers and Billabong Creek). This includes the establishment of bores and instrumentation.

• Section 4: Undertake field studies at the six sites to determine the connection status of the rivers

and infiltration rates by using a range of techniques. (Led by CSIRO.)

• Section 5: Undertake hydrogeophysical investigations, including waterborne and cross-sectional

resistivity imaging and down-hole electromagnetic induction logging, in order to scale up the findings of the site studies.

• Sections 6 and 7: Take the lessons learned from Flinders University and information from the field

studies to modify existing regional-scale models to a local scale, and compare the results using the different scales.

• Sections 8 and 9: Use the technical findings to explore implications for water management and monitoring and to formulate recommendations.

Comprehensive details of the major components of the project (Sections 2, 4 and 6) are reported in a series of published and unpublished papers and reports referenced throughout the text and listed

separately in Appendix A.



2. Conceptual and numerical modelling

2.1 Background and purpose

Flinders University was asked to investigate the conceptual and numerical modelling of the disconnected condition at a fundamental level. The purpose was to test the assumptions made by

MODFLOW about the unsaturated zone, and the suitability of regional MODFLOW models for estimating the river–aquifer water exchange where disconnection exists.

MODFLOW is the most common numerical model used to analyse basin-scale groundwater

hydrodynamics. Such models are used to explore the natural hydrodynamics of a system and the impact of water use and differing management options on the system’s behaviour. MODFLOW calculates water balances, including the quantum of the flux exchange between streams and

groundwater, thus allowing an examination of the relative impact of groundwater use on stream flow.

Several conceptual assumptions and approximations relating to unsaturated flow apply to the MODFLOW streamflow packages, and these have been assumed to be generally justified for a range

of applications. But it is important to know whether these assumptions matter when MODFLOW is used where disconnected conditions are found. In other words, how reliable are estimates of the regional water balance and river flux exchange where there are disconnected conditions?

MODFLOW and its streamflow packages were tested against a fully coupled model called Hydrogeosphere under a generic representation of streams and alluvial aquifers.

This section summarises the findings of these investigations and offers recommendations aimed at

developing more reliable numerical models. Critical issues were identified in the following areas:

1. Terminology: The term ‘disconnected’ is frequently misunderstood. This misunderstanding has far-reaching consequences ranging from the set-up of

numerical models to the development of water management policies.

2. Limitations of numerical models: The numerical model must be able to handle the physical processes that dominate the physical system under consideration. Model

choice is therefore a critical factor which deserves more attention than it has received to date in the general area of surface water–groundwater interaction.

3. Model calibration: Critical issues have been identified in the way models are

calibrated and how their reliability is demonstrated. In particular, model calibration does not guarantee that a model is reliable. By introducing more calibration parameters than available observations, even an incorrect model can be made to

reproduce observations.

4. Field data: No trustworthy model that claims to robustly and accurately simulate the interaction between surface water and groundwater can be constructed in the

absence of field data. Importantly, we determine the nature of the field data required to achieve appropriate model calibration with the specific questions of surface water–groundwater interaction and exchange fluxes in mind.

2.2 Disconnection: terminology, physics and implications

2.2.1 Terminology

The term ‘disconnected’ can be and often is misunderstood (Brunner et al. 2010b). A disconnected system is not a system where there is no exchange between the surface and the subsurface. In a

disconnected system, a river loses water to the groundwater. In fact, the infiltration rates of a



disconnected system are higher than those under a connected flow regime. The surface water body and the groundwater are disconnected only in the sense that changes in groundwater do not affect the infiltration rate.

However, the term ‘disconnected’ is the commonly used and accepted term in the literature, and the community either must clearly explain and define its use much more accurately and precisely than it has done to date, or seek an alternative, far superior term before attempting to replace it in the

literature. The latter is challenging, and a new superior term cannot be found at this time. Because of the confusion in the terminology, the Murray-Darling Basin Sustainable Yields Project used the term “maximum losing’ instead of disconnected (Rassam et al. 2008). However, we recommend that the

community continue to use the term, but make much more effort to explain it correctly, completely and unambiguously. The National Groundwater Assessment Initiative offers a unique opportunity to standardise the definition. Within the groundwater action plan (http://www.nwc.gov.au/www/html/350-

groundwater-action-plan.asp), a harmonisation of groundwater definitions and standards project is being conducted. On account of the water resource management implications of the term, a clear and unambiguous definition is required urgently. We would like to see the harmonisation project include

clear definitions of the key terms related to connectivity and disconnection.

Two important aspects need to be highlighted, as they have direct implications for our understanding of disconnected systems and in the set-up of numerical models aimed at simulating these phenomena:

• Lowering the water table under a disconnected river reach will not increase the infiltration rate at this specific point in space. However, most rivers are not disconnected along their entire length.

Rather, rivers often alter the flow regime between gaining and losing and losing-disconnected. In such systems, extracting groundwater under a disconnected reach will often extend the length of the river that is disconnected. Even though locally the infiltration rate is not increased, the river’s

water balance can still significantly be affected.

• Changes in the surface water body will change the infiltration rate of a disconnected system.

Changes in water depth or the width of the river affect the infiltration rate even if the river is disconnected.

There is often an implied and yet completely incorrect understanding or message communicated in documentation that a disconnected system means that the surface water and groundwater can be

managed as separate, non-interacting water resources. We cannot state more forcefully here how problematic this is, and that this misunderstanding requires urgent attention. It influences technical matters from approaches to modelling and field-based efforts required to elucidate losing and losing-

disconnected states, through to management decisions and policy.

To clarify the terminology and to explain the impact of disconnection on surface water management, some fundamental physical aspects of disconnection are discussed here in more detail.

2.2.2 Physics

Brunner et al. (2009a, b) describe the physics of disconnection. In summary:

1) A disconnected system can be conceptualised in terms of the build-up of a groundwater mound under a river and the development of a capillary zone above it. In transition and

disconnected systems, an unsaturated zone develops under a clogging layer, resulting in ‘suction’ (negative pressure) at the interface between the clogging layer and the aquifer. A clogging layer is a part of the streambed with a lower hydraulic conductivity than the

underlying aquifer. The origin of the clogging layer can be physical (accumulation of silt and clay-size particles), chemical (cementing of sand grains by carbonate mineral formation, etc) or biological (biofilms, organic detritus, etc.) If the water table is far enough below the

clogging layer, the suction under the clogging layer is essentially independent of changes in


http://www.nwc.gov.au/www/html/350-groundwater-action-plan.asp

http://www.nwc.gov.au/www/html/350-groundwater-action-plan.asp


the groundwater table, and hence the system is disconnected. For a complete understanding of the disconnection of surface water and groundwater, processes in both the unsaturated (vadose) and saturated (phreatic) zones have to be understood and quantified.

2) A clogging streambed is a precondition for disconnection: the hydraulic conductivity of the streambed must be smaller than that of the aquifer. Brunner et al. (2009a) developed an exact mathematical criterion that identifies whether a stream can disconnect or not (explained in

Figure 1):

Kc

Ka≤

hc

d + hc

This criterion helps one to choose a correct conceptual model. If a system cannot disconnect, a model without the capability of simulating unsaturated flow is sufficient.

3) The infiltration rate of a losing stream is given by the hydraulic gradient through the clogging

layer. The rate depends on the soil suction under the streambed and on the depth of the surface water body. Even though in a disconnected system the infiltration rate is not affected by changes in the groundwater table, it is affected by changes in the surface water body.

4) In wide rivers and lakes, a significant drop in the water table may be required to change the flow regime from connected to disconnected. However, transitional stages between connected and disconnected flow regimes exist, and they play an important role in surface

water-groundwater interaction. This situation challenges the common assumption that a system is either connected or disconnected.

5) The spatial distribution of seepage through a surface water body critically depends on the

state of connection. In a disconnected system, there is no spatial variation in infiltration (provided that Kc and river height are spatially constant). In connected and transitional systems, however, there is spatial variation in infiltration across the surface water body.

6) Lakes and wetlands are more likely to disconnect than rivers because of differences in the build-up of a groundwater mound in 2D and 3D. Therefore, lakes are more sensitive to changes in the regional groundwater table. This is of significance in assessments of the

impact of a changing groundwater table on surface water bodies.

7) The state of connection is a critical variable in dynamic systems. In a disconnected system, changes in the surface water body result almost immediately in a new steady infiltration rate.

In connected systems, in contrast, the new steady state is approached gradually. In connected systems, a reduction in surface water depth results initially in an infiltration rate below the steady-state solution, while an increase results in an infiltration flux above the steady state

solution. If the regional groundwater table changes rapidly, the assessment of the state of disconnection on the basis of a borehole reading can be biased.



Figure 1 The potential for a system to become disconnected is a function of stream depth (d), the thickness of the streambed clogging layer (hc), and the ratio of the hydraulic conductivity of the clogging layer (Kc) to that of the aquifer (Ka). Disconnection occurs where the factors limiting the feed of water from the river into the underlying strata (shallow river, hence low driving head) and thick or poorly permeable clogging layer prevent the replacement of water draining away from beneath the river. An unsaturated (or vadose) zone is then created. Where the water table occurs a small distance below the clogging layer, the system is in transition, and changes in groundwater table depth will still influence the infiltration rate. Where the water table is deeper, the system becomes disconnected, and the infiltration rate is no longer affected by further lowering of the water table.

2.2.3 Implications for modelling losing-disconnected systems

The facts that the exchange between surface water and groundwater in disconnected systems is not

zero, the flow regimes can change along a river, and the infiltration rates can change even in disconnected reaches have two important implications for setting up numerical models of disconnected rivers:

1) Because the state of connection can change along a river, a-priori knowledge of the state of connection is crucial to setting up a numerical model. Knowing the state of connection is as important as other fundamental characteristics that are routinely considered in setting up a

model, such as whether the aquifer is confined or unconfined.

2) Estimates of the rate of infiltration from surface water to groundwater are essential to model calibration and are a critical part of the modelling process, no matter whether the flow regime

is gaining, losing or losing-disconnected.



2.3 Limitations of numerical models

A variety of numerical models with different degrees of complexity are available to simulate surface water–groundwater interaction. The US Geological Service’s MODFLOW is one of the most commonly

used codes for groundwater modelling in Australia and internationally. MODFLOW is a fast and easy-to-use numerical model but is based on conceptual assumptions that can result in critical limitations, especially in the simulation of surface water–groundwater interaction, as we discuss below.

Nevertheless, MODFLOW is used routinely, often without a critical assessment of whether the code is able to simulate the dominant processes or not.

Brunner et al. (2010a) systematically analysed the conceptual assumptions of MODFLOW relevant to

the simulation of surface water–groundwater interaction in the case of losing and losing-disconnected systems, and identified situations where MODFLOW is not the appropriate choice:

1) MODFLOW does not simulate the unsaturated zone. In disconnected systems, ‘suction’

occurs at the bottom of the streambed. Therefore, MODFLOW underestimates infiltration fluxes for disconnected rivers. Neglecting the unsaturated zone can have a significant impact in the simulation of shallow, disconnected streams because in such systems the exchange

between the surface and the subsurface is largely controlled by processes of the unsaturated zone and the application of MODFLOW to simulate such a stream is problematic. In the range of the simulations tested in Brunner et al. (2010a), the increase of flux as a result of suction

was small (5 per cent or less) if the sum of the river depth and the thickness of the clogging layer was greater than 1 m.

2) A mismatch between river width and the grid cell size results in an error in the water table,

because the model distributes the exchange rates between surface water and groundwater over the area of the grid cell. In reality, however, the water table depends on the distribution of infiltration fluxes across the river and should not be related to the size of the grid cell. The

finite difference conceptualisation of MODFLOW poses severe limitations on varying the resolution within a model. Even though techniques such as local grid refinement are available, the more general approach of finite elements used in all advanced numerical models allows

greater flexibility. Therefore, modelling large catchments with small to medium-sized rivers in MODFLOW can be very challenging. In the model tested in Brunner et al. (2010a), changes in grid cell size resulted in differences of over 2 m of head under the river.

