Long-term annual groundwater storage trends in Australian catchments

Advances in Water Resources 74 (2014) 156–165

Contents lists available at ScienceDirect

Advances in Water Resources journal homepage: www.elsevier.com/locate/advwatres

Long-term annual groundwater storage trends in Australian catchments Lu Zhang a,⇑, Wilfried Brutsaert b, Russell Crosbie c, Nick Potter a a

CSIRO Water for a Healthy Country National Research Flagship, CSIRO Land and Water, Christian Laboratory, Canberra, ACT 2601, Australia School of Civil and Environmental Engineering, Cornell University, Ithaca, NY 14850, USA c CSIRO Water for a Healthy Country National Research Flagship, CSIRO Land and Water, Glen Osmond, SA 5064, Australia b

a r t i c l e

i n f o

Article history: Received 27 December 2013 Received in revised form 1 September 2014 Accepted 3 September 2014 Available online 16 September 2014 Keywords: Base flow Groundwater storage Climate change Trends Groundwater level

a b s t r a c t The period of direct groundwater storage measurements is often too short to allow reliable inferences of groundwater storage trends at catchment scales. However, as groundwater storage sustains low flows in catchments during dry periods, groundwater storage can also be estimated indirectly from daily streamflow based on hydraulic groundwater theory; this idea was applied herein to 17 selected Australian catchments to examine their long-term (half a century or longer) groundwater storage trends. On average, over past 45 years, groundwater storage exhibited negative trends in all the selected catchments, except in the Katherine River catchment located in the Northern Territory. These negative trends persisted over longer periods, close to 100 years in some catchments and the strongest decreasing trend of 0.241 mm per year was observed in the Barron River catchment in New South Wales. However, groundwater storage exhibited different trends over the different shorter periods. Thus, while during the period of 1997–2007, 15 out of the 17 catchments showed negative trends in groundwater storage, during the period of 1980–2000, 12 out of the 17 catchments exhibited positive trends in groundwater storage; this underscores the fact that record lengths of one or even two decades are inadequate to derive meaningful trends. Strong consistencies in the trends exist across most catchments, indicating that groundwater storage is affected by large-scale climate factors. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Groundwater represents a large proportion of readily available freshwater resources on a global scale and is the only available water resource in some areas [1]. During prolonged dry periods, groundwater sustains low flows in streams and changes in climate and land use can significantly affect groundwater storage and low flow regimes [2,3]. Characteristics such as the magnitudes and durations of low flows have been widely used to determine water allocation and ecological water requirements [2]. It is predicted that increasing human consumption and climate change will have profound effects on future water resources, although the predictions are associated with large uncertainties [4–7]. Most studies of these impacts on hydrology have focused on surface water and much less is known about the impact on groundwater [8,9]. In any event, the identification of trends in groundwater requires long time series of observations (viz. at least half a century to a century). Unfortunately, long-term direct measurements of groundwater levels are rare and often too short to make reliable inferences of groundwater trends at regional scales. This is certainly the case

⇑ Corresponding author. http://dx.doi.org/10.1016/j.advwatres.2014.09.001 0309-1708/Ó 2014 Elsevier Ltd. All rights reserved.

for many regions in Australia, where groundwater monitoring started in the 1970s and 1980s with discontinuous measurements. There is also the issue of scale; measurements of groundwater level in single isolated wells represent local scale, whereas water resources management requires information at the catchment scale, so that a wide network of many wells would be needed but is rarely available. As noted, during dry periods low flows of a river derive primarily from water released by the upstream groundwater aquifers. Consequently, on the basis of hydraulic groundwater theory, this principle has been applied [10] to develop a method that estimates groundwater storage changes using daily streamflow data from a catchment. Because streamflow measurements typically commenced much earlier than groundwater records, this method can provide regional estimates of groundwater storage for much longer time periods than for those obtainable from groundwater wells. Applications have already been made to catchments in Mongolia [11], Japan [12] and USA [3,13] to study trends in groundwater storage. Over the past 50 years, many catchments in Southern Australia have experienced declines in annual streamflow, mostly due to reductions in annual rainfall [14–16] but also due to reductions in groundwater storage [17]. Significant changes in low flows have been observed in these catchments with some perennial streams

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165

even becoming ephemeral [15]. In Australia many regions are dependent on groundwater resources for irrigation and domestic water use, and a number of the regions comprise significant groundwater-dependent ecosystems. A better insight into how groundwater systems in Australia have been changing will be critical in developing sustainable water resources management plans. The objectives of this study are (1) to test the groundwater storage trends derived from baseflow observations [10] against observed groundwater level data and (2) to examine long-term groundwater storage trends during the past century in selected catchments in Australia. 2. Catchment description and data This study selected 17 catchments that have at least 45 years of unregulated daily streamflow records and the area of the catchments ranges from 196 to 8358 km2 (Fig. 1). Unregulated streamflow is defined as streamflow that is not affected by human control or diversion. The selection of the catchments was also based on consideration of availability of continuous daily streamflow records, if possible though not absolutely necessary, groundwater level data, and a distribution of hydroclimatic conditions. The daily streamflow data were made available to us thanks to the state agencies and quality checks were performed. The first step in the quality check was to scan data quality codes for any missing and poor-quality data. Once identified, missing values were calculated from streamflow data observed on previous and following days. The second step was to plot values of daily rainfall and streamflow to check for inconsistencies and to identify errors in the streamflow data (e.g. irregular or unusual spikes, identical streamflow value over a significant period of time). The data quality controls ensured that only good quality streamflow data were considered. Spatially averaged annual rainfall was calculated for each catchment using gridded SILO daily rainfall [18]. The gridded daily rainfall data were obtained by interpolating point measurements from over 6000 rainfall stations across Australia. The spatial pixel size of the rainfall data is approximately 5  5 km. The spatial coverage of the rainfall stations is reasonably good, particularly in the southeastern Australia. Ordinary kriging technique was applied to interpolate monthly rainfall data. The method takes into account rainfall variations with elevation. Then daily rainfall data for each SILO pixel were generated from the monthly rainfall data based on daily rainfall distribution in the nearest station [18]. Catchment average rainfall was obtained by aggregating the SILO interpolated rainfall surfaces.

