Field comparison of methods for estimating groundwater discharge by evaporation and evapotranspiration in an arid-zone playa

Journal of Hydrology 527 (2015) 1073–1083

Contents lists available at ScienceDirect

Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol

Field comparison of methods for estimating groundwater discharge by evaporation and evapotranspiration in an arid-zone playa Margaret Shanafield a,⇑, Peter G. Cook a,b,1, Hugo A. Gutiérrez-Jurado a,1, Ralph Faux c, James Cleverly c, Derek Eamus c a

National Centre for Groundwater Research and Training, Flinders University, GPO Box 2100, Adelaide, SA 5001, Australia Water for a Healthy Country National Flagship, Commonwealth Scientific and Industrial Research Organisation, Division of Land and Water, Glen Osmond, Adelaide, SA 5064, Australia c National Centre for Groundwater Research and Training and School of the Environment, University of Technology Sydney, PO Box 123, Broadway, NSW 2007, Australia b

a r t i c l e

i n f o

Article history: Received 2 January 2015 Received in revised form 16 May 2015 Accepted 2 June 2015 Available online 7 June 2015 This manuscript was handled by Peter K. Kitanidis, Editor-in-Chief, with the assistance of Markus Tuller, Associate Editor Keywords: Evaporation Evapotranspiration Isotopes Playa Salt pan Groundwater discharge

s u m m a r y Evaporative losses typically play a substantial role in the water balances of arid regions. However, they are often poorly understood due to low flux rates and difficulty in direct measurement. We compared six field methods to quantify groundwater discharge due to evaporative and evapotranspirative fluxes from Stirling Swamp, a playa in central Australia; Bowen ratio–energy balance (BREB), maximum entropy production (MEP), chloride and stable isotope profiling, change in groundwater level, and 14C profiles within the aquifer. The latter method has not been previously used to determine groundwater discharge. Evaporative groundwater discharge estimates varied between 0 and 300 mm/y, partly due to variability in spatial and temporal scales captured by the individual methods. Within playa systems where evapotranspiration within the soil is negligible but the depth to groundwater is small, land surface energy balances were found to have the advantage of integrating over hundreds of metres, and when upscaled to annual estimates they agreed well with expected evaporative flux values. Soil profile methods yielded a wide range of results depending on the values of several constants that must be assumed, and the assumption of steady state was found to be a disadvantage. Groundwater methods also had the advantage of integrating over some distance within the aquifer; however, advective transport in the subsurface may have led to under-estimation of evaporative flux with these methods. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction Use of groundwater in arid regions is increasing worldwide, in response to a growing global population (Seely et al., 2003). To determine the impacts of this groundwater use on the long-term viability of the resource, as well as on the flora and fauna in these fragile arid ecosystems, accurate water budgets must be developed. Many arid environments are found in closed basins, where water lost to the combined processes of transpiration by plants and evaporation from bare surfaces can account for up to 95% of the total annual rainfall (Wilcox and Thurow, 2006). In playas or salt pans where water tables are shallow, diffuse discharge of groundwater is typically considered to be a large component of the water balance (Thorburn et al., 1992; Holland, 2002). However, evaporative loss through these features can be difficult ⇑ Corresponding author. Tel.: +61 8 82012193. 1

E-mail address: Margaret.shanafield@flinders.edu.au (M. Shanafield). Tel.: +61 8 82012193.

http://dx.doi.org/10.1016/j.jhydrol.2015.06.003 0022-1694/Ó 2015 Elsevier B.V. All rights reserved.

to quantify due to harsh conditions and low evaporative fluxes (Tyler et al., 1997). Due to these difficulties, relatively few field studies have been conducted in natural playa ecosystems to determine evaporative fluxes. Tyler et al. (1997) estimated mean groundwater evaporation from the playa surface at Owens Lake, California to be in the range of 88–104 mm/y using lysimeters and eddy correlation, but only 17–22 mm/y using chloride profiles. Malek et al. (1990) reported groundwater evaporation of 229 mm/y from a playa surface in eastern Utah. At the dry, saline bed of Lake Frome in South Australia, an early study using isotope and chloride profiles estimated an annual evaporation rate of 50–240 mm/y (Allison and Barnes, 1985). At the nearby dry, saline Lake Eyre, Ullman (1985) estimated a lower annual evaporation rate of only 9–28 mm/y using chloride and bromide soil profiles. Ullman (1985) attributed the low evaporative flux at Lake Eyre to the mulching effect and higher albedo associated with the salt crust at the soil surface. Also in this general region, Costelloe et al. (2014) estimated evaporation rates of approximately 7–79 mm/y using soil profiles. In a

1074

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

study of the soil water in the Sahara, Fontes et al. (1986) estimated even lower evaporation rates of 1–2 mm/y of evaporative flux from the groundwater. Australia represents the driest inhabited continent on earth, with more than 70% of the land area considered arid or semi-arid (James et al., 1999). The scarcity of water across vast areas of the Australian continent is a limiting factor for the development of human and economic activities and a permanent challenge to the equilibrium of fragile desert ecosystems. It has also provided the ideal setting for some seminal studies on dry lake or playa evaporation (Allison and Barnes, 1983, 1985; Ullman, 1985; Woods, 1991). As in other parts of the world, many of the dryland basins of inland Australia are being considered for further development. One example is the Ti Tree Basin north of Alice Springs in the Northern Territory of Australia. The basin is largely undeveloped pasture land; however, increased horticultural development is being encouraged. Previously developed water budgets for the basin suggest that groundwater is largely recharged in the southern and western margins of the basin, and that the water balance is regulated by evaporative fluxes through the Stirling Swamp, a sparsely vegetated playa at the northern extent of the basin (Harrington et al., 2002; DIPE, 2002). However, the magnitude of these fluxes has not been measured. In this study, we examine evaporative fluxes through Stirling Swamp using six different methods. The objective of the study is to compare well established techniques of measuring evaporation with novel techniques that have not previously been compared and to evaluate the usefulness of these methods for estimating groundwater discharge through evaporation at the Swamp. This analysis includes methods from three standpoints: land surface energy balances (Bowen ratio–energy balance and maximum entropy production methods), unsaturated soil profiles (chloride and stable isotope profile methods), and groundwater measurements (14C profiles within the aquifer and temporal groundwater level variation). The selection of methods offers a comparison of estimates covering a wide range of spatial and temporal scales, and highlights the advantages and disadvantages of each method for use in arid, playa environments.