3) Because MODFLOW does not simulate the unsaturated zone, grid cells can fall dry if the groundwater table falls below the bottom elevation of the model layer. In this case, the cells are deactivated. The methods available in MODFLOW to reactivate dry cells cause numerical

problems and are not suitable for most applications. To avoid dry model cells in MODFLOW, a coarse vertical discretisation of the aquifer is often used. It is therefore assumed that within a grid cell the hydraulic head does not vary vertically. However, under an infiltrating river this

might not be the case. If the vertical flow component is significant, it must be calculated using a fine vertical grid; in these cases MODFLOW is not a good choice, owing to the unresolved conceptual issues of dry cells. For the model simulated in Brunner et al. (2010a), the

differences in head under the river were up 4 m between different vertical discretisations.

To avoid the construction of numerical models using codes that cannot simulate the relevant processes, we recommend that the choice of a numerical model be considered carefully for every

single catchment. A standardised procedure for model choice should be listed in widely accepted modelling guidelines such as the Groundwater Flow Modelling Guidelines (Middlemis et al. 2000), issued by the Murray–Darling Basin Commission. This point is absolutely essential. Such a procedure

should provide the information required to identify in which cases a simple model such as MODFLOW is acceptable as outlined above. The findings and criteria developed by Brunner et al. (2009c), but



also those noted in Brunner et al. (2009a, b), should form a basis for formulating model selection criteria for losing and losing-disconnected streams. We recommend that the commonly used guidelines by Middlemis et al. (2000) should be revised according to the key findings from this work.

Community, stakeholder and legal issues arise here because MODFLOW is considered the standard accepted code. Serious discussion is required to ensure that model popularity not override technical considerations in choosing models. This is not trivial.

2.4 Model calibration

How reliable a numerical model is depends on how it was calibrated. This section outlines critical issues related to the calibration and verification of numerical models, describes some basic principles

on how to calibrate a model, and documents key indicators of model reliability.

When quantifying surface water–groundwater exchange, model non-uniqueness is a critical problem when models are only calibrated against aquifer head data. Without observations of infiltration along

the river, there may be an infinite number of hydrogeologic configurations (even ignoring the recharge component from diffuse recharge and evapotranspiration away from the river) that lead to an apparent match with head data in the aquifer. Without infiltration rate data along the river, the hydraulic

conductivity of the riverbed and that of the aquifer can be simultaneously calibrated to achieve a match to borehole data. Thus, the rates of exchange between surface water and groundwater are critical data components required to adequately constrain the model calibration. Such data must be

included routinely in the model calibration process if the surface water–groundwater interaction drives the focus of the model, or else the model is not accurate for making predictions in this area because there is no basis upon which to conclude that it is either reasonably calibrated or making reliable

predictions. Clearly, water management agencies will require funding and resources to implement this recommendation. There may also be a time lag in incorporating such data in decision making.

The reliability of a calibrated model is routinely demonstrated by a high correlation between calculated

and observed parameters that were used for calibration. That such a fit indicates a well calibrated model is also proposed by Middlemis et al. (2000). However, a good fit is a necessary but not sufficient condition for model reliability. For all of the abovementioned reasons, revised calibration

guidelines should be developed and used. Such guidelines should include the following aspects:

1) General principles of calibration.

a. A model calibration can only be unique when the number of calibration parameters is

less than the number of observations. Therefore, the number of calibration parameters must be smaller than the number of observations if its predictions are to be reliable. However, this is never the case with practical numerical models.

Consequently, predictions will have a degree of uncertainty that must be evaluated.

b. Highly correlated parameters such as recharge and hydraulic conductivity or streambed conductance and hydraulic conductivity of the aquifer should not be

calibrated jointly.

c. A model calibrated in the absence of observations of a particular physical process (e.g. stream–aquifer interaction) cannot be regarded as reliable without verification

against that process. Consequently, the sensitivity of predictions must be assessed.

d. Advanced calibration techniques allow calibration of a large number of parameters simultaneously. A model’s predictive uncertainty can be quantified on the basis of

parameter uncertainty calculated during calibration. These techniques allow insight into model uncertainty, even if the models were calibrated with a large number of



parameters. It is important to note, however, that if models are calibrated with a large number of parameters, a full analysis of their predictive uncertainty is essential.

2) Scientifically based indicators that demonstrate model reliability.

a. Every model should be assessed by a comparison between observed and calculated data that were not used as part of the calibration. Only a model that adequately describes the relevant physical processes is able to predict system behaviour in

response to changing forcing functions.

b. The correlation between calibrated model parameters can be used to assess the uniqueness of the model. For example, if only head observations are used during

calibration, increasing both recharge and hydraulic conductivity in the same ratio will usually result in a similar fit to the data. In such cases, the correlation between the calibration parameters (recharge and hydraulic conductivity) is high. A high correlation

between calibrated parameters illustrates that the model is non-unique.

Including such principles in the calibration and modelling guidelines by Middlemis et al. (2000) would be of great use in improving modelling practice more generally. Rassam et al. (2008) discuss further

improvements to current modelling methods in the context of surface water–groundwater interactions in the MDB.

2.5 Field data

The state of connection, the infiltration rates, the transient behaviour of the river and observations of the water table are essential in order to understand and model surface water–groundwater interaction. When the key questions of surface water–groundwater connectivity, interaction, and exchange fluxes

are considered, it is important to site observation boreholes and piezometers close to the river, as the influence of other hydrogeologic factors such as heterogeneity, recharge and regional flow increases with increasing distance from the river.

Since the measurement of exchange fluxes between surface water and groundwater is a critical component, this can be done either by using hydraulic head measurements or through estimates of infiltration losses, for example by measuring the difference in flow rates between river gauging

stations. A recent review of methods to estimate infiltration rates is given by Kalbus et al. (2006). Infiltration rates calculated from the head gradient between surface water and adjacent piezometers must be based on measurements of the hydraulic properties of the aquifer as well as of the

streambed. Such hydraulic methods to estimate infiltration rates cannot rely on the calibration of streambed properties. It is important to keep in mind that measurements of streambed properties are point measurements and may not necessarily represent the properties of the stream.

For many catchments in Australia, field data relevant to surface water–groundwater interaction that allow the quantification of infiltration rates are absent or limited. Suggestions for improving available field data are:

1) Increase the density of stream monitoring stations. Variables measured must include discharge and river stage measured continuously. Ideally, the network must be dense enough to correctly estimate the infiltration between two monitoring stations. Just as important is

documenting the surface water extractions and considering evaporative losses along a reach.

2) A network of observation boreholes/piezometers along a stream is critical. Ideally, observation boreholes should be co-located at the river gauging stations so that river stage height can be

directly compared with the groundwater levels immediately adjacent to the river.



3) At gauging stations where observation boreholes are located close to the river, measurements of streambed and aquifer properties will be useful. Such measurements will allow the estimation of infiltration rates by both water balance and hydraulic methods.

4) These points suggest the need for greater groundwater bore/piezometer density, but also surface water gauge (volumetric flow or stage height) data—and importantly their co-location in space for measurements to be meaningful.

Estimating infiltration rates from measurements of discharge between two stations can be challenging and uncertain. For example, the uncertainty of discharge measurement at gauging stations during low-flow conditions can be of the magnitude of the infiltration rates. It is therefore strongly recommended

not to rely solely on one particular method to estimate infiltration. A comparison of infiltration rates obtained using different methods will provide insight into how reliable and consistent the data are and will allow for comparisons to be made at different scales of measurement.

2.6 Recommendations

Actions are recommended in the following areas:

• Terminology: Considering the uncertainty of the meaning of ‘disconnection’, a clear definition and

description must be included not only in the modelling guidelines, but also in the formal regulations relevant to water management practices. Most obviously, inclusion of this term and a clear

understanding of what it means in the National Groundwater Assessment Initiative’s harmonisation of groundwater definitions and standards project would be beneficial.

• Choice of numerical model: MODFLOW is often used without assessment of whether it can handle the critical processes or not. We have identified situations relevant to surface water–

groundwater interaction in which MODFLOW is inappropriate. To avoid the construction of numerical models that cannot simulate the relevant processes, we recommend that the choice of numerical model be considered carefully for every catchment. A standardised procedure allowing

selection of the appropriate model should be listed in widely accepted modelling guidelines.

• Calibration of numerical models: The two critical issues related to model calibration are (1)

Principles of model calibration are often ignored, resulting in highly non-unique models and (2) The reliability of calibrated models is not demonstrated adequately, resulting in unjustified

confidence in model robustness and reliability. A standardised protocol listing the principles of model calibration and measures to assess and demonstrate the reliability of the calibrated models should be included in widely accepted modelling guidelines. The use of infiltration flux data in

model calibration is critical—especially when the model is being used for simulation of surface water–groundwater interaction and exchange processes.

• A revision of the Groundwater Flow Modelling Guidelines to include:

o a clear definition of the term ‘disconnection’

o a step-by-step method to assess whether MODFLOW is appropriate or not

o a revision of the indicators of model reliability and uniqueness.

• Field data: The amount of available field data relevant to surface water–groundwater interaction are not sufficient to construct reliable numerical models. We recommend increased investment in

monitoring infrastructure in order to collect more field data that pertain to surface water–groundwater interaction and connectivity. An increase in the density of the surface water gauging stations and groundwater bores/piezometers is required in order:

o to quantify infiltration rates along rivers, ideally using more than one method

o to continuously record river stage information at monitoring stations



o to record groundwater head data close to the river; observation boreholes should be

located at the same locations as the river gauging stations so that direct comparisons can be made, and exchange estimates can therefore be reliably made.



3. Establishment of field study sites

As discussed in Section 2, it is important to calibrate groundwater models in losing systems by

determining the nature of the connection and the infiltration rates along rivers. However, there are currently few independent measurements of the connection status or infiltration rate along losing rivers in the MDB. Assessments of connectivity at the catchment and regional scales have been made

(Parsons et al. 2008; Ivkovic 2009), but these are not foolproof, especially in their capacity to differentiate between losing-connected and losing-disconnected conditions. There are few estimates of infiltration rate from losing streams in the MDB other than those obtained through the calibration of

groundwater models.

This section describes the establishment of field study sites set up to determine the nature of the hydraulic connection between river and aquifer and to estimate infiltration rates. It documents site

selection on river reaches, drilling, piezometer design and construction, and the establishment of instrumentation. Section 4 goes on to describe the field study methodology and results.

3.1 Regional setting and site selection

The region chosen for the study lies within the NSW portion of the MDB where the inland, generally westerly flowing rivers move out onto the basin plain, in an area underlain by large productive alluvial

fan aquifers. Locations were sought where the water table is below river levels to give losing river conditions. Such locations are found in a generally arc-shaped region which trends parallel to the eastern edge of the basin, some 100 to 500 km to the west.

Average annual rainfall ranges from 300 to 600 mm, decreasing progressively towards the west, and with a general summer bias to the north and a winter bias to the south. The main rivers are regulated mainly for irrigation supply and town water and experience highly altered flow regimes. The rivers

typically have low energy and meander across a very low-sloping landscape. They are frequently interrupted by low weirs which create backwatered or very slow flow conditions.

3.1.1 Reach selection

The criteria for the selection of the river reaches to study were:

• an overall mix of both losing-connected and losing-disconnected systems

• a geographic spread across NSW

• access to a regional surface and groundwater monitoring network

• water sharing plans in place or under development to allow findings to be integrated into water management

• existing calibrated groundwater flow models

• a high level of groundwater usage compared with entitlements

• a reasonably accessible 5 to 30 km stretch of river with several access points and potential

locations for the installation of piezometer transects

• availability of existing infrastructure, such as river gauges.

