Modelling climate-change impacts on groundwater recharge in the Murray-Darling Basin, Australia Russell S. Crosbie & James L. McCallum & Glen R. Walker & Francis H. S. Chiew Abstract A methodology is presented for assessing the average changes in groundwater recharge under a future climate. The method is applied to the 1,060,000km2 Murray-Darling Basin (MDB) in Australia. Climate sequences were developed based upon three scenarios for a 2030 climate relative to a 1990 climate from the outputs of 15 global climate models. Dryland diffuse groundwater recharge was modelled in WAVES using these 45 climate scenarios and fitted to a Pearson Type III probability distribution to condense the 45 scenarios down to three: a wet future, a median future and a dry future. The use of a probability distribution allowed the significance of any change in recharge to be assessed. This study found that for the median future, climate recharge is projected to increase on average by 5% across the MDB but this is not spatially uniform. In the wet and dry future scenarios the recharge is projected to increase by 32% and decrease by 12% on average across the MDB, respectively. The differences between the climate sequences generated by the 15 different global climate models makes it difficult to project the direction of the change in recharge for a 2030 climate, let alone the magnitude. Keywords Australia . Climate change . Groundwater recharge/water budget
Received: 23 July 2009 / Accepted: 14 June 2010 Published online: 6 July 2010 © Springer-Verlag 2010 R. S. Crosbie ()) : J. L. McCallum : G. R. Walker CSIRO Water for a Healthy Country National Research Flagship, CSIRO Land and Water, PMB 2, Glen Osmond, SA 5064, Australia e-mail:
[email protected] Tel.: +61-8-83038751 Fax: +61-8-83038750 F. H. S. Chiew CSIRO Water for a Healthy Country National Research Flagship, CSIRO Land and Water, GPO Box1666, Canberra, ACT 2601, Australia Hydrogeology Journal (2010) 18: 1639–1656
Introduction Climate change has been defined as “a change of climate which is attributed directly or indirectly to human activity that alters the composition of the global atmosphere and which is in addition to natural climate variability observed over comparable time periods” (UNFCCC 1992). This definition is wider than the often used ‘global warming’ as it encompasses all aspects of climate, e.g. temperature, precipitation, solar radiation, humidity etc. Global climate models (GCMs) are used to predict change in climate from increased CO2 (and other greenhouse gases) in the atmosphere. There is considerable uncertainty about the predictions made by GCMs and it is a common practice to use the outputs of multiple models to assess scenarios. Climate drives the hydrological cycle and any change in climate will have consequential changes in the hydrological cycle. Precipitation is the largest term in the water balance of a catchment and as such the hydrological cycle is most sensitive to changes in rainfall. In sub-humid, semi-arid and arid regions, evapotranspiration (ET) is the second largest component of the water balance and so is affected not only by the availability of water but also by any changes in other aspects of climate (e.g. temperature). Recharge is often the smallest component of the water balance and is often calculated as the residual after subtracting ET and runoff from precipitation. Recharge is the addition of water to the groundwater store. There are many different mechanisms of groundwater recharge, including: leaking pipes in urban areas, the leaching fraction of water applied in irrigation areas, losing streams, and recharge from rainfall infiltration through both diffuse and preferential pathways. This study only considers dryland diffuse recharge. For this study, dryland diffuse recharge is defined as water that infiltrates the soil from rainfall (not irrigation), passes through the root zone and soil matrix, and then crosses the plane of the water table. There are few techniques available to estimate recharge on a regional basis. Scanlon et al. (2002) showed that of the field based techniques, tracers had the greatest spatial scale. Tracers produce an integrated estimate of recharge over the life of the tracer so this could be an estimate of recharge on the order of millennia depending upon the DOI 10.1007/s10040-010-0625-x
1640
tracer chosen. Management of groundwater resources requires that the recharge to the groundwater be known on finer scales; therefore, modelling is often the chosen method of estimating recharge. Many different modelling approaches have been used to estimate regional recharge. At its simplest, recharge can be modelled as a fixed proportion of rainfall. More complex methods include the inverse modelling of numerical groundwater models (D’Agnese et al. 1999), analysis of streamflow records (Delin et al. 2007) and geographic information system (GIS) based approaches (Szilagyi et al. 2005). The most complex models for estimating recharge are soil-vegetation-atmosphere-transfer (SVAT) models; these have the advantage of being process-based models that simulate the growth of vegetation and the routing of water through the soil zone. SVAT models are waterbalance models; they take rainfall (some models include irrigation) as inputs to the land surface and partition this input into evapotranspiration, runoff and recharge. SVAT models vary widely in their complexity with regard to how they treat vegetation growth and soil physics. Most SVAT models are point or one-dimensional (1-D) models. To be useful in a spatial context their use needs to be extended to all points of the landscape. Some approaches rely on 1-D models embedded in a GIS system to allow the 1-D model to be run on a grid (e.g. Beverly et al. 2005; Littleboy et al. 2003), while fully 3-D models are rare (e.g. Tuteja et al. 2005). SVAT models run spatially are computationally expensive and require a large amount of data. Another alternative is to upscale the 1-D model results from the point scale to the working scale using co-variates. This approach is more often used with field measurements (Brunner et al. 2004; Cook et al. 1989; Sophocleous 1992). Most studies investigating the climate-change impact on groundwater recharge have been small in scale and so do not represent more than a pixel in a GCM (100– 400 km in the horizontal direction).There are few studies greater than 5,000 km 2 reported in the scientific literature. Jyrkama and Sykes( 2007) describe a study in the 7,000-km2 Grand River Watershed in Ontario, Canada. In that study, the general predictions from the (IPCC 2001) were used to find that an increase in rainfall leads to an increase in recharge, although this increase is not spatially uniform due to differences in soils and land use. Serrat-Capdevila et al. (2007) evaluated the outputs from 17 GCMs for four different scenarios to select five climate sequences to model recharge and its effect on the riparian system of the 7,560-km2 San Pedro Basin (Arizona/Sonora, USA/Mexico), this study found that decreased rainfall leads to decreased recharge and decreased baseflow in the river. In the 15,000-km2 Edwards BFZ (Balcones Fault Zone) Aquifer in Texas (USA), Loáiciga et al.( 2000) used a single pixel from a single GCM to derive a scaling factor for runoff that was then applied to recharge (in the aquifer, recharge is derived from losing streams). The study found that the predicted decrease in streamflow and increase in demand would place excessive stress on the aquifer leading to decreased security Hydrogeology Journal (2010) 18:1639–1656
of water supply, decreased water quality and a detrimental impact on groundwater dependent ecosystems. The largest area reported for climate-change impacts on recharge is the 450,000-km2 Ogallala aquifer in central USA (Rosenberg et al. 1999). Due to the large scale of the study, the same GCM could predict an increase in precipitation in one river basin and a decrease in another river basin. Overall, every scenario investigated resulted in less recharge leading to an accelerated decline in aquifer pressure in an aquifer already being mined. This study aims to provide a methodology for estimating the change in diffuse groundwater recharge with a change in climate and to use this methodology to estimate the impact of climate change on the groundwater recharge to the MDB. All 15 GCMs, with model runs archived by the Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4; IPCC 2007) that have daily rainfall outputs readily available, have been used in this study.
