Climate change and runoff in south-western Australia

Climate change and runoff in south-western Australia

Journal of Hydrology 475 (2012) 441–455 Contents lists available at SciVerse ScienceDirect Journal of Hydrology journal homepage: www.elsevier.com/l...

3MB Sizes 0 Downloads 147 Views

Journal of Hydrology 475 (2012) 441–455

Contents lists available at SciVerse ScienceDirect

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

Climate change and runoff in south-western Australia R.P. Silberstein a,⇑, S.K. Aryal a, J. Durrant b, M. Pearcey b, M. Braccia b, S.P. Charles a, L. Boniecka b, G.A. Hodgson a, M.A. Bari c, N.R. Viney d, D.J. McFarlane a a

CSIRO Floreat Laboratories, Private Bag 5, PO Wembley, Western Australia 6913, Australia Department of Water, 168 St. Georges Terrace, Perth, Western Australia 6000, Australia c Bureau of Meteorology, PO Box 1370, West Perth, Western Australia 6872, Australia d CSIRO Land and Water, GPO Box 1666, Canberra, ACT 2601, Australia b

a r t i c l e

i n f o

Article history: Available online 17 February 2012 Keywords: Water resources Hydrological modelling Rainfall Stream flow Model

s u m m a r y This paper presents the results of computer simulations of runoff from 13 major fresh and brackish river basins in south-western Australia (SWA) under climate projections obtained from 15 GCMs with three future global warming scenarios equivalent to global temperature rises of 0.7 °C, 1.0 °C and 1.3 °C by 2030. The objective was to apply an efficient methodology, consistent across a large region, to examine the implications of the best available projections in climate trends for future surface water resources. An ensemble of rainfall-runoff models was calibrated on stream flow data from 1975 to 2007 from 106 gauged catchments distributed throughout the basins of the study area. The sensitivity of runoff to projected changes in mean annual rainfall is examined using the climate ‘elasticity’ concept. Averaged across the study area, all 15 GCMs project declines in rainfall under all global warming scenarios with a median decline of 8% resulting in a median decline in runoff of 25%. Such uniformity in projections from GCMs is unusual. Over SWA the average annual runoff under the 5th wettest and 5th driest of the 45 projections of the 2030 climate declines by 10 and 42%, respectively. Under the 5th driest projection the runoff decline ranges from 53% in the northern region to 40% in the southern region. Strong regional variations in climate sensitivity are found with the proportional decline in runoff greatest in the northern region and the greatest volumetric declines in the wetter basins in the south. Since the mid 1970s stream flows into the major water supply reservoirs in SWA have declined by more than 50% following a 16% rainfall reduction. This has already had major implications for water resources planning and for the preservation of aquatic and riparian ecosystems in the region. Our results indicate that this reduction in runoff is likely to continue if future climate projections eventuate. Ó 2012 Published by Elsevier B.V.

1. Introduction Assessment of the effects of climate change on catchment runoff is crucial for future management of water supplies, however undertaking these assessments on a regional scale with sufficient resolution to be useful to water resources engineers and planners is a challenge. The current suite of global climate models (GCMs) greatly assist the process, but their grid cells are large and interpreting the results at the scale of water resources catchments requires downscaling of projections to scales appropriate for catchment heterogeneity and management. This approach allows the future impacts of rainfall and the consequences for water resources and extreme hydrological events to be assessed using suitable catchment hydrological models. The impact of climate trends over the latter half of the 20th Century is well documented in numerous scientific studies reviewed in ⇑ Corresponding author. Tel.: +61 8 9333 6000. E-mail address: [email protected] (R.P. Silberstein). 0022-1694/$ - see front matter Ó 2012 Published by Elsevier B.V. doi:10.1016/j.jhydrol.2012.02.009

the Inter-governmental Panel on Climate Change Fourth Assessment Report (IPCC AR4, IPCC, 2007a). With a warming global climate there is a general trend for an increasing difference between regions of high and low rainfall. For example, in Iran GCM projections have indicated increasing rainfall in wetter areas and lower rainfall in drier areas and with a greater uncertainty in drier regions (Abbaspour et al., 2009). In Britain drier catchments were found more sensitive to climate change than wetter catchments (Arnell, 1992; Arnell and Reynard, 1996), and catchments where rainfall is most strongly concentrated in winter were projected to be even more so. Countries with a Mediterranean type climate, and particularly those with semi-arid climates, appear to have the greatest impact of climate change on water resources (Ragab and Prudhomme, 2002). D’Agostino et al. (2010) found a 5–10% reduction in rainfall would produce a 16–23% reduction in stream flow in the Candelaro catchment in southern Italy, based on GCM projections with a temperature rise of about 1.5 °C by 2050. In North Algeria (Jean-Pierre et al., 2010), Tunisia (Abouabdillah et al., 2010), Brazil (Montenegro and Ragab, 2010) and the USA (Parajuli, 2010)

442

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

et al., 2010). The proportion of rainfall that becomes stream flow (the runoff coefficient), already relatively low on a global scale (McMahon et al., 2007), has dropped nearly two-thirds from an average of around 10% in Perth water supply catchments (Silberstein et al., in preparation-b). The population of SWA is growing rapidly and is estimated to increase by about 40% by 2030 (currently 1.9 million inhabitants), placing further pressure on existing water resources. This paper presents projections of future rainfall and runoff across all major fresh river basins in SWA using climate projections developed from the 15 GCMs with readily accessible daily output used in the IPCC AR4 (IPCC, 2007a). These climate projections were used to develop modified daily sequences (Chiew et al., 2009) for input to rainfall-runoff models. Companion papers assess the impact on groundwater levels and recharge (Ali et al., this issue), the impacts of stream flow and groundwater level projections on water dependent ecosystems (Barron et al., this issue) and projections of future water yields and demands (McFarlane et al., this issue). There may be lessons learned in this region that will assist other jurisdictions in anticipating changes to water availability and in managing the consequences.

2500

Annual Total 1975 to 2010 (ave 1039mm) 2001 to 2010 (ave 965mm)

2000

1911 to 1974 (ave 1251mm) 1997 to 2010 (ave 985mm)

1500 1000 500

1000 900 800 700 600 500 400 300 200 100 0

2010

2000

1990

1980

1970

1960

1950

1940

1930

1920

1910

0

1970

1980

2010

1960

2000

1950

1990

1980

1940

1970

1940

1930

1960

1930

1920

1950

1920

Annual Total 1911 to 1974 (ave 338GL) 1975 to 2010 (ave 148GL) 1997 to 2010 (ave 101GL) 2001 to 2010 (ave 84GL)

1910

Total Annual Inflow to Perth Dams (GL)

Total Annual Rainfall at Jarrahdale (mm)

studies have found that projections of future climate are for reduced rainfall in semi-arid regions and a significant impact on water resources. Thomson et al. (2005) found that in semi-arid regions in the western USA, where irrigation water was already in short supply, reductions in water yields may exceed 50% of current levels. Piao et al. (2010) found that China has had increasing temperatures, longer droughts and lower stream flows in recent decades but projections into the future of the impact on water resources and the implications for agriculture are still unclear, and there is a clear need to improve the regional simulations of climate changes and their consequences. The impact on water resources of recent climate trends has been particularly acute in south-western Australia (SWA) and the IPCC AR4 identified the region as one that had experienced amongst the greatest impact on divertible water resources in the world (Hennessy et al., 2007). There has been a clear rising trend in temperature of nearly 1 °C through the 20th Century (Fig. 1), which has increased water demand as yields have declined. Since the mid 1970s a 16% reduction in rainfall has resulted in a decline of over 50% in stream flows into the major reservoirs in the region that supply the city of Perth, the State capital (Fig. 1 and Petrone

Temperature anomaly (°C)

1 0.5 0 -0.5 -1 -1.5 1910

1990

2000

Fig. 1. Annual rainfall at Jarrahdale, a representative site near the water supply reservoirs, and annual inflow to the water supply reservoirs in south-west Western Australia, with the temperature anomaly (difference from the mean) averaged over the region. A year is taken as May to April (Dam inflow data courtesy of the Water Corporation; rainfall and temperature data from the Bureau of Meteorology).