157

Table 1 lists the area, location, climatic and aquifer characteristics of the selected catchments, which represent different conditions and groundwater systems. The mean annual rainfall ranges between 664 and 1616 mm and the aridity index (i.e. potential evaporation divided by rainfall) varies from 0.64 to 2.07. For the selected catchments, the mean slope ranges from 0.6° to 8.6° and the surface soil types include sandy loam, loam, and clay with large local differences in saturated hydraulic conductivity and water holding capacity. The vegetation in the catchments includes crops, grass, woodlands, and forest. According to Coram et al. [19], local flow systems are the dominant groundwater systems in the selected catchments with intermediate flow systems existing in some of them (Table 1). The geology of the catchments includes Tindall limestone, Silurian limestone and sandstone, Atherton basalts, Mareeba granite and various other metamorphics and alluvials. In 12 of the 17 catchments, observation wells had been installed so that groundwater table levels have been observed for some time; these well observations allow a comparison with the results of the present method. The groundwater observation wells were selected based on consideration of their locations (i.e. proximity to streamflow gauging stations), length of records, and effect of pumping. However, as listed in Table 1, these available records are much shorter than those of the streamflow records. The potential evaporation data listed in Table 1 were obtained from the SILO data base based on Morton’s modification [20] of the Priestley and Taylor method [21]; these catchment values were calculated from the gridded SILO data base in the same manner as the listed catchment rainfall values. 3. Methods 3.1. Groundwater storage trends estimated from base flow The low flow hydrograph can be expressed as a function of time, as:

Q ¼ QðtÞ

ð1Þ

where Q is the rate of flow ½L3 T1  and t is the time ½T. For convenience of comparison with other fluxes in the water cycle such as rainfall and evaporation, in what follows, Q is transformed to flow per unit of drainage area ½LT1 , and denoted by y ¼ Q =A, in which A is the area of the catchment. Probably the most commonly used functional form of yðtÞ in hydrology is of the exponential type, namely

y ¼ y0 expðt=KÞ

ð2Þ

where K is the characteristic time scale of the catchment drainage process [T], also commonly known as the storage coefficient, and y0 is the value of y at the selected time origin t = 0. Eq. (2) was originally proposed on the basis of empirical evidence; however, some useful insight in its physical nature can be gained by examining how it can be derived by means of hydraulic groundwater theory. As shown elsewhere (e.g. [10,22]), the ‘‘long-time’’ solution of the linearized Boussinesq equation describing the outflow from a homogeneous and horizontal aquifer with the appropriate boundary conditions, yields the following for the characteristic drainage time scale

K ¼ 0:10ne

Fig. 1. Location map of the catchments. The size of circles indicates the relative catchment areas.

.

D2d k0 g0



ð3Þ

where k0 ½LT1  is the hydraulic conductivity, ne the drainable porosity, g0 is the average vertical thickness [L] of the layer in the soil profile occupied by flowing water, Dd ¼ L=A is the drainage density, where L is the total length of upstream channels in the catchment. Eq. (3) is based on the assumptions that the river channel network

1973–2006 (13620203) 1974–2008 (GW036108) – – 1974–2008 (GW036116) 1953–1993 (GW026956) – 1988–2007 (GW036757) 1984–2007 (RN022397) – – 1991–2006 (GW032176) 1977–2007 (GW042338) 1977–2007 (GW042327) 2000–2007 (20084434) 1992–2007 (9611921) 1910–2007 1945–2007 1945–2007 1913–2007 1931–2007 1940–2007 1941–2007 1938–2007 1962–2007 1921–2007 1924–2007 1927–2007 1933–2007 1947–2007 1957–2007 1952–2007 4755 1632 1667 3402 196 736 610 1553 8358 276 374 1889 391 668 1786 774 Mary River (138001) Gloucester River (208003) Manning (208005) Goulburn River (210006) Williams River (210011) Hunter River (210018) Little Plains (222004) Boorowa River (412029) Katherine River (814001) Florentine (304040) North Esk (318076) Murrumbidgee (410033) Adjungbilly Ck (410038) Goobarragandra (410057) Kent (604053) Donnelly (608151)

25.95 32.0 31.5 32.2 32.2 31.8 37.1 34.4 13.9 42.5 41.5 36.1 35.0 35.3 34.9 34.3

152.5 151.8 151.9 150.0 151.6 151.3 149.0 148.8 132.9 146.5 147.5 149.0 148.2 148.3 117.1 115.8

1243 1195 1067 664 1190 1086 862 674 1122 1509 1143 895 1038 1168 739 929

1121 1307 1288 1376 1325 1252 1106 1299 1898 942 986 1187 1229 1236 1254 1340

(0.16%) (3.7%) (3.0%) (7.3%) (6.4%) (9.1%) (2.8%) (14.1%) (3.2%) (0.0%) (0.8%) (2.6%) (4.1%) (1.82) (5.4%) (2.4%)

1952–2003 (11009999) 1926–2007 (2.1%) 239 Barron River (110003)

17.3

145.5

1616

1637

Period of groundwater record (Bore ID) Period of streamflow record (% missing data) PET (mm a1) Rainfall (mm a1) Long Lat Area (km2) Catchment (Gauging ID)

Table 1 Characteristics of selected catchments.

Regional flow systems in Cainozoic and Mesozoic volcanic plains/plateaus; Local flow systems in Palaeozoic rocks or Mesozoic intrusives Local flow systems in Palaeozoic rocks or Mesozoic intrusive Local flow systems in Palaeozoic rocks or Mesozoic intrusive Local flow systems in Palaeozoic rocks or Mesozoic intrusive Local flow systems in Cainozoic volcanics or Mesozoic sediments/volcanic Intermediate and local flow systems in Palaeozoic rocks or Mesozoic intrusives Intermediate and local flow systems in Palaeozoic rocks or Mesozoic intrusives Intermediate and local flow systems in Palaeozoic rocks or Mesozoic intrusives Intermediate and local flow systems in Palaeozoic rocks or Mesozoic intrusive Local flow systems in Precambrian rocks Local flow systems in Palaeozoic rocks or Mesozoic intrusive Local flow systems in Palaeozoic rocks or Mesozoic intrusive Local flow systems in Palaeozoic rocks or Mesozoic intrusives Intermediate and local flow systems in Palaeozoic rocks or Mesozoic intrusives(60%) Local flow systems in Palaeozoic rocks or Mesozoic intrusives Local flow systems in Precambrian rocks Local flow systems in Precambrian rocks

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165

Groundwater flow systems & geology of aquifer

158

within the catchment exhibits self-similarity, and that the parameters of (3) represent characteristics of the entire upstream basin. In the absence of groundwater recharge, the movable groundwater stored in a river basin S is the volume of water ½L3  that has not yet been released as baseflow y and as evapotranspiration from groundwater ew , that is:

S¼

Z

1

ðy þ ew Þdt

ð4Þ

t

While locally groundwater evapotranspiration can be substantial, as will be shown below in the Appendix A, at the catchment scale it can usually be neglected. With this assumption, upon integration with (2), (4) yields a linear relationship between groundwater storage and rate of flow:

