The Great Artesian Basin (GAB) in Australia, the largest artesian basin in the world, is rich in groundwater resources. This study analyzed the spatio-temporal characteristics of terrestrial water storage (TWS) in the GAB for 2003–2014 using satellite (Gravity Recovery and Climate Experiment, GRACE) data, hydrological models’ outputs, and in situ data. A slight increase in TWS was observed for the study period. However, there was a rapid increase in TWS in 2010 and 2011 due to two strong La Nina events. Long-term mean monthly TWS changes showed remarkable agreements with net precipitation. Both GRACE derived and in situ groundwater disclosed similar trend patterns. Groundwater estimated from the PCR-GLOBWB model contributes 26.8% (26.4% from GRACE) to the total TWS variation in the entire basin and even more than 50% in the northern regions. Surface water contributes only 3% to the whole basin but more than 60% to Lake Eyre and the Cooper River. Groundwater, especially deeper than 50 meters, was insensitive to climate factors (i.e., rainfall). Similarly, the groundwater in the northern Cape York Peninsula was influenced by some other factors rather than precipitation. The time-lagged correlation analysis between sea surface height and groundwater storage indicated certain correlations between groundwater and sea level changes.

INTRODUCTION

Groundwater plays a vital role in the indigenous industrial, domestic, and environmental water uses. In order to utilize groundwater resources precisely, the spatio-temporal characteristics of different components of terrestrial water storage (TWS) should be comprehensively understood. For a long time, the detection of regional water storage change has been mainly dependent on in situ observation and modelling approaches. Due to the limited spatial coverage of observing sites, it is a big challenge for large scale analysis and evaluation. Although hydrological models can estimate large scale water storage variations, the lack of observational data is the main limit to using these models.

Since the launch of the Gravity Recovery and Climate Experiment (GRACE) satellite in 2002, an unprecedented satellite-based approach to estimating TWS has been springing up. The GRACE products have widely been used in hydrology, including water balance research (Syed et al. 2005; Swenson & Wahr 2009), evapotranspiration (ET) estimates (Syed et al. 2014; Billah et al. 2015; Wan et al. 2015), extreme weather analysis (Long et al. 2014a; Thomas et al. 2014; Abelen et al. 2015), and groundwater storage (GWS) change (Döll et al. 2014; Famiglietti 2014; Chen et al. 2015; Richey et al. 2015).

Previous studies mostly focused on the world's large basins (Famiglietti 2014; Getirana et al. 2014; Long et al. 2015; Zhang et al. 2015) and groundwater over-exploitation areas (Rodell et al. 2009; Munier et al. 2012; Feng et al. 2013; Huang et al. 2015). Most of these areas are located in the northern hemisphere. There are a few studies for the whole of Australia (Brown & Tregoning 2010; van Dijk et al. 2011; Seoane et al. 2013; Wang et al. 2014), the Murray–Darling Basin (Leblanc et al. 2009, 2011; Brown & Tregoning 2010; Frappart et al. 2011), and the Canning Basin (Munier et al. 2012; Richey et al. 2015). However, no relevant studies have been reported on TWS and its spatio-temporal characteristics in the Great Artesian Basin (GAB). The GAB is a huge underground water reservoir in Australia. Low rainfall and limited supplies of surface water result in over usage of groundwater for irrigating, mining, industrial, and domestic sectors, which creates a number of problems from the last century, including reduced flow rates and pressures in the wells as well as stoppage of many artesian springs. A study on the spatio-temporal characteristics of TWS in the GAB will be of great importance in providing information on hydrological footprints, regional water resources management, sustainable water use and water resources variations in arid regions. In order to draw a clear picture of the TWS in the GAB, this study attempts to highlight the spatio-temporal characteristics of TWS for a 12-year period (2003–2014).

STUDY AREA

Location

The GAB is the largest artesian basin in the world, covering an area of about 1.7 million km2. It is mostly located in Queensland and partly in South Australia, New South Wales and the Northern Territory, stretching from Cape York Peninsula in far north Queensland, further west than the Lake Eyre Depression and east to the Great Dividing Range. The elevation in the basin ranges from a low of −23 m in the Lake Eyre Depression to a high of 1,347 m in the Great Dividing Range, with most parts of the basin located at lower elevations (an average of 174 m), as shown in Figure 1.

Climate

Average annual temperatures of 18–24 °C prevail in most parts of the basin. The area having a substantial higher elevation is mostly located on the eastern margin, as shown in Figure 1. The high temperature, low elevation, and geological features in the basin jointly result in an average low rainfall. The GAB lies predominantly under arid and semi-arid conditions, with an average annual rainfall of 425 mm. More than 55% of the basin's annual rainfall is under 400 mm, mainly in the southern and western regions. Along the eastern margin, average annual rainfall is above 800 mm, increasing to more than 1,800 mm in the far north of Cape York Peninsula. More than 50% of rainfall occurs in austral summer (December–February) and less than 10% occurs in austral winter (June–August). Annual actual ET ranges from less than 100 mm in the south-western parts to more than 1,100 mm in Cape York Peninsula, with an average of about 384 mm (data from Bureau of Meteorology (BOM), Australia). The climate is also strongly influenced by the El Nino Southern Oscillation and the Indian Ocean Dipole.

Water resources

Australia has significant groundwater resources, which occur beneath one-quarter of the country. These groundwater resources predominantly lie in the GAB. The GAB is a huge groundwater tank, which has been laid down for millions of years, and stores about 65,000 km3 of water. The recharging of the groundwater aquifer takes place mostly from the eastern high lands, namely the Great Dividing Range, and a very small amount of recharge from the western high lands. In the study area, most of the rivers, lakes, and streams are not perennial and do not flow throughout the year. Thus, the GAB grounder water is mostly used to feed springs and wetlands, and the indigenous people also rely on this.