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

2. Description of the study area The study area lies in south-western Australia (SWA) (Fig. 2) and includes all fresh (<500 mg/L) to brackish (<3000 mg/L) surface water basins from Gingin in the north-west to Albany in the southeast that are suitable for diversion for human consumption, irrigation, and industry requiring near potable quality water. 2.1. Geology, topography and land use South-western Australia has outcrops of some of the oldest rocks in the world and a complex geology which greatly affects the topography, soils and hydrology of the region. Almost all rivers rise on the dissected uplands of the Darling Plateau, with an average elevation of about 350 m Australian Height Datum (AHD), before dropping down the escarpment and entering the flatter coastal plains and then the ocean, usually through estuaries. The Darling Plateau forms the eastern part of the study area, separated from the Perth Basin by the Darling Fault follows approximately the same line as the eastern boundary of the groundwater area discussed by Ali et al. (2011). Harvesting surface water in the flat, sandy coastal catchments is not possible to any significant extent and only occurs in the uplands. The surface water supply catchments have retained much of their native forest cover (67% of the surface water basin area) and are managed for water quality, environmental and timber production purposes. The dry sclerophyllous native vegetation is mainly jarrah (Eucalyptus marginata) and marri (E. calophylla) in the uplands and Banksia spp. heath and woodland on the coastal plains. Commercial plantations of pine and eucalyptus cover only 1% of the area. Rapid economic development has resulted in total water use rising by about 50 GL/year since the 1960s, with new dams constructed and a substantial increase in groundwater

443

abstraction (McFarlane et al., this issue). Agriculture consumes only about 31% of the total water used in SWA, slightly less than half the average of the whole country (NWC, 2005). Dryland agriculture is extensive on the Swan Coastal Plain and in the Northern Perth Basin but there is little agricultural or industrial land use within the surface water supply catchments. Irrigated agriculture covers about only 1050 km2 (around 2%) of the study area and is either self-supplied mainly from groundwater or from surface water schemes (DoW, 2003). Variations of rainfall, geology, topography, occurrence of water features, soil characteristics and soil depth across the study area have resulted in a rich speciation of plants adapted to specific ecological niches. Climate change will have an impact on the local vegetation which will in turn affect the local surface and groundwater hydrology. There are six wetlands listed under the Ramsar Convention within the study area and their values may clearly be compromised if the drying trend continues (Barron et al., this issue). 2.2. Climate South-western Australia experiences a Mediterranean type climate, with a Köppen Classification of temperate in the south and sub-tropical in the north of the study area (http://www.bom. gov.au/cgi-bin/climate/cgi_bin_scripts/clim_classification.cgi). Up to 80% of annual rainfall occurs from May to October, with the majority produced by cold fronts that come through in the mid-latitudes (30°S to 55°S) (Gentilli, 1972). Since 1975, rainfall has ranged from around 500 mm in the north of the study area to 1230 mm along the highest parts of the Darling Range and in southern coastal areas (Fig. 3). The mean annual potential evaporation (PE), calculated using Morton’s wet environment reference evapotranspiration (ET) formulation (Chiew and Leahy, 2003; Morton, 1983) for the period 1975 to 2007 with data from the SILO Data Drill

Fig. 2. Study area in south-western Australia showing the surface water basins in regions, and the main land cover types.

444

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

Fig. 3. Spatial distribution of mean annual (1975–2007) rainfall, potential evaporation (PE) and ‘‘rainfall deficit’’ (the difference between the two) across the study area.

(, Jeffrey et al., 2001), is 1350 mm averaged across the study area and ranges from 1530 mm in the north to 1200 mm in the south. It exceeds rainfall throughout the study area with the exception of a small region on the south coast, and about 70% of PE occurs between October and March. The winter rainfall in this region has been previously considered amongst the most consistent and reliable in Australia (Nicholls et al., 1997). However, south-west of a line joining 30°S, 115°E and 35°S, 120°E, the region has had a substantial and prolonged decline in rainfall since the early 1970s (see Fig. 1 and Bates et al., 2008; CSIRO and Australian Bureau of Meteorology, 2007).

ter across the region (Silberstein et al., in preparation-b; Silberstein et al., 2011). Through the last decade average rainfall has been 24% lower than the previous decade and runoff to reservoirs 56% lower.

2.3. Surface hydrology

3.1. Rainfall runoff models

The river basins modelled extend a maximum of 150 km inland (Fig. 2). Rivers that extend further are too saline for consumptive use and four major saline rivers, the Swan-Avon, Murray, Blackwood and Frankland, are excluded from this study. The regolith structure has a major controlling influence on the hydrology of the region. The surface soils are highly permeable and local rainfall intensities rarely exceed the infiltration capacity (Sharma et al., 1987). Stream flow generation in the catchments is by saturation excess runoff in low-lying areas adjacent to streams or shallow groundwater discharge either directly to streams or by lateral throughflow perched on the subsurface clay or lateritic caprock (Silberstein et al., in preparation-b; Stokes and Loh, 1982). Since rainfall is low through the summer months in the northern half of the study area, unless rivers have significant groundwater connection they typically cease to flow through summer and autumn. This is less true in the southern catchments which receive some summer rainfall and have lower PE. Evapotranspiration typically averages 90% of rainfall across the region, and ranges from about 60 to nearly 100% in individual catchments (Ruprecht and Stoneman, 1993). As a consequence, the region has a relatively low average runoff coefficient by global comparisons and with high variability (Peel et al., 2001). The recent trend for a drier and hotter climate is reflected in the annual runoff and stream flow characteristics (Fig. 1) with progressively lower runoff coefficients over the last 50 years (Petrone et al., 2010). These catchments have had few changes in land management within them and the declining runoff trends are due to changes in the climate and forest density and to falling groundwa-

Five simple conceptual rainfall-runoff models are used to estimate runoff generated in the catchments under four scenarios. The models are SIMHYD (Chiew et al., 2002; Chiew et al., 2008), Sacramento (Burnash et al., 1973), IHACRES (Jakeman et al., 1990; Littlewood et al., 1997), AWBM (Boughton, 1996) and SMARG (Goswami et al., 2007; Goswami et al., 2002). Implementation of all models is as independent cells distributed across the catchments on a 0.05° longitude  0.05° latitude grid (25 km2 per grid cell). The model inputs are daily rainfall and PE on each grid cell and runoff depth is calculated for each cell and the mean of all cells within a catchment given as the catchment runoff. The flows from all catchments upstream of a terminal point, such as a dam, lake or stream outlet, are summed to give the flow to that point, and flows from all catchments within a basin summed to determine the total flow for the basin. Since our results are analysed as average annual totals, a simple summation of subcatchment flows to give larger catchment and basin flows is appropriate. There is no regulated flow on these rivers and they are short enough that explicit river routing was deemed unnecessary.

3. Methods and data Stream flow was simulated for each catchment at 204 defined points within the 13 river basins under scenarios of a continuation of the historical and recent climates and for projected future climate.

