Contents lists available at ScienceDirect

Journal of Hydrology: Regional Studies

journal homepage: www.elsevier.com/locate/ejrh

Future climate and runoff projections across South Asia fromCMIP5 global climate models and hydrological modelling

Hongxing Zhenga,⁎, Francis H.S. Chiewa, Steve Charlesb, Geoff Podgera

a CSIRO Land and Water, GPO BOX 1666, Canberra ACT 2601, Australiab CSIRO Land and Water, Perth, WA, Australia

A R T I C L E I N F O

Keywords:Climate changeProjectionsRunoffPrecipitationWater securitySouth Asia

A B S T R A C T

Study region: The South Asia.Study focus: This paper presents future climate and runoff projections for the South Asia regionunder the RCP8.5 scenario with climate change informed by 42 CMIP5 GCMs. Runoff is projectedfor 0.5° grids using hydrological models with future climate inputs obtained by empiricallyscaling the historical climate series.New hydrological insights: The modelling results indicate that future runoff will increasethroughout most of the region except in the far north-east and far north-west. The medianprojection shows increases in mean annual runoff of 20–30% in the Indian sub-continent for2046–2075 relative to 1976–2005. The change in runoff is driven mainly by the change inprecipitation, moderated (in wetter futures) or intensified (in drier futures) by higher tempera-ture and potential evaporation. The paper also investigates the uncertainties of the projection dueto scaling methods and selection of GCMs. The difference in runoff projections from differentscaling methods is small relative to the large uncertainty in precipitation projections from theGCMs. Sub-selecting only the “better” performing GCMs shows marginal difference in the un-certainty range of projected runoff. For the broad scale projections presented here, it is best to useprojections informed by all the GCMs to provide an indication of the full uncertainty range.

1. Introduction

Water security in South Asia will be under increasing stress owing to socio-economic growth and global climate change. It isforecasted that the population in South Asia will be more than 2.3 billion by 2050 (United Nations, 2015). The growth of populationand expanding economies will result in an increase in water demand. Meanwhile, there is strong evidence that many parts of SouthAsia are experiencing long-term warming trends that will continue into the future (Hijioka et al., 2014). The changes in climate willhave significant impacts on water availability. Moreover, the warmer future climate will increase evapotranspiration and henceincrease demand for water in irrigated agriculture, urban centres and water-dependent ecosystems. Exacerbated by climate change, itis estimated that water scarcity could cost up to 6% GDP loss in South Asia by 2050 (World Bank, 2016).

As water impacts practically all sectors (people, agriculture, industries and ecosystems), there have been considerable researchefforts on predicting or projecting water availability under future climate conditions (e.g. Devkota and Gyawali, 2015; Gupta andDeshpande, 2004; Immerzeel et al., 2010; Immerzeel and Bierkens, 2012; Nepal and Shrestha, 2015) to inform the development ofeffective adaptation options. For South Asia, most studies of hydrological response to climate change only investigate limited regions

https://doi.org/10.1016/j.ejrh.2018.06.004Received 19 December 2017; Received in revised form 26 April 2018; Accepted 19 June 2018

⁎ Corresponding author.E-mail address: hongxing.zheng@csiro.au (H. Zheng).

Journal of Hydrology: Regional Studies 18 (2018) 92–109

2214-5818/ © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

T

and climate scenarios. For example, Akhtar et al. (2008) simulated the hydrological response to climate change for three river basins(Astore, Gilgit and Hunza) in the upper Indus using two downscaling approaches driven by the HadAM3P GCM. One of the down-scaling approaches was a form of statistical downscaling using the ‘delta change’ approach and another was dynamical downscalingusing the PRECIS regional climate model at 25 km resolution. Lutz et al. (2014) assessed potential changes of water availability forthe large basins in northern South Asia, with increased runoff projected for 2050 for the upper Ganges, Brahmaputra, Salween andMekong primarily due to projected precipitation increases and for the Indus due to accelerated glacier melt in the upper basin.

The hydrological response to climate change is generally predicted using downscaled future climate projections to drive a hy-drological model (Chiew et al., 2009a; Teng et al., 2012a). One of the key challenges in factoring climate change into water resourcesmanagement lies in the uncertainty in the projections (Jiménez Cisneros et al., 2014; Lopez et al., 2009). The sources of the projectionuncertainties could be from the GCMs, the downscaling approaches, or the hydrological models (Chen et al., 2011; Teng et al., 2012b;Vetter et al., 2016). The performance of GCMs over the South Asia region have been investigated by quite a few researchers (e.g.,Chaturvedi et al., 2012; Freychet et al., 2015; Jourdain et al., 2013; Ogata et al., 2014; Palazzi et al., 2015; Saha et al., 2014). Forexample, Saha et al. (2014) found that the majority of the CMIP5 GCMs fail to simulate the post-1950 decreasing trend of Indiansummer monsoon rainfall, as they did not capture the weakening monsoon associated with the warming of southern Indian Oceanand strengthening of cyclonic formation in the tropical western Pacific Ocean. Some studies have suggested placing more weight onor using only projections from the better performing GCMs As noted by Palazzi et al. (2015), however, it is challenging in selectingbetter performing GCMs for the region as none of GCMs can reproduce all the salient features (e.g. seasonal and annual rainfallamounts, distribution and trend, or the large scale atmospheric-oceanic drivers of rainfall in the region). Like the GCMs, the latestdynamic downscaling runs in the CORDEX regional climate model experiments also do not capture the observed monsoon pre-cipitation trends or the correct magnitude of observed warming (Ghimire et al., 2015; Mishra et al., 2014). In any case, the un-certainty in climate projections (from GCMs and from downscaling approaches) must be adequately represented within the specificcontext and objectives of any hydrological modelling and integrated water resources management study.