The GAB consists of predominantly arid and semi-arid areas, where surface water resources are extremely scarce. This means that groundwater is an important, reliable source of water for farms and stocks, tourism, and domestic uses. The GAB's groundwater provides water to over 200,000 people. However, a number of problems are associated with the use of the GAB's groundwater, e.g., over-extraction and declining natural flows, with pressure declines of up to 100 meters (Murray-Darling Basin Commission (Australia) 1999).

DATA AND METHODS

TWS

TWS represents an aggregate of all forms of water stored above and underneath the land surface, including snow and ice, surface water, soil moisture, and groundwater. There are two major TWS products: direct satellite monitoring datasets and hydrological model simulation datasets.

The GRACE satellite mission provides a unique opportunity to monitor TWS from space (Swenson & Wahr 2002). It senses the TWS in all components (e.g., surface water storage, soil moisture, and groundwater). For this reason, GRACE data have been used in many hydrological applications at both global and regional scales. In this study, the GRACE JPL-RL05 monthly one-degree TWS products, accessed from the NASA JPL website (http://grace.jpl.nasa.gov/data/get-data/monthly-mass-grids-land/), were used, covering the period from January 2003 to December 2014. Due to the sampling and post-processing of GRACE observations, surface mass variations at small spatial scales tend to be attenuated (Swenson & Wahr 2006). The scaling factor was multiplied to restore signal loss. In the months with missing values, the TWSs were filled using simple linear interpolation. In addition, the long-term mean of the TWS (January 2003 to December 2014) was removed from each monthly TWS to obtain the monthly TWS anomalies (TWSA).

In addition to direct TWS monitoring, there are many hydrological models, including WGHM, GLDAS, and PCR-GLOBWB, providing a TWS component. Among them, GLDAS and PCR-GLOBWB were used in the present study.

GLDAS assimilates large quantities of observational meteorological data to constrain modelled outputs, resulting in accurate estimates of many hydrological processes (Rodell et al. 2004). GLDAS drives four land surface models (LSMs): Mosaic, Noah, Community Land Model (CLM), and Variable Infiltration Capacity (VIC). Monthly one-degree four LSMs of GLDAS version 1, available at NASA's website (http://disc.sci.gsfc.nasa.gov/services/grads-gds/gldas), were used in this study covering the same period as the GRACE data. The GLDAS TWS components include different depths of soil moisture, canopy water, and snow water equivalent. However, the groundwater and surface water are not included. The GLDAS TWS was constructed as the sum of all the available components.

PCR-GLOBWB is a large-scale hydrological model intended for global to regional studies. PCR-GLOBWB provides a grid-based representation of terrestrial hydrology with a typical spatial resolution of 0.5 × 0.5 globally (Wada et al. 2010, 2011; van Beek et al. 2011). Because the PCR-GLOBWB TWS (PCR TWS) covers all storage components vertically from surface water to groundwater, the component contributions (CCs) to the TWS variation can be calculated. In this study, PCR-GLOBWB version 1 was used for a comparative analysis. The data covering the years 2003–2012 were obtained from Global Earth Observation for Integrated Water Resource Assessment (eartH2Observe, https://wci.earth2observe.eu/). Note that these data are in different time periods from other datasets.

Precipitation and ET

The Tropical Rainfall Measuring Mission (TRMM) is a joint NASA and JAXA mission to monitor and study tropical rainfall and climate. The validation of the TRMM product has been carried out throughout the world (Adeyewa & Nakamura 2003; Dinku et al. 2007; Karaseva et al. 2011; Duan et al. 2012; Ji & Chen 2012; Yatagai et al. 2013). The monthly precipitation data of TRMM 3B43 Version 7 (http://disc.sci.gsfc.nasa.gov/precipitation/), covering the period from 2003 to 2014, were used in this study. This dataset was obtained by combining satellite information from the passive microwave imager (TMI) and precipitation radar onboard TRMM, the Visible and Infrared Scanner onboard the Special Sensor Microwave Imager, and rain gauge observations. This product provides the rainfall estimates at a 0.25 × 0.25 grid.

The Moderate Resolution Imaging Spectroradiometer (MODIS) ET products, MOD16 (http://www.ntsg.umt.edu/project/mod16), were used to estimate the regional actual ET. The areal actual ET products from BOM, Australia (http://www.bom.gov.au), were also used for comparison. The BOM ET product is based on the Morton's method (Morton 1983) using a series of adjustments with regional annual rainfall data (Chiew & Leahy 2003).

In situ groundwater levels

Groundwater level data were obtained from the Australian National Groundwater Information System. The monitoring bores are unevenly distributed across the GAB, covering all or part of the periods from 2002 to 2014, and almost all the available bores are located in the Murray–Darling basin, southeastern GAB. Only those bores (about 1389) were selected (Figure 1) that have at least six monitoring records in 7 years. These monitoring bores were further used for the detection of trend in ground water level at point scale. The study period (2002–2014) was divided into three sub periods for detailed analysis: 2003–2006, 2007–2010, and 2011–2014.

Trend analysis and correlation analysis

The least square fitting analysis was performed to establish a straight line fit over time series of hydrological variables at basin and pixel level to identify the change in their trends.

The Pearson correlation coefficient was used to understand the strength of agreement between GRACE-derived TWS and hydrological model-derived TWS. A regression fitting analysis was also applied to understand the relationship and the variability between time series of GRACE-derived TWS and hydrological model-derived TWS.

Contribution of TWS components

In order to assess the role of individual storage components in TWS variations, the CC was calculated by the equation 
formula
1
where and are the individual components and TWS variations at time t, respectively; T is the total length of the time series.