3.2. Model calibration and parameter estimation methodology 3.2.1. Model calibration The rainfall-runoff models are calibrated against daily observed stream flow data from 1975 to 2007 from 106 gauged catchments within the 13 basins in the study area. The catchments ranged in area from 10 to 4000 km2, and have at least 10 years of recent, good quality and continuous stream flow record. The length of

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

3.2.2. Parameter estimation in ungauged catchments The model parameter values are determined by the input climate and catchment characteristics on which they are calibrated. In addition to the 106 gauged catchments used in calibration, flow estimations are required in 98 ungauged catchments. Since there is no proven method to estimate these parameters for ungauged catchments, the stream flows for these catchments are simulated using model parameters transferred from calibrated catchments based on a combination of proximity and hydrological similarity to the calibrated catchments (SKM, 2008; Zhang and Chiew, 2009). This assessment of similarity had been made in a previous study for all of the calibration catchments and considered proximity, mean annual rainfall, perimeter of upstream catchment, mean annual actual evaporation, winter potential evaporation, percentage of woody vegetation and the mean elevation of the catchment. 3.3. Climate and development scenarios The assessments of current and future water availability are undertaken by considering five climate scenarios to 2030. The scenarios projected for 2030 are:

‘recent’ climate (1997–2007) – one sequence generated as the years 1997 to 2007 repeated three times; three ‘future’ (2030) climates – a wet, median and dry variant selected from 45 climate sequences based on 15 GCMs for each of low, medium and high global warming scenarios. The historical gridded daily rainfall and other meteorological variable sequences were taken directly from the SILO Data Drill (, Jeffrey et al., 2001) on the 0.05° by 0.05° grid for the 33 year period between 1st January 1975 and 31st December 2007, and potential evaporation (PE) sequences calculated using Morton’s (Morton, 1983) wet areal reference evaporation formulae, as described by Charles et al. (2010, Appendix B). This period was chosen because it follows a significant post-1974 climate shift which resulted in substantially reduced stream flows (as illustrated in Fig. 1, and see Cai and Cowan, 2007; Charles et al., 2010; IPCC, 2007a; Petrone et al., 2010). All the projected climate sequences (including the historical and recent scenarios) were constructed with this historical sequence to begin with, as ‘warm up’, and the projected sequences following on for the 33 year projection period. For the ‘recent’ climate scenario the 33-year climate sequence was generated from the rainfall and PE characteristics of the past 11 years (1997–2007) repeated three times. Although 11 years is too short to capture the variability in the regional climate, the average rainfall and runoff generated through this period was significantly lower than previous decades (Petrone et al., 2010) and accordingly local water management agencies have used this in their scenario planning (Water Corporation, 2005). The ‘future’ climate sequences were generated by rescaling the ‘historical’ sequence to reflect changes in rainfall and PE derived from the 45 GCM projections to 2030. The future climate scenarios used in the study were produced by applying a daily scaling approach (see Chiew et al., 2009) based on pattern scaling (Mitchell et al., 2003) with additional scaling to account for the projected changes in daily rainfall intensity (Mpelasoka and Chiew, 2009). The scenarios were developed on a ‘per degree of global warming’ basis. Monthly rainfall and other climate variables (as used in the PE calculation) for 1870–2100 were obtained from 15 GCMs (selected for their availability of daily rainfall and other data) used in the Intergovernmental Panel on Climate Change 4th Assessment Report (IPCC, 2007a). For each GCM, season, and GCM grid point these outputs were linearly regressed against simulated global

(b) 1.0 Nash-Sutcliffe Efficiency

(a)

‘historical’ climate (1975–2007) – one sequence based on the historical climate, hence assuming that there is a continuation of the climate of the period 1975 to 2007;

0.9 0.8 0.7 0.6 0.5

616216 616040 616011 616002 614036 614005 613008 612034 612019 612001 610014 610003 609002 607220 607004 606002 603012 603001

stream flow records varies so the calibrations on each catchment are for only partially concurrent periods. Parameters are optimised by automatic calibration to maximise the Nash–Sutcliffe (Nash and Sutcliffe, 1970) efficiency (NSE) of daily runoff together with a constraint that the total modelled runoff over the calibration period is within 5% of the total recorded runoff. The genetic algorithm, shuffled complex evolution, and uniform random sampling optimisation methods are applied and each is followed by the Rosenbrock local optimisation. The representative parameter set for each model in each catchment is from the optimisation algorithm that gives the highest objective function value. Within a calibration catchment the optimised parameter values are identical for all grid cells (Chiew et al., 2009). During the calibration period, logging, fire, the spread of dieback disease, mining and reforestation, and private farm dams may have affected runoff in some catchments but it is assumed the average impact of these activities is minor and is included in the optimised parameter set. Calibration downstream of major dams assumes that there had been no flow released from the reservoirs, based on information supplied by the Water Corporation, and all flow downstream was generated as runoff from the catchment area below the dam.

445

Catchments from north-west to south-east Fig. 4. (a) Frequency distribution of the Nash–Sutcliffe Efficiency for daily runoff for the Sacramento and IHACRES models and the adopted model being the average of the two, and (b) the distribution of NSE for each catchment arranged from north to south with the mean indicated by the dashed line.

446

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

Wet future climate

Dry future climate

Change in mean annual runoff

Change in mean summer runoff

Change in mean winter runoff

Change in runoff exceeded 10% of the time

Change in low flow days with runoff less than 0.1mm

Fig. 5. Comparison of relative changes to selected hydrological statistics under the wet and dry scenarios as simulated by the two selected models (Sacramento and IHACRES) for all gauged catchments.

average surface air temperature to give the proportional change in each variable per degree of global warming. Daily scaling factors for rainfall intensity were then obtained from GCM simulated changes to daily rainfall percentiles (also by GCM, season and GCM grid point) expressed as percent change per degree of global warming (Chiew et al., 2009). These seasonal and daily scaling factors were

multiplied by our low, medium and high global warming scenarios, 0.7 °C, 1.0 °C and 1.3 °C, to account for the full range of the IPCC AR4 projections for 2030 (relative to a 1990 climatological baseline). The historical time-series for all 0.05° grid cells within the study area were multiplied by the 45 sets of seasonal and daily scaling factors (15 GCMs  3 global warming scenarios) to produce an ensemble of

447

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455 Table 1 Mean annual rainfall, runoff and selected statistics under projected climate scenarios. Climate scenario

Study area averages 34,942 km2

Mean annual rainfall (mm) Mean annual runoff (mm) Stream flow volume (GL) Runoff Coefficient ‘‘Natural’’ elasticity*

Recent

837 98 3411 12% 2.95

818 819 769 91 88 74 3172 3068 2575 11% 11% 10% 1.95 2.96 2.82 Change relative to the historical scenario 2% 2% 8% 7% 10% 25%

717 57 1986 8% 2.97

738 754 692 40 43 33 396 422 320 5% 6% 5% 1.95 2.96 2.82 Change relative to the historical scenario 4% 1% 10% 13% 8% 30%

643 22 217 3% 2.97

783 803 742 108 112 93 663 689 569 14% 14% 12% 0.70 2.73 2.87 Change relative to the historical scenario 4% 2% 9% 11% 7% 23%

694 72 443 10% 2.88

871 858 818 111 103 89 2113 1957 1686 13% 12% 11% 2.12 3.02 2.59 Change relative the historical scenario 1% 3% 7% 4% 12% 24%

763 70 1326 9% 2.98

Mean annual rainfall Mean annual runoff Northern region (Gingin to Murray) 9848 km2

Mean annual rainfall (mm) Mean annual runoff (mm) Stream flow volume (GL) Runoff Coefficient ‘‘Natural’’ elasticity*

766 46 457 6% 2.95

Mean annual rainfall Mean annual runoff Central region (Harvey to Preston) 6134 km2

Mean annual rainfall (mm) Mean annual runoff (mm) Stream flow volume (GL) Runoff Coefficient ‘‘Natural’’ elasticity*

818 121 742 15% 2.76

Mean annual rainfall Mean annual runoff Southern region (Busselton to Denmark) 18,960 km2

Mean annual rainfall (mm) Mean annual runoff (mm) Stream flow volume (GL) Runoff Coefficient ‘‘Natural’’ elasticity*