S ¼ Ky

ð5Þ

Recall that as specified earlier, the quantity S is not the total storage, but only the storage above the water table level which results in zero-flow. In other words, S can be considered as the average thickness of water stored above the zero-flow level of the water table over the basin. Based on (5), the temporal trend of groundwater storage can be determined from the temporal trend in the base flow as follows:

dS dy ¼K dt dt

ð6Þ

3.2. Estimation of the characteristic drainage timescale The use of (6) for estimating groundwater storage trends requires estimates of K and dy=dt. Eq. (2) could in principle be used to estimate the characteristic drainage time scale K from observed hydrograph recessions. However, this and other expressions to describe baseflow as a function of time, yðtÞ, are sensitive to the start of baseflow, i.e. t ¼ 0, and impractical to use, because it is nearly impossible to define a new time origin for each recession in a long-term streamflow hydrograph, which goes through successive highs and lows. To avoid this difficulty in general for any functional form yðtÞ, Brutsaert and Nieber [23] proposed a method in which the time variable t is eliminated by expressing (1) in differential form:

dy ¼ uðyÞ dt

ð7Þ

where uðyÞ is a function that is characteristic for a given basin. Thus, in the case of (2), for the present purpose (7) becomes:

dy y ¼ dt K

ð8Þ

It can be assumed that the decrease in flow from groundwater and soil moisture storage in unconfined aquifers is much slower than that of surface runoff generated directly from rainfall. A second point is that evapotranspiration, as an additional depletion mechanism, tends to accelerate the decline of the river flow rate, and thus produce a larger jdy=dtj. This means that low flows during periods of minimal evapotranspiration can be associated with the smallest jdy=dtj for a given flow rate y, that is the lower envelope of the data cloud. This lowest envelope also yields the largest flow rate y for a given jdy=dtj, and is thus likely to represent (8) under conditions when the entire catchment is contributing to the flow. For each basin in this study, K was estimated by implementing this method of Brutsaert and Nieber [23] as follows. Eq. (8) describes the low flow hydrograph from a basin and use of the relationship for daily streamflow data requires determination of the streamflows that took place under low-flow conditions. As suggested by Brutsaert [10], data points in the streamflow record that meet the following criteria were excluded from the analysis:

159

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165

 flow data associated with positive and zero values of dy=dt;  suddenly anomalous data points, i.e. data points with dy=dt that are much larger than the previous one (i.e. twice or more);  2 data points after the last positive and zero dy=dt (i.e. after a maximum), and all subsequent data points until dy=dt ceases decreasing (i.e. until the inflection point on the recession curve);  at least one data point before a zero or new positive value of dy=dt (i.e. before a minimum flow data followed by a rise in flow);  flow data in a recession that are suddenly followed by a data point with a larger value of dy=dt. During a recession, dy=dt is expected to decrease monotonically and a recession should last at least days to contain useful information. Moreover, since concurrent precipitation data were available, in the present implementation of the procedure, careful scrutiny of these data allowed exclusion of any direct storm runoff on days of rainfall, and also a number (1 to 4 depending on A) of days after the rainfall event. After low flow data points were determined, they were plotted as logðdy=dtÞ versus logðyÞ relationships for the selected catchments. A straight line lower envelope with unit slope was fitted to the data. Because the data points show scatter, the determination of the exact position of the lower envelope is somewhat subjective. Therefore, as suggested by Troch et al. [24], the positions of the lower envelopes were determined such that 5% of the flow data points were eliminated.

4. Results and discussion 4.1. Characteristic drainage time scale The values of the catchment drainage time scale K estimated by means of the method described in the previous section are listed in Table 2. Figs. 2 and 3 show examples for the Barron River and Hunter River. For the 17 selected catchments, the drainage time scale K varied between 25 and 101 days with an average value of 51 days. These values are consistent with those obtained in other studies (e.g. [3,10,11], [13, Fig. 1]), where it was found that K was on average 48 days, with an uncertainty of less than 15 days. This nearconstancy of K, in many different basins of sufficiently large size A, was attributed to river channel networks obeying some universal similarity laws and to the fact that different aquifer material properties and topography tend to average out to some typical values. A possible exception to this general tendency is the Katherine River catchment, where K was estimated to be 101 days. The upstreams of the Katherine River are in the escarpment country of Arnhem Land and the Nitmiluk and Kakadu National Parks to the north. While the Katherine River is subject to high flows in wet seasons and occasional floodings, it also sustains relatively high dry season base-flows from groundwater discharge in parts of the basin; among the more important and substantial groundwater resources within the Katherine River catchment is the karstic Tindall Limestone Aquifer. 4.2. Annual characteristic low flows

3.3. Estimation of drainable porosity ne The drainable porosity can be defined as the volume of water extracted from the groundwater per unit area for a unit reduction in the water table height. Following Brutsaert [10], the drainable porosity for each catchment can be calculated from groundwater storage trends (dS=dt) and the corresponding groundwater table trends ðdh=dtÞ observed in the wells, as follows

ne ¼

dS=dt dh=dt

ð9Þ

A basic assumption in this procedure is that ðdh=dtÞ in each well is representative of the trend of the mean water table height in the entire catchment, (dhhi=dt).

In the application of (6), beside K, also a knowledge of ðdy=dtÞ is required; although other choices are possible to characterize the baseflow during any given year, in the present study this was estimated from the annual lowest seven-day flows, denoted as y7 . The annual lowest 7-day flow in a river is a common variable used in drought statistics, as a more robust measure of the low flow characteristics than, say, the lowest one-day flow, which is more subject to error. The main reasons for this choice in the present study are first, that the lowest annual flow reflects the groundwater storage, which is the most sustainable in the course of that year and which can be depended upon for next year; and second, that the lowest flow normally occurs when groundwater evaporation is most likely to be negligible so that (5) is valid. An average was

Table 2 Estimates of drainage time scale (K) and drainable porosity (ne ) for selected Australian catchments. Also shown are groundwater table (h) and its trends (dh/dt) expressed in m/yr and as a percentage of estimated groundwater storage per year, correlation coefficient of groundwater storage (S) with measured groundwater water table (h). Catchment

Drainage time scale K (days)

Drainable porosity ne dS=dt ne ¼ dh=dt

Barron River Mary River Gloucester River Manning Goulburn River Williams River Hunter River Little Plains Boorowa River Katherine River Florentine North Esk Murrumbidgee Adjungbilly Ck Goobarragandra Kent Donnelly

58 34 38 59 56 25 48 55 29 101 59 51 34 67 68 25 39

0.00414 0.00965 0.00417 – – 0.00367 0.00100 – 0.00101 0.00565 – – 0.00281 0.00367 0.02129 0.00307 0.00112

Average groundwater table h (m)

Groundwater table trend (m/yr) dh/dt

Groundwater table trend (% yr1) dh/dt

Correlation coefficient of groundwater storage S and groundwater table h

19.87 ± 4.23 6.57 ± 0.37 7.61 ± 0.65

0.2298⁄ (3.65E08) 0.0225⁄ (0.0004) 0.0007 (0.9569) – – 0.0007 (0.9569) 0.0689⁄ (0.0009) – 0.10694# (0.0537) 0.0166 (0.5139) – – 0.45726⁄ (0.0130) 0.01362 (0.2748) 0.018361# (0.0778) 0.0372⁄ (0.0500) 0.0532⁄ (0.0250)

1.40⁄ (3.65E08) 9.74⁄ (0.0004) 0.02 (0.9569) – – 0.04 (0.9569) 0.83⁄ (0.0009) – 14.03# (0.0537) 04.13 (0.5139) – – 2.39⁄ (0.0130) 0.06 (0.2748) 0.02# (0.0778) 0.10⁄ (0.0500) 0.26⁄ (0.0250)

0.601 0.689 0.618

7.04 ± 0.35 7.45 ± 0.86 7.32 ± 0.86 14.87 ± 0.92

6.23 ± 0.72 5.03 ± 0.46 5.94 ± 0.83 6.90 ± 0.54 6.13 ± 0.37

0.252 0.620 0.294 0.594

0.683 0.292 0.331 0.639 0.745

An asterisk indicates a trend value for which the probability of being different from zero is at least 0.95, and a hash a probability of at least 0.90. The corresponding p values are shown in parentheses.