The Murray-Darling Basin The MDB is Australia’s largest and most important river system. The MDB is home to about 2 million people and provides the potable water supply to another 1 million outside the basin. The MDB provides the water for 75% of Australia’s irrigation industry and this underpins the basin providing 34% of Australia’s agricultural output (ABS 2003). The basin is also home to several wetlands of international importance that are protected under international treaties (DFA 1974; DFAT 1986; Ramsar Convention Bureau 1971). The MDB covers 1,060,000 km2 or roughly one seventh of the Australian continent (see inset, Fig 1). Its climate varies from sub-tropical in the north, Mediterranean in the south, alpine in the south-east and semi-arid in the west. The mean annual rainfall, areal potential evapotranspiration and runoff averaged over the entire MDB are 457, 1,443 and 27 mm respectively. There is a clear east–west rainfall gradient, where rainfall is highest in the south-east (mean annual rainfall of more than 1,500 mm) and along the eastern perimeter, and lowest in the west (less than 300 mm). The east–west runoff gradient is much more pronounced than the rainfall gradient, with runoff in the south-east corner (mean annual runoff of more than 200 mm) and the eastern perimeter (20–80 mm) being much higher than elsewhere in the MDB (less than 10 mm in the western half; Chiew et al. 2008a; 2008b). According to the most recent IPCC report (Christensen et al. 2007), the impact of climate change upon the MDB will be higher average temperatures, greater potential evapotranspiration, lower mean annual rainfall in the south, and greater daily extreme rainfall events. These conclusions have been reinforced in a recent Australian study (CSIRO and BOM 2007) and the Murray-Darling Basin Sustainable Yields project (Chiew et al. 2008a). DOI 10.1007/s10040-010-0625-x
1641
Fig. 1 Location of the Murray-Darling Basin (MDB) in southeastern Australia and regions used for reporting, showing average annual rainfall (1895–2006). Also shown are the 20 points (A–T) used in the detailed modelling with WAVES (see Table 1)
Methods The method used in this study to investigate the impact of climate change on groundwater recharge is to compare the recharge estimates generated using the historical climate data to the recharge estimates generated from future climate scenarios. The results are reported as a recharge scaling factor (RSF) which is the ratio of the future climate recharge to the historical climate recharge. The steps involved in this modelling are summarised in the subsequent list and detailed in the following sections:
&
&
&
Point scale modelling. WAVES (Zhang and Dawes 1998) is used to model recharge at a series of points (20) throughout the MDB for the historical climate and the future climate generated from 15 GCMs and three global-warming scenarios for every combination of soil type (11) and vegetation type (3). Upscaling. The point scale modelling is used to create regression equations between annual average rainfall and annual average recharge for each combination of climate (46), soil (11) and vegetation (3). These regression equations are used to upscale the recharge to the entire MDB using rasters of annual average rainfall, soil type and vegetation type. These recharge rasters are then used to calculate RSF rasters (45). Aggregation. The 45 RSF rasters are fitted to a weighted Pearson Type III probability distribution to enable the
Hydrogeology Journal (2010) 18: 1639–1656
number of scenarios to be condensed down to three: a wet future, a median future, and a dry future.
Historical climate data The historical data for this study comes from the SILO Data Drill of the Queensland Department of Natural Resources and Water (Jeffrey et al. 2001). The SILO Data Drill provides data for 0.05° (~5 km) grid cells interpolated from point measurements made by the Australian Bureau of Meteorology. Daily time series of rainfall, temperature, vapour pressure deficit and solar radiation are used in this study.
Future climate data The outputs from GCMs cannot be directly used in hydrologic models due to issues of scale and accuracy. Lim and Roderick (2008) presented a summary of the GCM runs archived by the World Climate Research Programme’s (WCRP’s) Coupled Model Intercomparison Project phase 3 (CMIP3) multi-model dataset (Meehl et al. 2007) for the historic period (1970–1999) and the future period (2070–2099). This summary for the historic period showed that the average rainfall across Australia varied between 191 and 1,059 mm yr–1 depending upon which model was investigated, for this period the actual rainfall DOI 10.1007/s10040-010-0625-x
1642
across Australia averaged ~450 mm yr–1. There is also the issue of scale; the grid resolution of GCMs is in the order of 100s of kilometres, whereas hydrologic models require much finer resolution data. To overcome these limitations of GCMs the future climate projections are generally downscaled in some way for use in hydrologic models. Stochastic (Charles et al. 2004) and dynamic downscaling (Nunez and McGregor 2007) are very computationally intensive and simply could not have been done within the timeframe of this study. Therefore the simpler ‘pattern scaling’ technique was used to downscale the future climate results from the GCMs (Chiew et al. 2008a; 2009). Forty-five future climate variants, each with 112 years of daily climate sequences for 0.05° × 0.05° grid cells across the MDB are used. The future climate variants come from scaling the 1895–2006 historical climate data to represent the climate of around 2030 relative to 1990, based on analyses of 15 global climate models (GCMs) and three global-warming scenarios. The 15 GCMs were chosen because this represents all GCMs that have daily rainfall time series archived (Meehl et al. 2007). A future climate around the year 2030 was chosen for investigation because that timeframe is a bit longer than the current management time frames considered in this region (10 – 15 years). The method used to obtain the future climate series is the same as that used recently to investigate future runoff in the MDB (Chiew et al. 2009) and is explained in detail in Chiew et al. (2008a); a summary is given in the following:
&
&
&
&
Three global-warming scenarios for ~2030 relative to ~1990 are used: high (1.60°C), medium (1.03°C) and low (0.45°C). These three scenarios are inferred from the IPCC AR4 (IPCC 2007) and the latest climatechange projections for Australia (CSIRO and BOM 2007). They take into account uncertainties associated with greenhouse gas emissions, global climate sensitivity to greenhouse gas and amplification of climate change due to carbon cycle feedbacks. Archived monthly simulations from 15 IPCC 4AR (IPCC 2007) GCMs are analysed to estimate the change in rainfall, temperature, humidity and solar radiation per degree of global warming. Each GCM is analysed separately. Data from each of the four seasons are also analysed separately. The percent changes in the climate variables per degree of global warming for each of the four seasons from the 15 GCMs are then multiplied by the three levels of global warming to obtain the 45 sets of ‘seasonal scaling’ factors. The ‘seasonal scaling’ factors are then used to scale the historical daily climate data from 1895 to 2006 to obtain the 45 future climate variants, each with 112 years of daily climate data. The changes in the daily rainfall distribution are also considered by scaling different daily rainfall amounts differently.
The overall changes in rainfall are presented in Fig. 2 for the high-global-warming scenario (the low and Hydrogeology Journal (2010) 18:1639–1656
medium-global-warming scenarios are not shown here). This shows that the projections of rainfall are different between the 15 GCMs with some projecting an increase in rainfall and others projecting a decrease in rainfall. There is some consistency along the southern extremity of the MDB with nearly all GCMs projecting a decrease in rainfall.