881 117 2212 13% 3.04

Mean annual rainfall Mean annual runoff *

Wet

Future scenarios Median

Historical

Dry

14% 42%

16% 53%

15% 40%

13% 40%

Elasticity is calculated foll owing (Sankarasubramanian et al., 2001).

daily rainfall and PE scenarios encompassing the projected range of climate change for the study area. Results and discussion of this approach are presented in detail by Charles et al. (2010) and Chiew et al. (2009), and full descriptions of the climate scenarios by Silberstein et al. (2010). From the 45 sequences, three were selected for reporting to represent the range of projected flows at each outflow point, the 5th highest, the median, and 5th lowest runoff, representing a ‘wet’, ‘median’ and ‘dry’ future scenario. Note that the wet future scenario is so-called because it is at the high runoff end of projections; it may still be, and generally was, drier than the historical sequence in any catchment in the study area.

elled catchment flows for the recent and wet, median or dry future climate scenarios and the historical scenario give changes to the catchment flow regimes and, hence, available resources. In this way specific dependency of results on model structure and initial conditions is minimised and the results are assumed dependent only on the impact of the climate scenarios. The runoff and statistics of projected flows reported are for the ‘adopted model’ determined from the calibration results described below. These results are used to determine yields and to estimate the ability of catchments to meet projected future demands and to examine implications for water dependent ecosystems discussed by McFarlane et al. (this issue) and Barron et al. (this issue).

3.4. Running model simulations 4. Results All models are used in predictive mode – that is, the assembled scenario climate sequences begin in 1975, with the historical sequence up to 2007, and then run through the projected climates (historical, recent, and 45 future) for a 33 year simulation. The results from the 33 year simulations are evaluated as mean annual runoff in response to representative climate conditions in 2030; that is the sequence is intended as a statistical representation of conditions under a 2030 climate, not specifically runoff in any particular year, and any one of the years of projection is equally likely and just as likely as its equivalent year in the historical 33 year period. The modelling results are assessed as differences in projections of future catchment runoff from the base case of a continuation of the historical climate for another 33 years. For example, the relative changes between the daily flow duration curves of mod-

4.1. Model calibration 4.1.1. Selection of the ‘Adopted Model’ The five models used were calibrated on 106 gauged catchments distributed throughout the study area. Although no single model gave universally the best calibration for all catchments, the Sacramento model was the best of the five in 74 of the 106 catchments and IHACRES was best in 25 cases (Silberstein et al., in preparation-a). To choose the best ensemble model for each catchment, the selected runs from the five models were combined as simple linear average combinations and the runoff from each of the 26 multiple combinations plus the five individual model runs was compared against the calibration data. Analysis of the objective function given by NSE of daily observed flow and averaged

448

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

Fig. 6. (a) Mean annual rainfall, and (b) mean annual runoff under the historical climate and the relative change from this under the recent and future climate scenarios.

modelled flow from the two models IHACRES and Sacramento was found to be superior to the other combinations and this was selected as the ‘adopted model’ (Silberstein et al., in preparation-a). The adopted model represents the flow with daily NSE ranging from 0.63 to 0.93 (mean = 0.84) in the calibration catchments (Fig. 4). Over 80% of catchments were calibrated with daily NSE greater than 0.8 and monthly NSE was greater than 0.8 in 94% of catchments. Despite the large gradient in rainfall and PE across the study area, there is consistently good performance of the adopted model with no apparent geographic trend in model performance. The consistent calibration results from the two models with different internal structures gives confidence that simulated runoff responses are most likely within the range of actual response (Ye et al., 1997). Furthermore, the rainfall inputs for simulation are within 20% of the value in the calibration period, a limit suggested by Vaze et al. (2010) and Ye et al. (1997) for meaningful future projection. However, as Chiew et al. (2009) pointed out there may be significant model biases for individual climate

scenarios. Comparing two of the conceptual models (that we also tested) for the Murray Darling Basin they found that the models gave substantially the same output under the wet scenario, but significantly different under the dry scenario. In our case, the results are somewhat different, with greater consistency under both scenarios (Fig. 5) giving additional support for the use of the two models selected. The two models give a very close average projection under both the wet and dry scenarios without the bias found for SIMHYD and Sacramento in the Murray–Darling basin catchments under the dry future scenario reported by Chiew et al. (2009). In only 10% of the catchments does the change in average total runoff differ by more than 10% between these two models. There is greater difference in the changes to summer flows, with the projected changes in 25% of the catchments differing by more than 25% under the dry future climate, not dissimilar to the finding of Chiew et al. Although we found a greater difference between the models in changes to summer flow under the wet future climate, on average there is much less flow in summer than winter and

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

hence this difference is less significant for overall water balance projections. Note that in our case all future projections, including the wet future, are drier than the historical period. 4.2. Projected climate The mean annual rainfall during the recent period (1997– 2007) was lower than the previous 22 years, and all future climate projections from the 15 GCMs have a further decrease in rainfall (Table 1, Fig. 6). The average declines in rainfall across the study area relative to the historical sequence are 2%, 2%, 8% and 14% under the recent, wet, median and dry future scenarios, respectively. However, the decline is not uniformly distributed with the greatest proportional declines in the northern region of the study area (Table 1), and some areas in the southern region experiencing a rise in rainfall under the recent climate. The greatest rainfall depth declines under the recent climate are in the central region (Fig. 6) where the major water supply reservoirs are located, and this and the southern part of the northern region have the greatest declines under the median and dry future climate projections. 4.3. Rainfall and runoff projections 4.3.1. The range of GCM projections Averaged across all surface water basins, all GCMs under all warming scenarios project less rainfall than under the historical climate (Fig. 7a). Averaged over the whole study area, rainfall is projected to be more than 10% lower than under the historical climate in 12 of the 45 future climate projections, and more than 5% lower in 38 of the 45 projections (Fig. 7a). Runoff is projected to be lower than under the historical climate under all 45 future climate projections, by more than 10% in all but three of the projections and by more than 40% under four of the 45 future climate projections (Fig. 7b). 4.3.2. Regional changes in runoff Regional differences are apparent, with the central region having the highest runoff coefficient in the study area, at 15% under the historical climate, reducing to 10% under the dry future climate scenario, and this equates to a reduction in mean annual reservoir inflow volumes of 82 GL. The Swan Coastal and Murray basins of

449

the northern region contain the most important surface water reservoirs for the metropolitan and regional water supplies, yet the rainfall and runoff coefficient in these river basins are the lowest in the study area. Under the projected median and dry future climate scenarios rainfall declines by 10% and 16% and runoff by 30% and 53%, respectively. In several of the northern reservoirs the stream flow under the dry projected climate is less than 40% of that under the historical climate. Large areas in the southern region support irrigated agriculture, particularly horticulture, with water collected in gully and hillside on-farm dams. There are also major national parks in the region with important ecosystems dependent on a reliable stream flow. Although the recent climate in some basins in the southern region has been slightly wetter than the historical period and this has protected the region somewhat against the overall drying trend, the future projections are for a significant reduction in stream flows relative to those under the historical climate (Figs. 6 and 8, Table 1). The greatest declines in quantity of runoff are in the central and southern regions although the greatest proportional reductions are in the northern region. Under the wet future climate a slight decline (2%) in rainfall across the study area results in 10% decline in average runoff. Under the median future climate an 8% decline in rainfall results in a 25% reduction in runoff across the study area, ranging from 23% to 30% decline in the central and northern regions, respectively (and in one catchment in the northern region runoff declines by 50%). There is a significant decline in the high (low frequency) flows under the median and dry future climates, particularly along the south coast and in the high yielding central region (Fig. 8a). These flows are significant for flooding, for major reservoir filling events and for over-bank events that may be important for some ecological functions. However, of probably greater significance ecologically (Barron et al., this issue) is the impact on the high frequency low flow events (in this case the 90th percentile exceedance probability, Fig. 8c). Large proportions (over 70%) of the study area have Q90 flows of below 0.05 mm/day and hence the change maps show that for many of these areas Q90 falls to zero and there are large areas with effectively no flow for at least 10% of the time.