Figure 2 shows the locations chosen on Billabong Creek and the Lachlan, Macquarie, Namoi, Gwydir

and Dumaresq (otherwise known as Border) rivers. All criteria were met in all reaches except on the Border Rivers, where the regional groundwater model was known to be poorly calibrated and groundwater usage was relatively low. A mix of both losing-connected and losing-disconnected

conditions was expected to be found on the Namoi and Lachlan reaches, disconnected conditions only on Billabong Creek, and losing-connected conditions only on the Macquarie, Gwydir and Border rivers.



Figure 2 Locations of the river reaches selected for the losing streams project.

3.1.2 Piezometer transect location

Each study reach was instrumented with two riparian piezometer transects. A primary consideration in the selection of transect location was accessibility. Drilling sites needed to be easily accessible from public roads and, ideally, located on public land. However, permission was sought and granted to

install some transects on private property. Ideally, transects were located where an existing river gauging station was in place or where temporary river gauging infrastructure could be practically installed. Some transects were co-located with existing riparian piezometers to reduce cost.

As far as practicable, locations where high-yield groundwater extraction would artificially affect the groundwater system were also avoided, although in practice this was not always possible. River morphology also influenced the locations of the piezometer transects. Because of the meandering

nature of the rivers and its influence on groundwater flow paths, relatively straight river sections were selected as much as feasible.

All rivers are regulated and contain numerous weirs along their length. Although it was considered

acceptable to choose sites affected by weir pools, these were avoided as much as feasible.



3.2 Drilling location and piezometer design

The piezometer network had to fulfil several functions, including:

• to map the piezometric surface in the vicinity of the river, in particular to evaluate the likelihood

that the river was connected or disconnected

• to follow the response of the water table to changes in stream stage over time, such as bank recharge–discharge cycles during floods

• to be at the right spatial scale to enable the estimation of infiltration rates from groundwater-dating

environmental tracers such as CFCs, SF6 and 222Rn.

The regional hydrogeologist from the NSW Office of Water surveyed potential sites for the installation

of the riparian piezometer network. Final site selection and the design of the network at each transect were done in collaboration with CSIRO staff. The monitoring network had the following general design:

• Two piezometer transects were installed on each study reach. For practical reasons and to reduce

instrumentation cost, transects were located on one side of the river only.

• In addition to the river gauge, each transect consisted of three nested piezometers. Ideally, these were located at 30 (± 10), 70 (± 20) and 150 (± 50) m away from the river’s edge under low flow

conditions.

• Each nest consisted of one ‘shallow’ and one ‘deep’ piezometer. The shallow piezometer was

designed to have the top of the screen 3 to 5 m below the water table and the deeper one to be 10 to 12 m below the water table. The piezometers occasionally had to be inserted deeper than

planned when no water-bearing unit was encountered during drilling.

• The piezometers usually consisted of 100 mm diameter Class 12 PVC, with a 2 m slotted screen

and a 1 m sump. The screens were surrounded by a gravel pack (at least 25 cm above and below the screens) and isolated from the annulus with either bentonite or concrete. At the surface, all

piezometers were protected by a lockable steel casing secured into a concrete base.

• Following installation, each piezometer was usually purged for 1 or 2 days.

All piezometer installation was supervised by a NSW Office of Water hydrogeologist who also

collected a bore log during drilling. The construction details for the piezometer transects are described in the site reports (Lamontagne et al. 2011a–f).

3.3 Hydrometric monitoring

Hydrometric monitoring was provided by the Hydrometric Services Unit of the NSW Office of Water. An additional staff member was recruited to work on the project full-time for more than 12 months. The

hydrometric monitoring included:

• pressure transducers and data loggers in all piezometers to record water level and temperature

• one barometric pressure transducer per transect

• a staff gauge and a temporary hydrographic station (surface water level only) at each transect

when an established station was not already present

• measurement of the elevation of each casing and hydrographic station

• measurement of the surface elevation cross-section at each transect (including the riverbed)

• collection of time-series water level and temperature data for about 12 months at each transect

• management of all time-series data, including processing, quality coding and archiving into the

Office’s corporate hydrometric database

• dissemination of data as required to NSW Office of Water staff and research partners.



All project-related and ongoing hydrographic data can be obtained from the NSW Office of Water on DVD or on the website at www.water.nsw.gov.au


http://www.water.nsw.gov.au/realtime-data/default.aspx


4. Determination of connection status and infiltration rates

4.1 Background and purpose

In this component of the project, the connection status and infiltration rates for the six river reaches were determined. The detailed goals of the field studies were:

• to develop a method to identify losing-connected or -disconnected conditions in the field and

determine the connection status at the 12 riparian transects

• to estimate the hydraulic conductivity of the streambeds, including the presence of clogging layers

• to estimate the rate of infiltration through the streambeds by the Darcy flux method using piezomanometry to determine vertical hydraulic head gradients

• to compare the hydraulic head response of losing-connected and -disconnected sites with water

level variations in the rivers

• to estimate the source of recharge to the alluvial aquifers and the infiltration rates by sampling for

a suite of environmental tracers in the riparian piezometer network

• to estimate reach-scale infiltration by differential gauging.

The key component of the experimental design was the installation of piezometer transects near the rivers (Figure 3). As suggested in the modelling review, pairing of piezometers close to the river with stream gauges is the principal tool available to managers to evaluate the type of connection along a

river and the infiltration rates. The comparison with the streambed-scale and reach-scale measurements provided an opportunity to test how effective near-stream piezometers are at indicating the type of connection and the infiltration rates along rivers.

4.2 Site reports

A considerable amount of information was collected during the project, and only a brief overview is

provided here. The complete information has been summarised in Lamontagne et al. (2010, 2011a–f). In general, water levels in the rivers and piezometers were monitored for a year. All sites were also visited once by CSIRO staff during 2009:

• to assess the connection status

• to measure the infiltration rate through the streambed

• to sample for environmental tracers in the piezometer network

• to evaluate the reach-scale infiltration rate by differential gauging.

The NSW Office of Water’s hydrometric team did additional differential gauging surveys on most rivers.

4.3 Connection status

The method developed to determine connectivity in the field is presented in detail in Lamontagne et al. (2010). The approach uses measurements of fluid pressure in the streambed to confirm whether it is connected or not (see also Brunner et al. 2009a, Cook et al. 2010). Briefly, fluid pressure will always

be greater than or equal to zero in connected streambeds and lesser than or equal to zero in transitional or disconnected ones.



Figure 3 Conceptual representation of the different scales of investigation, using the Lachlan River as an example. The study reach for each river (bottom left) was usually 5–30 km in length. Each reach was instrumented with two piezometer transects (Hillston Bridge and Gonowlia Weir on the Lachlan). Each piezometer transect (bottom right) consisted of three piezometer nests installed perpendicular to the river. The section of the river in front of the piezometers was used for the streambed-based measurements.

A mixture of connected and disconnected environments was found at the six river reaches (Table 1).

Billabong Creek, the Lachlan River and the Gwydir River were losing-disconnected at the transects. Consistent with Brunner et al. (2009a), these sites had a clogging layer in the streambed (usually clay over sand). However, at one each of the Lachlan and Gwydir transects, the clogging layer was at



depth, creating perched aquifers. The Namoi and Macquarie rivers were losing-connected. The river at the two Border Rivers transects was connected, but one location was gaining and the other losing.

As a first approximation, the depth to the water table in the riparian zone can be used to evaluate

whether a stream is losing-connected or -disconnected. At our sites, the water table in the riparian zone was 6 m below streambed level or deeper at the losing-disconnected sites but never deeper than 1 m below streambed level at the losing-connected ones. Thus, as a first approximation, when the

water table in the riparian zone is 6 m deeper than streambed level, the rivers are likely to be disconnected. In addition to the methodology used here (Lamontagne et al. 2010), Cook et al. (2010) describe alternative techniques to determine the connection status in the field used elsewhere.

Table 1 Assessment of connection status at the 12 riparian transects in 2009 using bank tests, streambed fluid pressure measurements, and depth of the water table in the riparian zone relative to streambed elevation. Bank tests (holes dug along the banks) were first made to preliminarily classify each site (wet holes = connected, dry holes = transitional or disconnected) and guide the choice of streambed fluid pressure (ψ) measurement technique. The detailed procedures are described in Lamontagne et al. (2010, 2011a, c).

River and transect Bank test Streambed

fluid pressure (ψ)

Riparian water table

minus streambed

elevation (m)

Streambed texture

Connection status

Billabong—East Dry ψ < 0a –12 Clay Disconnected

Billabong—West Dry ψ < 0a –8.4 Clay Disconnected

Lachlan—Hillston Dry ψ < 0 –22 Clay Disconnected

Lachlan—Gonowlia Wet – –25 Silty sand Disconnected b

Macquarie—Woodlands Wet ψ > 0 –0.69 Clay/silty sand and cobbles

Connected

Macquarie—Macs Wet ψ > 0 +1.0 Clay/silty sand Connected

Namoi—Old Mollee Wet ψ > 0 –0.03 Silty sand Connected

Namoi—Yarral East Wet ψ > 0 –0.03 Silty sand Connected

Gwydir—Brageen Dry No data –12 Gravelly clay Disconnected c

Gwydir—Yarraman Dry/wet ψ < 0 –6.6 Silty sand over clay

Disconnected b

Border—Site 1 Wet ψ > 0 +0.33 Silty sand and cobbles

Connected (gaining)

Border—Site 2 Wet ψ > 0 +0.62 Silty sand Connected

a Inferred from moisture profiles. b Perched aquifer. c Inferred from the depth to the water table in the riparian zone (river dry at the time of sampling).

4.4 Streambed hydraulic conductivity

The vertical hydraulic conductivity (Kv) of the streambed was measured by using constant-head and falling-head tests (Hvorslev 1951; Cardenas and Zlotnik 2003) and other techniques. A wide range of

streambed Kv values was found within and between sites (Figure 4). In general, at the site level the streambeds could be divided into low (with Kv ranging from 10–10 to 10–6 m s–1) and high hydraulic conductivity sites (with Kv ranging from 10–5 to 10–3 m s–1). The low Kv streambeds were lined with

clays while the high Kv ones were generally a mixture of silt, sand and coarser sediments.



Figure 4 Summary of vertical hydraulic conductivity measurements made using falling-head tests at the study sites. Additional measurements were also made using constant-head tests at the Macquarie, Namoi and Dumaresq sites. The falling-head test has a lower detection limit of 10–6 m s–1. At Billabong Creek, more accurate estimates of the hydraulic conductivity of the clays were also made by using evaporation experiments with intact streambed sediment cores (Lamontagne et al. 2011a).

4.5 Infiltration through the streambed

A Darcy flux approach was used to estimate infiltration rates through the streambed at the losing-connected sites. This required the measurement of both Kv (Section 4.4) and the vertical hydraulic

gradients (i) at multiple points in the streambed. The vertical hydraulic gradients were measured with a piezomanometer system developed for the project and described in Lamontagne et al. (2011c).

In general, there were large variations in streambed hydraulic gradient at most of the losing-connected

sites, with both gaining and losing conditions found (Figure 5). The presence of gaining streambed



sections on a losing river was probably the result of the ‘working’ nature of the rivers. During low flow conditions (when all measurements were made), flow was not steady but consisted of a series of pulses from releases of water from upstream reservoirs, primarily to provide water for irrigation. As a

result, there was continuous lateral exchange of water between the rivers and the permeable deposits in the alluvial plain surrounding them (sand banks, gravel bars etc.). Thus, ‘gaining’ streambed sections were caused by small-scale exchange with nearby permeable deposits rather than by a

regional process. For example, it was common to have gaining conditions along sand and gravel bars and losing conditions elsewhere in the streambeds.

Figure 5 Variations in vertical hydraulic gradient and hydraulic conductivity (from constant-head tests—KCH) at Dumaresq River site 2 (Border Rivers). Negative gradients indicate losing conditions. Stream flow was gradually receding during the measurement period. Horizontal hydraulic gradients between bank pits and the river also shown (negative values again indicate losing conditions).