2. Study area and field methods The Stirling Swamp is located north of the Tropic of Capricorn, at a latitude of 22.8°S and a longitude of 133.7°E. It forms the northern extent of the Ti Tree Basin, approximately 35 km north of the town of Ti Tree, in the Northern Territory, Australia (Fig. 1). Although labelled a ‘‘swamp’’, this feature would more typically be considered a salt pan or playa. Most of the area is covered with bare soil, with patches of low-lying vegetation appearing seasonally, including salt tolerant Frankenia species and Marsilea hirsute, a freshwater fern. Depth to groundwater is typically between 1 and 3 m, with soil and groundwater exhibiting very high salinity levels (average groundwater chloride concentration is 48 g/L). During the summer months the entire Swamp is typically inundated with water for some period of time, including all sites included in this study. There is no development on the Swamp and little or no regular groundwater pumping, although the land is used for cattle grazing. Basement rock (sandstone or conglomerate, often fractured in the top few metres) was observed in all bores drilled below 8–10 m, suggesting a total aquifer depth of approximately 10 m. About 1–1.5 m of compacted silt, very fine silty sand, and clay overlies sandstones and siltstones throughout the Swamp. There are no long-term precipitation records near the Swamp; however, depending on the station chosen and the period available for that station, long-term average precipitation estimates for areas

within the Ti Tree Basin range between 270 and 318 mm/y, (Harrington, 1999; DIPE, 2002; Fritz, 2007). Historical data at Territory Grape Farm, located approximately 75 km from Stirling Swamp, shows an average of 56 rain days per year over 124 years of record (1889–2013; Station 015643; BOM, 2014). Groundwater recharge has been estimated at between 0.1 and 500 mm/y, with a median of 0.9 mm/y (Harrington et al., 2002). Like many arid regions, precipitation is extremely variable spatially and temporally (Gutzler and Preston, 1997; Mudd, 2006). The basin falls within the low-latitude, desert climate classification, with the majority of precipitation (72–86%) typically occurring during storm events coming from the northeast in the hot summer months (Cleverly et al., 2013a,b). The area of the Swamp is approximately 10 km2, extending as a long, narrow tongue from the southeast to the northwest. It is bounded by sand dunes on the southwest side and low rises on the northeast side, both of which contribute local runoff during large rain events. The ephemeral Hanson River is located to the west, and a paleochannel of this distributary system may flow into the Swamp during very large rain events. To study the Swamp, a transect of 5 sites (labelled A–E in Fig. 1) was installed at approximately 1.5 km intervals. A Bowen ratio–energy balance (BREB) system was installed at site E on 10 April 2013. The setup consisted of differential temperature and relative humidity (Model HMP4C, Campbell Scientific, Logan, Utah), and water vapour measurements (Model Dew-10, General Eastern Corp, Watertown, MA) from sensors mounted 1 m and 2 m above ground surface. Mean soil heat flux at the soil surface was computed by adding the measured soil heat flux at a depth of 0.08 m with the energy stored in the layer above this measurement. This was achieved using two sets of thermocouples installed at depths of 0.02 and 0.06 m, and heat flux plates installed at 0.08 m (TCAV and HFP01 sensors, Campbell Scientific, Logan, Utah), as well as soil moisture probes installed at depths of 0.05 and 0.10 m (Theta probes, Campbell Scientific, Logan, Utah). Solar radiation was measured using a net radiometer (NR-LITE2, Campbell Scientific, Logan, Utah) mounted at a height of 1 m, and corrected for windspeed. All instruments were connected to a CR3000 data logger (Campbell Scientific, Logan, Utah) and set to record averages at 20 min intervals. After 4 September 2013, rainfall was recorded daily using a tipping bucket also connected to the datalogger. Vegetation near site E is sparse (short plants and grasses typically less than 30 cm tall and typically only active between December and April) or non-existent for at least 100 m in every direction (Fig. 2), creating uniform, almost bare soil conditions within a large fetch. These cover conditions are also typical of sites A–D. Cracking at the surface can be observed, especially at sites D and E, suggesting the prevalence of surface clays at these sites. This was also confirmed in drill core observations at the sites. A salt crust can appear seasonally on some parts of the Swamp. For the soil profiles, three soil cores were dug by hand using a hand-held auger on 13 June 2013 (sites A, C, and E; see Fig. 1), and one additional core was collected on 4 September 2013 (site D). Samples were sealed in air-tight containers and returned to the laboratory for analysis. Finally, five groups of four nested piezometers were installed in Stirling Swamp at sites A–E for the groundwater measurements. The piezometers were installed in 2012 (Fig. 1), with rotary air drilling, to depths of 3–12 m. The screen length of each piezometer was between 0.4 and 1.0 m, and depth below water at time of sampling ranged from 1.7 to 9.9 m (Table 1). Samples for both d13C and 14 C were collected using a submersible 12 V Whale pump (Whale; Bangor, Ireland) 26 June–1 July 2012. At this time the groundwater level was observed to decrease at a consistent slope from an elevation of 473.9 m above sea level at site E to 472.6 m above sea level at site A; a slope of 0.00036 across a direct line of 3.6 km between

1075

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

Fig. 1. Map of the study area, showing location of Ti Tree Basin (Google Earth imagery) within Australia, and location of Stirling Swamp (Google Earth imagery) and sampling sites within the Ti Tree Basin. The blue polygon indicates the approximate extent of the Ti Tree Basin aquifer; dashed arrows indicate the approximate direction of groundwater flow within the Basin.

Table 1 Bore data for Stirling Swamp sites. Each screened interval is located in a separate PVC piezometer nested within the bore. All depths are in metres below ground level.

Fig. 2. Vegetation near the BREB tower.

these sites. Vented pressure transducers (Insitu Troll 500, Ft. Collins, Colorado; accuracy 0.01 m) were installed in all

Ending depth

Site

Electrical conductivity (lS/cm)

Depth to water

Screened interval (s)

12.4 5.0 11.0 3.5 11.5 3.0 11.0 3.4 11.2 7.0

A A B B C C D D E E

105,000

2.3 2.3 1.1 1.1 1.4 1.4 1.2 1.2 0.8 0.8

6.06–6.46 4.3–4.7 4.8–5.2 2.4–2.8 4.72–5.53 1.9–2.4 3.6–4.2 2.2–2.7 8.5–8.9 2–3

106,400 129,000 142,000 130,000 126,000

9.5–9.9

11–11.5

7.1–8.1

9.1–9.6

6.6–7.0

9.7–10.2

6.0–7.0

9.9–10.4

10.3–10.8 5.5–6.5

piezometers on 29 June–2 July 2012. Water levels were measured at hourly intervals and were cross-checked against manual measurements collected every 3–6 months over the two year study period. As there was typically very good agreement between the transducer measurements and the manual readings, the transducer data was shifted but no slope adjustments were applied to match observed water levels. Based on a porosity of approximately 0.40 (as measured from soil samples) and observed water contents in the profile averaging approximately 0.20, a specific yield of 0.20 was considered appropriate.