4.3.3. Sensitivity of runoff to change in rainfall Under the historical climate the runoff coefficient is lower ( 6%) in the northern region of the study area than in the central

Fig. 7. Percentage change relative to the historical climate in (a) rainfall and (b) runoff under the 15 GCMs subject to three levels of global warming averaged across all surface water basins in the study area. The left end of the orange bars indicates the high warming, the right-hand end of the green bars the low warming and the central change in colour the medium warming. The 5th wettest (wet), median and 5th driest (dry) runoff scenarios are indicated by the symbols.

450

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

Fig. 8. The (a) 10th (b) 50th and (c) 90th percentile daily runoff under the historical climate and change in these under the recent and future climate scenarios.

(15%) and southern regions (13%). While the annual rainfall is not particularly variable (coefficient of variation (CV) of 0.12), the annual runoff is (CV0.4). In SWA, runoff coefficient tends to rise

with rainfall ranging from 1% in low rainfall areas to 25% or higher in high rainfall areas, and is strongly influenced by density of vegetation cover (Li et al., 2010; Schofield et al., 1989).

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

451

Fig. 9. ‘‘Natural’’ or ‘‘non-parametric’’ rainfall-runoff elasticity (Sankarasubramanian et al., 2001) across the study area for the historical and recent and future climate scenarios.

The sensitivity of stream flow to climate change can be investigated with a number of statistical measures, for example the change in runoff coefficient with change in rainfall (Table 1). This sensitivity can also be assessed as the ratio of the relative change in runoff to the relative change in rainfall under the scenarios, sometimes referred to as the rainfall-runoff elasticity (Chiew et al., 2006; Sankarasubramanian et al., 2001; Schaake, 1990). Schaake (1990) introduced the concept of a rainfall-runoff elasticity for climate change assessments, defined as:

ep ¼

dQ =Q dQ P ¼ dP=P dP Q

ð1Þ

where dQ/Q represents the proportional change in flow and dP/P the proportional change in precipitation under the changed climate relative to the unchanged, historical, climate. In this case, elasticity can be calculated from the relative changes in runoff and rainfall (Table 1) projected by the model. For our study area the mean values are 3.5, 3.2 and 2.9 for the wet, median, and dry scenarios, respectively, meaning for a given fractional change in rainfall streamflow will change by that fraction multiplied by these factors. The higher elasticity value for the wet scenario is due partly to the relatively small change in rainfall between the historical and wet scenarios with a larger relative change in runoff. Sankarasubramanian et al. (2001) presented a ‘‘non-parametric’’ elasticity representing the inherent sensitivity of catchments to changes rainfall:

ep ¼ median

Qt  Q P Pt  P Q

! ð2Þ

where Qt and Pt are flow and rainfall for each year, or time increment, in the series, and Q and P are the mean flow rainfall through the series. This elasticity does not depend on the model used and further can be used with historical observed sequences, hence we refer to it as a ‘‘natural’’ elasticity. It can be used to compare different time periods or, as we do here, projected changed sequences with historical ones. Examined spatially (Fig. 9) there is a large range across the study area, with an average value of 3.08 under the historical climate rising to 3.48 under the dry future climate and with 70% of the region having elasticity falling between 2.5 and 4.6 under the historical climate and between 2.7 and 5.3 under

the dry future climate (Fig. 9) with parts of the northern and southern regions being the most sensitive to changes in climate. Thus consequences for both ecological flows and for water resources are most significant in these areas (Barron et al., this issue). Also some areas increase in elasticity and some decrease, thus the responses are certainly not uniform across the study area. 4.3.4. Seasonal distribution Under the recent climate, the rainfall peaks in August, a month later than under the historical climate (Fig. 10a and b). This later onset of winter may accentuate the impact of lower rainfall as a cause of lower recent runoff as soils are drier at the peak of the wet season than they would have been when there was an earlier start to winter. However, the simple scaling used to generate the future climate sequences preserves the number of rain days in each season with a slight change in seasonal totals, hence the seasonal distribution of flows is not greatly different under the future climate projections from under the historical climate. 4.3.5. Rainfall and runoff variability Runoff variability is increased under the future climate scenarios thus reducing the reliability of total stream flow in addition to already lower stream flow due to drier future climate. The variability is more pronounced in the northern region with less overall runoff due to lesser projected rainfall. Under the historical climate annual rainfall averaged across the study area exceeds 900 mm, with runoff exceeding 130 mm, about once every 5 years (Fig. 10c and d). Under the wet future climate these remain about the same, however, under the median future climate, rainfall is projected to reach 900 mm only about once in 20 years, and under the dry future around once in 33 years. There is a significant reduction in annual rain and runoff duration curves for the area under the future climate scenarios relative to those under the historical climate. As well as a reduction in frequency of high rainfall and runoff years under the future climate projections there is a reduction in the area producing significant runoff. Under the historical climate the median rainfall within the surface water basins was 838 mm which reduced to 773 mm under the median future climate projection and to 716 mm under the dry future projection (Fig. 10e). The maximum mean annual runoff from any cell under the historical climate was 342 mm, reducing to 279 mm under the median future

452

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

(a)

(b)

(c)

(d)

(e)

(f)

Fig. 10. Mean monthly (a) rainfall and (b) runoff, annual (c) rainfall and (d) runoff duration curves, and spatial distribution of (e) mean annual rainfall and (f) mean annual runoff averaged across the study area, under the historical, recent and median future climate scenarios with range between the wet and dry future scenarios indicated.

climate and to 231 mm under the dry future projection (Fig. 10f). Under all future climate projections the coefficient of variation (both temporal and spatial) of annual runoff is projected to increase, which would have implications for both the reliability of supply and the security of environmental flows. 4.3.6. Impact on inflow to reservoirs The major public water supply reservoirs are located in the northern and central regions of our study area. The regional distribution of changes in stream flow under the recent climate of 13% and 11% in the northern and central region respectively, is similar to the measured decline that has occurred in the past 11 years (1997–2007). Under the median future climate projection, runoff declines by 30% and 23% in the northern and central regions,

respectively, relative to the historical scenario. This translates to reductions in stream flows of 137 GL and 173 GL, respectively. Under the dry future scenario these reductions are 240 GL and 299 GL in the northern and central regions respectively. Several reservoirs are projected to have a 20% reduction in inflow under a continuation of the recent climate, and a 40% and 70% reduction under the median future and dry future climate projections, respectively (Fig. 11 and Table 2). This clearly has major implications for future water resources. While the impact on the Stirling, Wellington and Harris reservoirs in the central region is of a more modest proportion, by virtue of the size of the annual inflow there is a major impact on the inflow volume in the Wellington Dam. In total the projected decline in inflow under the median future climate is 93 GL and under the dry future scenario is 156 GL.

453

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

Fig. 11. (a) Percentage change and (b) volume change in inflow to the major reservoirs under the recent climate and median and dry future climate projections relative to that under the historical climate.