The rates of infiltration through the streambed ranged from –0.44 (gaining) to 1.2 m day–1 (Table 2). However, the errors on these estimates were usually large, so the estimates should be used with caution. The site with the highest average infiltration rate and the highest variability (Namoi—Yarral

East) consisted of drying pools interspersed with dry riverbed sections at the time of sampling.

Estimating the rate of infiltration through the streambed in losing-disconnected streams is difficult and has seldom been attempted. A field and modelling strategy to measure infiltration in losing-

disconnected environments was proposed and trialled as a part of the project (see Lamontagne et al. 2011a, b for details). The specific infiltration rates obtained were in the 0.2–8 mm day–1 range, lower than at the losing-connected sites. However, the key result from these preliminary trials was that the

infiltration process appeared heterogeneous in space. In other words, there appear to be ‘hotspots’ in clay-lined streambeds where much of the infiltration takes place.



Table 2 Mean and 95 per cent confidence intervals for the streambed infiltration rate measurements at losing-connected sites, expressed as the specific infiltration rate (q, m day–1) and the lineal infiltration rate (QL, ML km–1 day–1). A negative infiltration means groundwater discharge to the river (because q = –Kvi). A conservative approach was used to estimate the confidence intervals (Lamontagne et al. 2011c, e, f).

River and site q

(m day–1)

QL

(ML km–1 day–1)

Border—1 –0.44 ± 1.66 –0.28 ± 1.24

Border—2 –0.03 ± 1.19 –0.10 ± 1.43

Namoi—Old Mollee 0.29 ± 0.37 2.9 ± 3.3

Namoi—Yarral East 1.2 ± 3.5 7.4 ± 22

Macquarie—Woodlands 0.16 ± 0.30 1.6 ± 2.0

Macquarie—Macs Reserve 0.28 ± 0.52 6.8 ± 17

The precision of these infiltration estimates could be improved by interpreting the data with more advanced statistical techniques. Because of the large spatial variability, a significant sampling effort is

clearly required to precisely estimate infiltration rates from point measurements. However, the spatial variability in streambed hydraulic gradients should be lower if measurements could be made following periods with a steady river stage. Steady flow conditions would reduce the significance of parafluvial

exchange processes, in particular bank discharge during flow recession. However, infiltration through the streambed does not necessarily equal recharge to the alluvial aquifer, especially along those streams where riparian forest transpiration is significant.

The detailed measurements of hydraulic gradients in streambeds demonstrated that there is a significant two-way exchange of water between streams and permeable deposits in alluvial plains regardless of whether the rivers are gaining or losing relative to the surrounding alluvial aquifers. This

process is significant in the context of the interpretation of environmental tracers in alluvial aquifers.

4.6 Temporal hydraulic head response

If the water tables in losing-connected and -disconnected systems respond differently to flow events

(floods and irrigation pulses), the monitoring of paired piezometers and stream gauges could be used to identify the connection status of a river. In theory, the hydraulic head response should be delayed in losing-disconnected sites because of the need for the flood pulse to propagate through the

unsaturated zone before reaching the water table.

In general, hydraulic heads in the piezometer transects responded readily to changes in stream stage at the losing-connected sites but not at the losing-disconnected ones (Figure 6). At some of the

losing-connected sites, flow events also generated bank recharge–discharge cycles with the riparian zones. Because of the more subdued and delayed response expected in losing-disconnected environments, the monitoring period may not have been long enough to capture their full hydraulic

head responses. In addition, many losing-disconnected sites had no significant flow event during the monitoring period.

Thus, while there was an apparent difference in hydraulic head response between the losing-

connected and -disconnected sites, further work is required in order to develop a diagnostic tool; such work is being continued by Flinders University. Barometric effects were observed at some of the losing-disconnected sites and could be used as a part of a diagnostic tool. At many sites hydraulic

heads were apparently affected by other factors, such as groundwater pumping, and these would need to be taken into account.



Figure 6 Changes in hydraulic head over time in the piezometer nest closest to the river at (A) a losing-connected site (Macquarie River at Macs Reserve) and (B) a losing-disconnected site (Gwydir River at Brageen Crossing). Green line, shallow piezometer; blue line, deep piezometer. Note the flow reversal event in the Macquarie River.

4.7 Environmental tracers

4.7.1 Background

One of the key purposes of the field activities was to evaluate the sources of recharge to the alluvial aquifers and the infiltration rates by using environmental tracers. Environmental tracers are any

substance or property of groundwater that provides some information about its source and its ‘age’ since it was recharged. Potential sources of water for typical MDB alluvial aquifers include vertical recharge during overbank floods, infiltration through river banks under low flow conditions, and diffuse

winter rainfall recharge (Allison et al. 1985; Herczeg et al. 2001; Lamontagne et al. 2005). This study used salinity and the stable isotopes of water as the main tools to evaluate the sources of recharge. There are few methods available for dating groundwater at the week to decade time-scales (the age

range expected for riparian groundwater; Cook and Herczeg 2000); 222Rn and CFCs were used as the principal age-dating methods. Radon is a naturally occurring radioactive gas produced in small quantities by all geological materials (Cecil and Green 2000). Owing to its short half-life (t1/2 = 3.8

days), 222Rn can be used to age groundwater less than 2 weeks old (Hoehn and Cirpka 2006;



Lamontagne and Cook 2007). CFCs are anthropogenic in origin and can date groundwater from ~1965 to the present (Busenberg and Plummer 1992; Plummer and Busenberg 2000).

Sulfur hexafluoride (SF6) and 4He were tested as alternative dating methods at the Border Rivers sites

in one of the first applications of these methods in Australia. The principle of SF6 dating is similar to that of CFCs, with the difference that precise ‘excess air’ measurements in groundwater are required (Plummer and Busenberg 2000). Excess air in groundwater was measured by using noble gases such

as neon, argon, etc. (Aeschbach-Hertig et al. 2000). 4He slowly accumulates in groundwater at geological time-scales (thousands of years) from the radioactive decay of uranium and thorium isotopes and is an ideal marker for old, regional groundwater (Solomon 2000).

4.7.2 Recharge source

There was a consistent pattern in salinity and stable isotopes of water at all field sites. Except at the Billabong Creek sites, alluvial groundwater was fresh, with total dissolved solids (TDS) concentrations of <350 mg L–1. As diffuse rainfall recharge in the MDB would tend to have higher salinities (Herczeg

et al. 2001), the principal recharge source at the sites was the rivers. At Billabong Creek, groundwater TDS was more variable, ranging from <400 mg L–1 near the river to >18,000 mg L–1 further away. Thus, some infiltration from the river occurred at Billabong Creek as well. Stable isotope signatures in

groundwater at most sites had a consistent pattern of enriched values close to the rivers and depleted ones further away (Figure 7). These enriched values were due to evaporative enrichment, indicating some recharge of river water under low flow conditions. Because the main source of water in these

tributaries under low flow conditions is releases from upstream reservoirs, surface water will tend to have an evaporation signal (Simpson and Herczeg 1991a, b).

Thus, the overall picture at most sites was a combination of flood recharge with some bank infiltration

under low flow conditions. If the isotopic signature in the rivers over time were known, it would be possible to quantify more precisely the contribution of each source of recharge to the alluvial aquifers. On the other hand, the infiltration process is probably not one way and gradual in this environment

(the ‘piston flow’ model). In particular, bank recharge–discharge cycles triggered by different flow events (floods and irrigation releases) will tend to mix different sources of water together, even if, overall, there is little net exchange of water between the surface and subsurface. This was illustrated

at the gaining site on the Dumaresq River, where an evaporation signal was still found in alluvial groundwater close to the river (Lamontagne et al. 2011f).



Figure 7 Patterns in the stable isotopes of water at the Namoi transects. (A) Change in deuterium (2H) concentration with distance from the river and (B) 2H–18O plot showing the evaporation line for the samples (dashed). ‘Pore water’ is groundwater from the streambed. The local meteoric water line (LMWL; solid) represents the expected range in isotopic signature for rainfall in a region. Isotopic fractionation during evaporation tends to push isotopic signatures to the right of the LMWL. In the case of the Namoi River sites, all samples plot on the same evaporation line suggesting that they originated from one source. Open symbols – Yarral East site; Closed symbols – Old Mollee site.

4.7.3 Groundwater dating

In general, 222Rn activities in alluvial groundwater were close to values expected under equilibrium with geological materials, indicating that groundwater was older than 2 weeks. On the other hand, infiltration rates could be measured from vertical 222Rn profiles collected from streambeds

(Lamontagne et al. 2011c). Beneath the streambed, radon activities should increase with depth (until the equilibrium value is achieved), and the infiltration rate can be determined from the rate of increase. (See Figure 8 for an example from the Macquarie sites.)

A mixture of groundwater ages was inferred from CFCs but there were frequent mismatches between ages derived from CFC-11 and CFC-12 concentrations. As observed elsewhere (Happell et al. 2003), partial CFC degradation probably occurred in the aquifers because of the generally suboxic to anoxic

conditions present. This was confirmed by the comparison with SF6-inferred ages at the Border Rivers sites (Lamontagne et al. 2011f). Thus, CFCs may only be qualitative dating tools in MDB alluvial environments, and SF6 dating may be preferable.

Unexpectedly, elevated 4He concentrations were found in surface water and some alluvial groundwater samples at the Border Rivers sites (Lamontagne et al. 2011f), indicating that the study reach received some regional groundwater discharge. Thus, this river section was apparently a ‘flow-

through’ system, gaining groundwater regionally and losing water locally to the alluvial aquifer. The origin of the regional groundwater was not clear but was hypothesised to be either rock aquifers abutting the alluvial aquifer or leakage from the underlying Great Artesian Basin.



Figure 8 222Rn-inferred ages in streambed profiles on the Macquarie River. Specific infiltration rates (q) are estimated by assuming a porosity of 0.3.

Owing to the uncertainties in CFC-derived dates, it was not possible to estimate infiltration rates at the losing streams sites from groundwater collected from the piezometer transects. On the other hand,

this should be achievable by complementing CFC dating with SF6 dating in the future. The CFC- and SF6-derived ages will also need to be interpreted in the context of the different types of recharge processes in this environment. Further testing of the SF6-dating technique by CSIRO and Flinders

University is under way at the Namoi sites.

4.8 River gauging

4.8.1 Background

Measuring the difference in river discharge between locations along a river is a further method for estimating the volume of water leaking from (or discharging into) a river. If we assume that evaporation is negligible, and are able to exclude or quantify other inflows and outflows, then the difference in the flow

recorded upstream and downstream is the volume of water lost (or gained) through streambed seepage between the two measuring points. As well as being conceptually straightforward, the method offers the advantage of obtaining a bulk measurement at a useful scale, and overcomes the potential problems

caused by the spatial heterogeneity in streambed hydraulic properties.

In practice, the method presents a number of challenges. In addition to the assumptions outlined above, it requires a stable river. Thus, measurement cannot be taken during periods of rain or when

river operations cause the flow to vary. The river systems studied here are all ‘working’ in that they are regulated by major storages and weirs, with multiple diversion activities along their length. It is generally rare to experience periods where river flows are maintained at a stable rate for any suitable

length of time. A further challenge is to find river reaches where all off-takes or influents can be identified and measured, or preferably eliminated. The infiltration losses can also be small relative to the error in discharge measurements.

4.8.2 Results

The estimation of infiltration by differential gauging was attempted in all study reaches apart from

Billabong Creek and the Namoi River (Table 3). Rain prior to gauging in the Billabong Creek caused



unstable river conditions during the site visit and restricted access to suitable gauging sites because of wet ground. No attempt was made in the Namoi River because of very low flow conditions.