1076

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

3. Theory

G¼

3.1. Land surface energy balance 3.1.1. Bowen ratio–energy balance (BREB) The BREB method solves for the energy balance within a unit area, given as:

Rn þ G þ H þ kE ¼ 0:

ð1Þ

1 BðrÞ Is HjHj6 r I0

ð6Þ

where Is and I0 are the thermal and apparent thermal inertia of the soil and air respectively, and B(r) is the reciprocal of the Bowen ratio calculated as:

! rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 11 1þ BðrÞ ¼ 6 r1 36

ð7Þ

The Bowen ratio, b, which is the ratio of sensible heat H to the latent heat kE, is obtained using differential measurements of temperature and humidity, as:

with a dimensionless parameter r characterizing the phase-change related state of the evaporating surface as:

H DT b¼ ¼c ; kE De

rðT s ; qs Þ ¼

ð2Þ

where c is the psychometric constant to account for units, and DT and De are the differences in temperature and humidity, respectively, measured between a lower and an upper height. If net radiation, Rn, and ground heat flux, G, are also measured, evaporation, E, can then be calculated as:

E¼

Rn  G : 1 þ cðDT=DeÞ

ð3Þ

To minimize error in evapotranspiration calculations, all data were subjected to the BREB failure criteria as described in Perez et al. (1999). Land surface energy balance methods such as BREB have the advantage of being robust with respect to their adaptability and ease of implementation in almost any type of terrain. The BREB method estimates ET not just at a point, but over a large fetch. This fetch is presumed to be 100 times the height of the sensors, although this distance–height ratio exceeds 100-to-1 over short vegetation (Leclerc and Thurtell, 1990) and when making measurements of scalars such as vapour density and temperature (Schmid, 1994). This spatially integrated estimate of ET is appropriate for playas or saltpans with relatively uniform cover. However, the large gaps introduced by the BREB failure criteria complicate attempts to integrate over time. In drylands, difficulties in using the BREB method may exist due to the typically small gradient in vapour pressure. Limited soil moisture in these regions can typically produce vapour pressure gradients during daytime that are significantly smaller than temperature gradients, hence resulting in large b and indicating that most of the energy from solar radiation escapes as sensible heat without contributing to evaporative fluxes (Perez et al., 1999). These small vertical gradients further impact the quality of BREB measurements by requiring a large separation between sensors to obtain detectable differences between levels and to overcome the resolution of the sensors (Cleverly et al., 2013a,b). 3.1.2. Maximum entropy production model of evapotranspiration Similar to BREB, the maximum entropy production model of evapotranspiration (MEP-ET) is an energy balance based method for calculating latent and surface heat fluxes (Wang and Bras, 2011). The MEP-ET method is an unconventional dynamic-statistical model built on non-equilibrium thermodynamics theory and the Max-Ent postulates (Dewar, 2005). Following Wang and Bras (2011), the MEP-ET model provides simultaneous solutions of H, G, and E, at a point (i.e. source) using only net radiation (Rn), and surface temperature (Ts), and specific humidity (qs), as inputs by solving the following:

Rn ¼ G þ H þ E

ð4Þ

E ¼ BðrÞH

ð5Þ

k2 qs cp Rv T 2s

ð8Þ

in which k is the vapourization heat of liquid water, Rv is the gas constant for water vapour, and cp is the specific heat of the air under constant pressure. Obtaining G with Eq. (6) requires parameterization of Is and I0 using soil property information and an extremum solution of the Monin–Obukhov similarity equations characterizing the mode of turbulent transport of water vapour and momentum in the boundary layer of the evaporating surface (see Wang and Bras, 2009 and 2011 for more details). Nevertheless, when actual measurements of G exist, the MEP-ET model simplifies greatly to:

E¼

Rn þ G BðrÞ BðrÞ þ 1

ð9Þ

Moreover, Gutierrez-Jurado et al. (2014) showed better constrained calculations of soil evaporation with the MEP-ET model using measured G values. In this work we used Rn and G from the BREB system described above, while qs and Ts were captured using iButton Hygrochron sensors (model DS1923, Maxim Integrated, San Jose, CA, temperature resolution 0.0625 °C, relative humidity 0.04%) suspended 5 cm above the soil surface within the footprint of the BREB as in Gutierrez-Jurado et al. (2014). The iButton sensors recorded data between 30 September and 29 October 2013. Because it uses several of the same parameters, the MEP method shares similar advantages with the BREB method. However, since the flux estimates in the MEP-ET method do not depend on the measurement of vapour and temperature gradients, the potential dryland ET estimation difficulties of the BREB method are avoided. Moreover, the MEP method has the large advantage of parsimonious data inputs; the relatively few data requirements of this method avoid the need for a large amount of expensive equipment in the harsh saline environment, where the presence of stock animals further complicated data collection. A drawback of the MEP-ET method is the point-scale nature of its estimates. This means that to obtain a representative estimate of the fluxes in the area, the location of the measurements should be carefully determined. Alternately, the low cost of the instruments required to implement this method may allow the deployment of extra temperature and humidity sensors (and optionally soil heat flux sensors) to characterize the spatial variability (Gutierrez-Jurado et al., 2014). 3.2. Unsaturated soil profiles Soil chloride and stable isotope profiles in areas of intense evaporation typically show enrichment of solutes near the soil surface, with concentrations decreasing with depth. This profile shape reflects the balance between the downward diffusive flux of solute and the upward advective flux of soil water (Ullman, 1985). Below the evaporation front, water movement occurs predominantly in the liquid phase, and the flux, F, can be written as a sum of the advective and diffusive processes as:

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

F ¼ qc þ D

@c @z

ð10Þ

where q is the evaporative driven flux of water, c is concentration of the solute (or isotope), and z is depth. D is often considered to be the diffusion coefficient of the solute in liquid pore water (Barnes and Allison, 1988), although it is clear that it should also include dispersion, particularly if the evaporative flux is high. At steady state, the solute flux is constant with depth (and equal to zero in the case of chloride). The solution of this equation is then (Barnes and Allison, 1988):

 q  c ¼ cres þ ðc0  cres Þ exp  z D

ð11Þ

where c is the solute concentration or isotope ratio at a given depth z (z = 0 at the evaporating front); cres and c0 are the solute concentrations or isotope ratios at depth in the soil and at the evaporating front, respectively, and:

D ¼ D0 hs þ aq

ð12Þ

where D0 is the free solution diffusion coefficient, h is the volumetric water content, s is the tortuosity (0 < s < 1) and a is dispersivity. Thus, if vertical profiles of chloride, 2H and 18O are obtained, and D and h are known, then q can be estimated. Within the soil profiles, matric potential was analysed using the filter paper method (Greacen et al., 1989), and gravimetric water content was obtained by oven drying approximately 45 g of sample for a minimum of 24 h at 105 °C (Fig. 3). Chloride content was measured on these dried samples colorimetrically, after extraction with distilled water, and chloride concentrations in soil water were calculated by dividing chloride content of dry soil by the gravimetric water content. Soil water for stable isotope analysis was extracted from the core samples using azeotropic distillation with kerosene (Revesz and Woods, 1990). The soil water was then analysed for 2 H and 18O using IRMS at the Reston Stable Isotope Laboratory (RSIL) of the U.S. Geological Survey using a hydrogen equilibration technique (Coplen et al., 1991; Revesz and Coplen, 2008), although only deuterium data are reported here. Evaporation was estimated from chloride and deuterium profiles assuming steady state had been reached. A free water diffusion coefficient of D0 = 1.6  109 m2/s was used for chloride (based on NaCl at 25 °C; Cook and Böhlke, 2000) and D0 = 1.45  109 m2/s was used for deuterium in saturated NaCl (Allison and Barnes, 1983). A vapour phase diffusion coefficient of 2.6  105 m2/s was also used for deuterium, together with a partition coefficient between liquid water and vapour of 4.3  105 m3/g at 25 °C (Cook and Böhlke, 2000). Gravimetric