Table 2 Inflows to the major reservoirs in south-western Australia under the historical, recent and future climate scenarios. Reservoir

Catchment

Inflow (GL/y)

Change in inflow from historical (GL/y) Future scenarios

Mundaring Weir Canning Dam Wungong Dam Serpentine Dam North Dandalup Dam South Dandalup Dam Stirling Dam Wellington Dam Harris Dam Total % of Historical % of Recent

Future scenarios

Area (km2)

Historical

Recent

Wet

Median

Dry

Recent

Wet

Median

Dry

1470 728 132 664 153 311 251 2509 321 6539

17.4 39.6 16.4 28.6 12.2 12.5 47.1 113.2 39.3 326 100 113

12.6 34.6 14.3 21.9 10.8 10.9 42.9 102.7 38.8 290 89 100

15.5 36.8 15.2 24.8 11.0 10.9 44.5 102.8 36.9 299 92 103

10.4 29.0 12.2 16.6 8.2 7.4 37.9 81.5 30.2 234 72 81

5.2 20.1 8.8 9.8 5.3 4.3 31.2 62.4 23.2 170 52 59

4.8 5.0 2.0 6.7 1.4 1.6 4.3 10.4 0.5 37 11 0

1.9 2.8 1.2 3.7 1.2 1.5 2.7 10.4 2.3 28 8 3

6.9 10.6 4.2 11.9 4.0 5.0 9.2 31.7 9.1 93 28 19

12.2 19.5 7.5 18.8 6.9 8.2 15.9 50.8 16.1 156 48 41

5. Discussion The global climate system underwent a fundamental shift in the mid 1970s with impacts on sea surface and atmospheric temperatures and ocean currents (Giese et al., 2002; Wainwright et al., 2008). The impacts of this on water resources are nowhere more evident than in SWA (Hennessy et al., 2007) and our results show the impacts on stream flow of the projections of a drier future in the region. It is unusual for the uniformity found amongst GCM projections for this region as there is usually a range with different models projecting rises and falls in rainfall in most areas that have been examined; for example in the Murray–Darling basin (Chiew et al., 2009). This gives confidence in the direction of rainfall projections, if not also the magnitude, and has major implications for water resources planners with inflow to the major reservoirs potentially severely reduced. Previous studies of the impact of climate change on stream flow in south-western Australia have reported a threefold change in runoff for every unit change in rainfall (Berti et al., 2004; Jones et al., 2006; Kitsios et al., 2009; Smith et al., 2009) and this is broadly supported by our findings. Applying the ‘‘natural’’ climate–runoff elasticity concept (Sankarasubramanian et al., 2001) we find the region with somewhat higher elasticity than other areas in Australia. Chiew et al. (2006) examined this elasticity for 219 catchments across Australia and found an average value of 2.38, with 70% of catchments between 2.0 and 3.5. Thus the elasticity for our study area is already large (relative to the values given

by Chiew et al. (2006), for example) and rises under the future drying climate indicating the region’s sensitivity to climate change. There is a high inverse correlation between elasticity and the runoff coefficient (Q/P) (not shown but pointed out by Chiew et al. (2006)) but this is due largely to its inclusion in the ep calculations. Runoff from a catchment is the integrated result of many processes and affected by many factors including total rainfall, rainfall intensity and seasonal distribution, potential evaporation, catchment vegetation, and soil water storage and characteristics. The effect of future climate in any specific catchment is therefore difficult to project, with the possibility that the hydrological system may not be stationary and model calibrations from historical conditions may be unreliable for future projections. However, in our case since the projected scenarios are not so far from the historical conditions used in calibration, with rainfall reductions less than 14%, it is reasonable to accept the projected results (Vaze et al., 2010). In about half of the 106 gauged catchments, the runoff coefficient has reduced significantly over the last 30 years. The cause of this reduction is the subject of investigation elsewhere (Petrone and Silberstein, 2009; Petrone et al., 2010; Silberstein and Van Niel, 2009), but is likely the result of a combination of a lower annual total and seasonal shift in rainfall, a reduction in rainfall intensity, and forest disturbance. Observations of declining groundwater and the declining runoff coefficient suggest that in some of these catchments hydrological processes may be non-stationary. However, Silberstein et al. (in preparation-a) show that the likely impact of climate change is much greater than the uncertainty in modelling results and hence,

454

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455

while there are uncertainties in the magnitude of reduction in runoff our assessment of the prevailing drying trend is valid. The study has shown that a large scale regional water yield study can be carried out within a limited time and resource constraints. A clever use of conceptual models was made taking advantage of automatic calibration techniques to optimally calibrate a large number of catchments simultaneously with high goodness of fit across the project area. Although use of such models can be argued for and against with equal fervour, even with the somewhat simple representation of hydrological processes, the models achieved excellent calibration results with a high Nash– Sutcliffe efficiency in all catchments. Finally, the study delivered regional projections of mean annual stream flows under climate projected for 2030. We draw the conclusions based on differences in output from the base case. In presenting our results as the average of 33-year climate projections no particular year should be compared directly with our results. This is particularly noteworthy with 2010 the driest year on record in the study area and with total stream flow to the regional reservoirs around 5% of the annual average and substantially less than our dry projection.

6. Conclusions All GCMs used in the IPCC AR4 report (IPCC, 2007b) project the climate in south-western Australia will be drier and hotter by 2030 than experienced in the last 33 years. In most areas where similar analysis has been undertaken some models predict increases and some decreases in rainfall. Our modelling indicates that the impact of climate on surface water resources across south-western Australia is likely to be profound and far outweigh that of any expected development in agriculture or forestry although with the development of carbon markets or other industries it is possible this may change in future. Across the surface water basins in south-western Australia, the median projection is for an 8% decline in rainfall, resulting in a 24% decline in stream flow with an inter-decile range (10th to 90th percentile) (wet to dry) of 2–14% reduction in rainfall, resulting in a 2– 42% reduction in stream flow across the region. This is in addition to the more than 50% reduction in stream flow already experienced relative to the long-term average prior to 1975. While all regions in the study area have about a 14% decline in rainfall under the dry future climate projection, the northern region is projected to have the greatest decline in runoff, of about 53%. This region contains the major reservoirs for the supply of drinking water to the city of Perth and such a decline in reservoir inflows will have a major impact on water resources planning, and indeed the value of the dams themselves. Under the median and dry future climate projections the decline in stream flows is 137 GL and 240 GL in the northern region alone, and this is particularly significant given that the average surface water supplied to Perth is around 100 GL/year. This study has used the best available techniques appropriate for regional scale assessments, and as such provides plausible projections for the impact of climate change on the water resources of the region. The approach of applying the conceptual models at the regional scale has proved extremely useful to examine regional impacts of climate change. A similar approach could also be used for more specific ecological analysis (for example, Barron et al., this issue) or extreme circumstances such as flooding or droughts and, with different models, fire frequency trends. While improvements could be made to the modelling, with finer resolution of catchments, more process representation in the catchment models and the climate sequence generation, the results of this study are valid for the regional analysis undertaken and are useful to highlight regional issues and catchments of particular concern.

Acknowledgements The study was undertaken with funding and direction from the Australian Government’s Department of Environment, Water, Heritage and the Arts and in collaboration between CSIRO and the Western Australian Department of Water. The Water Corporation (Western Australia), the Western Australian Department of Agriculture and Food, and the Forest Products Commission of Western Australia provided advice and data for the analysis of reservoir inflows, water resources planning and industry, forestry and agricultural development. We thank Francis Chiew for valuable discussions on many aspects of the project. Valuable feedback was received on this paper from David Post and Andrew Herczeg. We also thank two anonymous reviewers for helpful suggestions on improvement to the paper.