Deriving GWS from GRACE and hydrological model

Many researchers (Voss et al. 2013; Awange et al. 2014; Chen et al. 2016) have demonstrated that groundwater can be successfully isolated from the GRACE TWS. The estimation of GWS variation can be approximated using the water balance equation, given as below: 
formula
2
where is estimated groundwater storage; is GRACE observed terrestrial water storage; and , , , and are hydrological model simulated soil moisture, surface water storage, snow water equivalent, and canopy water storage, respectively. GRACE cannot measure exact water storage amounts from space, but relative changes in water storage (Eamus et al. 2015). For this reason, all the terms in Equation (2) are expressed as anomalies with respect to their long-term mean (2003–2014).
In the GAB, surface water and canopy water contribute only about 3% of TWS variations. Therefore, the contribution of surface water and canopy water storage can be neglected in computing GWS. Similarly, the snow water equivalent, which contributes just a little compared to other contributors, can be ignored. So Equation (2) can be simplified as: 
formula
3

RESULTS

Spatio-temporal TWS change derived from GRACE

Monthly regional mean TWSA were calculated from GRACE solutions, as shown in Figure 2. For the last 12 years (2003–2014), the linear trend of TWS changes (TWSCs, spatial mean) over GAB showed a slight increase of 6.28 mm/yr (an equivalent water height), which is about 10.68 km3/yr. Since the changes in TWS were not the same in the entire study period (2003–2014), the whole period was divided into three sub periods (i.e., 2003–2006, 2007–2010, 2011–2014) to explore more detail on the pattern of the changes. From January 2003 to December 2006, TWS showed a slight decreasing trend, with a loss of 3.51 mm/yr (5.97 km3/yr) in water storage. On the other hand, during 2007–2010, the TWSCs showed a rapid increase of 21.47 mm/yr (36.50 km3/yr) in the GAB. However, after the year 2011, a sharp decrease of −27.38 mm/yr (46.56 km3/yr) was observed in the study area.
Figure 2

Time series of GRACE TWSA, precipitation anomalies and Nino 3.4 indices.

Figure 2

Time series of GRACE TWSA, precipitation anomalies and Nino 3.4 indices.

Precipitation is the most important factor that can affect the TWS. By comparing the anomalies of precipitation and TWS (Figure 2), more negative precipitation anomalies were observed before 2010, resulting in a slight decreasing trend in TWS. On the other hand, during 2010–2012, several continuous positive precipitation anomalies were explored, which resulted in an increase in TWS. However, a rapid decrease in the TWS was observed after 2013 due to several continuous negative precipitation anomalies.

In early 2008 and at the end of 2010, there were two strong La Nina events in Australia (Figure 2), which resulted in some extreme storms during late austral spring (2010) and early austral summer (2011). As a result, rainfall increased significantly in the GAB, causing a sharp increase in water storage.

Figure 3 shows the spatial patterns of the TWSCs and their confidence levels over the whole period as well as in three sub periods over the GAB. During the whole study period, almost all of the area of the GAB, especially in the eastern mountainous parts, showed slight increasing trends in TWS. However, the increasing trends in the far north and west corner were not statistically significant at 95% confidence level (P < 0.05). For three sub periods, the overall spatial TWS changing trends showed the same patterns as the basin average. There was a slight decreasing trend in most areas during 2003–2006. The western margin area, piedmont of the Great Dividing Range, and the far north Cape York Peninsula, showed a rapidly decreasing rate, of 10–20 mm/yr, while the other areas showed a slight decrease of 0–10 mm/yr. A recovery trend was observed from 2007–2010. The eastern areas recovered faster than the western. In contrast, the far north regions still showed a decreasing trend. After 2010, TWS decreased sharply except in some areas in the southwest of the basin. The far north showed an even greater decreasing trend, at 60 mm/yr. It was observed that the confidence levels (Figure 3) of TWS changing trends most of the time were statistically significant at 95% confidence level (P < 0.05). However, the northern areas during all the periods (the entire period and three sub periods), and the western areas during the period of 2003–2006, showed insignificant trends in the GAB.
Figure 3

Spatial distribution of TWS change trends (top) and confidence levels (bottom).

Figure 3

Spatial distribution of TWS change trends (top) and confidence levels (bottom).

Intra-annual characteristics of GRACE-derived TWS

The TWSC has an obviously seasonal characteristic. The periodic signals can be obtained by analyzing the intra-annual distributions of the GRACE TWS, precipitation, and ET.

Figure 4 shows the maximum value of TWS in March and minimum in November. On the other hand, the maximum and minimum values of both precipitation and the BOM-derived ET occurred in January and August, respectively. The extreme (maximum and minimum) months of the TWS were 2–3 month lags behind the extreme months for precipitation. The ET values estimated from MODIS were smaller than BOM, especially from October to March, with a difference of even more than 15 mm. The net precipitation was calculated by subtracting ET from precipitation. The coefficients of determination R2 are 0.89 and 0.90 for two different ET products (MOD ET and BOM ET). The TWSC was also calculated by subtracting the TWS of the previous month from the TWS of the current month. The TWSC variability was consistent with net precipitation variability (Figure 4, right). Suppose a long term annual TWS is stable (TWSC equals 0), the average annual net outflow from the GAB to the boundary is about 454.92 km3 from MOD and 144.77 km3 from BOM. The net outflow from TRMM and MOD is apparently higher than that from TRMM and BOM. The probably reason is the MOD16 ET values are underestimated over the GAB. The MOD16 algorithm estimate the ET based on the vegetation information and it does not consider the water body (Alemu et al. 2014; Penatti et al. 2015). Similar results are shown by other researchers in other areas (Vinukollu et al. 2011; Long et al. 2014b; Mohammadi et al. 2015; Penatti et al. 2015). However, an accurate estimation of ET is still a great challenge.
Figure 4

Intra-annual variability of GRACE TWS, precipitation, and ET (left) and net precipitation as well as GRACE TWSC (right).