1077

water content was multiplied by an assumed density of 1500 kg/m3 to approximate volumetric water content. Although the use of normalised depth functions have been proposed to correct for variations in water content and diffusion coefficient with depth (Barnes and Allison, 1983; Woods, 1991; Costelloe et al., 2014), we have not used these because they did not significantly improved the model fits. The tortuosity was assumed to be equal to the volumetric water content. The dispersivity was assumed to be a = 0.005 m (Gelhar et al., 1992). The use of chloride and stable isotope profiles has the advantage of integrating over a long time scale but represents only a point estimate of ET. This method has received a majority of the attention in studies characterizing bare soil or playa lake evaporation. In a comprehensive review of the influence of many factors on the estimates of evaporation provided by solute or isotope profiles, Barnes and Allison (1988) provide a summary of the method’s advantages for use in unsaturated systems, including the ability to linearize flow equations and the stability of diffusivity estimates despite variability in soil water content.

3.3. Groundwater measurements 3.3.1. Carbon-14 14 C is produced in the upper atmosphere by cosmic radiation and decays with a half-life of approximately 5730 years. Profiles of 14C in the aquifer therefore reflect a balance between downward transport by advection and diffusion, and loss by radioactive decay. Although 14C is a well-established tracer for dating groundwater and estimating basin recharge, we are not aware of previous applications in estimating groundwater discharge. In the absence of recharge, exchange of CO2 at the water table and downwards diffusion of dissolved 14C is responsible for the presence of 14C in the groundwater (Walker and Cook, 1991). However, as discharge rate increases, there is less 14C in the groundwater because it must diffuse downwards against the upward flow of water (Walker and Cook, 1991). This method has the advantage of providing a very long-term average evaporative flux estimate, which is spatially integrated along a groundwater flow line. Scanlon et al. (2002) suggest that estimates of recharge derived from 14C data may integrate over spatial scales of 1–100 km. Walker and Cook (1991) presented an analytical solution for 14C depth profiles in a two-dimensional flow field as a function of groundwater recharge or discharge rate. Assuming steady state conditions, negligible longitudinal diffusion, a no flux boundary condition at the bottom of the aquifer and constant concentration

Fig. 3. Matric potential (red) and gravimetric water content (black) measured in soil cores collected at Stirling Swamp sites (Fig. 1) 13 June 2013 (A, C, and E) and 4 September 2013 (D). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

1078

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

at the top of the aquifer (C = Co at z = zo), the mass balance for 14C concentration as a function of depth in the aquifer can be written:

0¼

R @C @2C ðH  zÞ þ D 2  khC H @x @z

ð13Þ

where R is the recharge rate, C is the 14C concentration in the groundwater, z is the depth within the aquifer, H the aquifer thickness, D the transverse dispersion coefficient (see Eq. (12)), h is the aquifer porosity, k is the radioactive decay constant (k = 1.21  104/y). Although this ignores the elevated atmospheric activity of 14C that has occurred during the past fifty years, this period of time is relatively small relative to 14C decay and so can be ignored in discharge or low recharge environments. This mass balance is then solved by series solution as:

C¼

N X C i ð1  z Þi

ð14Þ

fluctuates seasonally (or over several years) between recharge and discharge, then it is possible to estimate the discharge rate from the rate of water table decline over time multiplied by the specific yield. In the absence of vegetation, this method provides a sort of longterm water balance, with recharge by precipitation balancing discharge by evaporative flux. Although this method will underestimate the discharge rate due to inflow of groundwater from surrounding areas, the degree of underestimation will be small if the area of discharge is large. The spatial scale of these estimates will be related to the aquifer transmissivity, and is likely to be several tens of metres.

4. Results 4.1. Land surface energy balance

i¼0

where

C iþ2 ¼

 C i ð1  iR Þ  DT ði þ 1Þði þ 2Þ

ð15Þ

and z*, R*, and DT* are non-dimensionalized depth, recharge, and transverse diffusion, respectively, given as z* = z/H, R* = R/(hkH), and DT* = DT/(hkH2). Water samples from the nested piezometers were analysed for 14 C and d13C in dissolved inorganic carbon at Rafter Radiocarbon (Institute of Geological and Nuclear Sciences Ltd, New Zealand) using accelerator mass spectrometry. For the 14C simulations, Co was set at 100 pMC for the top of the aquifer, h was fixed at an estimated porosity of 0.40 (based on observed water contents measured from soil samples), tortuosity was 0.66 (Penman, 1940), the depth of the aquifer (H) was fixed at 10 m, the free water diffusion coefficient for carbonate in solution was assumed to be D0 = 2.2  1010 m2/s and dispersivity was a = 0.5 (Walker and Cook, 1991). Vertical profiles of 14C at each study site were then fit by iteratively calibrating R. 3.3.2. Temporal groundwater level variation Following the method of White (1932), ET can be calculated from diurnal patterns of groundwater level change. This method is based on the assumption that vegetation extracts groundwater predominantly during daylight hours. This causes the water table to fall during the day and then recover during the night-time hours when uptake of groundwater is much reduced. This daily pattern creates a sinusoidal signal in the local groundwater level that has been widely used to predict phreatophytic ET (White, 1932; Loheide et al., 2005; Lautz, 2008). Change in storage is measured from daily drawdown multiplied by specific yield. If the system

To produce accurate estimates of latent and sensible heat fluxes, the BREB relies on a well-defined ratio of the vapour pressure (De) and temperature (DT) gradients. When either gradient becomes too small or too large with respect to the other, unrealistic estimates of E and H result. Similarly, when De is zero, Eq. (3) is undefined. In Stirling Swamp, this was found to occur during the short lag period between the onset and cessation of solar radiation and the corresponding DT response. Analysis of the BREB data showed that approximately one-third of the data failed the quality control criteria suggested by Perez et al. (1999). These failures occurred when Rn  G > 0 but De < 0 (7.8% of data), and when Rn  G < 0 but De > 0 (26.6% of data). The majority of these conditions occurred at sunrise/sunset and during night hours, with only 5% during the daylight hours (9AM–7PM) failing the criteria. Therefore, the data were typically good during the periods where ET was expected to be largest. After this quality control, daily estimates of ET were produced summing all twenty-minute estimates for each day (Fig. 4A). Excluding periods of rainfall and approximately 10 days thereafter, when high moisture contents near the soil surface result in higher ET, the baseline ET from groundwater discharge over the entire dry period April–November varied between approximately 1.3 and 4.6 mm/d, with an average of 1.8 mm/d. The 10 days after-rainfall exclusion periods are based on empirical evidence at the site. Concurrent time-lapse photographs of the soil surface at the location of the BREB and MEP-ET setups show complete desiccation of the topsoil 8–10 days after significant wetting events (Fig. 5). These time windows for rainfall–moisture sourced ET are consistent with time scales of soil moisture residence times after large rainfalls reported in other studies in arid and semi-arid areas (Kurc and Small, 2004; Gutiérrez-Jurado et al., 2013). Nevertheless, to upscale this baseline daily ET estimate into an annual flux due to groundwater