River discharge will progressively decrease downstream under losing conditions and increase under

gaining conditions. On the Lachlan, Gwydir and Border rivers, no consistent trend either way can be seen in the results, and no further analysis was attempted.

On the Macquarie River, during the measurement period there was a consistent trend of decreasing

flow downstream (Figure 9). There were also minimal diversions along the reach and river stage was relatively stable (but not perfectly; see Figure 10). Discharge decreased by just under 70 ML day–1 from the uppermost to the lowermost gauging location. Over a 30-km river flow distance, lineal

infiltration rates averaged around 2.4 ML day–1 km–1. This compares to lineal infiltration rates estimated from streambeds tests of 1.6–6.8 ML day–1 km–1 (Table 2, Section 4.5) and of 0.14–0.18 ML day–1 km–1 at median discharge by regional- and local-scale groundwater models (Table 5, Section

6.3.2.2).

Table 3 Differential gauging results from site surveys. The results for each river are shown in an upstream to downstream sequence.

River Site Discharge (ML day–1)

Distance between sites

(river km)

Cumulative distance (river km)

Segment loss (ML day–1 km–

1)*

Cumulative loss (ML day–1 km–1)

Dumaresq 1 39.4 0 0

Dumaresq 2 35.7 9.0 9.0 0.42 0.42

Dumaresq 3 36.8 23.1 32.1 –0.05 0.08

Dumaresq 4 39.2 11.2 43.3 –0.22 0.00

Dumaresq 5 40.2 18.5 61.8 –0.05 –0.01

Dumaresq 6 42.7 10.1 71.9 –0.25 –0.05

Dumaresq 7 39.1 4.7 76.6 0.76 –0.00

Dumaresq 8 42.1 7.2 83.8 –0.41 –0.03

Dumaresq 9 43.3 24.6 108.4 –0.05 –0.04

Gwydir 1 23.8 0 0

Gwydir 2 24.1 8.6 8.6 –0.04 –0.04

Gwydir 3 25.3 16.7 25.3 –0.07 –0.06

Gwydir 4 28.0 11.0 36.3 –0.25 –0.12

Macquarie 1 694.3 0 0

Macquarie 2 669.5 10.7 10.7 2.32 2.32

Macquarie 3 693.6 6.0 16.7 –4.02 0.04

Macquarie 4 661.0 6.4 23.1 5.09 1.44

Macquarie 5 624.5 6.2 29.3 5.90 2.38

Lachlan 1 31.7 0 0

Lachlan 2 27.0 2.2 2.2 2.09 2.09

Lachlan 3 31.5 0.3 2.5 –14.78 0.07

Lachlan 4 32.5 0.9 3.4 –1.17 –0.26

Lachlan 5 31.9 13.8 17.2 0.04 –0.02

Lachlan 6 35.7 0.9 18.1 –4.23 –0.23

Lachlan 7 28.9 2.8 20.9 2.44 0.13

* Minus = gain.



Figure 9 River gauging sites on the Macquarie River, 23–24 September 2009.

Figure 10 Hydrograph of river discharge (ML day–1) on the Macquarie River at Baroona (site 421127). The shaded area represents the differential gauging measurement period at the study reach.



5. Geophysics

Geophysics was used in an attempt to scale up the above findings from site scale to reach scale, and

to provide further interpretation of conditions at the bore transects. Three methods were used:

1) Waterborne resistivity imaging—compared between study sites at the reach scale.

2) Down-hole electromagnetic (EM) induction and gamma logging—to determine bulk

groundwater salinity and lithology in the formation surrounding the bore.

3) Cross-sectional resistivity imaging along the bore transects.

5.1 Reach-scale geophysics: waterborne resistivity imaging

When the interaction between surface water and groundwater resources is the focus of investigation, surveying should be conducted from the surface water bodies rather than on adjacent land, where

relevant features and water movements might be missed. For this reason, geophysics done on the water gives the most reliable, as well as the most detailed, information on stream–aquifer connectivity.

Waterborne resistivity imaging was conducted on:

• the Dumaresq River between Glenarbon and Texas (Border Rivers)

• the lower Gwydir River between Brageen Crossing and the Newell Highway, Moree

• the lower Namoi River between Wee Waa and Narrabri (Gunidgera Creek Weir and Mollee Weir)

• the lower Macquarie River near Narromine.

• the lower Lachlan River near Hillston.

• Billabong Creek between Hartwood Weir and Jerilderie.

A versatile amphibious Argo craft was used for the surveys to cope with the often log-jammed and vegetation-clogged rivers. A 100-m-long geo-electric streamer was readily pulled through all river obstacles and achieved approximately 30 m depth of investigation, including very good near-surface

resolution. Along all but the most clogged reaches, the geo-electric system achieved about 10 km of survey in a day.

The objective of the surveys was to map losing river reaches, either connected or disconnected. As

the resistivity of soil is very sensitive to saturation, saturated sediments underlying a losing system should have a higher resistivity than that of the infiltrating river water (by virtue of dependence on porosity), and unsaturated sediments should have a much higher value. (The presence of clay

complicates this view, as it degrades the sediment resistivities.) Low resistivities usually indicate freshwater-saturated sediments beneath a river. Unsaturated zones in a losing-disconnected system are expected to have very high resistivities relative to the infiltrating water resistivity. Imaging was

presented with a water depth trace overlay, obtained from sonar, so that surface water resistivity comparison could occur.

All six rivers had losing reaches for part of their length (Allen 2009). No extremely high resistivities

were encountered. Peak resistivities decreased in the order of Dumaresq - Macquarie - Gwydir - Namoi - Lachlan - Billabong Creek. Genuine disconnected systems were not identified, possibly because river leakage might maintain a nearly saturated column of sediments immediately beneath

the river, with rapid lateral transition to true unsaturated conditions beyond the reach of the geophysical sensors. However, clay could have mitigated any high resistivities. Nevertheless, losing conditions were identified readily, without clear establishment of connected or disconnected status.

Confining clay layers were readily resolved in most cases. They were indicated by resistivities at the high end of the spectrum, but could be confused with saline groundwater if the resistivities are very low. In ambiguous cases, independent information is required to differentiate between the two



possibilities. However, both of these scenarios indicate where only relatively small volumes of fresh water are escaping from rivers. The clearest example of a confining layer in the imaging is associated with the Gwydir River.

Beneath the Macquarie and Dumaresq rivers, basement rock highs are evident from their concave-down resistivity anomalies and some surface outcrops. The highs control groundwater flow and may affect the extrapolation of point test results. Conductive features within the highs identify likely sources

of saline inflow to either the rivers or the underlying aquifers.

The higher resistivities in the bed of the Dumaresq and Macquarie rivers indicate losing status. The Gwydir, Namoi and Lachlan rivers reveal mainly mid-resistivities and occasional high resistivities

which most often indicates clay. The Gwydir River had the highest incidence of near-surface clays. Billabong Creek was the only river with a pervasive highly conductive base, which is due to clay and brackish groundwater (Allen 2009).

5.2 Site-scale geophysics

Several geophysical methods were used to assist in determining the hydrogeological conditions along the bore transects between the piezometers. Of particular interest were the changes in the

groundwater salinity and absolute salinity levels near the surface.

The deeper piezometer in all bores was logged using down-hole EM and gamma logging in both transects at all six sites. Resistivity imaging was conducted along each bore transect.

The electrical conductivity of sediment is a function of many variables, including porosity, degree of saturation, clay content (grain size) and the salinity of groundwater. It represents a lumped average electrical conductivity value influenced by these factors. Changes in any one variable may cause the

conductivity to change.

The convention in geophysics is to use electrical resistivity (the inverse of electrical conductivity), measured in ohm-metres (Ωm). However, surface-based EM techniques record data as conductance

(siemens), even though the physical property being measured is the same. Measurements of electrical conductivity are presented as mS/m. The conversion from one value to the other is achieved by dividing into 1000: thus, 10 mS/m = 100 Ωm (1000/10 = 100); 1 Ωm = 1000 mS/m (1000/1 = 1000).

All values were analysed as mS/m.

5.2.1 Down-hole electromagnetic induction and gamma logging

Each of the deeper of the paired piezometers was logged using an EM probe and a gamma probe. To allow correction to formation EM, temperature was also logged. The probes ‘see through’ the PVC

casing and provide data on the bulk water salinity and the lithology of the surrounding formations. The principles of the EM induction logging method and basic interpretation theory are outlined in Williams and Beckham (1995).

The gamma logs appear to have signatures that assist in defining geological marker beds in what are otherwise monotonous sandy facies that can be difficult to categorise, especially given the drilling method used (rotary mud). The logs were used to provide a depth calibration for subsequent

monitoring runs of the down-hole EM probe.

5.2.2 Cross-sectional resistivity imaging

The resistivity imaging technique consists of deploying equally spaced electrodes in a linear array and systematically choosing electrode pairs as transmitters and receivers to obtain a high density of data

representing the subsurface. This work was performed using an Advanced Geosciences Incorporated Sting/Swift resistivity system, which has a high level of automated data acquisition. The technique is an




extension of the traditional ‘DC resistivity’ technique that has received widespread recognition, over a long period, for its ability to provide information about the subsurface resistivity distribution (Burger 1992).

Data processing involves the inversion techniques introduced by Loke and Barker (1996) to transform

the measured (apparent resistivity) data into estimated values of ‘true’ ground resistivity as a function of depth along the survey line (which lies directly along the bore transect).

Resistivity depth sections taken at a 4-m electrode spacing were used over all survey lines. The

survey lines are measured from the water’s edge and are parallel to the line of the piezometer nests.

For each survey line, two forms of depth section are produced:

• Apparent resistivity or pseudo-sections—these measure apparent resistivity along each survey

line as a function of ‘approximate depth’ (determined as a function of the spread geometry).

• Interpreted resistivity-depth sections—these represent the interpreted true ground resistivity. The percentage root-mean-square error is a measure of the degree of fit of the inverted model against

the field data.

The sections use contour intervals that are not evenly spaced; rather, full colour stretches are presented. The intervals have been selected, somewhat subjectively, to accentuate very low resistivity

(high conductivity) features. All sections use the same contour scheme to allow a consistent evaluation over the entire survey.

For each of the bore sites, the down-hole EM formation conductivities were converted to resistivities

and overlain on the cross-sectional EM resistivities to allow comparison. Figures 11 and 12 show the Namoi depth sections as an example.

The upstream section (Figure 11) shows a relatively poor correlation, although this may to some

degree be an artefact of the contouring of the resistivity imaging data. No shallow resistive layer was identified.

The downstream section (Figure 12) provides a relatively good depth slice. There is a high-resistivity

zone at each of the piezometer nests immediately above the water table whose physical nature is unclear.

The detail in both transects is still not sufficiently accurate to enable these geophysical methods to be

used to integrate the geophysical information for the assessment of stream losses at any site.


Figure 11 Cross-sectional resistivity imaging with overlain down-hole resistivity for the Namoi River upstream section.

Figure 12 Cross-sectional resistivity imaging with overlain down-hole resistivity for Namoi River downstream section.



6. Applied modelling research

6.1 Background and Purpose

The final component of the project investigated aspects of applied groundwater modelling, with the aims of improving our understanding of the dynamics of surface water–groundwater exchange, and of

providing improved estimates of water loss from rivers.

The approach taken is to couple lessons taken from the studies of model conceptualisation and reassessment by Flinders University (Section 2) and the site-scale data and findings reported by

CSIRO (Section 4) and develop site-scale models to meet the following objectives:

1) Further improve our understanding of surface water–groundwater interaction within each river–aquifer system.

2) Quantify river infiltration in each region.

3) Compare the infiltration flux estimated by fine-scale (‘child’) sub-models with estimates from the coarse-scale (‘parent’) regional models.

4) Study the effect of time and spatial scales on estimates of river–aquifer water exchanges.

The work was performed in two parts:

1) Modification of three existing regional models to explore research questions relating to spatial

and temporal scales.