This paper aims to present future climate (precipitation, temperature and potential evaporation) and runoff projections acrossSouth Asia, and the associated uncertainties and robustness of the projections from different considerations or treatments of climatechange projections and hydrological modelling. The future runoff is projected using hydrological models with future climate inputsinformed by 42 GCMs reported in the Intergovernmental Panel on Climate Change Fifth Assessment Report (IPCC AR5) (IPCC, 2013).The paper investigates (i) the sensitivity of runoff to changes in the climate variables, (ii) the projected runoff change resulting fromdifferent hydrological modelling approaches, (iii) the projected runoff change using different empirical downscaling approaches, (iv)the uncertainties of runoff projection within and across the GCMs, and (v) whether using a sub-set of the ‘better’ performing GCMscan reduce the uncertainties in the projection. The structure of the paper is as follows – Section 2 outlines the data and methods used;Section 3 presents and discusses the projected changes in the climate variables across South Asia; Section 4 presents the climatechange impacts on runoff and water supply security, and also discusses results and implications from different modelling con-siderations; Section 5 presents the summary and conclusions.

2. Data and methods

In this study, the South Asia region refers to the domain between 5.25–38.00 °N and 65.25–93.75 °E. To facilitate presentation anddiscussion of results, nine sub-regions based on the freshwater ecoregion map of Abell et al. (2008) are used. The nine sub-regions, asshown in Fig. 1, are North-West region (NW), part of the Tibetan Plateau (TP), Indus River Basin (IND), Ganges-Brahmaputra RiverBasin (GA), part of the Arakan Coast (AC), Narmad-Tapti River Basin (NT), Deccan Plateau (DP), Ghats Coast (GH), and Sri LankaIsland (SL).

2.1. Baseline observations and CMIP5 database

The baseline period of 1976–2005 is used for this study (‘baseline’, ‘present’ and ‘historical’ are used interchangeably throughoutthe paper, they all refer to the period 1976–2005). The daily gridded climate data of precipitation, temperature, wind speed, airpressure and radiation are obtained from the Princeton Global Forcing (PGF) dataset with spatial resolution of 0.5°×0.5° (Sheffieldet al., 2006). Fig. 1 shows the annual mean of precipitation, temperature, potential evaporation and aridity index for the baselineperiod across the South Asia region. The precipitation and temperature comes directly from the PGF dataset. Potential evaporation isestimated from the PGF climate variables using Morton’s formulation of wet environment (or equilibrium or areal potential) eva-poration (Morton, 1983; Chiew and McMahon, 1991). Aridity index is defined as mean annual potential evaporation divided by meanannual precipitation.

The future climate scenarios are derived from the CMIP5 database (http://cmip-pcmdi.llnl.gov/cmip5/), which archives transientclimate experiments from more than 20 climate modelling groups around the world. All the ensemble runs of 42 CMIP5 GCMs withboth historical and future outputs available on 15 March 2013 (the same date as that adopted by IPCC AR5) are used here. The resultspresented in this study are for a future period over 2046–2075 relative to the baseline period (1976–2005) for RCP8.5 (high re-presentative greenhouse gas concentration pathway corresponding to radiative forcing of 8.5W/m2 by 2100 relative to pre-industrialvalue, see Moss et al., 2010; Taylor et al., 2012). It is worth noting that the discussions and results presented throughout this paperare applicable for any future global warming pathway or scenario, but the projected changes would be smaller for lower RCPs (likeRCP4.5, RCP2.6 and RCP6.0).

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

93

2.2. Empirical downscaling of GCM climate variables

Downscaling is the process of translating (or modelling) the coarse resolution outputs from GCMs to catchment scale climate thatis required for impact-adaptation-vulnerability studies and hydrological modelling. There are three types of downscaling approaches(Fowler et al., 2007; Chen et al., 2011; Ekström et al., 2015): (1) Empirical downscaling, or change-factor method, which applies thechange informed by GCMs (for a future period relative to a baseline period) to perturb the baseline observations to reflect a futureclimate, (2) Statistical downscaling, which develops a statistical relationship between GCM atmospheric variables and observedclimate variables, and then use this relationship to derive future climate variables from future GCM variables; and (3) Dynamicaldownscaling, where regional climate models (RCMs) are used (Kumar et al., 2013) to model the physical atmospheric and landsurface processes at smaller spatial scale informed or constrained by a larger scale GCM.

Many downscaling comparisons have been conducted (Chen et al., 2011; Fowler et al., 2007) and generally conclude that no

Fig. 1. Mean annual precipitation, temperature, potential evaporation and aridity index across South Asia for the baseline period (1976–2005)derived from Princeton Global Forcing dataset. Map also shows the nine regions (North-West (NW), part of the Tibetan Plateau (TP), Indus RiverBasin (IND), Ganges-Brahmaputra River Basin (GA), part of the Arakan Coast (AC), Narmad-Tapti River Basin (NT), Deccan Plateau (DP), GhatsCoast (GC), Sri Lanka Island (SL)) and two locations/basins (Koshi and Brahmani-Baitarni) where results are discussed in more detail.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

94

single downscaling method is better, and this is largely because, like GCMs, different downscaling methods are tailored for differentapplications. Chiew et al. (2010) found that results from empirical downscaling typically lie within the range of other regionallyavailable statistical and dynamical downscaling methods. Frost et al. (2011) recommends using empirical downscaling methods forregional water resource planning applications, due principally to the method’s robustness.