Point scale modelling The model chosen for the recharge modelling in this study was WAVES. WAVES has been shown to be able to reproduce the water balance of field experiments in many studies in the MDB (Crosbie et al. 2008; Slavich et al. 1999; Zhang et al. 1999), the rest of Australia (Dawes et al. 2002; Salama et al. 1999; Xu et al. 2008) and throughout the world (Wang et al. 2001; Yang et al. 2003; Zhang et al. 1996). WAVES is a 1-D daily time-step model that simulates the fluxes of mass and energy between the atmosphere, vegetation and soil systems. It achieves a balance in its modelling complexity between soil physics, plant physiology, energy and solute balances. The energy-balance component of the model partitions the available energy into canopy and soil for plant growth and evapotranspiration. The water-balance component of the model incorporates infiltration, runoff, evapotranspiration (using the Penman-Monteith equation), soil-moisture redistribution (using Richards’ equation), drainage and interactions with the water table (not used here). The carbon balance component of the model calculates carbon assimilation using an integrated rate methodology (Wu et al. 1994) and dynamically allocates carbon to leaves, stems and roots. Some changes were made to the model code to tailor its use for this study; specifically, these changes revolve around making CO2 concentration a variable rather than a hard-coded parameter. These changes are detailed in (McCallum et al. 2010). The WAVES model requires three different data sets— the climate, soil and vegetation inputs. The input files for the model were created for each combination of climate, soil and vegetation found in the MDB. A 4-m deep soil profile was modelled with a free draining lower boundary condition. It was assumed that the deep drainage from the bottom of the model was equivalent to groundwater recharge and did not become lateral flow within the unsaturated zone. The assumption was made that diffuse recharge in dryland areas was not affected by depth to groundwater; this assumption will result in errors where the water table is close to the surface. With each run of the WAVES model taking up to 2 min to complete, it was impractical to model the entire basin at the same scale as the climate data (~5 km grid). A series of points were selected across the basin to reflect the mean annual rainfall gradient, the seasonal changes in rainfall and a bias toward the areas with high groundwater use. Three transects were selected to cover the rainfall gradient from the highest rainfall along the Great Dividing Range to the semi-arid western boundary of the basin. There was one transect in the north to account for the summerdominated-rainfall climatic region, the second transect DOI 10.1007/s10040-010-0625-x
1643
Fig. 2 Change in average annual rainfall projected by the 15 GCMs for a ~2030 climate for the high-global-warming scenario. See Table 3 for model names
was in the equi-seasonal-rainfall climatic region and the third transect was in the south where there is winterdominated rainfall. Additional points were added in some high priority groundwater use areas to arrive at a total of 20 points that were selected for detailed modelling (Table 1; Fig. 1). The Broadbridge-White equation (Broadbridge and White 1998) for soil moisture retention is used in WAVES. To calculate hydraulic conductivity (K) and matric potential (Ψ) as a function of moisture content (θ), the equation requires five parameters: saturated hydraulic conductivity (Ks, m day–1), saturated moisture content (θs , cm3 cm–3), residual moisture content (θr , cm3 cm–3), inverse capillary length scale (α, m) and an empirical constant based on soil properties (C, –). Q¼
q qr qs qr
Hydrogeology Journal (2010) 18: 1639–1656
ð1Þ
where Θ is the relative moisture content (scaled between 0 and 1). K ðQÞ ¼ KS
ðC 1Þ Q2 CQ
ð2Þ
1 1Q 1 CQ þ ln YðQÞ ¼ a Q C QðC 1Þ
ð3Þ
The ASRIS 1 database (Johnston et al. 2003) is the best soil data set that covers the entire MDB. ASRIS has data layers for soil type (Isbell 2002), Ks, and plant-available water capacity (PAWC) for two soil layers. The PAWC is defined as being the difference in volumetric moisture content between matric potentials of 0.1 and 15 bar. A map of soil types is shown in Fig. 3. DOI 10.1007/s10040-010-0625-x
1644 Table 1 Location of points selected for detailed unsaturated zone modelling (The reporting region is given as a guide to location; there is not meant to be a modelled location within each reporting region) Location Code
Longitude
Latitude
Reporting region
A B C D E F G H I J K L M N O P Q R S T
139.5 142 143.6 144.65 144.85 144.95 145 145.1 145.45 145.65 146.3 147.85 147.85 147.9 148.4 149.65 149.7 150.15 150.35 151.95
–35.2 –31.6 –35.95 –36.25 –35.55 –27.6 –30.05 –34.55 –36 –31.85 –36.1 –27.7 –32.05 –33.65 –36.15 –32.15 –29.4 –27.75 –30.35 –27.85
Eastern Mt Lofty Ranges Murray Loddon Avoca Campaspe Murray Paroo Barwon-Darling Murrumbidgee Murray Barwon-Darling Ovens Condamine-Balonne Macquarie-Castlereagh Lachlan Murrumbidgee Macquarie-Castlereagh Gwydir Moonie Namoi Condamine-Balonne
The Ks and PAWC of the topsoils and subsoils were averaged across the basin by soil type, and θs and θr were determined using Eqs. (1) and (3). The C and α parameters were estimated based on soil texture and Ks. The soil parameters used are shown in Table 2. Within the point scale modelling, not every point within the basin was going to be modelled and so not every vegetation type could be included. For simplicity, a
land use map (BRS 2008) was reclassified to three vegetation classes: annuals, perennials and trees (Fig. 4). It was assumed that the land use did not change between the historical climate scenario and the future climate scenarios. The vegetation parameters required by WAVES were taken from the WAVES user manual (Dawes et al. 2004). The annuals (including crops) were modelled as an annual pasture, the perennials were modelled as a perennial pasture, and the trees were modelled as an overstorey of Eucalypts with an understorey of perennial grass. A CO2 concentration of 378 ppm is used to represent current conditions, and CO2 concentrations of 455, 446 and 437 ppm are used to represent the 2030 conditions for the high, medium and low global warming respectively. The modelling results are relatively insensitive to this small variation in CO2 concentration (McCallum et al. 2010). As the WAVES model was not calibrated to field estimates of recharge, a sensitivity analysis of the soils and vegetation input parameters was undertaken to assess the uncertainty in the baseline historical climate scenario. The sensitivity analysis was conducted separately for the soils and vegetation parameters. For the baseline scenario, the geometric mean of the hydraulic conductivity for each soil type was calculated from the information contained within ASRIS. For the sensitivity analysis one standard deviation either side of the mean was chosen for the low recharge (low Ks) scenario and the high recharge (high Ks) scenario. In a similar way, the PAWC was incorporated into the sensitivity analysis using one standard deviation either side of the arithmetic mean for the low recharge scenario (high PAWC) and the high recharge scenario (low PAWC). The vegetation parameters, selected from the WAVES user manual, are based upon many studies where WAVES has been calibrated to field data (many of these in the MDB). In the user manual, a range of values is given for each parameter as high, medium and low. Generally the baseline case uses the medium value for each parameter. For this sensitivity analysis, the three vegetation parameters that recharge is most sensitive to (Xu et al. 2008) were varied within the range given in the user manual. These parameters were: the rainfall interception coefficient, light extinction coefficient and the maximum carbon assimilation rate.