Figure 4

Intra-annual variability of GRACE TWS, precipitation, and ET (left) and net precipitation as well as GRACE TWSC (right).

Figure 5 demonstrates the months having maximum and minimum GRACE TWS values in different parts of the GAB. It is clear that different areas reached their TWS peaks in different months. The northern regions reached their maximum TWS in February and March, and southern areas in April, June, and July. The months having minimum TWS showed similar patterns as maximum with the following order: the northern, eastern and southern areas. The maximum TWS values mostly occurred in March and the minimum TWS was in January.
Figure 5

Spatial distribution of maximum (a) and minimum (b) GRACE TWS occurring months.

Figure 5

Spatial distribution of maximum (a) and minimum (b) GRACE TWS occurring months.

Comparison with hydrological models

The pixel-based Pearson correlation coefficients were analyzed between the GRACE TWS and TWS derived from five hydrological models (i.e., CLM2.0, Mosaic, Noah2.7, VIC, and PCR-GLOBWB) and are shown in Figure 6. Generally, the hydrological model-derived TWS correlated well with the GRACE TWS in most areas. In the northern and south-western regions, the correlation coefficients between GRACE TWS and GLDAS models showed high correlations, with correlation coefficients greater than 0.6. The area with low correlation coefficients derived from GLDAS models extended from the central GAB to the south-eastern Murray–Darling basin. The PCR-GLOBWB derived TWS correlate well with GRACE TWS for the entire GAB. Except for some central, western, and northern areas, the other places showed high correlations.
Figure 6

Spatial distributions of correlation coefficients between TWS derived from GRACE and five hydrological models.

Figure 6

Spatial distributions of correlation coefficients between TWS derived from GRACE and five hydrological models.

Figure 7 shows the comparison of the monthly spatial averaged TWS from GRACE and hydrological models. The TWS from the Noah model showed the best correlation with the GRACE TWS. The correlation coefficients between the GRACE TWS and the hydrological models, CLM2.0, MOSAIC, NOAH2.7, VIC, and PCR-GLOBWB, were 0.73, 0.79, 0.86, 0.67, and 0.85, respectively.
Figure 7

Correlation coefficients between monthly basin averaged TWS derived from GRACE and five hydrological models.

Figure 7

Correlation coefficients between monthly basin averaged TWS derived from GRACE and five hydrological models.

GWS derived from GRACE and hydrological model

In the present study, it was assumed that the changes in surface water (e.g., lakes, reservoirs, and rivers) were negligible. The GWS anomalies were obtained by subtracting the GLDAS-Noah TWSA from the GRACE TWSA. So in this section, GWS was estimated from GRACE and the GLDAS-Noah model using Equation (3) because the GLDAS-Noah model showed the best correlations with the GRACE TWS.

Figure 8 illustrates the time series of the regional averaged GRACE-TWSA (the deviation from the study period mean), the GLDAS-Noah total water storage anomalies, and the estimated GWS. The GRACE-TWSA and GLDAS-Noah total water storage anomalies showed similar changing trends, and the GWS showed a slight fluctuation in the GAB during the study period. The estimated GWS displayed a slight increase of 4.77 mm/yr (8.11 km3/yr).
Figure 8

Time series of the regional averaged GRACE TWSA, the GLDAS-Noah total water storage anomalies and estimated GWS.

Figure 8

Time series of the regional averaged GRACE TWSA, the GLDAS-Noah total water storage anomalies and estimated GWS.

Figure 9 shows the linear trends of estimated GWS from GRACE and PCR-GLOBWB. Both models showed an overall increasing trend in the GAB. However, the PCR-GLOBWB model did not cover the years 2013 and 2014. Both GWS estimations indicated an increase in the east and decrease in the west. The GWS estimated from GRACE showed a more increasing trend in the piedmont of the Great Dividing Range but a decreasing trend in Lake Eyre and the southern regions. On the whole, these trends were statistically significant in most parts of the GAB, except in the northern region.
Figure 9

Spatial distribution of estimated GWS change trends (left) and confidence levels (right). Note that the PCR-GLOBWB only covers the period of 2003–2012.

Figure 9

Spatial distribution of estimated GWS change trends (left) and confidence levels (right). Note that the PCR-GLOBWB only covers the period of 2003–2012.

In order to elaborate the contributions of individual components (i.e., surface water) to the variations of total TWS, the CCs were investigated using Equation (1). For the whole basin, the estimated GWS contributed about 26.4% to the total TWS variations, and soil moisture contributed about 73.6%. This means that the shallow soil moisture contributes to the total TWS about three times more than GWS.

The PCR-GLOBWB model outputs include all vertical water storage layers of the earth surface, so it is a good choice to analyze the different CCs to the total TWS variations (Figure 10). It was concluded that the soil moisture's contribution was dominant, especially in most parts of the western arid regions, in surface water and groundwater. Groundwater rarely contributed to TWS variations in the west; by contrast, it greatly contributed in the eastern mountainous areas and even much more in the northern coastal areas. In the east, precipitation was greater compared to the west, and this is an important groundwater recharge area, where the groundwater contributes more to TWS variations. The great groundwater contributions over the northern coastal area were mainly due to more precipitation and sea water seepage. Surface water contributed greatly in the central Warburton River, the Cooper River, the Eyre River, and the eastern Darling River. Over the GAB as a whole, 70.1% of TWS change was due to soil moisture, 26.8% due to groundwater, and 3.1% due to surface water. These were consistent with GRACE estimates of CCs.
Figure 10

CCs to the total TWS variations. Soil moisture (a), groundwater (b), surface water (c). Blue lines in (c) represent major rivers. The full colour version of this figure is available in the online version of this paper, at http://dx.doi.org/10.2166/ws.2016.136.