This study therefore uses empirical scaling to scale the baseline climate observations, informed by GCM outputs for a future periodrelative to the baseline period, to reflect a future climate (Zheng et al., 2015). For the empirical downscaling here, the scaling factorfor precipitation and potential evaporation is estimated as,

= X XSF /f h (1)

where Xf and Xh are the GCM simulation for the future (2046–2075) and baseline (1976–2005) periods respectively. Once derived,the scaling factor is then used to scale the baseline climate observations (1976–2005) to produce a future climate time series. As theGCMs spatial resolutions are different to the observations, and different GCMs have different spatial resolutions, the scaling factorsderived for each of the GCMs are resampled to match with the 0.5° x 0.5° grid in Princeton dataset.

In Eq. (1), if Xf and Xh are annual means, then the scaling approach is defined as ‘annual scaling’, where all the daily values ofprecipitation (or potential evaporation) in the 30-year baseline period (1976–2005) are scaled by this single annual scaling factor toprovide a 30-year future (2046–2075) daily precipitation (or potential evaporation) series. Most of the results presented here, de-scribing runoff sensitivity to climate inputs (Section 4.1) and comparing different modelling approaches (Section 4.2) are for ‘annualscaling’.

The study here (in Section 4.3) also explores the difference in runoff projection where the climate variables in the differentseasons are scaled by factors estimated separately for the different seasons (‘seasonal scaling’), and where daily precipitation amountsor percentiles are scaled differently (‘daily scaling’). In the seasonal scaling, the original seasonal scaling factors are firstly estimatedfor each of the four seasons (Dec-Jan-Feb, Mar-Apr-May, Jun-Jul-Aug, Sep-Oct-Nov) informed by the GCM outputs. These factors arethen rescaled such that the eventual future climate series has the same change to that from “annual scaling” approach at the meanannual level. The rescaled seasonal scaling factors are estimated as,

′ =∑

SF SF SF XSF Xi i

a h

i h,i (2)

where SFi and SFi’ are the original and rescaled seasonal scaling factors for season i respectively, SFa is the annual scaling factor, andXh,i is the observed baseline mean for season i.

The ‘daily scaling’ is carried out only for precipitation, mainly to reflect potentially different climate change impact signal in thehigher daily precipitation intensity. Here, for each season, 53 daily scaling factors are calculated representing the changes in thehighest 50 daily precipitation values (as reflected by the GCM), and changes in the average of the 51–100th daily precipitation values,101–200th daily precipitation values, and 200th onward daily precipitation values. Similar to the rescaling of seasonal factors, theoriginal daily scaling factors derived from the GCMs are rescaled for each season to ensure the seasonal change is the same when thecorresponding rescaled seasonal scaling factor is used.

The ‘annual scaling’ therefore uses a single factor to scale all the baseline daily precipitation (or potential evaporation) to reflect afuture climate. The ‘seasonal scaling’ uses four different factors to scale the precipitation (or potential evaporation) in the differentseasons. The ‘daily scaling’ scales the daily precipitation in accordance with 53×4 scaling factors. The rescaling procedure men-tioned above ensures that the change in mean annual precipitation (or potential evapotranspiration) in the ‘annual scaling’, ‘seasonalscaling’, and ‘daily scaling’ are the same, as reflected by the GCM at the mean annual scale, and allows results from the differentempirical scaling methods to be directly compared. The intention here is not to find the best method, but to explore potentialuncertainties owing to different empirical scaling treatments to obtain the future climate series.

Different to precipitation and potential evaporation, the scaling factor of temperature is defined as the absolute change oftemperature and is expressed as,

= −T TSF f h (3)

where Tf and Th are the GCM simulated annual/seasonal mean daily temperature for the future and baseline periods respectively.Both annual and seasonal scaling factors of temperature are derived herein.

In the above empirical scaling methods, the future daily climate sequence is the same as the baseline climate sequence, scaled bydifferent factors. The annual, seasonal and daily empirical scaling are applied to precipitation. For temperature and potential eva-poration, only the annual and seasonal scaling are applied.

To explore climate change impact on water supply reliability (in Section 2.4 and Section 4.4), the potential change in futureclimate sequence is also considered. Here, a quantile mapping bias correction approach is used, where the GCM daily precipitationvalue at a given percentile/rank in the baseline period is matched to the observed daily precipitation value (at the Princeton 0.5° grid)at the same percentile/rank, and this relationship is then used to translate the future GCM daily precipitation series to an 0.5° gridfuture daily precipitation series. This quantile mapping bias correction is processed as,

=′ −X F F X[ ( )]fut obs his fut1

(4)

where Fobs is the cumulative distribution function of the historical observation, Fhis is the cumulative distribution function of thehistorical GCM simulation, Xfut is the future daily GCM precipitation value and X’fut is the future 0.5° grid precipitation value. Thecumulative distribution function is considered separately for each of the four seasons, and only precipitation amounts greater than

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

95

0.01mm are considered. Like in the empirical scaling methods, the future precipitation series here is rescaled such that the change inseasonal means are the same as those in the empirical seasonal scaling and the change in the annual mean is the same as that in theempirical annual scaling. It should be noted that the distribution of daily precipitation simulated by GCMs can be very different to thedistribution of observed daily precipitation (daily precipitation is smaller in GCMs, and there are more rain days in GCMs), and thequantile mapping bias correction can introduce large errors and artificially corrupt future GCM projected trends and characteristics(Hay and Clark, 2003; Maraun, 2013; Piani et al., 2010).