Upscaling to the entire Murray-Darling Basin The output of the WAVES modelling was used to create regression equations between average annual rainfall and average annual recharge for each combination of soil type, vegetation and climate. The form of the relationship was a power function: Fig. 3 Soil types across the Murray-Darling Basin from Johnston et al. (2003) Hydrogeology Journal (2010) 18:1639–1656
RS ¼ aPS b
ð4Þ DOI 10.1007/s10040-010-0625-x
1645 Table 2 Soil parameters used in Broadbridge-White model for soil moisture retention and hydraulic conductivity in WAVES Soil type
Calcerosols Chromosols Dermosols Ferrosols Hydrosols Kandosols Kurosols Rudosols Sodosols Tenosols Vertosols
Topsoil Ks (m day–1)
C (-)
α (m)
θs (cm3 cm–3)
θr (cm3 cm–3)
Subsoil Ks (m day–1)
C (-)
α (m)
θs (cm3 cm–3)
θr (cm3 cm–3)
2.3 1.6 1.5 1.5 1.4 3.6 2.4 2.0 0.75 0.25 0.04
1.01 1.02 1.05 1.02 1.02 1.01 1.02 1.02 1.4 1.01 1.4
0.04 0.05 0.05 0.05 0.05 0.02 0.03 0.03 0.13 0.04 0.35
0.21 0.24 0.26 0.25 0.22 0.21 0.22 0.23 0.28 0.23 0.45
0.07 0.1 0.1 0.1 0.1 0.06 0.08 0.08 0.13 0.07 0.3
0.12 0.08 0.4 0.08 0.18 0.42 0.20 0.24 0.02 1.3 0.01
1.5 1.3 1.45 1.3 1.5 1.45 1.48 1.48 1.5 1.02 1.5
0.2 0.2 0.15 0.2 0.18 0.15 0.15 0.18 0.5 0.05 0.5
0.33 0.34 0.28 0.38 0.32 0.29 0.28 0.30 0.45 0.22 0.47
0.2 0.2 0.15 0.25 0.18 0.15 0.18 0.18 0.35 0.1 0.35
where Rs is the recharge for a given scenario, Ps is the rainfall for a given scenario and a and b are fitting parameters. The fitting parameters were determined using a least squares routine between the 20 model runs (i.e. one for each of 20 control points) for every combination (1518) of soil (11), vegetation (3) and climate (46). The regression equations were applied with an upper rainfall limit of 1651 mm (the highest average annual rainfall of any of the control points) to prevent recharge being predicted to be greater than rainfall when the regression equation was applied to areas that had rainfall greater than the control points. Using the set of regression equations (Eq. 4) and the set of annual average rainfall rasters for each climate
scenario (Chiew et al. 2008a), a series of annual average recharge rasters on a 0.05° grid were developed for the historical climate scenario and the 45 different future climates. The future climate results are reported as a recharge scaling factor (RSF): RSF ¼
RS RH
ð5Þ
where Rs is the recharge for a given scenario and RH is the annual average recharge for the historical period. The RSF was calculated at a pixel level by dividing each future climate scenario recharge raster by the historical climate scenario recharge raster.
Fig. 4 Vegetation types across the Murray-Darling Basin (BRS 2008) Hydrogeology Journal (2010) 18: 1639–1656
DOI 10.1007/s10040-010-0625-x
1646
Aggregation of recharge estimates from 15 global climate models The 45 different future climate outputs were used to investigate the differences between GCMs and their projections of how the future climate could impact upon groundwater recharge. The 15 rasters of RSF from a given global-warming scenario were fitted to a Pearson Type III distribution. This probability distribution was chosen because it allows the incorporation of both positive and negative skewness of the distribution of RSFs that are calculated here. This method allows a comparison to be made of the 10th, 50th and 90th percentiles of the RSFs at a pixel scale when the probability distribution is weighted for GCM performance or unweighted if all GCMs are assumed of equal validity. For the current analysis, the ability of each GCM to replicate the historical climate was used as an indicator of its ability to predict a future climate. Post et al. (2009) presented the results of different studies examining the performance of the different GCMs as was relevant to Australia. Table 3 shows the GCMs used and the model weight factor, defined as the proportion of comparison studies in which that model was above the median of all models considered. The highest weighted GCM was assessed to be GFDL which scored above the median in nine out of 10 studies and the lowest weighted was NCAR_PCM which scored below the median in all studies. This means that, for this analysis, the results derived from the NCAR_PCM model are ignored and the results derived from the GFDL model carry the most weight.
The Pearson Type III distribution uses three parameters: the mean RSF (RSF ), standard deviation (s) and the skewness (g): N P
RSF ¼
wi :RSFi
i¼1 N P
ð6Þ wi
i¼1
where RSFi is the ith observation of RSF, wi is the weight associated with the ith observation and N is the number of observations. 2
3 N 2 0:5 P N w : RSF RSF i i 6 7 6 7 s ¼ 6 i¼1 7 N P 4 5 ðN 1Þ wi
ð7Þ
i¼1
N2 g¼
N P
3 wi : RSFi RSF
i¼1 N P
wi ðN 1ÞðN
ð8Þ 2Þs3
i¼1
To determine the RSF for a given probability of exceedance, a frequency factor (KY) is calculated from the standard normal deviate (z); Eq. (9) is valid for –1< g <1 (Pilgrim 1987): o3 2 n g g z þ1 1 g 6 6
Table 3 Table of GCMs and the weights used in the Pearson Type III probability distribution. The weights are calculated as 1 minus the failure rate reported in (Post et al. 2009)
KY ¼
Model
Organisation
Weight
CCCMA t47 CCCMA t63 CNRM CSIRO GFDL GISS AOM IAP
Canadian Climate Centre - Canada
0.40
Canadian Climate Centre - Canada
0.50
The exceedance probabilities to be calculated are 10, 50 and 90%, and z for these probabilities are –1.280, 0 and 1.282 respectively. The fitted RSF for the given exceedance probability is calculated by:
Meteo-France - France CSIRO - Australia Geophysical Fluid Dynamics Lab - USA NASA/Goddard Institute for Space Studies USA LASG/Institute of Atmospheric Physics China Institute of Numerical Mathematics - Russia Institut Pierre Simon Laplace - France Centre for Climate Research - Japan Meteorological Institute of the University of Bonn/Meteorological Research Institute of KMA – Germany/Korea Max Planck Institute for Meteorology DKRZ - Germany Meteorological Research Institute - Japan National Centre for Atmospheric Research USA National Centre for Atmospheric Research USA
0.37 0.50 0.90 0.37
INMCM IPSL MIROC MIUB MPI MRI NCAR CCSM NCAR PCM
Hydrogeology Journal (2010) 18:1639–1656
0.33 0.50 0.11 0.80 0.75 0.70
RSFY ¼ RSF þ KY s
ð9Þ
ð10Þ
To assess whether the RSF is statistically significantly different to 1 (i.e. is there a change in recharge?) a 95% confidence interval is calculated for the fitted RSF of a given exceedance probability (USGS 1982): KY;P
0:5 KY KY 2 a:b ¼ a
ð11Þ
0.70 0.62 0.00
a ¼1
z2 2ðN 1Þ
ð12Þ DOI 10.1007/s10040-010-0625-x
1647
b ¼ KY 2
z2 N
ð13Þ
with heavy soils and perennial vegetation and can be greater than 100 mm yr–1 in high rainfall areas with light textured soil and annual vegetation.