160

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165

Fig. 2. Relationship between logðdy=dtÞ and logðyÞ for the Barron River. The lower envelope line has a unit slope and the estimate of the characteristic drainage timescale K = 58 days.

Fig. 3. Relationship between logðdy=dtÞ and logðyÞ for the Hunter River. The lower envelope line has a unit slope and the estimate of the characteristic drainage timescale K = 48 days.

taken over 7 days to minimize error, and in our calculations the lowest 7-day flow was determined as the lowest value of the 7day running averages for each year of record. Across the catchments, the magnitude of the y7 values ranges between 0.03 and 1.02 mm per day; examples of the evolutions of these flows are shown in Figs. 4 and 5 for the Barron River and Katherine River catchments, respectively. It can be noted in Fig. 5 that the annual lowest 7-day flows are much larger over the period of 1999– 2004 and this is likely caused by above average rainfall over this period in the Katherine River catchment. For example, the longterm average annual rainfall is 1092 mm, whereas the average annual rainfall over the period of 1999–2004 was 20% higher than this long-term average. Added to this, is the fact that annual rainfall in 2000 was 1619 mm, 50% higher than the long-term average. 4.3. Validation of the method: comparison between estimated lowest groundwater storage and observed groundwater table The annual lowest groundwater storage levels were determined by means of (5) with the annual lowest 7-day flows as described and with the K values listed in Table 2. The correlation coefficients between the time series of the values thus obtained and those of

Fig. 4. Evolution of the annual lowest 7-day flows of the Barron River catchment. The straight line represents the regression of the flow flows over the period of record with a significant decreasing trend (p = 1.396E09). The trend of the low flows over the period is 0.000968 mm d1 a1.

Fig. 5. The annual lowest 7-day flows for the Katherine catchment during the period of 1962–2007. The straight line represents the regression of the flow flows over the period of record with a significant increasing trend (p = 2.19415E07). The trend of the low flows over the period is 0.000457 mm d1 a1.

the observed annual lowest groundwater levels measured in the observation wells in each of the catchments are presented in Table 2. As examples, Figs. 6 and 7 show the comparisons for the Barron River catchment and the Hunter River catchment. It can be seen that the two independent time series exhibited a high correlation coefficient and similar patterns over the periods of available groundwater records. For the 12 catchments with available groundwater table measurements, the correlation coefficient between estimated groundwater storage and observed groundwater table ranges between 0.252 and 0.745 (see Table 2) with an average value of 0.530 ± 0.175; for 7 out of the 12 catchments the correlation is of the order of 0.6 or larger. These results are consistent with the findings of Brutsaert [10] for the Illinois River and Rock River basin in the US. In addition, for these 12 catchments the temporal trends of lowest annual groundwater table over the period of groundwater well record were estimated using linear least square regression. The groundwater table exhibited negative trends ranging from 0.0007 to 0.457 m/yr in 10 catchments and a positive trend in only two catchments (i.e. Katherine and Goobarragandra) over the period of groundwater record (Table 2). The groundwater table trends are also expressed as a percentage of estimated groundwater storage per year and range from 0.02% per year (Gloucester River) to 14.03% per year (Boorowa

161

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165

in 11 out of the 12 catchments groundwater storage trends derived from daily streamflow are consistent with trends in groundwater table (Tables 2 and 3). It should be repeated that in the comparison, only one observation well in each catchment was available to represent the regional groundwater table and the lowest observed groundwater table levels were used to represent the situation each year. Given the uncertainties in both streamflow, groundwater well measurements and their placement, the correlation coefficients listed in Table 2 provide strong support for the use of the Brutsaert [6] method to estimate catchment scale groundwater storage. 4.4. Estimates of drainable porosity ne Fig. 6. Comparison of the evolution of the annual groundwater storage S (mm) for the Barron River catchment with the evolution of the lowest groundwater table h (m) measured at a monitoring well within the catchment. The trends obtained by linear least square regression are dS/dt = 0.2385 mm a1 and dh=dt = 0.2299 m a1. The correlation coefficient between the two time series is 0.601.

Fig. 7. Comparison of the evolution of the annual groundwater storage S (mm) for the Hunter River catchment with the evolution of the lowest groundwater table h (m) measured at a monitoring well within the catchment. The trends obtained by linear least square regression are dS/dt = 0.00692 mm a1 and dh=dt = 0.0689 m a1. The correlation coefficient between the two time series is 0.6208.

River). Recall again that the estimated groundwater storage is not the total storage, but only the storage above the water table level which results in zero-flow in a catchment. It is encouraging that

The drainable porosity was estimated using Eq. (9) and the results are given in Table 2. The estimated drainable porosity values range between 0.001 and 0.01296 with an average of 0.0051 ± 0.0031; these values are somewhat smaller than the ones of Buckland et al. [25] and McGuire et al. [26] who reported drainable porosity values of the order of 0.05, and those of Brutsaert and Lopez [27], Brutsaert [10] and Brutsaert and Hiyama [28], who suggested that ne is of the order of 0.01; but they are larger than those of Mendoza et al. [29] who derived values between 0.00002 to 0.0001. Again, it is impossible to draw a general conclusion here because each of the values of ne is based measurements of dh/dt at a single well, which was supposed to represent the entire catchment. Nevertheless in general, drainable porosities obtained from laboratory and local pumping tests are larger than those determined from model inversion with larger scale field data. Recall in this context that the drainable porosity is not really a physical property of the aquifer material, but merely a convenient parameter introduced to compensate for the neglect of the partly saturated flow zone, when the water table is assumed to be a true free surface [22]. Therefore, its value tends to depend not only on the actual aquifer characteristics but also on the method of its estimation and the assumed underling physical model; it is best determined by inverse calculations with measurements on the basis of the specific physical or computational model, for which it is intended for future prediction. 4.5. Annual groundwater storage trends The temporal annual groundwater storage trends were determined from the annual low flows as described by Brutsaert [10]

Table 3 Groundwater storage trends in mm/yr derived from the low flow measurements for different periods. Catchment

Trend period of record

Trend 1957–2007

Trend 1997–2007

Trend 1980–2000

Barron River Mary River Gloucester River Manning Goulburn River Williams River Hunter River Little Plains Boorowa River Katherine River Florentine North Esk Murrumbidgee Adjungbilly Ck Goobarragandra Kent Donnelly