2.3. Modelling climate change impact on runoff

The climate change impact on runoff presented and discussed in various parts of Section 4 come from three sources: (1) Directsimulations from the GCMs, (2) Budyko relationship estimates for the future relative to the baseline (using annual means with scalingfactors of precipitation and potential evaporation), and (3) H08 hydrological modelling for the future relative to the baseline (usingdaily climate inputs with scaling factors of precipitation and temperature). The similarities and differences in the runoff projectionsfrom the different sources and methods are explored.

The Budyko relationship (Budyko, 1974) is the most widely used equation to describe basin or regional scale long-term energyand water balance (Oudin et al., 2008; Sivapalan et al., 2003; Teng et al., 2012b; Zhang et al., 2004). The Budyko relationship iscommonly presented as,

= − − =EP

ϕ ϕ ϕ ϕ PET P[ tanh(1/ )(1 exp( ))] , /1/2(5)

where E is the mean annual actual evapotranspiration, P is the mean annual precipitation, PET is the mean annual potential eva-potranspiration and ø is the aridity index. Over a long enough time period (several years), the mean annual runoff (Q), can beestimated from the above equation as the difference between mean annual precipitation and mean annual actual evaporation (Q=P- E). This simple Budyko relationship is remarkably robust for estimating large-scale energy and water balance and has been used inliterally thousands of studies, including adaptations to account for landscape features (particularly vegetation) and for local con-ditions, and to represent sub-annual and inter-annual time scales. In this study, the simple basic Budyko relationship above is used todirectly estimate the baseline (1976–2005) and future (2046–2075) mean annual runoffs from mean annual values of precipitationand potential evaporation.

The H08 model used here is one of the main global hydrological models used in regional and global water modelling studies(Hanasaki et al., 2008; Haddeland et al., 2011; Warszawski et al., 2014; Zhang et al., 2016). The runoff simulation scheme in H08 isbased on the bucket model concept (Manabe, 1969) but modified by using a “leaky bucket’’ formulation in which subsurface runoffoccurs continually as a function of soil moisture. The algorithm used in H08 for PET estimation is the mass-transfer method (Dalton,1802; Penman, 1948), which relates PET to vapour pressure deficit and wind speed. The H08 runs on a daily time step and consists ofsix main modules (land surface hydrology, river routing, crop growth, reservoir operation, environmental flow, and anthropogenicwater withdrawal). This study uses only the land surface hydrology module, which conceptualises each grid or catchment as in-terconnected stores with equations describing the hydrological processes (including snowmelt processes) and movement of waterbetween stores. The daily climate inputs needed to run the model are precipitation, temperature, solar radiation, humidity, and windspeed. The suggested or default parameters of H08 (Hanasaki et al., 2008) are used for the modelling here without further calibrationagainst local data. Global hydrological models like H08 has limitation in simulating reliably the actual runoff values at a particularlocation, but they can reasonably simulate the runoff variability over annual and longer time scales and the trend in runoff, as well asrelative changes in runoff resulting from changes in the climate inputs (Haddeland et al., 2011; Zhang et al., 2016).

Unlike the simple Budyko relationship, the H08 hydrological model runs on a daily time step and dynamically models thedifferent hydrological fluxes and stores. The H08 model is therefore likely to better represent the hydrology of the region (comparedto the Budyko relationship) as well as take into account changes in the climate inputs at the sub-annual scale. The H08 modellinghere, at 0.5° grid resolution, therefore provides a consistent and broad-scale estimation of the climate change impact on runoff acrossthe South Asia region, but cannot replace the need for more detailed and intensive modelling efforts (like parameter calibration andmodel validation) to fully inform specific management, planning or adaptation in local catchments or basins.

2.4. Modelling climate change impact on water supply security

The climate change impact on water availability will be reflected not only in the average streamflow volume but also in otherriver flow characteristics. The seasonality in river flow is likely to increase with wet seasons becoming wetter and dry seasonsbecoming drier (Ashfaq et al., 2009; Sivakumar and Stefanski, 2011). This enhanced hydrologic variability will change the reliabilityof water supply, thus compounding water management and climate adaptation challenges.

To investigate the impact of climate change on water supply security, a storage analysis (McMahon and Adeloye, 2005; McMahonet al., 2007) is carried out, where the dynamic monthly water balance of a hypothetical reservoir storage is modelled as,

= + − ≤ ≤+S S Q D S S, (0 )t t t t max1 (6)

where St+1 and St are the reservoir storages in month t and t+ 1 respectively, Qt is the runoff (or inflow into the storage) in month t,and Dt is water demand or withdrawal in month t. For the hypothetical modelling here, the storage capacity (Smax) is set equal to themean annual runoff and the monthly water demand (assumed to also include evaporation from the reservoir) is set to 0.9 times the

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

96

mean monthly runoff. To express the storage modelling results, a reservoir or storage reliability is defined here as 100% minus thepercentage time (or months) that the demand cannot be adequately met (i.e. when St+1 is negative and has to be set to zero).