(z=1.645 for a 95% confidence interval)
Upscaling the point estimates to the entire basin RSFY;P ¼ RSF þ KY;P s
ð14Þ
If RSF Y;P 1 RSF Y;Pþ then the calculated RSF is not significantly different from 1, this means that there is no detectable change in the recharge from the historical scenario to the scenario under investigation. Only three variants of future climate had the results reported:
& & &
90th percentile of recharge as calculated from the highglobal-warming scenario (wet future) 10th percentile of recharge as calculated from the highglobal-warming scenario (dry future) 50th percentile of recharge as calculated from the medium-global-warming scenario (median future)
The median future was selected because the median RSF from the medium-global-warming scenario represents the middle estimate of recharge under a future climate in this study. The greatest variability in estimates of recharge change comes from the high-global-warming scenario. Therefore the dry future scenario comes from the 10th percentile of the high-global-warming scenario and wet future scenario comes from the 90th percentile of the high-global-warming scenario. The average RSFs for these three scenarios were reported at the scale of the reporting regions (Fig. 1). The calculated recharge varied greatly depending upon soil type and land use but could give very similar RSFs. Therefore the average RSF over a reporting region was calculated as the area-weighted average of the recharge for the scenario divided by the area-weighted average recharge for the historical scenario: RSF ¼
Rs RH
ð15Þ
Results Point scale modelling At each climate point, a range of vegetation and soil types were modelled (Fig. 5). The trends between model runs were consistent with those shown by Abbs and Littleboy (1998); recharge is greater under annuals than perennials and greater under lighter textured soils than heavy clays. Although the model parameters were not calibrated to each point in the MDB the magnitude of the results is consistent with field studies (Petheram et al. 2002). Recharge can be less than 1 mm yr–1 in low rainfall areas Hydrogeology Journal (2010) 18: 1639–1656
The regression equations developed for each combination of climate, soil and vegetation type (Eq. 4) are shown graphically in Fig. 5 for the historical climate scenario (the 45 future climate scenarios give similar results and are not shown here). For every soil type, the relationship between rainfall and recharge under annual vegetation shows comparatively more scatter than under the other two vegetation types. Although there is scatter in the relationships developed, the curve fitted is statistically significant for every combination of climate, soils and vegetation. The upscaled recharge (R) averaged across the MDB is 25 mm yr–1 (Fig. 6). It is highly spatially variable with over 100 mm yr–1 of recharge in the south east of the MDB where the rainfall is highest and less than 1 mm yr–1 on the vertosol soils. The results of the sensitivity analysis of the WAVES input parameters are also shown in Fig. 6. Shown are the results of the sensitivity analysis on the soil parameters as these gave a greater range of uncertainty than the vegetation parameters. For the low-recharge case, the recharge averaged across the MDB is 12 mm yr–1. For the high recharge case, the recharge averaged across the MDB is 85 mm yr–1. The sensitivity analysis of the vegetation parameters resulted in a smaller range that was 18 mm yr–1 for the low-recharge scenario and 55 mm yr–1 for the high-recharge scenario. The reporting of the future climate results as a RSF is the way chosen for this study to deal with the uncertainty in the baseline historical climate scenario. The RSF rasters derived from the 15 GCMs for the high-global-warming scenario are shown in Fig. 7 (the medium- and low-global-warming scenarios are not shown here). These show a broad agreement with the change in rainfall rasters (Fig. 2); the three GCMs that project the greatest decrease in rainfall (IAP, MRI and CNRM) also show the smallest RSF. The reverse is also true with the GCMs that project the greatest increases in rainfall (CCCMA t63 and NCAR_PCM) also have RSFs derived from them that project an increase in recharge.
Aggregation To demonstrate why the Pearson Type III probability distribution was chosen, the model results at two points for the high-global-warming scenario and a given soil and vegetation type were fitted to three different probability distributions. A box plot of the results from the 15 models runs is displayed in Fig. 8a. The box plot for point R shows that the 15 RSFs are positively skewed; this means that the mean is higher than the median and the longer tail is on the high side of the box. The box plot for point D shows that the 15 RSFs are negatively skewed; this means that the mean is lower than the median and the longer tail DOI 10.1007/s10040-010-0625-x
1648
Fig. 5 Results of the WAVES modelling for the historical climate scenario for each combination of soils and vegetation type. The regression equations are used in the upscaling to create a raster of recharge across the MDB
Hydrogeology Journal (2010) 18:1639–1656
DOI 10.1007/s10040-010-0625-x
1649
Fig. 6 The results of the upscaling of the recharge (R) modelling for the historical climate scenario is shown as the centre plot baseline. The Low R and High R rasters are the results of an uncertainty analysis of the input parameters to the WAVES modelling giving a low recharge and a high recharge range
is on the low side of the box. If the results at these two points were normally distributed then the mean and median would be equal and the tails on both sides of the box would be equal, they would also plot as a straight line in a normality plot (Fig. 8b). A log transformation of the data is often used if the data is log-normally distributed. In this case, a log transformation results in an improved fit for point R which is positively skewed, but a worse fit for point D which is negatively skewed (Fig. 8c). To overcome the problem of data which can be both positively or negatively skewed, a probability distribution is needed which accounts for the skewness. By fitting the model results at these two points to a Pearson Type III distribution the fit is improved in both cases (Fig. 8d). For each global-warming scenario there are 15 rasters of RSF that show a range of climate-change impacts upon groundwater recharge. The projections show that recharge could be increased or decreased by over 50% (Fig. 7). To assess these differences, the RSF rasters have been fitted to a weighted Pearson Type III distribution at a pixel scale. The three parameters used by the weighted Pearson Type III distribution are the mean, standard deviation and skewness of the 15 RSFs—Eqs. (6), (7) and (8); these are shown in Fig. 9 for the high global-warming scenario (the low and medium global-warming scenarios are not shown here). This plot shows that the mean RSF across the MDB is not spatially uniform. The southern extremity of the MDB has a mean RSF of below 1, meaning a decrease in recharge for the future climate scenario when compared to the historical climate scenario, whereas most of the rest of the MDB has a RSF of above 1, meaning an increase in recharge for the future climate scenario compared to the historical scenario. The standard deviation plot shows that there is a greater spread in the results from the climate sequences derived from the 15 GCMs for the north and west of the MDB than the south and east of the MDB. The skewness plot shows that the north of the MDB has a Hydrogeology Journal (2010) 18: 1639–1656
positive skewness for the RSF derived from the 15 GCMs and the south of the MDB has a negative skewness. The probability distribution can be used to estimate exceedance probabilities from the data fitted to it. The 50% exceedance represents the median from the fitted probability distribution. For the high global-warming scenario, about half of the MDB does not have a statistically significant (p<0.05) RSF for the 50% exceedance (Fig. 9). Of the areas that do have a statistically significant RSF for the 50% exceedance probability, the very southern extremity of the MDB has a statistically significant RSF of less than 1 and a band following the agricultural regions has a statistically significant RSF of above 1. The recharge calculated for the future climate scenario is lower than the historical scenario along the southern extremity of the MDB because almost all the GCMs project less rainfall for these areas (Fig. 2). The areas that have a statistically significant RSF of above 1 are very closely aligned with the areas of annual vegetation (Fig. 4). The reason for the increase in recharge has several causes that include an increase in rainfall intensity and a decrease in the water use efficiency of the vegetation; these causes are explored further in a companion paper (McCallum et al. 2010). The 10% probability of exceedance from the fitted probability distribution represents the wet end of the distribution. For the high-global-warming scenario almost all of the MDB has a statistically significant RSF of greater than 1 except for the very southern extremity of the MDB (Fig. 9). This indicates that for most of the MDB under a wet future climate, recharge is projected to increase. The 90% probability of exceedance from the fitted probability distribution represents the dry end of the distribution. For the high-global-warming scenario most of the MDB has a statistically significant RSF of less than 1 except for a band that is closely aligned with the annual vegetation (Fig. 9). This indicates that for most of the DOI 10.1007/s10040-010-0625-x
1650
Fig. 7 Rasters of RSF as calculated from climate sequences derived from each GCM for a high-global-warming scenario
MDB under a dry future climate, recharge is projected to decrease.