Figure 10

CCs to the total TWS variations. Soil moisture (a), groundwater (b), surface water (c). Blue lines in (c) represent major rivers. The full colour version of this figure is available in the online version of this paper, at http://dx.doi.org/10.2166/ws.2016.136.

In situ groundwater level observations

The groundwater over-extraction and declining natural flows are more serious in the southern Queensland portion of the Murray–Darling Basin (Murray-Darling Basin Commission (Australia) 1999). The in situ groundwater bores are mostly located in these areas (Figure 1). Although these areas only cover the south-eastern region, it is meaningful to analyze the in situ groundwater level change trend.

The whole period (2003–2014) and three sub periods (i.e., 2003–2006, 2007–2010, and 2011–2014) were analyzed to investigate the changing trends. The monitored groundwater change trends showed the same patterns as the GRACE derived TWS, as illustrated in Table 1 and Figure 11. For the whole period (2003–2014), 74% of bores showed increasing trends in the observed groundwater levels. On the other hand, 68% and 51% of bores showed decreasing trends in 2003–2006 and 2011–2014, respectively, and 68% of the bores showed increasing trends in 2007–2010. These bores corresponding to the periods were mostly statistically significant except for the period of 2011–2014. More than 40% of the bores showed changing trends between −0.2 m and 0.2 m for all the periods. Assuming a specific yield of 0.1, the equivalent water heights of groundwater change trends might range between −20 mm/yr and 20 mm/yr. It was also noted that the GRACE derived TWS was also within this range. This indicates that GRACE can detect groundwater water storage changes in this area.
Table 1

Statistics of point-scale groundwater changing trends

 Periods
2003–20142003–20062007–20102011–2014
All used bores (AUB) 1,380 1,369 1,368 1,187 
Increasing trend bores (ITB) 1,022 437 930 578 
Decreasing trend bores (DTB) 358 932 438 609 
Statistically significant increasing trend bores (SITB) 870 181 698 373 
Statistically significant decreasing trend bores (SDTB) 241 640 292 297 
Increasing trend bores ratio, ITB/AUB 74% 32% 68% 49% 
Decreasing trend bores ratio, DTB/AUB 26% 68% 32% 51% 
Statistically significant increasing trend bores ratio, SITB/ITB 85% 41% 75% 65% 
Statistically significant decreasing trend bores ratio, SDTB/DTB 67% 69% 67% 49% 
 Periods
2003–20142003–20062007–20102011–2014
All used bores (AUB) 1,380 1,369 1,368 1,187 
Increasing trend bores (ITB) 1,022 437 930 578 
Decreasing trend bores (DTB) 358 932 438 609 
Statistically significant increasing trend bores (SITB) 870 181 698 373 
Statistically significant decreasing trend bores (SDTB) 241 640 292 297 
Increasing trend bores ratio, ITB/AUB 74% 32% 68% 49% 
Decreasing trend bores ratio, DTB/AUB 26% 68% 32% 51% 
Statistically significant increasing trend bores ratio, SITB/ITB 85% 41% 75% 65% 
Statistically significant decreasing trend bores ratio, SDTB/DTB 67% 69% 67% 49% 
Figure 11

Point-scale observed groundwater level changing trends (a) and their trends distributions (b).

Figure 11

Point-scale observed groundwater level changing trends (a) and their trends distributions (b).

The relations between bore depths and groundwater changing trends are explained in Figure 12. It was explored that groundwater showed a trend tendency only in a bore depth of less than ∼50 meters. The groundwater trends showed normal distribution when the bore depths were greater than ∼50 meters. This indicates that shallow groundwater is more sensitive to climate change and anthropogenic activities.
Figure 12

Relations between groundwater bore depths and groundwater level change trends.

Figure 12

Relations between groundwater bore depths and groundwater level change trends.

DISCUSSION

In 1999, the Australian Government commenced a new program called the Great Artesian Basin Sustainability Initiative (Dowsley, 2013), whose main aim is to repair uncontrolled artesian bores and replace open earthen bore drains with piped water reticulation systems. The monitoring results indicate that this program has made some remarkable achievements (Great Artesian Basin Sustainability Initiative Phase 3 Mid-Term Review). However, these results are based on a limited amount of monitoring data and lack of overall assessment. The GRACE estimates, the main objective of this study, provide a clear picture of the distribution of TWS and GWS in the GAB.

This study highlights that the GRACE data, in combination with hydrological model and in situ groundwater monitoring data, play an important role in regional water resources assessment. Different datasets can verify each other and can get relatively convincing results. This study used GRACE and five hydrological models to estimates TWS and GWS; they mostly showed consistency with each other in TWS and GWS changing trends and spatial distribution. Although in situ groundwater level data only cover a small part of the GAB, the results also showed similar patterns to those from GRACE. Although estimates of TWS and GWS may subject to some biases from a number of sources (e.g., the GRACE data processing method, hydrological model simulation accuracy, in situ bore spatial coverage, and in situ measurement error), this study can be a good reference for related research and water resources management in the GAB.