2) Development of six fine-scale local-area models calibrated against the site-specific data collected by CSIRO (Section 4).

The results are reported in more detail in Heritage Computing (2009, 2010).

6.2 Stream–aquifer interaction sensitivities by modelling scale

6.2.1 Spatial scale

The following possible research questions relate to spatial scale:

• Are coarse-scale regional models good enough for water budget estimation?

• Does fine scale matter for water resource allocation?

• Does fine scale matter for proper simulation of surface water–groundwater interaction processes?

• How many layers are needed to simulate surface water–groundwater interaction processes?

• Are telescoped or local grid refinement (LGR) sub-models required?

• Are 2D section models sufficient to capture detail?

• How can sub-model or section model findings be scaled up to a catchment?

The spatial scale issue was investigated by extracting a fine-scale (child) sub-model from the lower Lachlan coarse-scale (parent) regional model, with refinement of spatial scale from 1,000 m to 90.9 m per cell. This was done using LGR and telescopic mesh refinement (TMR) tools. The LGR method

proved inappropriate for the Lachlan, owing to the presence of unsaturated cells within the sub-model domain. The LGR method requires a constant head in each boundary cell of the sub-model.

One major finding of this study was that when model cell size is reduced, the calibrated permeabilities

of the regional model are no longer valid at the groundwater abstraction sites. As the drawdown in a finite-difference model is pertinent to a radius of about 20 per cent of the cell size, a finer cell will



report a larger drawdown for the same hydraulic conductivity. As a consequence, recalibration will be necessary when sub-models are developed from regional models.

The key finding, however, for the spatial scale issue, is that there appears to be a measurable

overestimation of stream leakage (on the order of 5 per cent) according to the current practice of using monthly stress periods in regional models. The regional model consistently estimates higher fluxes than the sub-model (Figure 13).

Figure 13 Temporal variations in stream leakage estimated by regional (parent) model and sub-model (TMR).

0 2 4 6 8 10 12 14 16 18YEARS

1200

1400

1600

1800

2000

2200

2400

2600

2800

3000

STR

EAM

INFLO

W (kL

/day)

LEGENDParent Model

TMR Model

1200

1400

1600

1800

2000

2200

2400

2600

2800

3000

[TMR] TMRvPARENT.grfRIVleakage.xls!Comparison26Sep2009

6.2.2 Temporal scale

The following possible research questions relate to temporal scale:

• Are monthly stress periods good enough for water budget estimation?

• Are daily stress periods essential for proper simulation of surface water–groundwater interaction

processes?

• Are long-term averages the same regardless of the stress period length?

• Are run times for smaller time-steps prohibitive?

The time scale issue was investigated by running the lower Macquarie model for 2000 with rainfall and

river levels specified for daily, weekly and monthly stress periods (Figure 14). Groundwater abstraction remained as a constant daily rate over a month. Use of a daily or weekly modelling time scale adds significantly to the modelling effort and effectively reduces the period of time that can be

considered for groundwater model calibration and prediction. The error in long-term average leakage incurred by adopting a weekly or monthly timescale was found to be no more than 0.5 per cent for the whole length of the Macquarie River. It follows that, on the strength of this example, the current

practice of using monthly stress periods in regional models is appropriate for groundwater resource assessment.



Figure 14 Macquarie River losses estimated from groundwater models at daily, weekly and monthly time scales.

0

50000

100000

150000

200000

250000

300000

350000

1 51 101 151 201 251 301 351

Time (day)

Flu

x (m

3/d

ay)

0

20

40

60

80

100

120

Rai

nfa

ll (

mm

)

Rainfall (mm) Daily Sum Reach (1-5) Weekly Sum Reach (1-5) Monthly Sum Reach (1-5)

6.3 River site models

Six fine-scale groundwater models were developed for the local area around the study sites. The

objectives were to quantify the rate of water loss from the losing-disconnected river system to the groundwater in each area, and to compare the rate of loss estimated by the sub-model with estimates from the regional model.

Local sub-models were extracted and refined from the regional models by reducing cell size from 500–2500 m to 50–100 m, subdividing the upper model layer into two or three layers, using a time scale of 1 day instead of 1 month, and calibrating the sub-model over a period of 3 to 11 months

during 2009 (according to the availability of simultaneous river stage and groundwater level data).

6.3.1 Regional models—river loss calculations and connection status

Each of the six study areas has existing regional groundwater models. The models differ according to software base, lead modeller and time since the last update. Most models were developed within the

NSW Office of Water, but some were developed externally under contract. In the case of Billabong Creek, a new regional model had to be compiled by stitching together two separate regional models before the sub-model could be extracted.

The regional models were interrogated to provide estimates of river losses, presented as a lineal rate of loss (or occasionally gain) in terms of flux per day per kilometre of river (kL day–1 km–1), for given river flow exceedance probabilities.

The estimates of lineal river losses at median river flow (50th percentile exceedance probability) and high river flow (10th percentile exceedance probability) are shown in Table 4. The losses at median flows are similar at each site, ranging from 31 to 150 kL day–1 km–1. Exceptions are the Border Rivers,

where the regional model wrongly has a permanent gaining condition (see discussion below), and the Billabong Creek where losses are relatively small. The more southerly sites seem to have lower loss rates. The rate of river losses at high flow differs considerably between rivers.



Table 4 Summary of river losses (kL day–1 km–1) and connection status (% of flow duration) — regional models.

River Losses - at median river flows (50%)

Losses – at high river flow (10%)

Gaining Losing-

connected Losing-

disconnected

Border Rivers –54* –20* 99 1 0

Gwydir 150 300 0 3 97

Namoi 110 1,400 8 11 81

Macquarie 140 670 1 38 61

Lachlan 31 260 0 0 100

Billabong 40** 85** 0 0 100

• *Gaining. See discussion below.

• ** Note the values originally published were in error and have been amended

The method used to determine these estimates is described in Heritage Computing (2010).

The relative river levels, river bed elevations and groundwater levels in the modelled river cells were examined to classify the river-aquifer system as gaining, losing-connected or losing-disconnected. When the water table was modelled to be lower than the river bed, it was assumed that the system is

disconnected. We have learned from earlier in the project that this assumption does not necessarily hold true in nature, but this is the way disconnection is defined in MODFLOW river algorithms. See Heritage Computing (2010) for details and variations in modelling approaches between the rivers.

Table 4 shows that, with the exception of the Border Rivers, all systems are predominantly losing-disconnected according to the models. The development of the sub-model for the Border Rivers shows that the regional model is unreliable owing to substantial errors in river levels. Accordingly, the

interpreted ‘gaining’ status for the Border Rivers is not correct.

6.4 Local (reach-scale) models—river loss calculations and connection status

6.4.1 Model conversion

Each regional model was converted to a finer local-scale model. The first step was to interrogate the models to assess, according the Flinders University findings, whether division of the regional model’s

layer 1 and simulation of variably saturated flow were advisable when converting a model to a finer scale. The University’s Figures 3 and 6 are reproduced here as Figures 15 and 16. The assessment required analysis of typical water depth (d) and the magnitude of normalised river leakage (q/Ka).

The figures suggest that:

• there is no need to simulate the unsaturated zone if river water depth d > 1 m

• there is no need to split layer 1 of the regional model if the normalised flux q/Ka < 0.05 m–1.

An example of the water depth assessment is given in Figure 17; full details are given in individual technical reports and Heritage Computing (2010). The probability plots show that water depth exceeds 1 m more than 90 per cent of the time at all sites except the lower Gwydir, where it exceeds 1 m about

70 per cent of the time. We conclude that there is no compelling reason, on theoretical grounds, to simulate the unsaturated zone beneath a river bed. The standard RIV algorithm in MODFLOW should be sufficient for all rivers except the lower Lachlan, where the SFR package is used by the regional

model, because layer 1 is frequently dry, and RIV cells in layer 1 would deactivate automatically.



Figure 15 The graph from Brunner et al. (2009a) shows that as river depth approaches 1 m (dotted line), the difference between flux estimates obtained by explicitly simulating unsaturated flow between the river bed and the water, and not, become negligible.

Figure 16 MODFLOW model accuracy for vertical layer discretisation. The graph shows that as the normalised river leakage (q/Ka) decreases, the differences between the calculated water table mount height beneath the river, estimated by explicitly simulating the unsaturated flow between the river bed and the water, and not, become small.



Figure 17 River height probability distribution for the Namoi River as an example.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7RIVER WATER DEPTH d [m]

0

10

20

30

40

50

60

70

80

90

100

PR

OB

AB

ILIT

Y [%

]

0

20

40

60

80

100

[Lower_Namoi][Interrogation]WaterDepth.grf

LNamoi_River_Interrogation.xls!Probability

An example of the river leakage assessment is given in Figure 18; full details are given in individual technical reports and Heritage Computing (2010). The probability plots show that normalised river

leakage is much less than 0.05 m–1 at all sites all of the time. We conclude that there is no compelling reason, on theoretical grounds, to subdivide the uppermost layer of the regional model. However, layer 1 did have to be subdivided in this study, because nested piezometers were installed within the

limits of regional layer 1. Otherwise, it would not be possible to separately calibrate against nested piezometers, owing to MODFLOW’s assumption of non-varying head with depth across a model layer at a given location (Dupuit assumption: depth-averaged heads).

Figure 18 River leakage probability distribution for the Namoi River as an example.

1E-007 1E-006 1E-005 0.0001 0.001NORMALISED RIVER LEAKAGE q/Ka [1/m]

0

10

20

30

40

50

60

70

80

90

100

PR

OB

AB

ILIT

Y [%

]

0

20

40

60

80

100

[Lower_Namoi][Interrogation]NormalisedLossRate.grf

LNamoi_River_Interrogation.xls!Probability




6.4.2 Results and comparison with regional models

For each of the sub-models, local river loss over time from the nearest river cell at each transect was estimated. As sub-model cell sizes are small (50–100 m), it is instructive to estimate the river loss from the stretch of river between each pair of transects (typically 30 km). Accordingly, lineal loss rates were

calculated and plotted. An example for the Macquarie River is given in Figure 19.

Figure 19 Lineal river loss in the Macquarie River.

The sub-model river flow (and hence lineal river loss) probability distributions are much narrower than the regional ones, as expected, because of the much shorter assessment period and the lack of high river flows in 2009. Accordingly, the probability results do not give a measure of the uncertainty in

leakage estimates, but do compare the magnitude and variability between the six sites. The river losses estimated by the regional models and sub-models are not dissimilar in the mid-probability range (~200 kL day–1 km–1), except for the Gwydir, which has anomalously high values in the sub-

model (recalibration is recommended).

Table 5 compares the model estimates of river leakage. The Namoi, Macquarie and Lachlan regional models and sub-models are in good agreement at the median level. The Billabong Creek regional

models and sub-models models differ significantly however.

Table 5 Summary and comparison of river losses (kL day–1 km–1)—local models.

River Local losses—

median flow (50%) Local losses—high

flow (10%) Regional losses—median flow (50%)

Regional losses—high flow (10%)

Border Rivers 53 86 –54* –20*

Gwydir 1045 1402 150 300

Namoi 150 280 110 1,400

Macquarie 177 317 140 670

Lachlan 48 57 31 260

Billabong 15 16 40** 85**

• * Gaining.

• ** Note the values originally published were in error and have been amended



6.5 Findings

The applied modelling studies reported the following main findings:

• Use of a daily time scale allows better model replication of dynamic groundwater levels.

sted

if

s are taken from a regional model in a TMR (telescopic mesh refinement) l.

rvations within the model area.

y

ent metering practice, by which spot readings are

al area model.

en models of different scales.

s rates between the

transects, owing to limited exposure

erences in model estimates of average

eas

by the regional models would

• Finer vertical discretisation allows better model replication of vertical hydraulic gradients in ne

shallow piezometers.

• Finer lateral discretisation allows better model representation of river morphology in a finite-difference model.