3. Projected climate change across South Asia from CMIP5

The plots in Fig. 2–4 show the range of projected changes in the daily average temperature, potential evaporation and pre-cipitation respectively across South Asia, presented for each of the four seasons and annual values. The plots show the median value,25–75th percentile range and 10–90th percentile range from the 42 CMIP5 GCMs for RCP8.5. The range of temperature projections in

Fig. 2. Projected change in daily average temperature (median, 10th, 25th, 75th and 90th percentiles from the 42 GCMs) for RCP8.5 for 2046–2075relative to present/baseline (1976–2005) for the four seasons and the annual scale.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

97

Fig. 2 (and to a large extent precipitation projections in Fig. 4) are consistent with those in the IPCC AR5 atlas of global and regionalclimate projections (IPCC, 2013). However, the projections here are presented at higher temporal and spatial resolutions and comefrom a much larger and complete set of CMIP5 GCMs, to adequately reflect the range of uncertainty in hydrological modelling toinform impact-vulnerability-adaptation assessments of water resources and related sectors across South Asia.

As shown in Fig. 2, there is reasonable agreement between the GCMs in the temperature projections. For example, all the GCMsproject increases in temperature, compared to precipitation where projection can range from large increases to large decrease(Fig. 4). Averaged across the South Asia region, the median projection for RCP8.5 is an increase in daily average temperature of 2.9 °Cby 2046–2075 relative to the baseline period (1976–2005), with a 25–75th percentile range of 2.6 to 3.5 °C and 10–90th percentilerange of 2.3 to 4.0 °C. The projected increase in daily minimum temperature is slightly higher and the projected increase in daily

Fig. 3. Projected percentage change in potential evaporation (median, 10th, 25th, 75th and 90th percentiles from the 42 GCMs) for RCP8.5 for2046–2075 relative to present/baseline (1976–2005) for the four seasons and the annual scale.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

98

maximum temperature is slightly lower than the projected increase in daily average temperature (not presented here). Seasonally, theprojected temperature increase is slightly higher in winter (DJF) than in summer (JJA). Spatially, the projected temperature increaseis noticeably higher in the north (high altitude regions) than in the south, which is consistent with that found by Mountain ResearchInitiative EDW Working Group, (2015).

There is also general agreement in the potential evaporation projections (Fig. 3), where the potential evaporation for each GCMgrid is estimated from the solar radiation, maximum and minimum temperatures, and actual vapour pressure data using Morton’s wetenvironment (or equilibrium or areal potential) evaporation formulation (Morton, 1983). These projected increases in potentialevaporation are driven mainly by the increase in temperature. Most GCMs also show a slight increase in vapour pressure deficit(which will also increase potential evaporation), but there is little change and agreement between GCMs in the direction of change of

Fig. 4. Projected percentage change in precipitation (median, 10th, 25th, 75th and 90th percentiles from the 42 GCMs) for RCP8.5 for 2046–2075relative to present/baseline (1976–2005) for the four seasons and the annual scale.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

99

other variables (solar radiation and wind speed) that influence potential evaporation. Averaged across the South Asia region, themedian projection for RCP8.5 is an increase in mean annual potential evaporation of 6.2% by 2046–2075 relative to baseline, with a25–75th percentile range of 4.8 to 7.6% and 10–90th percentile range of 3.4 to 8.8%. At the extreme percentile, some of the pro-jections show a decrease in potential evaporation in summer due to higher relative humidity, and possible decrease in solar radiation.Seasonally, the projected percentage increase in potential evaporation is considerably greater in winter (DJF) than in summer (JJA).Spatially, the projected percentage increase in potential evaporation tends to be greater in the north-west.

There is much greater uncertainty in the precipitation projections, with large variations between GCMs, and in the differentseasons and regions (Fig. 4). Nevertheless, a higher proportion of GCMs project an increase in precipitation, particularly in the north-east and much more so in the monsoon summer (Jun-Jul-Aug) than in winter (Dec-Jan-Feb). The higher monsoon precipitation undera warmer climate is also consistent with those reported in climate processes and climate modelling studies briefly described in theIntroduction above. Averaged across the southern half of the South Asia region (latitude below 25 °N), the median projection forRCP8.5 is a change in annual precipitation of +11% (with a large 10–90th percentile range of -2% to +41%) by 2046–2075 relativeto the 1976–2005 baseline, and averaged across the northern half of the region (latitude above 25 °N), the median projection forRCP8.5 is a change in precipitation of +7% (with a 10–90th percentile range of -9% to +35%). The projections also suggest possibleintensification in the high extreme precipitation (i.e. the highest 0.1 percentile daily precipitation) particularly in the southern parts(not shown here).

4. Climate change impact on water across South Asia

4.1. Climate elasticity of runoff across South Asia

The plots in Fig. 5 show the climate elasticity of runoff, which is a useful concept that describes the sensitivity of runoff to changesin precipitation, potential evaporation and temperature, as simulated by the Budyko relationship and the H08 hydrological model.The climate elasticity of runoff is defined as the percentage change in runoff for a one percent change in precipitation or potentialevaporation, or a one degree change in temperature (Schaake, 1990; Sankarasubramanian and Vogel, 2001; Chiew et al., 2006;Chiew, 2006).

To estimate the climate elasticity of runoff, the climate data of the baseline period is perturbed from -10% to +10% for pre-cipitation and potential evaporation (with gradations of 1%) and from -2.0 °C to+ 2.0 °C for temperature (with gradations of 0.1 °C).