Fasullo et al. (2016) comprehensively analyzed the anthropogenic and climate driven TWS changes from multi-decadal timescales, and they concluded that climate change is the main contributor to TWS variability. In this study, both total TWS (Figure 2) and GWS (Figure 8) derived from GRACE showed a slight increase in the GAB. TWS showed a slight fluctuation before 2010 and a rapid increase in 2010 and 2011. The main reason for this sharp fluctuation was two strong La Nina events in Australia in 2010 and 2011. The intensive precipitation due to La Nina events caused a rapid recovery of TWS. It confirmed that climate factors dominate the TWS variability over human activities in the GAB, thus it is inappropriate to link the TWS trends to the anthropogenic factors in this area.

Tangdamrongsub et al. (2016) estimated that GWS contributes 71.1% to the total TWS in the Tonlé Sap basin, Cambodia, and surface water contributes about 4.4% to total TWS. Pokhrel et al. (2013) estimated that soil and groundwater contribute 71% to the total TWS in the Amazon, and surface water contributes 29%. In contrast to the big proportion in these studies, the groundwater estimated from PCR-GLOBWB only contributes 26.8% (26.4% from GRACE) of total TWS variations in the GAB. The discrepancies are mainly due to the lack of rainfall in arid or semi-arid regions like the GAB. The Tonlé Sap basin and the Amazon are located in a wet area and a floodplain area, rainfall has more chance and time to enter the subsurface by seepage when it falls on the ground. In the GAB, most of the rainfall has evaporated before it enters deep underground due to the high temperature and low humidity. Therefore, in arid and semi-arid regions, groundwater contributes a relatively small fraction of total TWS variations.

The GRACE-TWS changing trends in the northern Cape York Peninsula were not statistically significant in the study period and all sub periods. The basin averaged TWS reached their peaks (maximum and minimum) always starting from the northern Cape York Peninsula and extending gradually to the south. Groundwater contributed more than 50% to the total TWS variations in the Cape York Peninsula. One possible reason for these abnormal phenomena is that there is more precipitation in this region, and this may have some relation with sea water changes. In order to figure out this possibility, sea surface height (SSH) data (http://sealevel.colorado.edu/) were used to investigate correlations with the groundwater derived from GRACE (Figure 13). The maximum correlation coefficient between SSH and GWS was more than 0.5 in the northern and central GAB, with 0–6 month lags from SSH. This indicates, though, that sea level in all directions (i.e., the southeastern Tasman Sea) does not have these relations (the figures are not given), GWS in the GAB has a certain relation with the neighboring sea level changes. However, an extensive study is needed on this in the future.
Figure 13

Maximum correlation coefficients (a) and its corresponding month lags (b) between monthly SSH and GWS. The red box represents the area used for monthly average SSH. The full colour version of this figure is available in the online version of this paper, at http://dx.doi.org/10.2166/ws.2016.136.

Figure 13

Maximum correlation coefficients (a) and its corresponding month lags (b) between monthly SSH and GWS. The red box represents the area used for monthly average SSH. The full colour version of this figure is available in the online version of this paper, at http://dx.doi.org/10.2166/ws.2016.136.

CONCLUSIONS

In this study, the temporal and spatial characteristics of the TWS were analyzed based on GRACE data, hydrological model simulations, and in situ groundwater level records in the GAB for a 12-year period (2003–2014).

The monthly basin averaged TWS showed a slight increase of 6.28 mm/yr equivalent water height, or 10.68 km3/yr in volume, over the study period. However, there was a rapid increase in the TWS in the years 2010 and 2011 due to two strong La Nina events. Before 2010 and after 2012, the TWS showed a slight fluctuation. Spatially, almost all the area of the GAB, especially in the eastern mountainous parts, showed slight increasing trends in TWS. In a year, the basin averaged TWS reached its maximum capacity in March and minimum in November, with 2–3 months' lag to precipitation. The TWSCs correlated very well with net precipitation (e.g., R2 = 0.89 and R2 = 0.90 for two different products). Different regions reached their peaks differently, for example the northern and eastern regions most of the time reached their peaks earlier than the southern and western regions.

From the comparison with hydrological models, GRACE-derived TWS showed good correlations with hydrological model estimates both temporally and spatially. The Noah and PCR-GLOBWB models showed the best correlations, with an R2 of 0.86 and 0.85, respectively.

The GWS showed slight fluctuation, with an increased rate of 4.77 mm/yr (8.11 km3/yr) for 2003–2014 from GRACE estimates over the GAB. Both the GWSs estimated from GRACE and PCR-GLOBWB showed a slight increase in the east and decrease in the west.

During the CC investigation, it was found that, in the whole GAB, the groundwater derived from GRACE only contributed 26.4% to the total TWS variations, while 26.8% was simulated from PCR-GLOBWB. On the other hand, soil moisture almost contributed three-fourths of the TWS variations, but surface water was estimated to contribute only 3.1% to the TWS. For most parts of the GAB, especially the arid west region, surface water was the dominant contributor to the TWS variations. On the other hand, groundwater was the main contributor in the north Cape York Peninsula and surface water contributed more than 60% in Lake Eyre and the Cooper River.

Estimates from in situ bore data (only covering the south-eastern region) showed similar trends to the GRACE-derived GWS. The relations between bore depth and groundwater changing trends explored that the shallow groundwater, less than 50-meter in depth, was more likely to have a trend tendency than the deep groundwater.

The northern Cape York Peninsula showed insignificant trends in the TWS. More groundwater contribution to reach monthly peaks indicated that this region might be influenced by some other factors than precipitation. The time-lagged correlation analysis between SSH and GRACE-derived GWS showed some strong linkages between groundwater and sea level changes.

ACKNOWLEDGEMENTS

This study was supported by National Natural Science Foundation of China (41471463). GRACE land are available at http://grace.jpl.nasa.gov, supported by the NASA MEaSUREs Program. Last but not least, the authors offer gratitude to the reviewers and the editor for their valuable comments to improve the quality of this article.