Discussion Implications of climate change on groundwater recharge in the Murray-Darling Basin
The results of fitting the 45 future climate scenarios to the weighted Pearson Type III probability distribution can be simplified to a median estimate (50% exceedance from the medium-global-warming scenario) and the extreme wet and dry estimates (10 and 90% exceedances Hydrogeology Journal (2010) 18:1639–1656
from the high-global-warming scenario respectively) of RSF (Fig. 10). These are shown in Table 4 as averages for each catchment within the MDB (Fig. 1). This shows that for the median estimate there is projected to be a small increase in recharge in most catchments and this averages 5% across the MDB. For the wet estimate, all catchments show an increase in recharge with an average basin wide increase of 32% and for the dry estimate all catchments show a decrease in recharge by an average of 12%. The areas predicted to have an increase in recharge will result in an increase in the water available for extraction. However, many areas within the MDB have saline groundwater which is of little beneficial use. Land and water DOI 10.1007/s10040-010-0625-x
1651
Fig. 8 a Box plot and b–d examples of fitting three different types of probability distribution, at a point scale to the 15 estimates of RSF from the high global-warming scenario for control point R for perennial vegetation on a sodosol soil and control point D for annual vegetation on a sodosol soil
salinisation is a major concern in the MDB and changes in groundwater recharge will have implications for salinity. The decrease in recharge along the southern margins of the basin will be good for salinity in dryland areas and the increases in recharge projected for the eastern margins of the basin will see an increase in the area affected by salinity. The effect of climate change upon salinity needs to be put into context of the magnitude of the change in recharge due to climate change and land use change. The clearing of native vegetation resulted in changes in recharge of more than an order of magnitude; by comparison, changes in recharge from climate change are only projected to be less than 50%. The upscaling routine used in this study was based upon three simplified vegetation classes: annuals, perennials and trees. It was assumed that the current (~2008) spatial distribution of these vegetation classes is appropriate to be used for the future climate (~2030) scenario when projecting changes in future recharge. Any changes in the future distribution of vegetation will have an impact upon recharge. The median future climate scenario shows that recharge is projected to increase in areas of annual vegetation (Fig. 10). This is due to the vegetation using less water, primarily due to being outside of their optimum temperature range. The model used does not predict grain yields but does project a decrease in leaf area in some cases. If this is interpreted as meaning that some areas of cropping may not be viable under a future climate, then the impact of land-use change upon recharge will be much greater than the impact of climate change upon recharge. Hydrogeology Journal (2010) 18: 1639–1656
It can be seen in Fig. 5 that below 600 mm yr–1 annual average rainfall the modelled recharge under annual vegetation can be an order of magnitude greater than the recharge under perennials or trees for all soil types. The land clearing for agriculture in the nineteenth and twentieth centuries resulted in the conversion of previously perennial or tree vegetation to annuals with a consequential increase in recharge (Petheram et al. 2002). If climate change forces large areas currently under annual cropping to alternate land uses then the recharge could decrease to levels closer to those pre–land clearing. Another cause of land use change under a future climate could be driven by policy. Currently growing trees is not economically viable in areas of low-to-medium rainfall and so the dominant land uses are cropping and grazing. If, through subsidies or an increase in the value of the product, forestry became profitable in previously agriculturally dominated areas, then there would be a decrease in recharge as trees are able to use more water than crops and pastures. The incorporation of future land-use scenarios along with future climate scenarios into future water-resources projections is an area of research that needs further exploration.
Predicting climate-change impacts upon groundwater recharge The reason for about half of the MDB not having a statistically significant change in recharge for the ‘median’ future climate scenario can be seen in a plot of the number DOI 10.1007/s10040-010-0625-x
1652
Fig. 9 Results from the high global-warming scenario. The mean, standard deviation and skewness are calculated from the 15 estimates of RSF from the climate sequences derived from the 15 different GCMs. Areas shaded black indicate that the change in recharge is not statistically significant (p<0.05) between the future climate scenario and the historical climate scenario
of model runs that project a decrease in recharge (Fig. 11). For rainfall, the majority of GCMs project a decrease for the entire basin and in the south of the basin almost all of the GCMs project a decrease. However, when the recharge is analysed in the same way, the majority of the basin shows that between one third and two thirds of the climate sequences derived from the 15 GCMs project a decrease in recharge, whereas it is only the southern margins of the basins where more than two thirds of the climate sequences derived from the 15 GCMs are consistent in projecting a decrease in recharge. For some of the areas under annual vegetation more than two thirds of the climate sequences derived from the 15 GCMs project an increase in recharge. When all the WAVES model runs are plotted together on a scatter plot of change in rainfall versus change in recharge, a linear regression line fitted through all the data does not pass through the origin but slightly above it (Fig. 12). For the three global-warming scenarios investigated, the regression lines are almost parallel, with a different offset in proportion to the degree of global warming. The low-global-warming scenario has an intercept of 3%, the medium 8% and the high 13%. This means that without a change in rainfall, an increase in recharge due to global warming is being projected. This Hydrogeology Journal (2010) 18:1639–1656
finding is counter-intuitive and has been explored further in a companion paper by (McCallum et al. 2010). They showed that this is due to the vegetation having a lower leaf area index with increased temperature leading to less interception and more infiltration of rainfall leading to more recharge. This observation helps to explain why the majority of GCMs predict a decrease in rainfall and yet the modelling results shown here suggest that in substantial areas of the MDB more than half the climate sequences generated from the 15 GCMs project an increase in recharge (Fig. 11). The lack of agreement between the climate sequences derived from the 15 GCMs on the change in recharge is highlighted in the standard deviation plots. The spatial arrangement of the standard deviation of the recharge scaling factors closely followed the distribution of rainfall in the basin (Fig. 9). The highest standard deviations were found in the west of the basin where the rainfall is lowest and lower standard deviations were found in the southeast where rainfall is highest. This is due to the importance of episodic recharge in semi-arid areas and its dependence upon the daily scaling of extreme rainfall events. As an extreme example, one of the WAVES model runs (data not shown) managed a doubling of the total recharge over 112 years when compared to the historical DOI 10.1007/s10040-010-0625-x
1653
Fig. 10 The difference between using a weighted and unweighted probability distribution for the 10 (wet) and 90 (dry) percent exceedance of RSF from the high-global-warming scenario and the 50 (median) percent exceedance of RSF from the medium-globalwarming scenario
climate sequence from a high-global-warming climate sequence that had less rainfall. (Although recharge doubled, it was still a very small number.) The extreme volatility of episodic recharge is damped in the upscaling process in this study but the spatial distribution of standard deviation shows that episodic recharge is an important process to be considered in the impact of climate change upon groundwater recharge. It is not uncommon for the RSFs calculated at a point to have extremes that project either a 50% increase in Hydrogeology Journal (2010) 18: 1639–1656
recharge or a 50% decrease in recharge depending upon which GCM was used to derive the climate sequence (Fig. 7). As can be seen from the 10 (wet) and 90 (dry) percent exceedance probability rasters of RSF (Fig. 10), collectively there is no agreement on the direction of change in recharge let alone the magnitude. The use of the weighted probability distribution has helped in reducing this range between the wet and dry rasters of RSF. When these rasters are created using both a weighted and unweighted probability distribution, it can be seen that DOI 10.1007/s10040-010-0625-x
1654 Table 4 Estimates of RSF when aggregated to the regional level. The dry estimate is the 90% exceedance from the high-global-warming scenario, the median estimate is the 50% exceedance from the medium-global-warming scenario and the wet estimate is the 10% exceedance from the high-global-warming scenario Region
Dry
Median
Wet
Warrego Condamine-Balonne Paroo Moonie Border Rivers Barwon-Darling Gwydir Namoi Macquarie-Castlereagh Murray Lachlan Murrumbidgee Eastern Mt Lofty Ranges Wimmera Loddon-Avoca Ovens Goulburn-Broken Campaspe
0.81 0.80 0.79 0.83 0.85 0.84 0.93 0.96 0.94 0.87 0.93 0.88 0.88 0.88 0.88 0.83 0.84 0.84
1.02 1.01 1.04 1.01 1.01 1.06 1.07 1.09 1.09 1.06 1.09 1.05 1.04 1.08 1.07 1.02 1.02 1.00
1.39 1.36 1.44 1.32 1.31 1.45 1.38 1.38 1.39 1.28 1.33 1.28 1.20 1.27 1.24 1.24 1.18 1.12
Fig. 12 All WAVES model runs as a scatter plot of change in rainfall versus change in recharge for the high, medium and lowglobal-warming scenarios when compared to the historical scenario
the unweighted rasters have comparatively low RSFs for the 50 and 90% exceedances. The reason for this is that at the basin scale the three rasters with the lowest RSFs (Fig. 7) are amongst the lowest weighted GCMs (Table 3). Previous studies have not found such variation in the average groundwater levels or recharge when comparing climate scenarios produced by different GCMs. Loáiciga et al. (2000) used six GCMs and found all produced
Fig. 11 Number of model results that show a decrease in recharge out of 15 GCM-derived climate sequences Hydrogeology Journal (2010) 18:1639–1656
DOI 10.1007/s10040-010-0625-x
1655
spring flows less than the historical scenario within a narrow band. Rosenberg et al. (1999) used three GCMs and found that all predicted less recharge although the magnitude varied. Brouyère et al. (2004) used three GCMs and found all resulted in decreased groundwater levels. Serrat-Capdevila et al. (2007) found that recharge was reduced in four of the five GCM-derived climate sequences selected. It is unclear whether the consistency in groundwater projections from previous studies when compared to this study is due to the selection of the GCMs that produced the climate projections or the inherent variability in Australia’s climate that makes predictions difficult.