Fig. 5. Climate elasticity of runoff across the South Asia. The elasticity of runoff to precipitation and PET is dimensionless (left). The elasticity ofrunoff or PET to temperature (far right column) is in the unit of % change per oC.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

100

The perturbed climate data is then used as inputs into the Budyko relationship and H08 hydrological model, and the modelled runofffor the perturbed climate is compared with the modelled runoff for the baseline climate to estimate the climate elasticity of runoff.The mean of the estimated elasticity basing on all the perturbations are then presented in Fig. 5. As discussed earlier, potentialevaporation for the Budyko relationship is estimated here using Morton’s areal potential evaporation formulation, whilst the H08model estimates potential evaporation directly using the algorithms in the model.

The H08 modelling results in Fig. 5 indicate that the precipitation elasticity of runoff generally range from about 1.5 or 2.0 (i.e. a10% change in precipitation is amplified as a 15–20% change in runoff) in the wetter regions in the west coast, east coast and north-east to greater than 2.0 (i.e. a 10% change in precipitation is amplified as a 20% change in runoff) in the drier regions in the north-west and inland south. As expected, both the H08 hydrological model and the Budyko relationship show the similar spatial trends inthe precipitation elasticity of runoff (Fig. 5). However the runoff sensitivity to changes in precipitation is higher in the Budykorelationship, with precipitation elasticity of runoff of 2.2 for regions with aridity index ( =ϕ PET P/ ) of 1 and 3.0 for aridity index of2. It is worth mentioning that the Budyko relationship is developed and parameterised from global energy and water datasets and willgive the same results for regions with the same aridity index. For cold regions like the North-West and the Tibetan Plateau, theBudyko relationship may not lend itself to extrapolation into the future where higher temperatures will affect snow, glacier andpermafrost processes. In contrast, the H08 model simulates the hydrological dynamics and fluxes and storages at a daily time scale,and therefore reflect in more detail the hydroclimatology of the region, for the baseline climate data here and the hydrology asconceptualised and parameterised in H08.

Both the H08 hydrological model and Budyko relationship also show the similar spatial trends in the potential evaporationelasticity of runoff (Fig. 5). The runoff sensitivity to potential evaporation is slightly greater in the H08 model (particularly in thewestern arid regions) with 1% increase in potential evaporation resulting in 2–5% decrease in runoff, compared to the Budykorelationship indicating a 2–3% decrease in runoff for a 1% increase in potential evaporation. Because of the complementarity in therunoff sensitivity to precipitation and runoff sensitivity to potential evaporation (as can be seen in the Budyko relationship, see alsoArora, 2002; Dooge et al., 1999; Zheng et al., 2009), the regions with high (positive) precipitation elasticity of runoff also have high(negative) potential evaporation elasticity of runoff.

The sensitivity of runoff to higher future temperature reflects clearly the increase in potential evaporation from the increase intemperature causing the decrease in runoff. The spatial trends in the temperature elasticity of runoff and potential evaporationelasticity of runoff are therefore the same. In most regions across SA, the H08 hydrological model indicates that a 1 °C increase intemperature will result in about 5% increase in potential evaporation, which in turn leads to a 5–10% decrease in runoff. Theelasticity is close to the Clausius-Clapeyron rate, which means an increase in the water holding capacity of air (the saturation watervapour pressure) of approximately 7% per degree Celsius rise in temperature (Held and Soden, 2006).

4.2. Modelled climate change impact on mean annual runoff across South Asia

The plots in Fig. 6 show the percentage change in mean annual runoff for the RCP8.5 scenario for 2046–2075 relative to thebaseline (1976–2005), modelled by the H08 hydrological model, Budyko relationship and global climate models (GCMs). The futurerunoff for H08 and Budyko are modelled using ‘future 2046–2075′ precipitation, potential evaporation and temperature (daily seriesfor H08 and annual values for Budyko) obtained by scaling the baseline climate using empirical annual scaling factors derived fromthe 42 GCMs. The plots in Fig. 6 present the median as well as the 10th and 90th percentile of annual runoff change across the SouthAsia. The percentiles of H08 and Budyko come from the 42 future climate projections informed by the 42 GCMs, while the percentilesof the GCM modelled runoffs come from only 17 models where runoff outputs are readily available. Nevertheless, the percentiles ofH08 and Budyko estimated from the same subset of 17 GCM show relatively minor differences compared to those estimated from allthe 42 GCMs.

The median projections from the H08 hydrological modelling indicate that runoff will be greater in the future across most ofSouth Asia except in the North-West (where projections show a 12% decline in mean annual runoff in 2046–2075 relative to1976–2005 baseline). Elsewhere, the median H08 modelling projections indicate increases in mean annual runoff of about 10% inIndus region (IND), Tibetan Plateau region (TP) and Arakan Coast region (AC), about 15% in Ganges-Brahmaputra region (GA),Deccan Plateau region (DP) and Ghats Coast region (GH), and a bit more than 20% in Narmad-Tapti region (NT) and Sri Lanka (SL).There is large uncertainty in the projections, with the 10th percentile and 90th percentile estimates of percentage change in futuremean annual runoff differing by more than 50% in most regions and more than 100% in some regions. Nevertheless, there is goodagreement in the direction of projected change in future precipitation (Fig. 4) and future runoff (Fig. 6) over the south, central andnorth-east of the South Asia region, with practically all the modelling results projecting an increase in future runoff in the regions. Incontrast, in the north-east, there is considerable disagreement between the GCMs in the direction of change in future precipitationand therefore future runoff.