• Use of a smaller model extent causes a model to be heavily reliant on boundary conditions;

boundary conditionprocess, the accuracy of the sub-model depends heavily on the accuracy of the regional mode

• Use of a smaller model extent requires higher topographic accuracy than is usually sought in a regional model.

• Use of a smaller model extent denies incorporation of external pumping stresses that might

impinge on obse

• If pumping stresses are evident, use of a daily time scale requires detailed knowledge of dail

pumping schedules; this is inconsistent with currtaken at intervals of around 3 to 12 months.

• Use of a daily time scale allows better estimation of variations in river leakage rate with time in

losing river–aquifer systems.

• Considerable recalibration of aquifer properties and river conductances is required when moving

from a regional model to a loc

• Regional model errors in local topographic levels, river levels or simulated groundwater levels compromise the validity of comparisons betwe

• The time duration of observations in this study (<1 year) has limited exposure of the aquifer

system to high river flows; this affects the comparison of estimates of river lossub-models and regional models over a longer time period.

• Models of different scale are in general not dissimilar in their median lineal loss rates (expressed as kL day–1 km–-1) aggregated between research transects.

• Models of different scale were found to give quite different extreme values of lineal loss rates

(expressed as kL day–1 km–1) aggregated between research to high river flows during the model calibration period.

• Owing to substantial differences between the characterisation of regional models and sub-models and different stress fields, it is inconclusive whether diff

leakage rates are due to differences in temporal or spatial scales.

• Given the confounding effects of proximal boundary conditions, the limited sizes of the study ar

and the limited duration of the study, water balance estimates givenbe more appropriate for groundwater management purposes.


7. Discussion

In general, there has been significantly less research on losing than on gaining stream environments.

As a result, the methodological ‘toolbox’ used to define the connection status or to measure infiltration rates of losing streams is incomplete. The project has significantly advanced the theory underpinning the behaviour of losing streams and trialled a number of methodologies to assess connectivity and

infiltration rates in the field. Some of the findings presented in previous sections are further discussed below.

7.1 Disconnection status Very few studies have attempted to differentiate between losing-connected and losing-disconnected river sections (see review in Cook et al. 2010). Regional-scale assessments evaluate the likelihood of

disconnection on the basis of the expected depth of the water table relative to the stream stage (obtained by groundwater modelling or other approaches). However, because of potentially steeply sloping water tables near streams (from recharge mounds), these approaches are not foolproof.

For comparison, the field-based methods to assess connectivity were compared to water table maps derived from groundwater models or other approaches. The methods did not assess the connection status at the same temporal and spatial scales and therefore will not necessarily agree with one

another because of the dynamic nature of connectivity. This is because at any given site, the connection can change over time; and at any given time, gaining, losing-connected and losing-disconnected river sections can be found along a given reach.

As expected, there was some agreement and some disagreement between methods (Table 6). As the site investigations were made during an extended regional drought (2009), they were anticipated to represent the most likely conditions for disconnection: those rivers that could disconnect should have

become disconnected at the time of the study. The methods agreed generally that the Billabong Creek and Lachlan River study reaches were losing-disconnected, consistent with relatively deep regional water tables. Most methods agreed that the Border Rivers reach was primarily connected, but

disagreed on the extent of losing and gaining sections.

The methods did not agree on the Namoi, Macquarie and Gwydir rivers (Table 6). However, it is not possible to assess which method was ‘right’ because, as mentioned above, they assessed

connectivity at different spatial scales. A mixture of losing-connected and losing-disconnected river sections may have been present along these reaches. This possibility could be further evaluated by longitudinal surveys of streambed fluid pressure along the study reaches, a relatively foolproof but

tedious method.

Table 6 Comparison of the connection status for the study reaches determined by different approaches.

River Regional water table (Parsons et al. 2008)

Groundwater model

(from Table 4)

Riparian water table and streambed fluid

pressure Geophysics Consensus

Billabong Creek Transition area from

LC to LD LD LD LC LD

Lachlan Transition area from

LC to LD LD LD LC? LD

Macquarie LC LC&D LC LD? LC&D?

Namoi LC LD LC LC LC&D?

Gwydir LC LD LD L (undiff.) LD

Border Rivers G&LC N/A G&LC LD G&LC

Key: L = losing, G = gaining, C = connected, D = disconnected.



There was a high uncertainty associated with the assessment of connectivity using geophysical techniques. Resistivity-based geophysical measurements respond to clay content, water content and salinity, all of which can vary substantially in losing environments. Thus, geophysical techniques could

benefit from being specifically targeted. For example, different approaches could be used to map clogging layers (a function of clay content) or the position of the water table (a function of water content) in streambeds. Because of their potential for reach- and regional-scale mapping of key losing

stream features, further development of geophysical techniques in losing environments is warranted.

7.2 Disconnection potential Brunner et al. (2009a) developed a criterion for determining which stream reaches will become

disconnected when surrounding water tables are lowered (Section 2.2.2). All sites that were found to be losing-disconnected in this study satisfied the Brunner et al. (2009a) criterion. In particular, all losing-disconnected sites had well defined clogging layers (clay units 0.5 to 2 m thick) in the

streambed. Thus, reach-scale assessments of the potential for disconnection should be possible if we map the presence and thickness of clay units in MDB streambeds. The potential for rivers to disconnect may be beneficial under some circumstances because a surface pool will remain for longer

as the water table drops. In contrast, losing-connected rivers will tend to dry up more quickly, because the infiltration flux will keep increasing with a further decline in the water table.

The Brunner et al. (2009a) criterion considers a clogging layer with a uniform thickness across the

streambed. However, field measurements suggest that recharge ‘hotspots’ occur in clay-lined streambeds, presumably where the clogging layer is thinner or absent. The criterion could be further evaluated by including the case of streambeds with clogging layers of variable thickness.

7.3 Infiltration rates The regional groundwater models showed that the study rivers all had significant seepage losses, with

the exception of the Border Rivers where the existing model requires further calibration and the magnitude of the seepage losses is unknown. The estimation of infiltration by groundwater age-dating was partially successful. 222Rn can age groundwater only at the scale of days to weeks and was not

suitable for age-dating at the riparian scale (tens to hundreds of metres). However, it proved useful for estimating infiltration at the streambed scale. CFC dating was found to be unreliable because of unsuitable geochemical conditions in the aquifers. However, limited SF6 measurements made during

the study suggested that SF6 could replace CFCs as a groundwater dating tool. Another potential tracer, post-bomb tritium (3H; Morgenstern et al. 2010), was not trialled but could also be considered.

The estimation of infiltration by differential gauging and the Darcy flux through the streambed had

significant uncertainties because of the unsteady flow conditions in the rivers during the measurement periods. To improve the accuracy of these methods would require stabilising river flows for several weeks ahead of the measurement period, which would require a fair amount of coordination among

the agencies responsible for managing the rivers. Several weeks of steady flow are probably required in order to stabilise exchanges with the wetlands and parafluvial zones bordering the rivers. There are precedents for stabilising river flows prior to and during measurement periods in the MDB to improve

accuracy. For example, the salt load to the South Australian section of the River Murray is estimated using run-of-river salinity surveys, which require flows to be stabilised for several weeks ahead of the measurement period (Porter 2001). The development of an improved differential gauging protocol

suitable for MDB rivers is a priority.

Both gross and net infiltration rates can be estimated by combining differential gauging with the release of in-stream tracers (Ruehl et al. 2006; Payn et al. 2009). Differential gauging only measures

net infiltration rates. In cases where both groundwater discharge and infiltration takes place between gauging points, the total (or gross) infiltration rate is underestimated by differential gauging. This is not



a concern along losing-disconnected river sections, where only infiltration can occur (i.e. gross = net infiltration). Where significant differences between gross and net infiltration rates could occur would be for groundwater flow-through rivers (like the Dumaresq) or where there is local discharge to rivers from

irrigation mounds. The use of in-stream tracers is logistically complex in larger rivers because large amounts of tracers have to be added and subsequent measurements have to be made over larger distances and time periods. Many practical tracers for larger rivers (such as dyes and radioisotopes)

are not recommended for waterways used for irrigation or drinking water supplies. Nevertheless, when knowing the gross and net infiltration rates matters, and adding tracers to a river reach is feasible, differential gauging should be combined with in-stream tracer experiments.

Longer-term water balances along study reaches can also be used to estimate the infiltration into alluvial aquifers. However, estimates of the infiltration rates in MDB rivers by the water balance method tend to have a large error (Holz et al. 2010). Possible sources of error include poorly

constrained sink terms (surface water pumping) and errors in stage–discharge relationships, especially at higher flows. There clearly remain some technical challenges to the estimation of infiltration rates along MDB rivers.



8. Conclusions

In summary, several outcomes can be used immediately to inform water management and policy:

• Infiltration still occurs when rivers become disconnected. Disconnected does not mean that there is no exchange between rivers and aquifers.

• While further reduction in water table depth will not change the infiltration rate at a specific point in

disconnected streams, infiltration may increase at the reach scale if other parts of the stream switch from being connected to disconnected.

• Once a stream is disconnected, the infiltration rate is still affected by surface water management. At higher stream stages the infiltration rate will increase because of a greater hydraulic head in the

streambed and the increase in streambed wetted perimeter.

• Some river reaches currently considered to be losing in the MDB could be groundwater flow-

through systems, such as in the case of some of the alluvial aquifers overlying the GAB. Failure to account for inter-aquifer exchange could lead to errors in the water balance of alluvial aquifers.

8.1 Recommendations Several recommendations can also be made to guide the calibration process for regional groundwater models used to manage alluvial aquifers:

• The criterion developed by Brunner et al. (2009a) can be used to evaluate which stream sections

have the potential to become disconnected.

• A method was developed to test the appropriateness of MODFLOW for creating regional models that include stream–aquifer interactions. This method, along with the correct definition of the term

‘disconnection’, should be included in the groundwater modelling guidelines currently used by the industry.

• Existing reach-scale groundwater models do not necessarily provide better estimates of infiltration than regional ones. Likewise, the use of time steps shorter than a month is not recommended nor

is currently possible with some monitoring data (such as pumping extraction rates).

• Field assessments of connectivity and infiltration rates should be included as a component of the

calibration process for regional groundwater models

8.2 Future studies A discussion is also required between research and management agencies to scope a research agenda in the area of surface water–groundwater interaction in the MDB. A key challenge in the

context of the MDB and many other river systems in Australia is the difficulty to measure exchange processes at a large spatial scale. A combination of approaches may be required that would combine site, reach and regional scale measurements. It may not be necessary to instrument or characterise

whole river networks evenly because regional groundwater models could be more sensitive to change or uncertainty along some river sections than others. In summary, some of the research topics to consider for future studies include:

• The development of a hierarchical approach to investigate surface water–groundwater exchange at the site, reach and regional scale. This includes the design of sensitivity analyses using existing

or new groundwater models in order to identify river reaches where additional monitoring or field measurements would provide the most benefits to model calibration.

• To develop methodologies to map streambed hydraulic conductivity and (in disconnected rivers) depth to the water table below streambeds using geophysical techniques.



• To develop a protocol for differential gauging suitable for MDB rivers. The development of in-

stream tracer release techniques should also be encouraged in order to measure both gross and net infiltration rates along losing rivers when these are required.

• To continue the development of environmental tracer techniques to identify recharge mechanisms

and measure infiltration rates at the riparian and floodplain scale. As an alternative to piezometer networks, investigate if direct push drive point technologies could be used to provide snap-shot surveys of environmental tracers in riparian and floodplain groundwater.

The project was too short in duration to investigate the relative significance of infrequent but large overbank flow events to the water balance of MDB alluvial aquifers. A hierarchical approach could also

be used to estimate recharge to alluvial aquifers during exceptional events. This would combine large scale estimates of flood recharge using the latest flood mapping technologies and smaller scale field-derived estimates of recharge using geophysical or environmental tracer techniques.