Fig. 4. (A) Evapotranspiration rates calculated from BREB method. No data were recorded for the period 13 November–14 December 2013. 24 h rainfall measured at an eddy covariance tower approximately 35 km from Stirling Swamp for 1 April–1 September 2013, and at the BREB tower from 1 September 2013–12 February 2014. Data from the shaded period 9/5/2013–10/27/13 are shown in (B), comparing the BREB and MEP-ET methods.

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

1079

Fig. 5. Time-lapse photograps taken from the BREB tower showing conditions on (A) 18 January 2014 at 9:00 AM immediately following a rain event and (B) 28 January 2014 at 9:00 AM.

Table 2 Summary of groundwater discharge estimates from six evaporation/evapotranspiration methods. Daily evaporation estimates obtained with BREB and MEP and groundwater level methods have been scaled to annual discharge assuming 125 days per years of groundwater discharge. Soil profile estimates assume a dispersivity of 0.005 m; bounds reflect order of magnitude uncertainty in these parameters. Estimated average annual groundwater discharge (mm/y) Method BREB & MEP Chloride profiles Stable isotope profiles 14 C aquifer profiles Groundwater level

Site A

Site B

2 (2–20) 3.3 25

Site C

4.2. Unsaturated soil profiles Site D

Site E

4 (2–20)

225 30 (10–300) 11 (9–15)

4.6 25

2.7 25

3.5 (2–20) 0.1 63

50

slight dips in the daily ET estimates by the MEP method, as opposed to the spurious spikes observed in the daily BREB estimates. Overall, the estimated evaporation using the land surface energy balance methods was approximately two-thirds of the available energy (Rn  G).

discharge, considerations regarding the average number of days with rain must be taken into account. Historical data from Territory Grape Farm were grouped into an average of 24 rainfall events per year over 124 years of record (1889–2013; Station 015643; BOM, 2014). Assuming each time a rainfall event occurs, ET is predominantly a result of moisture from the unsaturated zone instead of upwards flux of groundwater for 10 days, and given an average daily ET rate of 1.8 mm/d calculated from the April– November BREB data, this yields an approximate annual rate of 225 mm of groundwater discharge due to ET (Table 2). This is, however, a conservative estimate based on the mean baseline ET rate at the driest time of the year, when all latent heat fluxes in the swamp are sourced from the shallow groundwater. The MEP-ET results closely match those from BREB (Fig. 4B), suggesting that BREB was estimating groundwater discharge fluxes (Table 2). Notable discrepancies between the results occur during periods of rainfall, when spurious spikes are typically observed in the BREB results and dips in the MEP results (Fig. 4B). As examples, 0.2 mm was recorded on 15 September, 7 mm of rain was recorded in the three days 28–30 September, and 0.4 was recorded on 15 October, but each event led to similar behaviour. Our results show that the inconsistencies in the BREB method observed by Ohmura (1982) did not affect the MEP method; that is, rain events caused

Gravimetric water contents measured to a depth of 0.1 m in the unsaturated soil ranged between 0.005 and 0.30 g/g, with the lowest values in each profile occurring within the upper 0.02 m. Matric potentials were also lowest immediately below the soil surface, with values between 117 and 92 MPa at 0.01–0.02 m depth (Fig. 3). Matric potential gradients clearly indicated upward water movement at all sites. Chloride concentrations in soil water were 4–372 g/L, while concentrations in groundwater samples collected 1–10 m below the water table were 11–59 g/L (Fig. 6). Soil profiles D and E both displayed high chloride concentrations immediately below the soil surface, consistent with steady state evaporation. Soil profile E also showed enrichment of deuterium in the upper 0.3–0.4 m, with the most enriched value at 0.04 m depth (Fig. 7, deuterium was not measured on profile D). In contrast profiles A and C showed enrichment in deuterium but no evaporative concentration of chloride. The lack of chloride enrichment at these sites presumably indicates a very low rate of evaporation, or that evaporation has only been occurring for a relatively short period of time (insufficient for significant chloride accumulation), and so profiles were not in steady state. It is noteworthy that profile A displayed the lowest water contents of all sites (<0.01 g/g above 0.1 m); this low water content soil at the surface might act like a mulch and reduce evaporative loss at this site. The total chloride mass in the upper 0.5 m of each profile was 1140, 2980, 4710 and 8150 g/m2 at sites A, C, D and E, respectively. The evaporating front at each site was assumed to be at 0.04 m depth, determined as the depth of the maximum (most enriched) deuterium value at all sites. Estimates of evaporation vary depending on the value of dispersivity used. For a = 0.005 m, evaporation rates for sites D and E were estimated to be 4 mm/y and 30 mm/y, respectively, based on chloride profiles. However, if we consider an order of magnitude uncertainty for dispersivity and factor-of-two

1080

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

Fig. 6. Chloride concentrations measured in four soil cores collected at Stirling Swamp 13 June 2013 (A, C, and E) and 4 September 2013 (D).

Fig. 7. Deuterium values measured in soil cores collected at Stirling Swamp 13 June 2013 (A, C, and E).

uncertainty for tortuosity, we get evaporation estimates between approximately 2 and 20 mm/y at site D and between 10 and 300 mm/y at site E. Based on the total chloride in the upper 0.5 m of each profile, and a mean chloride concentration in groundwater of 48 g/L, profiles D and E represent a cumulative evaporation of 94 and 160 mm, respectively (ignoring back diffusion of chloride). At the discharge rate obtained from the meteorological methods (200 mm/y), this would take less than one year. Fluxes were also calculated from the three deuterium profiles. Using a = 0.005 m, we get evaporation rates of 2, 3.5, and 11 mm/y for profiles A, C and E, respectively. Considering uncertainty in dispersivity of 0.001 m < a < 0.02 m, gives 2 < E < 20 mm/y (Fig. 7). Note that the lowest flux from the three 2 H profiles (A) also showed no chloride enrichment. 4.3. Aquifer profiles of 14