References Abbaspour, K.C., Faramarzi, M., Ghasemi, S.S., Yang, H., 2009. Assessing the impact of climate change on water resources in Iran. Water Resour. Res. 45, W10434. doi:10.1029/2008WR007615. Abouabdillah, A., Oueslati, O., De Girolamo, A.M., Lo Porto, A., 2010. Modeling the impact of climate change in a Mediterranean catchment (Merguellil, Tunisia). Fresenius Environ. Bull. 19 (10A), 2334–2347. Ali, R., McFarlane, D.J., Varma, S., Dawes, W.R., Emelyanova, I., Hodgson, G.A., this issue. Climate change impacts on groundwater resources of south-western Australia. J. Hydrol. Arnell, N.W., 1992. Factors controlling the effects of climate change on river flow regimes in a humid temperature environment. J. Hydrol. 132 (1–4), 321–342. Arnell, N.W., Reynard, N.S., 1996. The effects of climate change due to global warming on river flows in Great Britain. J. Hydrol. 183 (3–4), 397–424. Barron, O., McFarlane, D.J., Silberstein, R.P., Ali, R., et al., this issue. Climate change and water dependent ecosystems in south-western Australia. J. Hydrol. Bates, B.C., Hope, P., Ryan, B., Smith, I., Charles, S., 2008. Key findings from the Indian Ocean Climate Initiative and their impact on policy development in Australia. Clim. Change 89, 339–354. doi:10.1007/s10584-007-9390-9. Berti, M.L., Bari, M.A., Charles, S.P., Hauck, E.J., 2004. Climate Change, catchment Runoff and Risks to Water Supply in the South-West of Western Australia. Department of Environment. Boughton, W.J., 1996. AWBM Water Balance Model Calibration and Operation Manual, Report to CRC for Catchment Hydrology, Australia. Burnash, R.J.C., Ferral, R.L., McGuire, R.A., 1973. A Generalised Streamflow Simulation System – Conceptual Modelling for Digital Computers. US Department of Commerce, National Weather Service and State of California, Department of Water Resources. Cai, W., Cowan, T., 2007. SAM and regional rainfall in IPCC AR4 models: Can anthropogenic forcing account for southwest Western Australian winter rainfall reduction? Geophys. Res. Lett. 33(L24708). doi:10.1029/2006GL028037. Charles, S.P., Silberstein, R.P., Teng, J., Fu, G., Hodgson, G.A., Gabrovsek, C., Crute, J., Chiew, F.H.S., Smith, I.N., Kirono, D.G.C., Bathols, J.M., Li, L.T., Yang, A., Donohue, R.J., Marvanek, S.P., McVicar, T.R., Van Niel, T.G., Cai, W., 2010. Climate Analyses for the South-West Western Australia Sustainable Yields Project. A Report to the Australian Government from the CSIRO South-West Western Australia Sustainable Yields Project, CSIRO Water for a Healthy Country Flagship, Australia. . Chiew, F.H.S., Leahy, C., 2003. Comparison of evapotranspiration variables in Evapotranspiration Maps of Australia with commonly used evapotranspiration variables. Aust. J. Water Resour. 7, 1–11. Chiew, F., Peel, M., Western, A., 2002. Application and testing of the simple rainfallrunoff model SIMHYD. In: Singh, V.P., Frevert, D.K. (Eds.), Mathematical Models of Small Watershed Hydrology and Applications. Water Resources Publication, Littleton, Colorado, pp. 335–367. Chiew, F.H.S., Peel, M.C., McMahon, T.A., Siriwardena, L.W., 2006. Precipitation elasticity of streamflow in catchments across the world. In: Demuth, S., Gustard, A., Planos, E., Scatena, F., Servat, E. (Eds.), Climate Variability and Change – Hydrological Impacts. IAHS Publication, Int. Assoc. Hydrological Sciences, Wallingford, pp. 256–262. Chiew, F.H.S., Vaze, J., Viney, N.R., Jordan, P.W., Perraud, J.-M., Zhang, L., Teng, J., Young, W.J., Peña-Arancibia, J., Morden, R.A., Freebairn, A., Austin, J.M., Hill, P.I., Wiesenfeld, C.R., Murphy, R.E., 2008. Rainfall-Runoff Modelling Across the Murray–Darling Basin, A Report to the Australian Government From the CSIRO Murray-Darling Basin Sustainable Yields Project, Canberra, Australia. Chiew, F.H.S., Teng, J., Vaze, J., Post, D.A., Perraud, J.M., Kirono, D.G.C., Viney, N.R., 2009. Estimating climate change impact on runoff across southeast Australia: method, results, and implications of the modeling method. Water Resour. Res. 45, W10414. doi:10.1029/2008wr007338. CSIRO and Australian Bureau of Meteorology, 2007. Climate Change in Australia. Technical Report. CSIRO, Australia. .

R.P. Silberstein et al. / Journal of Hydrology 475 (2012) 441–455 D’Agostino, D.R., Trisorio, L.G., Lamaddalena, N., Ragab, R., 2010. Assessing the results of scenarios of climate and land use changes on the hydrology of an Italian catchment: modelling study. Hydrol. Process. 24 (19), 2693–2704. doi:10.1002/hyp.7765. DoW, 2003. State Water Strategy. Department of Water, Government of Western Australia, Perth, Australia. . Gentilli, J., 1972. Australian Climate Patterns. Nelson, Melbourne. Giese, B.S., Urizar, S.C., Fuckar, N.S., 2002. Southern Hemisphere origins of the 1976 climate shift. Geophys. Res. Lett. 29 (2), 1014. doi:10.1029/2001gl013268. Goswami, M., O’Connor, K.M., Shamdeldin, A.Y., 2002. Structures and performances of five rainfall-runoff models for continuous river-flow simulation. In: 1st Biennial Meeting of International Environmental Modeling and Software Society, vol. 1, Lugano, Switzerland, pp. 476–481. Goswami, M., O’Connor, K.M., Bhattarai, K.P., 2007. Development of regionalisation procedures using a multi-model approach for flow simulation in an ungauged catchment. J. Hydrol. 333 (2–4), 517–531. Hennessy, K., Fitzharris, B., Bates, B.C., Harvey, N., Howden, S.M., Hughes, L., J.S., Warrick, R., 2007. 2007: Australia and New Zealand. climate change 2007: impacts, adaptation and vulnerability. In: Parry, M.L., Canziani, O.F., Palutikof, J.P., van der Linden, P.J., Hanson, C.E. (Eds.), Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK, pp. 507–540. IPCC, 2007a. Climate Change 2007: Impacts, adaptation and vulnerability. In: Parry, M.L., Canziani, O.F., Palutikof, J.P., van der Linden, P.J., Hanson, C.E. (Eds.), Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom; New York, NY, USA. IPCC, 2007b. Climate Change 2007: The physical science basis. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Contributions of Working Group 1 to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom; New York, NY, USA. Jakeman, A., Littlewood, I., Whitehead, P., 1990. Computation of the instantaneous unit hydrograph and identifiable component flows with application to two small upland catchments. J. Hydrol. 117, 275–300. Jean-Pierre, L., Philippe, G., Mohamed, A., Abdelmatif, D., Larbi, B., 2010. Climate evolution and possible effects on surface water resources of North Algeria. Curr. Sci. 98 (8), 1056–1062. Jeffrey, S.J., Carter, J.O., Moodie, K.B., Beswick, A.R., 2001. Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environ. Model. Softw. 16, 309–330. Jones, R.N., Chiew, F.H.S., Boughton, W., Zhang, L., 2006. Estimating the sensitivity of mean annual runoff to climate change using selected hydrological models. Adv. Water Resour. 29 (10), 1419–1429. doi:10.1016/j.advwatres.2005.11.001. Kitsios, A., Bari, M.A., Charles, S.P., 2009. Projected Impacts of Climate Change on the Serpentine Catchment, Western Australia: Downscaling from Multiple General Circulation Models. Water Resource Technical Series, WRT 36, Department of Water. Li, M., Wallace, J.F., Campbell, E., 2010. Forest Change and Water Yield Response: A Statistical Data Integration Analysis, CSIRO: Water for a Healthy Country National Research Flagship. . Littlewood, I., Down, K., Parker, J., Post, D., 1997. IHACRES – Catchment-scale Rainfall-Streamflow Modelling (PC Version) Version 1.0, The Australian National University, Institute of Hydrology and Centre for Ecology and Hydrology, Canberra, Australia. McFarlane, D.J., Stone, R., Marten, S., Thomas, J., Silberstein, R.P., Ali, R., this issue. Climate change impacts on water yields and demands in south-western Australia. J. Hydrol. McMahon, T.A., Vogel, R.M., Peel, M.C., Pegram, G.G.S., 2007. Global streamflows – Part 1: characteristics of annual streamflows. J. Hydrol. 347 (3–4), 243–259. doi:10.1016/j.jhydrol.2007.09.002. Mitchell, D., Williams, K., Desmond, A., 2003. Swan coastal plain 2 (SWA 2 – Swan Coastal Plain Sub-region). In: May, J.E., McKenzie, N. (Eds.), A Biodiversity Audit of Western Australia’s 53 Biogeographical Subregions. Department of Conservation and Land Management, Perth, pp. 606–623. Montenegro, A., Ragab, R., 2010. Hydrological response of a Brazilian semi-arid catchment to different land use and climate change scenarios: a modelling study. Hydrol. Process. 24 (19), 2705–2723. doi:10.1002/hyp.7825. Morton, F.I., 1983. Operational estimates of areal evapotranspiration and their significance to the science and practice of hydrology. J. Hydrol. 66 (1–4), 1–76. Mpelasoka, F.S., Chiew, F.H.S., 2009. Influence of rainfall scenario construction methods on runoff projections. J. Hydrometeorol. 10 (5), 1168–1183. doi:10.1175/2009jhm1045.1. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models. Part 1 – a discussion of principles. J. Hydrol. 10, 282–290. Nicholls, N., Drosdowsky, W., Lavery, B., 1997. Australian rainfall variability and change. Weather 52, 66–72. NWC, 2005. Baseline Assessment of Water Resources (Australian Water Resource 2005). National Water Commission, Australian Government, Canberra.