0.28603⁄ (1.40E9) 0.00035 (0.876) 0.01623 (0.453) 0.02000 (0.385) 0.00378⁄ (0.034) 0.00123 (0.793) 0.00061 (0.977) 0.03465⁄ (0.006) 0.00116# (0.079) 0.05535⁄ (2.19E7) 0.10912⁄ (0.005) 0.08858⁄ (1.60E8) 0.01147⁄ (0.001) 0.00400 (0.831) 0.12956⁄ (0.038) 0.00014 (0.320) 0.02291⁄ (0.001)

0.24153⁄ (1.17E5) 0.01527⁄ (0.012) 0.03394 (0.150) 0.04200 (0.130) 0.00503⁄ (0.008) 0.00525 (0.357) 0.04253⁄ (0.018) 0.03190# (0.088) 0.00058 (0.433) 0.05535⁄ (2.19E7)a 0.212⁄ (0.003) 0.02684 (0.289) 0.01421⁄ (0.001) 0.05931# (0.088) 0.10395 (0.205) 0.00014 (0.320) 0.01311⁄ (0.001)

0.36280 (0.484) 0.04775# (0.085) 0.00396 (0.984) 0.41610⁄ (0.045) 0.02836 (0.241) 0.00924 (0.852) 0.01102 (0.920) 0.07478 (0.465) 0.00257 (0.207) 0.04908 (0.701) 0.52240 (0.109) 0.25230 (0.145) 0.03818⁄ (0.011) 0.18607 (0.416) 0.66361 (0.419) 0.00112# (0.075) 0.02217⁄ (0.026)

0.01589 (0.925) 0.00173 (0.879) 0.12322 (0.058) 0.09500 (0.366) 0.01447⁄ (0.027) 0.01111 (0.582) 0.02187 (0.717) 0.10340 (0.165) 0.00174 (0.646) 0.03568 (0.220) 0.648⁄ (0.048) 0.08500 (0.307) 0.01663 (0.168) 0.09371 (0.410) 0.36623 (0.268) 0.00057 (0.287) 0.01141# (0.081)

An asterisk indicates a trend value for which the probability of being different from zero is at least 0.95, and a hash a probability of at least 0.90. The corresponding p values are shown in parentheses. a Period of 1962–2007.

162

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165

Fig. 8. Groundwater storage trends expressed as a percentage of average storage per year (% yr1) derived from the low flow measurements for different periods.

using the K values of Table 2. The results are listed in Table 3 for four different periods: the period of record of each gauging station, the common period for all the catchments (i.e. the period of 1957– 2007), the recent drought period (1997–2007) in southeast Australia [16,30], and the period of 1980–2000, which is considered as a baseline period in climate change impact studies for Australian catchments [31]. The groundwater storage trends are expressed as mm per year. Over their entire period of record, which varies from 45 to 97 years, 13 out of the 17 catchments showed negative trends in groundwater storage (Table 3). The Mary River catchment showed a negative trend in groundwater storage over the past century (i.e. 1910–2007). Over a similar period (i.e. 1913–2007), a negative trend in groundwater storage was also observed in the Goulburn River catchment (see Fig. 8). A positive groundwater storage trend was detected in the Gloucester River catchment over its period of record (i.e. 1945–2007). Two catchments in Western Australia showed opposite trends: a statistically significant negative trend in Donnelly and a positive trend in Kent (Table 3). Inspection of Table 3 reveals that 10 out of the 17 catchments have statistically significant groundwater storage trends at least at the 0.1 level of significance. Over the common period of 1957–2007, all the catchments showed negative trends in groundwater storage, except for the Katherine River and Kent River catchments (see Table 3 and Fig. 8). In 10 out of the 17 catchments, the groundwater trends are statistically significant at least at the 0.1 significance level. The strongest negative trend was obtained for the Barron River catchment (Fig. 9), namely 0.241 mm per year. On average, the negative trend in groundwater storage is 0.06 mm per year. The magnitudes of the groundwater storage trends are similar to those reported by Brutsaert [3] for the Mid Atlantic, the Great Lakes, the Tennessee, the Lower Mississippi, the Souris-Red-Rainy, the Missouri, and the Arkansas-White-Red regions, but in the opposite direction. However, over a similar period (i.e. 1959–2006), the Kherlen River basin in Mongolia showed a negative trend in groundwater storage [11]. Not surprisingly, these catchments in different parts of the world showed different trends in groundwater storage mostly due to local hydro-climatic conditions, namely rainfall and evaporation. Over the period of 1997–2007, 15 out of the 17 catchments showed negative trends in groundwater storage (Fig. 8). Most of the trends over this period are not statistically significant (Table 3) and this is likely caused by the shortness of these records used. The average groundwater storage over this period was 40% below the long-term average with the exception of the Katherine River catchment where groundwater storage was 80% above the longterm average. The period of 1997–2007 largely coincides with wide-spread drought in south-eastern Australia. During this period, annual rainfall reduction was 4% compared with the long-term

Fig. 9. Evolution of the annual groundwater storage over the periods of 1957–2007, 1980–2000, and 1997–2007 for the Barron River catchment. The catchment scale drainage timescale is 58 days. The linear line represents the regression over the period of 1957–2007, indicating a significantly (p = 1.17E5) decreasing trend of 0.241 mm per year. The catchment area at this gauging station is 239 km2.

Fig. 10. Evolution of the annual groundwater storage over the periods of 1962– 2007, 1980–2000, and 1997–2007 for the Katherine River catchment. The catchment scale drainage timescale is 101 days. The linear line represents the regression over the period of 1962–2007, indicating a significantly (p = 2.19E7) increasing trend of 0.055 mm per year. The catchment area at this gauging station is 8358 km2.

average, while the annual streamflow reduction was 21% for the Murray–Darling Basin in south-eastern Australia [16]. However, the Katherine catchment in Northern Australia did not experience any drought during this period and the average annual rainfall was 17% above the long-term average. Changes in annual rainfall amount and seasonal distribution are undoubtedly the causes of the changes in streamflow and groundwater storage in these catchments. Over the period of 1980–2000, 12 out of the 17 catchments exhibited positive trends in groundwater storage, but only the Goulburn

163

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165 Table 4 Catchment scale groundwater consumption by evapotranspiration. Catchment (ID)

Length of streams (km)

Average riparian width (m)

Basin area (km2)

Riparian area as fraction

Catchment scale GWU ew (mm/d)

hy7i(mm/d)

(100 * ew/hy7i) (%)

Barron River Mary River Gloucester River Manning Goulburn River Williams River Hunter River Little Plains Boorowa River Katherine River Florentine North Esk Murrumbidgee Adjungbilly Ck Goobarragandra Kent Donnelly