REFERENCES

REFERENCES
Abelen
S.
Seitz
F.
Abarca-del-Rio
R.
Güntner
A.
2015
Droughts and floods in the La Plata basin in soil moisture data and GRACE
.
Remote Sensing
7
(
6
),
7324
7349
.
Adeyewa
Z. D.
Nakamura
K.
2003
Validation of TRMM radar rainfall data over major climatic regions in Africa
.
Journal of Applied Meteorology
42
(
2
),
331
347
.
Awange
J. L.
Gebremichael
M.
Forootan
E.
Wakbulcho
G.
Anyah
R.
Ferreira
V. G.
Alemayehu
T.
2014
Characterization of Ethiopian mega hydrogeological regimes using GRACE, TRMM and GLDAS datasets
.
Advances in Water Resources
74
,
64
78
.
Billah
M. M.
Goodall
J. L.
Narayan
U.
Reager
J. T.
Lakshmi
V.
Famiglietti
J. S.
2015
A methodology for evaluating evapotranspiration estimates at the watershed-scale using GRACE
.
Journal of Hydrology
523
,
574
586
.
Chen
J. L.
Famiglietti
J. S.
Scanlon
B. R.
Rodell
M.
2015
Groundwater storage changes: present status from GRACE observations
.
Surveys in Geophysics
37
(
2
),
397
417
.
Chen
J. L.
Wilson
C. R.
Tapley
B. D.
Scanlon
B.
Güntner
A.
2016
Long-term groundwater storage change in Victoria, Australia from satellite gravity and in situ observations
.
Global and Planetary Change
139
,
56
65
.
Chiew
F. H. S.
Leahy
C. P.
2003
Comparison of evapotranspiration variables in Evapotranspiration maps for Australia with commonly used evapotranspiration variables
.
Australian Journal of Water Resources
7
(
1
),
1
11
.
Dinku
T.
Ceccato
P.
Grover-Kopec
E.
Lemma
M.
Connor
S. J.
Ropelewski
C. F.
2007
Validation of satellite rainfall products over East Africa's complex topography
.
International Journal of Remote Sensing
28
(
7
),
1503
1526
.
Dowsley
K.
2013
Great Artesian Basin Sustainability Initiative Phase 3 Mid-Term Review. Australian government, Canberra
.
Duan
Z.
Bastiaanssen
W. G. M.
Liu
J.
2012
Monthly and annual validation of TRMM Multisatellite Precipitation Analysis (TMPA) products in the Caspian Sea Region for the period 1999–2003
. In:
Geoscience and Remote Sensing Symposium (IGARSS), 2012 IEEE International
, pp.
3696
3699
.
Eamus
D.
Zolfaghar
S.
Villalobos-Vega
R.
Cleverly
J.
Huete
A.
2015
Groundwater-dependent ecosystems: recent insights from satellite and field-based studies
.
Hydrology and Earth System Sciences
19
(
10
),
4229
4256
.
Famiglietti
J. S.
2014
The global groundwater crisis
.
Nature Climate Change
4
(
11
),
945
948
.
Fasullo
J. T.
Lawrence
D. M.
Swenson
S. C.
Fasullo
J. T.
Lawrence
D. M.
Swenson
S. C.
2016
Are GRACE-era terrestrial water trends driven by anthropogenic climate change?
Advances in Meteorology, Advances in Meteorology
2016
,
e4830603
.
Frappart
F.
Ramillien
G.
Leblanc
M.
Tweed
S. O.
Bonnet
M.-P.
Maisongrande
P.
2011
An independent component analysis filtering approach for estimating continental hydrology in the GRACE gravity data
.
Remote Sensing of Environment
115
(
1
),
187
204
.
Getirana
A. C. V.
Dutra
E.
Guimberteau
M.
Kam
J.
Li
H.-Y.
Decharme
B.
Zhang
Z.
Ducharne
A.
Boone
A.
Balsamo
G.
Rodell
M.
Toure
A. M.
Xue
Y.
Peters-Lidard
C. D.
Kumar
S. V.
Arsenault
K.
Drapeau
G.
Leung
L. R.
Ronchail
J.
Sheffield
J.
2014
Water balance in the Amazon Basin from a land surface model ensemble
.
Journal of Hydrometeorology
15
(
6
),
2586
2614
.
Huang
Z.
Pan
Y.
Gong
H.
Yeh
P. J.-F.
Li
X.
Zhou
D.
Zhao
W.
2015
Subregional-scale groundwater depletion detected by GRACE for both shallow and deep aquifers in North China Plain
.
Geophysical Research Letters
42
(
6
),
1791
1799
.
Karaseva
M. O.
Prakash
S.
Gairola
R. M.
2011
Validation of high-resolution TRMM-3B43 precipitation product using rain gauge measurements over Kyrgyzstan
.
Theoretical and Applied Climatology
108
(
1–2
),
147
157
.
Leblanc
M. J.
Tregoning
P.
Ramillien
G.
Tweed
S. O.
Fakes
A.
2009
Basin-scale, integrated observations of the early 21st century multiyear drought in southeast Australia
.
Water Resources Research
45
(
4
),
W04408
.
Leblanc
M. J.
Tweed
S. O.
Ramillien
G.
Tregoning
P.
Frappart
F.
Fakes
A.
Cartwright
I.
2011
Groundwater change in the Murray basin from long-term in-situ monitoring and GRACE estimates
. In:
Climate Change Effects on Groundwater Resources (Vols 1–0)
,
CRC Press
.
Long
D.
Shen
Y.
Sun
A.
Hong
Y.
Longuevergne
L.
Yang
Y.
Li
B.
Chen
L.
2014a
Drought and flood monitoring for a large karst plateau in Southwest China using extended GRACE data
.
Remote Sensing of Environment
155
,
145
160
.