Evaluation of the methods used The method developed here for estimating the impact of climate change on recharge on a regional scale was appropriate for the questions being investigated. The results of the future climate scenarios were presented as a recharge scaling factor rather than as a depth or volume of recharge. In this way, the effects of the considerable uncertainty in the baseline historical climate scenario estimates of recharge (Fig. 6) were minimised when evaluating the future climate scenarios. If the results of the recharge estimation were to be used directly in a water-resource assessment, then there is considerable uncertainty in the estimates of recharge leading to considerable uncertainty in the estimates of water availability. This uncertainty is compounded in areas that do not meet the assumptions made in the modelling. These include areas of very shallow water tables where the free draining lower boundary condition of the model is not appropriate and areas with very shallow soils where the assumption of a 4-m soil column is not appropriate.
Conclusions For the Murray-Darling Basin, the median projection of the change in recharge due to climate change for a future climate around 2030, compared to the historical climate, is for an increase in recharge of 5%. This is not spatially uniform with some areas along the southern extremity of the MDB projecting a decrease in recharge and other areas under annual vegetation projected to have a larger increase in recharge. The variability in the recharge response to climate change can be seen in the extreme wet and dry projections; in the wettest case, the recharge could increase by 32% on average across the basin, while in the driest case, recharge could decrease by 12%. Such variability in recharge estimates using different climate sequences means that making recommendations for water-resources management based upon this approach is highly uncertain. Acknowledgements This paper is based upon work conducted as a small portion of the Murray-Darling Basin (MDB) Sustainable Yield (MDBSY) Project (http://www.csiro.au/partnerships/MDBSY. html) funded by the National Water Commission. Hydrogeology Journal (2010) 18: 1639–1656
References Abbs K, Littleboy M (1998) Recharge estimation for the Liverpool plains. Aust J Soil Res 36(2):335–357 ABS (2003) Year Book Australia 2003. Australian Bureau of Statistics, Canberra, Australia Beverly C, Bari M, Christy B, Hocking M, Smettem K (2005) Predicted salinity impacts from land use change: comparison between rapid assessment approaches and a detailed modelling framework. Aust J Exp Agric 45(11):1453–1469 Broadbridge P, White I (1998) Constant rate rainfall infiltration: a versatile non-linear model: I. analytical solution. Water Resour Res 24:145–154 Brouyère S, Carabin G, Dassargues A (2004) Climate change impacts on groundwater resources: modelled deficits in a chalky aquifer, Geer basin, Belgium. Hydrogeol J 12(2):123–134 BRS (2008) Integrated Vegetation Cover 2008. Bureau of Rural Sciences, Canberra, Australia Brunner P, Bauer P, Eugster M, Kinzelbach W (2004) Using remote sensing to regionalize local precipitation recharge rates obtained from the chloride method. J Hydrol 294(4):241–250 Charles SP, Bates BC, Smith IN, Hughes JP (2004) Statistical downscaling of daily precipitation from observed and modelled atmospheric fields. Hydrol Process 18(8):1373–1394 Chiew FHS, Teng J, Kirono D, Frost AJ, Bathols JM, Vaze J, Viney NR, Young WJ, Hennessy KJ, Cai WJ (2008a) Climate data for hydrologic scenario modelling across the Murray-Darling Basin. A report to the Australian government from the CSIRO Murray-Darling Sustainable Yields Project, CSIRO, Canberra, Australia Chiew FHS, Vaze J, Viney NR, Jordan PW, Perraud J-M, Zhang L, Young WJ, Penaarancibia J, Morden RA, Freebairn A, Austin J, Hill PI, Weisenfeld CR, Murphy R (2008b) Rainfall-runoff modelling across the Murray-darling basin. A report to the Australian government from the CSIRO Murray-Darling Sustainable Yields Project, CSIRO, Canberra, Australia Chiew FHS, Teng J, Vaze J, Post DA, Perraud JM, Kirono DGC Viney NR (2009) Estimating climate change impact on runoff across southeast Australia: method, results, and implications of the modeling method. Water Resour Res 45, W10414, 17 pp Christensen JH, Hewitson B, Busuloc A, Chen A, Gao X, Held I, Jones R, Kolli RK, Kwon WT, Laprise R, Magana Rueda V, Mearns L, Menendez CG, Ralsanen J, Rinke A, Sarr A, Whetton P (2007) Regional climate projections. In: Solomon S et al (eds) Climate change 2007: the physical science basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge Cook PG, Walker GR, Jolly ID (1989) Spatial variability of groundwater recharge in a semiarid region. J Hydrol 111:195–212 Crosbie RS, Wilson B, Hughes JD, McCulloch C, King WM (2008) A comparison of the water use of tree belts and pasture in recharge and discharge zones in a saline catchment in the Central West of NSW, Australia. Agric Water Manage 95 (3):211–223 CSIRO and BOM (2007) Climate change in Australia, CSIRO and BOM, Canberra, Australia D’Agnese FA, Faunt CC, Hill MC, Turner AK (1999) Death Valley regional ground-water flow model calibration using optimal parameter estimation methods and geoscientific information systems. Adv Water Resour 22(8):777–790 Dawes WR, Gilfedder M, Stauffacher M, Coram J, Hajkowicz S, Walker GR, Young M (2002) Assessing the viability of recharge reduction for dryland salinity control: WANILLA, Eyre Peninsula. Aust J Soil Res 40(8):1407–1424 Dawes W, Zhang L, Dyce P (2004) WAVES v3.5 User Manual, CSIRO Land and Water, Canberra, Australia Delin GN, Healy RW, Lorenza DL, Nimmoc JR (2007) Comparison of local- to regional-scale estimates of ground-water recharge in Minnesota, USA. J Hydrol 334(1–2):231–249 DFA (1974) Agreement between the Government of Australia and the Government of Japan for the protection of migratory birds in DOI 10.1007/s10040-010-0625-x