14

C

C values were between 56.0 and 107.5 pMC across all sites and generally decreased with depth below the water table (Fig. 8). Sites B and C had 14C values greater than 100 (104.1 and 107.5 pMC, respectively) within 2 m of the water table. At sites A and C, 14C values were higher approximately 9 m below the water table than those measured at shallower depths within the aquifer. Overall, the 14C concentrations observed across all sites fit within expected values for evaporation rates on the order of 0–12 mm/y (Fig. 8). When individual profiles were fit to the model, fluxes of 2.7–4.6 mm/y were estimated, with an average of 1.6 mm/y. Only the profile at E fit an exponential decay curve well; this site was fit with a discharge of 2.7 mm/y and was the only site where the model suggested recharge. Values for d13C were between 3.8‰ and 19.6‰, with an average value of 7.4‰. The values

Fig. 8. 14C levels measured in water samples collected from nested piezometers at Stirling Swamp 26 June–1 July 2012. Each line represents a modelled groundwater discharge (mm/y, marked on the lines) in which negative values indicate recharge.

observed at Stirling Swamp were generally consistent with values observed throughout the basin (Wood et al., 2014), and indicated that water–rock interaction was not a significant factor for carbon isotopes. 4.4. Groundwater level fluctuations Examination of hourly vented pressure transducer data from 20 shallow, nested piezometers at sites A–E showed an unexpected pattern of two periods of drawdown each day (Fig. 9), making the diurnal groundwater fluctuation method unreliable for Stirling Swamp. These patterns are more related to atmospheric pressure forcing than local groundwater use by vegetation, and are the subject of further research. However, approximately two years of water level data are available at the Stirling Swamp piezometers, making it possible to discern both daily and long-term, seasonal patterns in groundwater level change (Fig. 10). Long-term (24–159 day), linear declines in the water levels within the piezometers were therefore examined to estimate average evaporative and evapotranspirative fluxes from the Swamp. During the first year of measurements (2012–2013), significant rainfall events were few, and from the groundwater level response, it is evident that only two periods of recharge occurred; on 15 December 2012 and 14 May 2013, although a slight rise (by 0.01 m) in the water level was observed at site E between July

1081

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

evaporative groundwater discharge occurred per year, the average daily flux of 0.3 mm/day yielded an average annual rate of 37.5 mm/y. When long-term trends were examined (ie decline over the entire period of record), yearly water table decline, accounting for periods of recharge, was approximately 96 mm/y averaged over the two-year period, but varied from 32.7 to 200.4 mm/y across the five sites.

5. Discussion

Fig. 9. Depth to water (metres below ground) for weekly intervals starting on the days listed for a period of approximately five months at site E. The black dashed line denotes a hypothetical pattern expected by the White (1932) method of calculating ET.

Fig. 10. Contrasting groundwater level patterns over the two year study period at sites A and B.

and the beginning of August 2012 (Fig. 9). In the second year, a small recharge event occurred on 25 December 2013, followed by a large event commencing on 18 January 2014. Following these recharge events, the groundwater table rose quickly for several days and then declined, returning to pre-event levels in approximately 3 months in 2012/2013 and 5 months following the larger event in 2013/2014. These events were not felt evenly throughout the Swamp; there was large spatial variation in the magnitude of the response, even across a relatively small distance (Fig. 10). During the dry periods, the daily evaporative flux ranged across the five sites monitored, from a minimum of 0.2 mm/day to a maximum of 0.5 mm/day (Table 3), but was fairly consistent over the study period at each individual site resulting in an average evaporative flux of 0.3 mm/day (assuming specific yield is 0.2). Assuming, as in the BREB analysis, that on average 125 days of

The use of multiple methods for estimating evaporation at Stirling Swamp provided a range of flux estimates (Table 2). This can partly be attributed to differences in the spatial and temporal resolution of the methods (Table 3), and partly to variability between the sites, which illustrates the complex hydrology of the Swamp, where both recharge and discharge processes occur. Assuming an aerial extent of 10 km2, evaporative groundwater loss through the Swamp was estimated at 10–3000 ML/y across all methods (including uncertainty). If, as has been supposed in past modelling efforts, Stirling Swamp is the primary hydrological endpoint for the basin, a median estimate of basin recharge of 0.9 mm/y (Harrington et al., 2002) over a basin area of 5500 km2 indicates that a recharge of at least 4950 ML/y would be expected. Following this rationale, land surface energy balance methods are most in line with expected values. BREB and MEP-ET methods closely paralleled each other (Fig. 4B) on a daily basis and provided consistent estimates of groundwater fed evaporative fluxes during the dry season (in this case the only season when evaporative groundwater fluxes are expected to occur). Land surface energy balance methods measure total evapotranspiration, rather than groundwater discharge. To estimate groundwater discharge requires either a period of time in which evapotranspirative loss from soil can be assumed to be negligible, or other methods for differentiating between these sources (Yamanaka and Shimizu, 2007). Therefore, care should be taken when using these methods in arid regions with deeper water tables or more vegetation. Further, as the methods typically yield measurements over the course of one day, measurements collected only over short periods but upscaled to account for annual estimates may be unrealistic. When uncertainty in tortuosity and dispersivity are considered, the possible range of evaporative flux for one of the chloride profiles spans the evaporation estimate obtained from the land surface energy balances. However, the other profiles indicate lower evaporation rates, and all three deuterium profiles indicated lower evaporation rates than the land surface energy balances. The major drawbacks of using soil profiles are the need to assume steady state and uncertainty in parameters such as dispersivity and tortuosity. Following Barnes and Allison (1988), the characteristic time

Table 3 Comparative considerations of each evaporative flux method. Method

Land surface energy balances

Soil profiles

BREB

MEP

Chloride

Stable isotopes

Cost Requirements

High Bowen ratio system

Low Soil core, chloride analyses

Medium High Soil core, stable Monitoring wells, 14 isotope C analysis analyses

Most uncertain aspect

Vapour pressure gradient, upscaling in time Hourly to daily Fetch (100 m)

Low Net radiation, surface temperature, and specific humidity measurements Upscaling to annual fluxes

Temporal scale Spatial scale a

Low if monitoring wells already exist.

Point

Groundwater measurements

Time to reach steady state, dispersivity, upscaling in space Several years Point Point

14

C aquifer profiles

Aquifer thickness, porosity variations Thousands of years Aquifer flow lines (1–100 km)

Groundwater level Mediuma Monitoring wells; water level loggers

Groundwater inflow unknown Daily to yearly 10–100 m

1082

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083

for development of steady state evaporation from a profile, tc, will be:

that the evaporative flux within the basin is larger than the recharge from the regional aquifer.