Long
D.
Longuevergne
L.
Scanlon
B. R.
2014b
Uncertainty in evapotranspiration from land surface modeling, remote sensing, and GRACE satellites
.
Water Resources Research
50
(
2
),
1131
1151
.
Long
D.
Longuevergne
L.
Scanlon
B. R.
2015
Global analysis of approaches for deriving total water storage changes from GRACE satellites
.
Water Resources Research
51
(
4
),
2574
2594
.
Mohammadi
A.
Costelloe
J. F.
Ryu
D.
2015
Evaluation of remotely sensed evapotranspiration products in a large scale Australian arid region
,
Cooper Creek
,
Queensland. 21st International Congress on Modelling and Simulation, Broadbeach, Queensland
.
Morton
F. I.
1983
Operational estimates of areal evapotranspiration and their significance to the science and practice of hydrology
.
Journal of Hydrology
66
(
1
),
1
76
.
Munier
S.
Becker
M.
Maisongrande
P.
Cazenave
A.
2012
Using GRACE to detect groundwater storage variations: the cases of Canning basin and Guarani aquifer system
.
International Water Technology Journal
2
,
2
13
.
Murray-Darling Basin Commission (Australia)
1999
Murray-Darling Basin Groundwater: a Resource for the Future
.
The Commission
,
Canberra
.
Penatti
N. C.
de Almeida
T. I. R.
Ferreira
L. G.
Arantes
A. E.
Coe
M. T.
2015
Satellite-based hydrological dynamics of the world's largest continuous wetland
.
Remote Sensing of Environment
170
,
1
13
.
Pokhrel
Y. N.
Fan
Y.
Miguez-Macho
G.
Yeh
P. J.-F.
Han
S.-C.
2013
The role of groundwater in the Amazon water cycle: 3. Influence on terrestrial water storage computations and comparison with GRACE
.
Journal of Geophysical Research: Atmospheres
118
(
8
),
3233
3244
.
Richey
A. S.
Thomas
B. F.
Lo
M.-H.
Reager
J. T.
Famiglietti
J. S.
Voss
K.
Swenson
S.
Rodell
M.
2015
Quantifying renewable groundwater stress with GRACE
.
Water Resources Research
51
(
7
),
5217
5238
.
Rodell
M.
Houser
P. R.
Jambor
U.
Gottschalck
J.
Mitchell
K.
Meng
C.-J.
Arsenault
K.
Cosgrove
B.
Radakovich
J.
Bosilovich
M.
Entin
J. K.
Walker
J. P.
Lohmann
D.
Toll
D.
2004
The global land data assimilation system
.
Bulletin of the American Meteorological Society
85
(
3
),
381
394
.
Rodell
M.
Velicogna
I.
Famiglietti
J. S.
2009
Satellite-based estimates of groundwater depletion in India
.
Nature
460
(
7258
),
999
1002
.
Seoane
L.
Ramillien
G.
Frappart
F.
Leblanc
M.
2013
Regional GRACE-based estimates of water mass variations over Australia: validation and interpretation
.
Hydrology and Earth System Sciences
17
(
12
),
4925
4939
.
Swenson
S.
Wahr
J.
2002
Methods for inferring regional surface-mass anomalies from Gravity Recovery and Climate Experiment (GRACE) measurements of time-variable gravity
.
Journal of Geophysical Research: Solid Earth
107
(
B9
),
2193
.
Swenson
S.
Wahr
J.
2006
Post-processing removal of correlated errors in GRACE data
.
Geophysical Research Letters
33
(
8
),
L08402
.
Swenson
S.
Wahr
J.
2009
Monitoring the water balance of lake Victoria, East Africa, from space
.
Journal of Hydrology
370
(
1–4
),
163
176
.
Syed
T. H.
Famiglietti
J. S.
Chen
J.
Rodell
M.
Seneviratne
S. I.
Viterbo
P.
Wilson
C. R.
2005
Total basin discharge for the Amazon and Mississippi River basins from GRACE and a land-atmosphere water balance
.
Geophysical Research Letters
32
(
24
),
L24404
.
Syed
T. H.
Webster
P. J.
Famiglietti
J. S.
2014
Assessing variability of evapotranspiration over the Ganga river basin using water balance computations
.
Water Resources Research
50
(
3
),
2551
2565
.
Thomas
A. C.
Reager
J. T.
Famiglietti
J. S.
Rodell
M.
2014
A GRACE-based water storage deficit approach for hydrological drought characterization
.
Geophysical Research Letters
41
(
5
),
1537
1545
.
van Beek
L. P. H.
Wada
Y.
Bierkens
M. F.
2011
Global monthly water stress: 1. Water balance and water availability
.
Water Resources Research
47
(
7
),
W07517
.
Wada
Y.
van Beek
L. P. H.
van Kempen
C. M.
Reckman
J. W. T. M.
Vasak
S.
Bierkens
M. F. P.
2010
Global depletion of groundwater resources
.
Geophysical Research Letters
37
(
20
),
L20402
.
Wada
Y.
Van Beek
L. P. H.
Viviroli
D.
Dürr
H. H.
Weingartner
R.
Bierkens
M. F.
2011
Global monthly water stress: 2. Water demand and severity of water stress
.
Water Resources Research
47
(
7
),
W07518
.
Wang
H.
Guan
H.
Gutiérrez-Jurado
H. A.
Simmons
C. T.
2014
Examination of water budget using satellite products over Australia
.
Journal of Hydrology
511
,
546
554
.