1656 danger of extinction and their environment. Australian Treaty Series 1981 No. 6, Department of Foreign Affairs, Canberra, Australia DFAT (1986) Agreement between the Government of Australia and the Government of the People’s Republic of China for the protection of migratory birds and their environment. Australian Treaty Series 1988 No. 22, Department of Foreign Affairs and Trade, Canberra, Australia IPCC (2001) Climate change 2001: the scientific basis. Contribution of working group 1 to the third assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge IPCC (2007) Climate change 2007: the physical science basis. Contribution of working group 1 to the fourth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, 996 pp Isbell RF (2002) Australian soils classification. CSIRO, Collingwood, Victoria, Australia, 144 pp Jeffrey SJ, Carter JO, Moodie KB, Beswick AR (2001) Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environ Model Softw 16(4):309–330 Johnston RM, Barry SJ, Bleys E, Bui EN, Moran CJ, Simon DAP, Carlile P, McKenzie NJ, Henderson BL, Chapman G, Imhoff M, Maschmedt D, Howe D, Grose C, Schoknecht N, Powell B, Grundy M (2003) ASRIS: the database. Aust J Soil Res 41 (6):1021–1036 Jyrkama MI, Sykes JF (2007) The impact of climate change on spatially varying groundwater recharge in the Grand River watershed (Ontario). J Hydrol 338(3–4):237–250 Lim WH, Roderick ML (2008) Global water cycle atlas based on the IPCC AR4 Climate Models. The Australian National University, Canberra Littleboy M, Herron N, Barnett P (2003) Applying unsaturated zone modelling to develop recharge maps for the Murray-Darling Basin in New South Wales, Australia. In: Post DA (ed) Proceedings of International Congress on Modelling and Simulation, MODSIM 2003, Modelling & Simulation Society of Australia & New Zealand Inc., Perth, Australia Loáiciga HA, Maidment DR, Valdes JB (2000) Climate-change impacts in a regional karst aquifer, Texas, USA. J Hydrol 227 (1–4):173–194 McCallum JL, Crosbie RS, Walker GR, Dawes WR (2010) Impacts of climate change on groundwater in Australia: a sensitivity analysis of recharge. Hydrogeol J. doi:10.1007/s10040-0100624-y Meehl GA, Covey C, Delworth T, Latif M, McAvaney B, Mitchell JFB, Stouffer RJ, Taylor KE (2007) The WCRP CMIP3 multimodel dataset: a new era in climate change research. Bull Am Meteorol Soc 88(9):1383–1394 Nunez M, McGregor JL (2007) Modelling future water environments of Tasmania, Australia. Clim Res 34(1):25–37 Petheram C, Walker G, Grayson R, Thierfelder T, Zhang L (2002) Towards a framework for predicting impacts of land-use on recharge: 1. a review of recharge studies in Australia. Aust J Soil Res 40(3):397–417 Pilgrim DH (1987) Australian rainfall and runoff: a guide to flood estimation, 1. The Institution of Engineers, Barton, ACT, Australia Post DA, Chiew FHS, Teng J, Vaze J, Yang A, Mpelasoka F, Smith IN, Katzfey J, Marston F, Marvanek SP, Kirono D, Nguyen K, Kent D, Donohue RJ, McVicar TR (2009) Climate scenarios for Tasmania. Tasmania Sustainable Yields Project, A report to the Australian Government from the CSIRO Tasmania Sustainable Yields Project, CSIRO, Canberra, Australia
Hydrogeology Journal (2010) 18:1639–1656
Ramsar Convention Bureau (1971) Convention on wetlands of international importance especially as waterfowl habitat. Ramsar, Iran, 2 February 1971. UN Treaty Series No. 14583. As amended by the Paris Protocol, 3 December 1982, and Regina Amendments, 28 May 1987 Rosenberg NJ, Epstein DJ, Wang D, Vail L, Srinivasan R, Arnold JG (1999) Possible impacts of global warming on the hydrology of the Ogallala Aquifer region. Clim Change 42(4):677–692 Salama R, Hatton T, Dawes W (1999) Predicting land use impacts on regional scale groundwater recharge and discharge. J Environ Qual 28(2):446–460 Scanlon BR, Healy RW, Cook PG (2002) Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeol J 10:18–39 Serrat-Capdevila A, Valdés JB, Pérez JG, Baird K, Mata LJ, Maddock T (2007) Modeling climate change impacts - and uncertainty - on the hydrology of a riparian system: the San Pedro Basin (Arizona/Sonora). J Hydrol 347(1–2):48–66 Slavich PG, Walker GR, Jolly ID, Hatton TJ, Dawes WR (1999) Dynamics of Eucalyptus largiflorens growth and water use in response to modified watertable and flooding regimes on a saline floodplain. Agric Water Manage 39(2–3):245–264 Sophocleous M (1992) Groundwater recharge estimation and regionalization: the Great Bend Prairie of central Kansas and its recharge statistics. J Hydrol 137:113–140 Szilagyi J, Harvey FE, Ayers JF (2005) Regional estimation of total recharge to ground water in Nebraska. Ground Water 43(1):63– 69 Tuteja NK, Vaze J, Teng J (2005) The CLASS modelling framework: a platform for distributed eco-hydrological modelling, MODSIM, 2005, Melbourne, Australia, December 2005 UNFCCC (1992) United Nations framework convention on climate change. UN, New York USGS (1982) Guidelines for determining flood flow frequency. Bulletin #17B of the Hydrology Subcommittee, US Geological Survey, Reston, VA Wang HX, Zhang L, Dawes WR, Liu CM (2001) Improving water use efficiency of irrigated crops in the North China Plain: measurements and modelling. Agric Water Manage 48(2):151– 167 Wu H, Rykiel EJ, Hatton T, Walker J (1994) An integrated rate methodology (IRM) for multi-factor growth rate modelling. Ecol Model 73(1–2):97–116 Xu C, Martin M, Silberstein R, Smetten K (2008) Identifying sources of uncertainty in groundwater recharge estimates using the biophysical model WAVES. Water Down Under, Adelaide, Australia Yang YH, Watanabe M, Wang ZP, Sakura Y, Tang CY (2003) Prediction of changes in soil moisture associated with climatic changes and their implications for vegetation changes: WAVES model simulation on Taihang Mountain, China. Clim Change 57 (1–2):163–183 Zhang L, Dawes W (1998) WAVES: an integrated energy and water balance model. Technical Report No. 31/98, CSIRO Land and Water, Canberra, Australia Zhang L, Dawes WR, Hatton TJ (1996) Modelling hydrologic processes using a biophysically based model–application of WAVES to FIFE and HAPEX-MOBILHY. J Hydrol 185(1– 4):147–169 Zhang L, Dawes WR, Hatton TJ, Hume IH, O’Connell MG, Mitchell DC, Milthorp PL, Yee M (1999) Estimating episodic recharge under different crop/pasture rotations in the Mallee region. Part 2. Recharge control by agronomic practices. Agric Water Manage 42(2):237–249
DOI 10.1007/s10040-010-0625-x