t c ¼ D=E2

6. Conclusions

ð16Þ

Using this equation, evaporative fluxes of 2–300 mm/y suggest chloride characteristic times of 2.2–28 years (although as mentioned in the results, cumulative chloride concentrations in the soil profiles suggest shorter times are also possible). However, this equation is only valid when the profile is in steady state, making it difficult to actually estimate the time needed to achieve steady state conditions for this method. Allison et al. (1984) showed that it is possible to obtain profiles from soil columns that had undergone intermittent recharge but showed similar shapes to those expected under steady state conditions. However, they cautioned against the validity of evaporative flux estimates from these profiles because the evaporation rate will be under-estimated if the method is applied to profiles that are not in steady state. Although Allison and Barnes (1985) reported similar evaporation estimates using several chloride and isotope profile analyses, other authors have found discrepancies between evaporation rates calculated from chloride profiles and from various other techniques. Tyler et al. (1997) suggested that the low evaporation estimates from chloride profiles provided a minimum groundwater discharge estimate due to osmotic and thermal fluxes in the shallow profile that were not accounted for. Costelloe et al. (2014) reported that chloride profile estimates of evaporative flux were much lower than those from isotope profiles, and attributed this difference to the use of a constant tortuosity value that was more appropriate for isotopes than chloride. Groundwater methods have the advantage of integrating over some area of the aquifer. However, in the case of 14C this means that the method potentially integrates aquifer processes over thousands of years. Using a simple decay equation, the observed 14C values in the Swamp suggest groundwater ages of 0–3090 years. From Fig. 8, it is clear that low levels of modern carbon would be found at depth in the aquifer at evaporation rates greater than approximately 20 mm/y. Hence, the much higher levels of modern carbon observed suggest that either the historic evaporation rates were much lower; or more likely, that young water has been mixed into the aquifer. Further, the presence of younger water at depth within some of the profiles suggests either convective overturning within the shallow aquifer, or perhaps that the spatial scale of measurement is greater than the size of the playa. It is also possible that playa groundwater receives input of freshwater recharge from the dunes and rises, which then incompletely mixes with the older, saltier water flowing from the Ti Tree Basin through the Swamp. Similar dynamics were found in a playa system in Nevada, USA, where observed groundwater discharge by ET was primarily through vegetated surfaces, with low rates of bare soil playa ET stemming only from localized, ephemeral soil moisture and contributing a small fraction of overall valley groundwater discharge by ET (Garcia et al., 2015). These processes complicated efforts to use this method in Stirling Swamp and likely resulted in underestimation of modern evaporative fluxes. The use of groundwater levels provided a simple and in this case, cost-effective method of predicting evaporative fluxes. As shown by Sanderson and Cooper (2008), care must be taken using this method, as small differences in water table decline may actually signify larger changes in evaporative groundwater discharge. As mentioned previously, this method likely underestimated the true evaporative flux in the Swamp due to inflow of groundwater from surrounding areas. If Stirling Swamp is the discharge point of the greater Ti Tree Basin, it is expected that flowthrough would be continually recharging the Stirling Swamp groundwater levels. However, from the falling groundwater levels, it can be determined

 Land surface energy balance methods measure total evapotranspiration, rather than groundwater discharge. To estimate groundwater discharge requires either a period of time in which evapotranspirative loss from soil can be assumed to be negligible, or other methods for differentiating between these sources. In arid areas, the method will therefore be most successful when the water table is shallow (soil depth is small).  The chloride and stable isotope profile methods rely on an assumption of steady state. However, it is difficult to determine whether this assumption is valid, and if it is not then the flux will be underestimated.  The 14C method has not been previously used, and has potential for estimating long-term discharge. However, the possibility of advective transport caused by salinity-induced density gradients requires further consideration. This may cause discharge rates to be underestimated, or even for negative discharge (ie recharge) to be estimated.  The diurnal water table fluctuation method did not work at our site, as there was no clear diurnal water table pattern.  Discharge rates estimated from rates of groundwater table decline during periods without rainfall allow a minimum estimate of groundwater discharge to be obtained.

Acknowledgments Funding for this research was provided by the National Centre for Groundwater Research and Training, an Australian Government initiative, supported by the Australian Research Council and the National Water Commission. This research was also supported by funding from the National Collaborative Research Infrastructure Scheme. The authors would like to thank the owners of Pine Hill Station and Stirling Station. Nick White, Stephanie Villeneuve, Cameron Wood, and Randol Villalobos-Vega provided invaluable field assistance. References Allison, G., Barnes, C., 1983. Estimation of evaporation from non-vegetated surfaces using natural deuterium. Allison, G., Barnes, C., 1985. Estimation of evaporation from the normally ‘‘dry’’ Lake Frome in South Australia. J. Hydrol. 78 (3), 229–242. Allison, G., Barnes, C., Hughes, M., Leaney, F., 1984. Effect of climate and vegetation on oxygen-18 and deuterium profiles in soils. Isotopes Hydrology. Isotopes Hydrol. IAEA, Vienna, 105–122. Barnes, C., Allison, G., 1983. The distribution of deuterium and 18 O in dry soils: 1. Theory. J. Hydrol. 60 (1), 141–156. Barnes, C., Allison, G., 1988. Tracing of water movement in the unsaturated zone using stable isotopes of hydrogen and oxygen. J. Hydrol. 100 (1), 143–176. Cleverly, J., Boulain, N., Villalobos-Vega, R., Grant, N., Faux, R., Wood, C., Cook, P.G., Yu, Q., Leigh, A., Eamus, D., 2013a. Dynamics of component carbon fluxes in a semi-arid Acacia woodland, central Australia. J. Geophys. Res.: Biogeosci. 118 (3), 1168–1185. Cleverly, J., Chen, C., Boulain, N., Villalobos-Vega, R., Faux, R., Grant, N., Yu, Q., Eamus, D., 2013b. Aerodynamic resistance and Penman-Monteith evapotranspiration over a seasonally two-layered canopy in semiarid central Australia. J. Hydrometeorol. 14 (5), 1562–1570. Cook, P.G., Böhlke, J.-K., 2000. Determining timescales for groundwater flow and solute transport. Environmental Tracers in Subsurface Hydrology. Springer, pp. 1–30. Coplen, T.B., Wildman, J.D., Chen, J., 1991. Improvements in the gaseous hydrogen– water equilibration technique for hydrogen isotope-ratio analysis. Anal. Chem. 63 (9), 910–912. Costelloe, J., Irvine, E., Western, A., 2014. Uncertainties around modelling of steadystate phreatic evaporation with field soil profiles of d18O and chloride. J. Hydrol. 511, 229–241. Dewar, R.C., 2005. Maximum entropy production and the fluctuation theorem. J. Phys. A: Math. Gen. 38 (21), L371.