21 285 113 157 247 47 56 47 104 417 30 39 132 33 49 102 90

10 1 20 10 1 10 1 10 1 20 1 1 5 5 5 10 10

239 4755 1632 1667 3402 196 736 610 1553 8358 276 374 1889 391 668 1786 774

0.00173 0.00012 0.00277 0.00188 0.00015 0.00481 0.00015 0.00155 0.00013 0.00199 0.00022 0.00021 0.00070 0.00085 0.00073 0.00114 0.00231

0.003451571 0.000240054 0.005547787 0.003765011 0.000290945 0.009613039 0.000306028 0.003097676 0.000269109 0.003987509 0.000437507 0.000417052 0.00139477 0.001707621 0.001462126 0.002278208 0.004625426

0.2752 0.0134 0.0733 0.0636 0.0052 0.0537 0.0419 0.0694 0.0681 0.0819 0.4349 0.1876 0.5538 0.3263 0.0174 0.0171 0.0904

1.25 1.79 7.57 5.92 5.60 17.90 0.73 4.46 0.40 4.87 0.10 0.22 0.25 0.52 0.17 13.09 5.12

Note: GWU = Groundwater use calculated by assuming 2 mm/day in riparian areas based on O’Grady et al. [34]; hy7i = Average annual minimum seven-day flow.

River showed a statistically significant positive trend. Obviously, this period is very different from the other three periods discussed above in terms of groundwater storage trends (Fig. 8). The Kent and Donnelly River catchments in Western Australia exhibited consistent negative groundwater storage trends over the period of 1980–2000 and 1997–2007. These trends in groundwater storage may be caused by continued rainfall decline in Western Australia since the mid-1970s. The Katherine River catchment in the Northern Territory showed consistent positive trends over the three periods (Fig. 10); the maximal value of the positive trend is about 0.055 mm per year (Table 3). If, however, one selected the period of 1974–1997, the Katherine would exhibit a strong negative trend. All this shows, as also illustrated in Fig. 8, that the investigated catchments exhibited mostly similar trends over the longer periods, but vastly different trends in groundwater storage over the selected shorter periods; these results strongly suggest that time series over only one or two decades are usually not long enough to allow valid conclusions regarding climate change impact.

5. Conclusions This study examined groundwater storage trends in 17 selected Australian catchments over the last 45 to 97 years from measured daily streamflow. The method is based on the concept that base flow in a natural river system is directly controlled by groundwater storage and hence measured streamflow can provide quantitative estimates of catchment-scale groundwater storage evolution. The selected catchments have at least 45 years of continuous daily streamflow data and represent different hydroclimatic and aquifer characteristics. Base flow analysis showed that the drainage time scale K varied between 25 and 101 days with an average value of 51 days, consistent with estimates of roughly 45(±15) days by Brutsaert ([13], Fig. 1) for catchments in other parts of the world. Estimates of groundwater storage were compared with independent groundwater level data and high correlations between them were obtained in most cases. Estimates of drainable porosity range between 0.001 and 0.01296. These values are comparable with those of Brutsaert and Lopez [27], Brutsaert [10] and Brutsaert and Hiyama [28] and not inconsistent with those of, for example, Buckland et al. [25], Mendoza et al. [29], and McGuire et al. [26], all obtained by different methods in other catchments. Results from this study indicate that the method used is valid for estimating groundwater

storage changes and provide confidence in using the method to investigate long-term groundwater storage trends. Overall, groundwater storage exhibited negative trends over the past half century in all the selected catchments, except the Katherine River catchment in the Northern Territory. The negative trends in groundwater storage persisted over longer periods, close to a century in some catchments and the strongest negative trend of 0.241 mm per year was observed in the Barron River catchment. From 1997 to 2007, 15 out of the 17 catchments showed negative trends in groundwater storage, consistent with the trends in annual streamflow over this period. It is interesting to note that 12 out of the 17 catchments exhibited positive trends in groundwater storage over the period of 1980–2000, opposite of the trends in the period of 1957–2007. It can also be seen that the groundwater storage exhibited different trends over the different periods considered and these results indicate that short-term climate variability can significantly affect groundwater storage. Conversely, this shows how important it is to understand the role of aquifer properties in the response of catchments to short-term rainfall changes. It also shows that time series over only one or two decades are insufficiently long to draw valid conclusions regarding climate change. The identified trends only cover the analysis period which ended in 2007; whether the trends will continue on a similar trajectory or disappear or even reverse remains to be seen and will require further analysis of data since 2007 and into the future. But this is beyond the scope of the present study. The method of Brutsaert [10] should enable examination of the potential impact of climate change on groundwater storage over longer periods of time. One of the advantages of this method is that it uses daily streamflow data, which are more readily available over longer periods of record than most groundwater level data. Acknowledgments This study was supported by the CSIRO Water for a Healthy Country Flagship. We would like to thank Anthony O’Grady for discussion and suggestions on catchment scale evapotranspiration from groundwater table. We would like to acknowledge the Department of Primary Industries, Parks, Water and Environment, Tasmanian, Department of Environment and Primary Industries, Victoria, Department of Primary Industries, New South Wales, Department of Natural Resources and Mines, Queensland, Department of Land Resource Management, Northern Territory, and Department of Water, Western Australia for providing streamflow

164

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165

and groundwater data. We would also like to thank David Rassam for his comments on the manuscript.

Appendix A. Evapotranspiration from water table at catchment scales The effect of evapotranspiration on flow recession can be assessed by considering groundwater use by vegetation. Although no information on this is available for the seventeen catchments of this paper, a number of studies have addressed groundwater use by vegetation for the specific conditions of climate and vegetation in different parts of Australia. These have generally shown that deep-rooted trees have the potential to access groundwater and that depth to water table is a key factor in determining the extent of their groundwater use. For instance, Cook et al. [32], who studied the water budget of a tropical Eucalyptus savanna ecosystem in northern Australia, concluded that during the dry season with groundwater depths of around 6.5 m the overstory trees did not use groundwater for transpiration, but were sustained solely by the unsaturated zone. Lamontagne et al. [33] and O’Grady et al. [34] investigated groundwater use by riparian vegetation in the Daly River basin in the Northern Territory and found that trees that rely on groundwater were located over water tables with depths smaller than 5 m; trees at locations higher than 5 m above the river were principally using soil water. In south-eastern Australia Benyon et al. [35] studied the impact of tree plantations of common Eucalyptus and Pinus species on groundwater at 21 sites and showed that groundwater use by trees or transpiration from groundwater was negligible when the groundwater table was deeper than 7.5 m; they concluded that their results suggest that the maximum water depth for significant groundwater use must be somewhere between 6 and 8 m. In Western Australia Farrington et al. [36] measured groundwater consumption by Eucalyptus trees at three sites and concluded that the trees used groundwater where the ground water depth was 6 m, but not where it was 14 and 30 m. In south-western Australia Zencich et al. [37] determined seasonal water sources of Banksia trees growing on a coastal dune system overlying a shallow sandy aquifer. At their ‘‘dampland’’ site with a shallow 2.5 m water table in summer groundwater was nearly 100% of water used by the trees. At their ‘‘lower slope’’ site with water table depths of around 4 m, the trees showed evidence of some groundwater uptake but they obtained a significant portion of their total water needs from shallow soil moisture. At the ‘‘upper slope’’ site (9 m to water table), groundwater was not significant source of water for transpiration during the driest months, and trees used water mostly from the more shallow soil component. Thus, their results were consistent with those of Dodd and Bell [38] on Banksia trees with a similar depth to groundwater of 6 to 7 m. Zencich et al. [37] explained the reduced groundwater use by upper slope trees in summer with the root measurements of Farrington et al. [39], showing that these tree roots were present mainly down to 8 m, but almost absent below that. In an extensive report summarizing most available studies on groundwater use by vegetation in Australia, O’Grady et al. [40, p. 37] concluded that ‘‘there is some indication that, other than for two studies, groundwater uptake is very low when water table depths are deeper than 5 m’’. Nevertheless, they added that there is some evidence that locally individual tree roots of some species have the capacity to explore the soil profile to much greater depths down to 20 m and even 40 m in one case; but there was no evidence that this occurred on a catchment wide scale. All these studies show that by and large substantial groundwater use by vegetation occurs in areas with shallow water tables, namely with depths no larger than 5 to 7 m. As shown in Table 2, most of the observation wells showed water tables in excess of