It is comforting to see much similarity in the projected change in future runoff from the H08, Budyko and GCM modelling, in thespatial trends and the direction of change, and to a lesser extent also the modelled values. This gives some confidence in the broadscale direction and range of plausible future projections in the climate and runoff. However, despite the similarities and consistencyin the projections from the different modelling approaches, there are several key exceptions. First, the median projection from H08indicate runoff increase in the Indus region, but the median projection from Budyko show little change and the median GCMprojection indicate runoff decrease. Second, the future runoff projections from the GCMs, and to a lesser extent the Budyko re-lationship, are wetter in the south compared to the projections from H08 modelling (slightly wetter in the median projection andconsiderably wetter in the 90th percentile projection). The higher projected absolute runoff change by the Budyko relationship

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

101

compared to the H08 hydrological model (in all regions, drier in the north-west and wetter and elsewhere) occurs because of thehigher precipitation elasticity of runoff in the Budyko equation as discussed in Section 4.1. Third, the range and therefore uncertaintyin the projections from the GCMs is higher than in the H08 modelling. This is most likely because of the different land surfacehydrology conceptualisation and parameterisation in the GCMs compared to the single H08 parameterisation modelling the change inrunoff resulting from the change in the climate inputs.

The uncertainty in the future runoff projection mainly comes from the uncertainty in the future precipitation projection. This isillustrated in Fig. 7, which shows the range of percentage change in future runoff modelled by H08 from projected changes inprecipitation alone (the baseline temperature and potential evaporation series is used for these simulations) and from projectedincreases in temperature and potential evaporation alone (the baseline precipitation series is used for these simulations). The resultsin Fig. 7 indicate that the small range (or differences) in the projected increase in temperature and potential evaporation from the 42GCMs translates to a relatively small range in the projected decrease in the modelled future runoff. In contrast, the large range in theprojected change in precipitation (with GCMs disagreeing in the direction of change, particularly in the north-west) translates to alarge range in the modelled change in future runoff.

The modelled change in mean annual runoff is driven by the projected change in the climate variables (Section 3) and the runoffsensitivity to the climate variables (Section 4.1). However, the modelled change in future runoff is not a direct sum of the change due

Fig. 6. Modelled percentage change in mean annual runoff (median, 10th and 90th percentiles) for RCP8.5 for 2046–2075 relative to present/baseline (1976–2005) from the H08 hydrological model and Budyko relationship (informed by ‘annual scaling’ of climate projections from the 42GCMs) and directly from the GCMs (only 17 of the 42 GCMs where runoff outputs are readily available are used here).

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

102

to precipitation change alone and due to temperature and potential evaporation change alone. This is partly because of the non-linearity in the hydrological response to the different climate variables, and partly because of the different possible combinations ofchanges in the climate variables in the GCMs (for example, the median projected change in future runoff does not necessarily comefrom the median projected changes in all the different climate variables). In general, the H08 modelling results indicate that thefuture change in runoff by 2046–2075 across most of South Asia is driven mainly by the change in precipitation, and moderated(where a runoff increase is projected) or amplified (where a runoff decrease is projected) by the change in temperature and/orpotential evaporation.

4.3. Runoff projections from different empirical scaling treatments to obtain future climate change inputs

The plots in Fig. 8 show the H08 modelled changes in future runoff using future climate inputs derived using ‘annual scaling’,‘seasonal scaling’ and ‘daily scaling’ respectively. The results indicate that the future runoff projections from the seasonal scaling iswetter than those from the annual scaling, where the median projection from the seasonal scaling show 2–4% greater increase infuture mean annual runoff. This is because the seasonal scaling method reflects the proportionally higher projection of precipitationincrease in the summer monsoon (Jun-Jul-Aug), when higher runoff occurs, compared to winter (see Fig. 4). The seasonal scalingtreatment therefore enhances the summer monsoon runoff and subsequently the annual runoff compared to the annual scalingtreatment. Similarly, since daily scaling reflects the proportionally higher increases in the high daily precipitation amounts (thatgenerate high runoff) projected by the GCMs (particularly in the south), the future runoff projections from daily scaling is wetter thanthose from the seasonal scaling, except in the far north. The increases in future mean annual runoffmodelled with future precipitationinputs from the daily scaling, for the median projection, are 3–10% greater (more in the south) than in the seasonal scaling.

As also shown in Fig. 8, the 10th percentile, median and 90th percentile of the projection for each scaling method indicate theuncertainty in the future runoff projections resulting from the uncertainty or range in the future precipitation projected by GCMs. Theuncertainty of runoff projections owing to empirical scaling treatments are relatively smaller than that due to the range in the futureprecipitation projections. However, the difference of runoff projection from the three empirical scaling treatments can be significantand may need to be accounted for in detailed modelling studies to inform planning and adaptation, particularly when the projectedchanges at the seasonal and daily scale are supported by multiple lines of evidence from climate science and climate modelling, andwhere simulations of river flow characteristics other than just the long-term averages are required.

Fig. 7. Modelled percentage change in mean annual runoff (median, 10th and 90th percentiles) for RCP8.5 for 2046–2075 relative to present/baseline (1976–2005) by the H08 hydrological model for changes in precipitation alone (top row) and changes in temperature and potentialevaporation alone (bottom row).

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

103

Fig. 8. Modelled percentage change in mean annual runoff (median, 10th and 90th percentiles) for RCP8.5 for 2046–2075 relative to present/baseline (1976–2005) from the H08 hydrological model, with annual scaling, seasonal scaling and daily scaling treatments to obtain future climateinputs informed by climate projections from the 42 GCMs.

Table 1Modelled percentage change in reservoir storage reliability (median, 10th and 90th percentiles) for RCP8.5 for 2046–2075 relative to historical/baseline (1976–2005) with future climate inputs derived using annual scaling, seasonal scaling, daily scaling and bias corrected quantile mapping(informed by 42 CMIP5 GCMs).