M. Shanafield et al. / Journal of Hydrology 527 (2015) 1073–1083 DIPE, 2002. Ti-Tree Region Water Resource Strategy. Northern Territory Government Department of Infrastructure, Planning and Environment (DIPE). Fontes, J.C., Yousfi, M., Allison, G., 1986. Estimation of long-term, diffuse groundwater discharge in the northern Sahara using stable isotope profiles in soil water. J. Hydrol. 86 (3), 315–327. Fritz, L., 2007. In: Read, R.T. (Ed.), The Ti-Tree Basin Aquifer. Hydrogeology, Department of Natural Resources, Environment and the Arts, Northern Territory Government, Australia. Garcia, C.A., Huntington, J.M., Buto, S.G., Moreo, M.T., Smith, J.L., Andraski, B.J., 2015. Groundwater Discharge by Evapotranspiration, Dixie Valley, West-Central Nevada, March 2009–September 2011. U.S.D. o.t. Interior, U.S. Geological Survey, 89. Gelhar, L.W., Welty, C., Rehfeldt, K.R., 1992. A critical review of data on field-scale dispersion in aquifers. Water Resour. Res. 28 (7), 1955–1974. Greacen, E.L., Walker, G.R., Cook, P.G., 1989. Procedure for filter paper method of measuring soil water suction, CSIRO. Gutiérrez-Jurado, H.A., Vivoni, E.R., Cikoski, C., Harrison, J.B.J., Bras, R.L., Istanbulluoglu, E., 2013. On the observed ecohydrologic dynamics of a semiarid basin with aspect-delimited ecosystems. Water Resour. Res. 49 (12), 8263–8284. Gutierrez-Jurado, H., Guan, H., Wang, J., Wang, H., Bras, R.L., Simmons, G., 2014. Monitoring evapotranspiration and its partitioning over heterogeneous terrain and vegetation cover using the MEP model. Geophys. Res. Lett. Gutzler, D.S., Preston, J.W., 1997. Evidence for a relationship between spring snow cover in North America and summer rainfall in New Mexico. Geophys. Res. Lett. 24 (17), 2207–2210. Harrington, G., 1999. Recharge Mechanisms and Chemical Evolution in An Arid Groundwater System, Central Australia. Flinders University of South Australia (Doctor of Philosophy). Harrington, G.A., Cook, P.G., Herczeg, A.L., 2002. Spatial and temporal variability of ground water recharge in central Australia: a tracer approach. Groundwater 40 (5), 518–527. Holland, K.L., 2002. Blackbox (Eucalyptus largiflorens) Tree Health and Groundwater Discharge on the Lower River Murray Floodplain, South Australia: Linking Spatial and Temporal Scales. Flinders University, School of Chemistry, Physics and Earth Sciences. James, C.D., Landsberg, J., Morton, S.R., 1999. Provision of watering points in the Australian arid zone: a review of effects on biota. J. Arid Environ. 41 (1), 87–121. Kurc, S.A., Small, E.E., 2004. Dynamics of evapotranspiration in semiarid grassland and shrubland ecosystems during the summer monsoon season, central New Mexico. Water Resour. Res. 40 (9). Lautz, L.K., 2008. Estimating groundwater evapotranspiration rates using diurnal water-table fluctuations in a semi-arid riparian zone. Hydrogeol. J. 16 (3), 483– 497. Leclerc, M., Thurtell, G., 1990. Footprint prediction of scalar fluxes using a Markovian analysis. Bound.-Layer Meteorol. 52 (3), 247–258. Loheide, S.P., Butler, J.J., Gorelick, S.M., 2005. Estimation of groundwater consumption by phreatophytes using diurnal water table fluctuations: a saturated–unsaturated flow assessment. Water Resour. Res. 41 (7). Malek, E., Bingham, G.E., McCurdy, G.D., 1990. Evapotranspiration from the margin and moist playa of a closed desert valley. J. Hydrol. 120 (1), 15–34.

1083

Mudd, S.M., 2006. Investigation of the hydrodynamics of flash floods in ephemeral channels: scaling analysis and simulation using a shock-capturing flow model incorporating the effects of transmission losses. J. Hydrol. 324 (1), 65–79. Ohmura, A., 1982. Objective criteria for rejecting data for Bowen ratio flux calculations. J. Appl. Meteorol. 21 (4), 595–598. Penman, H., 1940. Gas and vapour movements in the soil: I. The diffusion of vapours through porous solids. J. Agric. Sci. 30 (03), 437–462. Perez, P., Castellvi, F., Ibanez, M., Rosell, J., 1999. Assessment of reliability of Bowen ratio method for partitioning fluxes. Agric. For. Meteorol. 97 (3), 141–150. Revesz, K.M., Coplen, T.B., 2008. Determination of the [delta](2H/1H) of Water, RSIL Lab Code 1574. Revesz, K., Woods, P.H., 1990. A method to extract soil water for stable isotope analysis. J. Hydrol. 115 (1), 397–406. Sanderson, J.S., Cooper, D.J., 2008. Ground water discharge by evapotranspiration in wetlands of an arid intermountain basin. J. Hydrol. 351 (3), 344–359. Scanlon, B.R., Healy, R.W., Cook, P.G., 2002. Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeol. J. 10 (1), 18–39. Schmid, H., 1994. Source areas for scalars and scalar fluxes. Bound.-Layer Meteorol. 67 (3), 293–318. Seely, M., Henderson, J., Heyns, P., Jacobson, P., Nakale, T., Nantanga, K., Schachtschneider, K., 2003. Ephemeral and endoreic river systems: Relevance and management challenges. Transboundary rivers, sovereignty and development: hydropolitical drivers in the Okavango River basin, Pretoria. Thorburn, P.J., Walker, G.R., Woods, P.H., 1992. Comparison of diffuse discharge from shallow water tables in soils and salt flats. J. Hydrol. 136 (1), 253–274. Tyler, S., Kranz, S., Parlange, M., Albertson, J., Katul, G., Cochran, G., Lyles, B., Holder, G., 1997. Estimation of groundwater evaporation and salt flux from Owens Lake, California, USA. J. Hydrol. 200 (1), 110–135. Ullman, W.J., 1985. Evaporation rate from a salt pan: estimates from chemical profiles in near-surface groundwaters. J. Hydrol. 79 (3), 365–373. Walker, G., Cook, P., 1991. The importance of considering diffusion when using carbon-14 to estimate groundwater recharge to an unconfined aquifer. J. Hydrol. 128 (1), 41–48. Wang, J., Bras, Rafael L., 2009. A model of surface heat fluxes based on the theory of maximum entropy production. Water Resour. Res. 45 (11). Wang, J., Bras, R., 2011. A model of evapotranspiration based on the theory of maximum entropy production. Water Resour. Res. 47 (3). White, W.N., 1932. A Method of Estimating Ground-Water Supplies Based on Discharge by Plants and Evaporation From Soil: Results of Investigations in Escalante Valley. US Government Printing Office, Utah. Wilcox, B.P., Thurow, T.L., 2006. Emerging issues in rangeland ecohydrology: vegetation change and the water cycle. Rangel. Ecol. Manage. 59 (2), 220–224. Wood, C., Cook, P.G., Harrington, G.A., Meredith, K., Kipfer, R., 2014. Factors affecting carbon-14 activity of unsaturated zone CO2 and implications for groundwater dating. J. Hydrol. 519, 465–475. Woods, P.H., 1991. Evaporative discharge of groundwater from the margin of the Great Artesian Basin near Lake Eyre, South Australia. Yamanaka, T., Shimizu, R., 2007. Spatial distribution of deuterium in atmospheric water vapor: diagnosing sources and the mixing of atmospheric moisture. Geochim. Cosmochim. Acta 71 (13), 3162–3169.