these depths. In natural catchments water tables smaller than 5 to 7 m are nearly always found near rivers in areas occupied by riparian vegetation. To explore this further, in what follows an estimate is made of the fractional area occupied by these areas in the catchments of the present study, in order to determine the basinscale evapotranspiration. For each catchment, the total length of permanent stream channels and also the average width of the vegetated riparian areas were determined. The former were estimated from digital elevation model based calculations with ‘‘Geofabric’’ [41]; the latter were derived from air photos. The resulting fractional areas are shown in Table 4. In several studies measurements were made of the groundwater consumption by vegetation with the water table at depths of around 5 to 7 m as found in riparian areas. For instance, O’Grady et al. [34] estimated this to be roughly between 60% and 75% of the total evapotranspiration, or 1.9 to 2.4 mm/day in the dry season along the Daly River on the basis of isotope measurements of the xylem sap; Benyon et al. [35] observed that on average during the year about 50% of the total evapotranspiration or 1.83 mm/day came from groundwater; Doody et al. [42] found that the average groundwater use during summer by the vegetation in the riparian area of the River Murray with a water table at approximately 5 m depth was of the order 0.86 mm/day. Adopting 2 mm/day as the typical largest value of groundwater consumption by riparian vegetation among these various estimates, multiplication with the fraction occupied by riparian areas yields the catchment scale evapotranspiration due to these areas with shallow water tables. The results of these operations are shown in Table 4. Also shown in the table are the average values of the annual seven day low flows for each basin. It can be seen in the last column of Table 4 that the groundwater consumption by these areas, relative to the base flows used in the present study, is very small and in all cases well within the error bounds of streamflow measurements. Indeed, careful analysis of the error structure of streamflow measurements by Harmel et al. [43] has shown that under typical scenarios in the field the uncertainty contributed to streamflow measurements ranges from 6% to 19%; even under ideal near-laboratory conditions (with a pre-calibrated flow control structure, a stable bed and channel, and a stilling well for stage measurement) the probable error range is still 3%. The values in the last column of Table 4 are mostly of the same order of magnitude or smaller than these error bounds; this strongly suggests that the effect of evapotranspiration on the magnitude of the base flows in the rivers of this study can be safely neglected.

References [1] Shiklomanov IA, Rodda JC. World water resources at the beginning of the twenty-first century. Cambridge University Press; 2003. 435 p. [2] Smakhtin VU. Low flow hydrology: a review. J Hydrol 2001;240(3–4):147–86. http://dx.doi.org/10.1016/S0022-1694(00)00340-1. [3] Brutsaert W. Annual drought flow and groundwater storage trends in the eastern half of the United States during the past two-third century. Theor Appl Climatol 2010;100(1):93–103. http://dx.doi.org/10.1007/s00704-009-0180-3. [4] IPCC. Climate change 2007 – the synthesis report: contribution of working groups I, II and III to the fourth assessment report of the intergovernmental panel on climate change, Geneva, Switzerland: IPCC; 2007. p. 104. [5] Chiew FHS, Teng J, Vaze J, Post DA, Perraud JM, Kirono DGC, Viney NR. Estimating climate change impact on runoff across southeast Australia: method, results, and implications of the modeling method. Water Resour Res 2009;45:w10414. http://dx.doi.org/10.1029/2008WR007338. [6] Kiem AS, Verdon-Kidd DC. Steps towards ’useful’ hydroclimatic scenarios for water resource management in the Murray–Darling Basin. Water Resour Res 2011;47:W00G06. http://dx.doi.org/10.1029/2010WR009803. [7] Gallant AJE, Kiem AS, Verdon-Kidd DC, Stone RC, Karoly DJ. Understanding hydroclimate processes in the Murray–Darling Basin for natural resources management. Hydrol Earth Syst Sci 2012;16:2049–68. http://dx.doi.org/ 10.5194/hess-16-2049-2012. [8] Taylor RG, Koussis A, Tindimugaya C. Groundwater and climate in Africa – a review. Hydrol Sci J 2009;54(4):655–64. http://dx.doi.org/10.1623/ hysj.54.4.655.