455

Parajuli, P.B., 2010. Assessing sensitivity of hydrologic responses to climate change from forested watershed in Mississippi. Hydrol. Process. 24 (26), 3785–3797. doi:10.1002/hyp.7793. Peel, M.C., McMahon, T.A., Finlayson, B.L., Watson, F.G.R., 2001. Identification and explanation of continental differences in the variability of annual runoff. J. Hydrol. 250 (1–4), 224–240. Petrone, K., Silberstein, R.P., 2009. Streamflow Decline in South-West Western Australia: Groundwater and Surface Water Threshold Response in a Changing Climate, Greenhouse 2009. CSIRO, Perth. Petrone, K.C., Hughes, J.D., Van Niel, T.G., Silberstein, R.P., 2010. Streamflow decline in southwestern Australia, 1950–2008. Geophys. Res. Lett. 37 (11), 11401. doi:10.1029/2010GL043102R. Piao, S.L., Ciais, P., Huang, Y., Shen, Z.H., Peng, S.S., Li, J.S., Zhou, L.P., Liu, H.Y., Ma, Y.C., Ding, Y.H., Friedlingstein, P., Liu, C.Z., Tan, K., Yu, Y.Q., Zhang, T.Y., Fang, J.Y., 2010. The impacts of climate change on water resources and agriculture in China. Nature 467 (7311), 43–51. Ragab, R., Prudhomme, C., 2002. Climate change and water resources management in arid and semi-arid regions: prospective and challenges for the 21st century. Biosyst. Eng. 81 (1), 3–34. doi:10.1006/bioe.2001.0013. Ruprecht, J.K., Stoneman, G.L., 1993. Water yield issues in the jarrah forest of southwestern Australia. J. Hydrol. 150 (2–4), 369–391. Sankarasubramanian, A., Vogel, R.M., Limbrunner, J.F., 2001. Climate elasticity of streamflow in the United States. Water Resour. Res. 37 (6), 1771–1781. Schaake, J.C., 1990. From climate to flow. In: Waggoner, P.E. (Ed.), Climate Change and U.S. Water Resources. John Wiley, New York, pp. 177–206 (Chapter 8). Schofield, N., Loh, I., Scott, P., Bartle, J., Ritson, P., Bell, R., Borg, H., Anson, B., Moore, R., 1989. Vegetation Strategies to Reduce Stream Salinities of Water Resource Catchments in South-West Western Australia, WS33, Water Authority of Western Australia. Sharma, M.L., Barron, R.J.W., Fernie, M.S., 1987. Areal distribution of infiltration parameters and some soil physical properties in lateritic catchments. J. Hydrol. 94 (1–2), 109–127. Silberstein, R.P., Van Niel, T., 2009. What is the connection between long-term changes in rainfall and stream flow in catchments in south-west Western Australia? In: Proceedings Greenhouse 2009, held in Perth (March). Silberstein, R.P., Aryal, S.K., Hodgson, G.A., McFarlane, D.M., Pearcey, M., Durrant, J., Bari, M.A., 2010. Rainfall-Runoff Modelling in South-west Western Australia. 2. Rainfall-Runoff Modelling Methodology, CSIRO Water for a Healthy Country Flagship, Australia. Silberstein, R.P., Macfarlane, C.K., Petrone, K.C., Hughes, J.D., Dawes, W.R., Lambert, P., Smart, N.F., Li, M., Wallace, J.F., Aryal, S.K., 2011. Stream Flow and Vegetation Dynamics Under a Changing Climate and Forest Management, Final Report to the WA Water Foundation on Project 041-05, Publ. CSIRO Water for a Healthy Country National Research Flagship. Silberstein, R.P., Aryal, S.K., Durrant, J., Braccia, M., in preparation-a. Model selection and performance for future runoff projections under a changing climate. J. Hydrol. Silberstein, R.P., Macfarlane, C.K., Dawes, W.R., Petrone, K.C., Hughes, J.D., in preparation-b. Stream flow, forest density and a changing climate. J. Hydrol. SKM, 2008. Estimation of Sustainable Diversion Limits for catchments in South West Western Australia. Regionalisation of Sustainable Diversion Limits for Catchments, Sinclair Knight Mertz for the Department of Water, WA. Smith, K., Boniecka, L., Bari, M.A., Charles, S.P., 2009. The Impact of Climate Change on Rainfall and Streamflow in the Denmark River Catchment, Western Australia. Surface Water Hydrology Series, HY 30, Department of Water. Stokes, R.A., Loh, I.C., 1982. Streamflow and solute characteristics of a forested and deforested catchment pair in south-western Australia. In: Proceedings of the First National Symposium on Forest Hydrology. Institution of Engineers Aust., pp. 60–66. Thomson, A.M., Brown, R.A., Rosenberg, N.J., Srinivasan, R., Izaurralde, R.C., 2005. In: Rosenberg, N.J., Edmonds, J. (Eds.), Climate Change Impacts for the Conterminous USA: An Integrated Assessment. Part 4. Water Resources. Springer, Netherlands, pp. 67–88. Vaze, J., Post, D.A., Chiew, F.H.S., Perraud, J.M., Viney, N.R., Teng, J., 2010. Climate nonstationarity – validity of calibrated rainfall-runoff models for use in climate change studies. J. Hydrol. 394 (3–4), 447–457. doi:10.1016/j.jhydrol.2010.09.018. Wainwright, L., Meyers, G., Wijffels, S., Pigot, L., 2008. Change in the Indonesian Throughflow with the climatic shift of 1976/77. Geophys. Res. Lett. 35, L03604(3). doi:10.1029/2007GL031911. Water Corporation, 2005. Integrated Water Supply Scheme (IWSS) Source Development Plan 2005. Planning Horizon 2005–2050, Perth, Western Australia, Publ. P. Water Corporation, Western Australia. . Ye, W., Bates, B.C., Viney, N.R., Sivapalan, M., Jakeman, A.J., 1997. Performance of conceptual rainfall-runoff models in low-yielding ephemeral catchments. Water Resour. Res. 33, 153–166. Zhang, Y.Q., Chiew, F.H.S., 2009. Relative merits of different methods for runoff predictions in ungauged catchments. Water Resour. Res. 45, W07412. doi:10.1029/2008wr007504.