9. References Allen D. 2009. Run-of-rivers geo-electric surveys of losing/disconnected streams in New South Wales.

Report to the NSW Office of Water by Groundwater Imaging Pty Ltd.

Aeschbach-Hertig W, Peeters F, Beyerle U, Kipfer R. 2000. Paleotemperature reconstruction from noble gases in ground water taking into account equilibration with entrapped air. Nature 405: 1040–1044.

Allison GB, Stone WJ, Leaney FWJ. 1985. Recharge in karst and dune elements of a semi-arid landscape as indicated by natural isotopes and chloride. Journal of Hydrology 76: 1–25.

Brunner P, Cook PG, Simmons CT. 2009a. Hydrogeologic controls on disconnection between surface water and groundwater. Water Resources Research 45, doi: 10.1029/2008wr006953.

Brunner P, Simmons CT, Cook PG. 2009b. Spatial and temporal aspects of the transition from connection to disconnection between rivers, lakes and groundwater. Journal of Hydrology 376: 159–169.

Brunner P, Simmons CT, Cook PG, Therrien R. 2010a. Modeling surface water–groundwater interaction with MODFLOW: some considerations. Ground Water 48: 174–180, doi: 10.1111/j.1745–6584.2009.00644.x.

Brunner P, Cook PG, Simmons CT. 2010b. Disconnected surface water and groundwater: from theory to practice. Ground Water 49, doi: 10.1111/j.1745–6584.2010.00752.x.

Burger H R, Exploration Geophysics of the Shallow Subsurface, Prentice Hall P T R, 1992.

Busenberg E, Plummer LN. 1992. Use of chlorofluorocarbons (CCL3F and CCL2F2) as hydrologic tracers and age dating tools: the alluvium and terrace system of central Oklahoma. Water Resources Research 28: 2257–2283.

Cardenas MB, Zlotnik VA. 2003. A simple constant-head injection test for streambed hydraulic conductivity estimation. Ground Water 41: 867–871.

Cecil LD, Green JR. 2000. Radon-222. In Cook PG, Herczeg AL [eds]. Environmental Tracers in Subsurface Hydrology, pp. 175–194. Kluwer, London.

Cook PG, Herczeg AL [eds]. 2000. Environmental Tracers in Subsurface Hydrology. Kluwer, London.

Cook PG, Brunner P, Simmons CT, Lamontagne S. 2010. What is a disconnected stream? Groundwater 2010. 31 Oct – 4 Nov, Canberra.

Happell JD, Price RM, Top Z, Swart PK. 2003. Evidence for the removal of CFC-11, CFC-12, and CFC-113 at the groundwater–surface water interface in the Everglades. Journal of Hydrology 358: 332–353.

Herczeg AL, Dogramaci SS, Leaney FWJ. 2001. Origin of dissolved salts in a large, semi-arid groundwater system: Murray Basin, Australia. Marine and Freshwater Research 52: 41–52.

Heritage Computing. 2009. Stream–aquifer interaction sensitivities for temporal and spatial modelling scales, and proximity to groundwater abstraction. Report HC2009/11. Report to National Water Commission by Heritage Computing.

Heritage Computing. 2010. Stream–aquifer interaction research models at six sites in New South Wales. Report HC2010/13. Report to National Water Commission by Heritage Computing.

Hoehn E, Cirpka OA. 2006. Assessing residence times of hyporheic ground water in two alluvial flood plains of Southern Alps using water temperature and tracers. Hydrology and Earth System Sciences 10: 553–563.

Holz L, Chowdhury S, McNeilage C. 2010. How useful is surface water unaccounted differences for quantifying groundwater–surface water inter-connectivity? Groundwater 2010, 31 Oct – 4 Nov, Canberra.

Hvorslev MJ. 1951. Time-lags and soil permeability in ground-water observations. Bulletin 36. Waterways Experiment Station. US Army Corps of Engineers, Vicksburg, Mississippi.



Ivkovic KM. 2009. A top-down approach to characterise aquifer–river interaction processes. Journal of Hydrology 365: 145–155.

Kalbus E, Reinstorf F, Schirmer M. 2006. Measuring methods for groundwater–surface water interactions: a review. Hydrology and Earth System Sciences 10: 873–887.

Lamontagne S, Cook PG. 2007. Estimation of hyporheic water residence time in situ using 222Rn disequilibrium. Limnology and Oceanography 5: 407–416.

Lamontagne S, Leaney FW, Herczeg AL. 2005. Groundwater–surface water interactions in a large semi-arid floodplain: implications for salinity management. Hydrological Processes 19: 3063–3080.

Lamontagne S, Cook PG, Taylor AR, Crosbie R, Brownbill R. 2010. Field assessment of the connection status in losing streams. Groundwater 2010, 31 Oct – 4 Nov, Canberra.

Lamontagne S, Taylor AR, Cook PG, Brownbill R. 2011a. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Billabong Creek Site Report. Water for a Healthy Country Flagship: Adelaide.

Lamontagne S, Taylor AR, Crosbie R, Cook PG, Kumar P. 2011b. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Lachlan River Site Report. Water for a Healthy Country Flagship: Adelaide.

Lamontagne S, Taylor AR, Cook PG, Hamilton S. 2011c. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Macquarie River Site Report. Water for a Healthy Country Flagship: Adelaide.

Lamontagne S, Taylor AR, Cook PG, Smithson A. 2011d. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Namoi River Site Report. Water for a Healthy Country Flagship: Adelaide.

Lamontagne S, Taylor AR, Cook PG, Barrett C. 2011e. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Gwydir River Site Report. Water for a Healthy Country Flagship: Adelaide.

Lamontagne S, Taylor AR, Cook PG, Gardner WP O’Rourke M. 2011f. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Border Rivers Site Report. Water for a Healthy Country Flagship: Adelaide.

Loke MH, Barker RD. 1996. Rapid least squares inversion of apparent resistivity pseudosections by a quasi-Newton method. Geophysical Prospecting 44: 131–152.

Middlemis H, Merrick NP, Ross JB. 2000. Groundwater flow modelling guideline, November 2000. Murray–Darling Basin Commission, Canberra.

Morgenstern U, Stewart MK, Stenger R. 2010. Dating of stream water using tritium in a post-bomb world: continuous variation of mean transit time with stream flow. Groundwater 2010, 31 Oct – 4 Nov, Canberra.

Parsons S, Evans R, Hoban M. 2008. Surface–groundwater connectivity assessment. Report to the Australian Government from the CSIRO Murray–Darling Basin Sustainable Yields project. CSIRO Publishing, Melbourne.

Payn RA, Gooseff MN, McGlynn BL, Bencala KE, Wondzell SM. 2009. Channel water balance and exchange with subsurface flow along a mountain headwater stream in Montana, United States. Water Resources Research 45: W11427, doi: 10.1029/2008WR007644.

Plummer LN, Busenberg E. 2000. Chlorofluorocarbons. In Cook PG, Herczeg AL [eds]. Environmental Tracers in Subsurface Hydrology, pp. 441–478. Kluwer, London.

Porter, B. 2001. Run of river salinity surveys – A method of measuring salt load accessions to the River Murray on a kilometre by kilometre basis. Proceedings of the 8th Murray-Darling basin Groundwater Workshop. 4 – 6 September 2001, Victor Harbor, SA.



Rassam D, Walker G, Barnett B. 2008. Recommendations for modelling surface–groundwater interactions based on lessons learnt from the Murray–Darling Basin Sustainable Yields project. A report to the Australian Government from the CSIRO Murray–Darling Basin Sustainable Yields project. CSIRO Publishing, Melbourne.

Ruehl C, Fisher AT, Hatch C, Los Huertos M, Stemler G, Shennan C. 2006. Differential gauging and tracer tests resolve seepage fluxes in a strongly-losing stream. Journal of Hydrology 330: 235–248.

Simpson HJ, Herczeg AL. 1991a. Stable isotopes as an indicator of evaporation in the River Murray, Australia. Water Resources Research 27: 1925–1935.

Simpson HJ, Herczeg AL. 1991b. Salinity and evaporation in the River Murray Basin, Australia. Journal of Hydrology 124: 1–27.

Solomon DK. 2000. 4He in groundwater. In Cook PG, Herczeg AL [eds]. Environmental Tracers in Subsurface Hydrology, pp. 425–440. Kluwer, London.

Williams RM, Beckham J. 1995. Stream/aquifer interactions along the lower Darling River, NSW. Dept. Water Resources Technical Services Report TS 95.022. Sydney.




Appendix A: Compendium of project publications Allen D. 2009. Run-of-rivers geo-electric surveys of losing/disconnected streams in New South Wales.

Report to the NSW Office of Water by Groundwater Imaging Pty Ltd.

Brownbill R J, Williams, R M 2010. Interconnection of Surface and Groundwater Systems – River losses from Losing/Disconnected Streams. Groundwater 2010, 31 Oct – 4 Nov, Canberra.

Brunner P, Cook PG, Simmons CT. 2009a. Hydrogeologic controls on disconnection between surface water and groundwater. Water Resources Research 45, doi: 10.1029/2008wr006953.

Brunner P, Simmons CT, Cook PG. 2009b. Spatial and temporal aspects of the transition from connection to disconnection between rivers, lakes and groundwater. Journal of Hydrology 376: 159–169.

Brunner P, Simmons CT, Cook PG, Therrien R. 2010a. Modelling surface water–groundwater interaction with MODFLOW: some considerations. Ground Water 48: 174–180, doi: 10.1111/j.1745-6584.2009.00644.x

Brunner P, Cook PG, Simmons CT. 2010b. Disconnected surface water and groundwater: from theory to practice. Ground Water 49, doi: 10.1111/j.1745-6584.2010.00752.x.

Cook PG, Brunner P, Simmons CT, Lamontagne S. 2010. What is a disconnected stream? Groundwater 2010, 31 Oct – 4 Nov, Canberra.

Heritage Computing. 2009. Stream–aquifer interaction sensitivities for temporal and spatial modelling scales, and proximity to groundwater abstraction. Report HC2009/11. Report to National Water Commission by Heritage Computing.

Heritage Computing. 2010. Stream–aquifer interaction research models at six sites in New South Wales. Report HC2010/13. A Report to National Water Commission by Heritage Computing.

Lamontagne S, Cook PG, Taylor AR, Crosbie R, Brownbill R. 2010. Field assessment of the connection status in losing streams. Groundwater 2010, 31 Oct – 4 Nov 2010, Canberra.

Lamontagne S, Taylor AR, Cook PG, Brownbill R. 2011a. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Billabong Creek Site Report. Water for a Healthy Country Flagship, Adelaide.

Lamontagne S, Taylor AR, Crosbie R, Cook PG, Kumar P. 2011b. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Lachlan River Site Report. Water for a Healthy Country Flagship, Adelaide.

Lamontagne S, Taylor AR, Cook PG, Hamilton S. 2011c. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Macquarie River Site Report. Water for a Healthy Country Flagship, Adelaide.

Lamontagne S, Taylor AR, Cook PG, Smithson A. 2011d. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Namoi River Site Report. Water for a Healthy Country Flagship, Adelaide.

Lamontagne S, Taylor AR, Cook PG, Barrett C. 2011e. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Gwydir River Site Report. Water for a Healthy Country Flagship, Adelaide.

Lamontagne S, Taylor AR, Cook PG, Gardner WP, O’Rourke M. 2011f. Interconnection of surface and groundwater systems—river losses from losing/disconnected streams. Border Rivers Site Report. Water for a Healthy Country Flagship, Adelaide.

Taylor AR, Lamontagne S, Cook PG. 2010. Comparison of techniques to measure the matric potential in sediment profiles below losing streams. Groundwater 2010, 31 Oct – 4 Nov, Canberra.

Interconnection of surface and groundwater systems – · PDF fileLeading policy and reform in sustainable water management Interconnection of surface and . groundwater systems –

Documents