L. Zhang et al. / Advances in Water Resources 74 (2014) 156–165 [9] Green TR, Taniguchi M, Kooi H, Gurdak JJ, Allen DM, Hiscock KM, et al. Beneath the surface of global change: impacts of climate change on groundwater. J Hydrol 2011;405(3–4):532–60. http://dx.doi.org/10.1016/ j.jhydrol.2011.05.002. [10] Brutsaert W. Long-term groundwater storage trends estimated from streamflow records: climatic perspective. Water Resour Res 2008;44(2):W2409. http://dx.doi.org/10.1029/2007WR006518. [11] Brutsaert W, Sugita M. Is Mongolia’s groundwater increasing or decreasing? The case of the Kherlen River basin. Hydrol Sci J 2008;53(6):1221–9. http:// dx.doi.org/10.1623/hysj.53.6.1221. [12] Sugita M, Brutsaert W. recent low-flow and groundwater storage changes in upland watersheds of the Kanto region. Jpn J Hydrol Eng 2009;14(3):280–5. http://dx.doi.org/10.1061/(ASCE)1084-0699(2009) 14:3(280). [13] Brutsaert W. Are the North American deserts expanding? Some climate signals from groundwater storage conditions. Ecohydrology 2012;5(5):541–9. http:// dx.doi.org/10.1002/eco.263. [14] Bates B, Hope P, Ryan B, Smith I, Charles S. Key findings from the Indian Ocean climate initiative and their impact on policy development in Australia. Clim Change 2008;89(3–4):339–54. http://dx.doi.org/10.1007/s10584-007-9390-9. [15] Petrone KC, Hughes JD, Van Niel TG, Silberstein RP. Streamflow decline in southwestern Australia, 1950–2008. Geophys Res Lett 2010;37(11):L11401. http://dx.doi.org/10.1029/2010GL043102. [16] Potter NJ, Chiew FHS, Frost AJ. An assessment of the severity of recent reductions in rainfall and runoff in the Murray–Darling Basin. J Hydrol 2010;381(1–2):52–64. http://dx.doi.org/10.1016/j.jhydrol.2009.11.025. [17] Hughes JD, Petrone KC, Silberstein RP. Drought, groundwater storage and stream flow decline in southwestern Australia. Geophys Res Lett 2012;39(3):L3408. http://dx.doi.org/10.1029/2011GL050797. [18] Jeffrey SJ, Carter JO, Moodie KB, Beswick AR. Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environ Modell Softw 2001;16(4):309–30. http://dx.doi.org/10.1016/S1364-8152(01)00008-1. [19] Coram JE, Dyson PR, Houlder PA, Evans WR. Australian groundwater flow systems contributing to dryland salinity. Bureau of rural sciences for the dryland salinity theme of the national land and water resources audit, Canberra, CD; 2000. [20] Morton FI. Operational estimates of areal evapotranspiration and their significance to the science and practice of hydrology. J Hydrol 1983;66(1– 4):1–76. http://dx.doi.org/10.1016/0022-1694(83)90177-4. [21] Priestley CHB, Taylor RJ. On the assessment of surface heat flux and evaporation using large-scale parameters. Mon Weather Rev 1972;100(2):81–92. http://dx.doi.org/10.1175/1520-0493(1972) 100<0081:OTAOSH>2.3.CO;2. [22] Brutsaert W. Hydrology: An Introduction. Cambridge University Press, 2005. 618 p. [23] Brutsaert W, Nieber JL. Regionalized drought flow hydrographs from a Mature Glaciated plateau. Water Resour Res 1977;13. http://dx.doi.org/10.1029/ WR013i003p00637. [24] Troch PA, De Troch FP, Brutsaert W. Effective water table depth to describe initial conditions prior to storm rainfall in humid regions. Water Resour Res 1993;29(2):427–34. http://dx.doi.org/10.1029/92WR02087. [25] Buckland GD, Harker DB, Sommerfeldt TG. Comparison of methods for determining saturated hydraulic conductivity and drainable porosity of two Southern Alberta soils. Can J Soil Sci 1986;66(2):249–59. http://dx.doi.org/ 10.4141/cjss86-027. [26] McGuire KJ, Weiler M, McDonnell JJ. Integrating tracer experiments with modeling to assess runoff processes and water transit times. Adv Water Resour 2007;30(4):824–37. http://dx.doi.org/10.1016/j.advwatres.2006.07.004.

165

[27] Brutsaert W, Lopez JP. Basin-scale geohydrologic drought flow features of riparian aquifers in the Southern Great Plains. Water Resour Res 1998:34. http://dx.doi.org/10.1029/97WR03068. [28] Brutsaert W, Hiyama T. The determination of permafrost thawing trends from long-term streamflow measurements with an application in eastern Siberia. J Geophys Res 2012;117(D22):D22110. http://dx.doi.org/10.1029/ 2012JD018344. [29] Mendoza GF, Steenhuis TS, Walter MT, Parlange JY. Estimating basin-wide hydraulic parameters of a semi-arid mountainous watershed by recessionflow analysis. J Hydrol 2003;279(1–4):57–69. http://dx.doi.org/10.1016/ S0022-1694(03)00174-4. [30] Verdon-Kidd DC, Kiem AS. Nature and causes of protracted droughts in Southeast Australia: comparison between the federation, WWII, and big dry droughts. Geophys Res Lett 2009;36(22):L22707. http://dx.doi.org/10.1029/ 2009GL041067. [31] Chiew FHS, Vaze J, Viney N, Jordan P, Perraud J, Zhang L, et al. Rainfall-runoff modelling across the Murray–Darling Basin. Canberra, Australia: CSIRO Land and Water; 2008. 62 p. [32] Cook PG, Hattona TJ, Pidsley D, Herczeg AL, Held A, O’Grady A, et al. Water balance of a tropical woodland ecosystem, Northern Australia: a combination of micro-meteorological, soil physical and groundwater chemical approaches. J Hydrol 1998;210:161–77. http://dx.doi.org/10.1016/S0022-1694(98)00181-4. [33] Lamontagne S, Cook PG, O’Grady A, Eamus D. Groundwater use by vegetation in a tropical savanna riparian zone (Daly River, Australia). J Hydrol 2005;310:280–93. http://dx.doi.org/10.1016/j.jhydrol.2005.01.009. [34] O’Grady AP, Eamus D, Cook PG, Lamontagne S. Groundwater use by riparian vegetation in the wet-dry tropics of northern Australia. Aust J Bot 2006;54:145–54. http://dx.doi.org/10.1071/BT04164. [35] Benyon RG, Theiveyanathan S, Doody TM. Impacts of tree plantations on groundwater in south-eastern Australia. Aust J Bot 2006;54:181–92. http:// dx.doi.org/10.1071/BT05046. [36] Farrington P, Greenwood EA, Bartle GA, Beresford JD, Watson GD. Evaporation from Banksia woodland on a groundwater mound. J Hydrol 1989;105:173–86. http://dx.doi.org/10.1016/0022-1694(89)90102-9. [37] Zencich SJ, Froend RH, Turner JV, Gailitis V. Influence of groundwater depth on the seasonal sources of water accessed by Banksia tree species on a shallow sandy coastal aquifer. Oecologia 2002;131:8–19. http://dx.doi.org/10.1007/ s00442-001-0855-7. [38] Dodd J, Bell D. Water relations of the canopy species in a Banksia woodland, Swan Coastal Plain. Western Australia. Aust J Ecol 1993;18:281–93. http:// dx.doi.org/10.1111/j.1442-9993.1993.tb00456.x. [39] Farrington P, Turner JV, Gailitis V. Tracing water uptake by jarrah (Eucalyptus marginata) trees using natural abundances of deuterium. Trees 1996;11:9–15. [40] O’Grady A, Carter J, Holland K. Review of Australian groundwater discharge studies of terrestrial systems. CSIRO: Water for a healthy country, National research flagship report; 2010. 60 p. [41] WIRADA. Water information research and development alliance: science symposium proceedings, Melbourne, Australia, 1–5 August 2011. CSIRO: Water for a healthy country, National research flagship; 2012. 425 p. [42] Doody TM, Holland KL, Benyon RG, Jolly ID. Effect of groundwater freshening on riparian vegetation water balance. Hydrol Process 2009;23:3485–99. http://dx.doi.org/10.1002/hyp.7460. [43] Harmel RD, Cooper RJ, Slade RM, Haney RL, Arnold JG. Cumulative uncertainty in measured streamflow and water quality data for small watersheds. Trans ASABE 2006;49(3):689–701. http://dx.doi.org/10.13031/2013.20488.