Grid Brahmani-Baitarni grid (21.25 °N, 85.75 °E) Koshi grid(27.25 °N, 86.25 °E)

Scaling method Annualscaling

Seasonalscaling

Dailyscaling

Bias correction Annualscaling

Seasonalscaling

Dailyscaling

Bias correction

Percentage change in storage reliability 10th percentile −7.7 −6.4 −8.4 −16.2 −9.0 −6.4 −5.5 −27.2Median 0.0 0.6 −0.7 −4.1 0.8 1.5 1.8 3.490th percentile 2.3 2.3 2.3 3.6 3.7 4.6 4.6 7.9

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

104

4.4. Modelling climate change impact on future water supply security

Table 1 shows the storage reliability for the two grid cells where the Brahmani-Baitarani and Koshi basins are located (see Fig. 1).The storage analysis is discussed in Section 2.4. The runoff or inflow inputs for the storage analysis come from the H08 modellingwith the baseline period (1976–2005) and the projected runoff of the future derived from the annual scaling, seasonal scaling, dailyscaling and bias correction methods.

Table 1 shows that the storage reliability from the modelling with baseline climate data (1976–2005) is 95% for the Brahmani-Baitarani grid and 91% for the Koshi grid. The results from the three empirical scaling treatments to obtain the future climate inputsare very similar (little change in Brahmani-Baitarani, and very small increase in storage reliability in Koshi). Accounting for thedifferent changes in the climate inputs through the year (in seasonal scaling and daily scaling) therefore made little difference to theprojection of future water supply security in these two grids compared to applying a uniform/constant percentage change through theyear (annual scaling), although the results could be quite different elsewhere.

However, the results from the bias correction using quantile mapping are considerably different to those from the three empiricalscaling methods. This is because the quantile mapping takes into account changes in the sequencing in the future climate series aswell as the inter-annual and multi-year variability characteristics reflected by the GCMs. This highlights the importance of con-sidering key climate drivers other than long-term averages that significantly influence the hydrologic metrics important for decisionmaking and planning (the storage or water supply security in this particular example). Nevertheless, as discussed earlier (end ofSection 2.2), simulations using bias corrected GCM (or even dynamically downscaled regional climate model) outputs must beinterpreted cautiously because of the often large bias corrections required for GCM and RCM precipitation. This limitation anduncertainty is also reflected in the much larger range of results from the bias corrected quantile mapping compared to the empiricalscaling methods (Table 1).

4.5. Uncertainty in runoff projections from uncertainty in GCM precipitation projections

Fig. 9 shows the changes in future mean annual runoff modelled by the H08 hydrological model for the Brahmani-Baitarani andKoshi grid cells (2046–2075 relative to 1976–2005) informed by projected future climate from all ensemble runs for all 42 CMIP5

Fig. 9. Changes in future mean annual runoff modelled by the H08 hydrological model for the Brahmani-Baitarani and Koshi grid cells (2046–2075relative to 1976–2005) informed by projected future climate from all ensemble runs for all 42 CMIP5 GCMs for RCP8.5 (for annual, seasonal anddaily empirical scaling). The first column (G00 in red) shows results from the first run of each of the 42 GCMs. The other columns (G01 to G42 inblue) show results for ensemble runs for each of the 42 GCMs.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

105

GCMs for RCP8.5. The first column (G00) shows results from the first run of each of the 42 GCMs, which was used for the presentationand discussion in the preceding sections. The other columns (G01 to G42) show the modelled changes in future runoff using climateprojections from all the available ensemble runs for each of the 42 GCMs. The plots show again the large range in runoff projectionsresulting from the uncertainty in future projections of precipitation (G00), as well as show that the uncertainty in the projectionswithin many GCMs can be as large as the differences in projections between GCMs.

The differences in the modelled future change in runoff using future precipitation projections from all 42 GCMs versus usingprojections from only the ‘better’ GCMs are also investigated, with the results shown in Fig. 10. Here, the GCM performance isevaluated by comparing the ranked annual precipitation simulated by the GCM versus the Princeton 0.5° grid ‘observed’ precipitation(30 values over 1975–2006). The root mean square error is calculated with a lower value indicating better agreement between theGCM and observed annual precipitation distributions and therefore a better GCM ability in reproducing the baseline precipitation.

The modelling results from using the best five, best 10, best 15, best 20 and all 42 GCMs are presented in Fig. 10. Overall, thereare no clear trends or conclusions from the results, although the ‘better’ GCMs tend to project slightly wetter futures (particularlydaily scaling for Brahmani-Baitarani). The uncertainty or range in the projections are also similar for the different sub-selections ofGCMs (except the best five GCMs for Koshi). To put this into context, there have been many studies on GCM sub-selection with moststudies indicating little correlation between future precipitation projections and GCM ‘performance’ (Arnell, 2011; Brekke et al.,2008; Chiew et al., 2009b; Palazzi et al., 2015). Many studies favour the use of projections from all GCMs to represent the full rangeof uncertainty (Chiew et al., 2009b; Palazzi et al., 2015), although some studies sub-select GCMs to either remove the clearly poorlyperforming GCMs or to consider only GCMs with different or independent representation of the physical processes (Stainforth et al.,2005, 2007; Evans et al., 2011).

It is also difficult to adequately assess and determine which are the better performing GCMs (e.g. assessment against simulation ofdifferent precipitation and climate characteristics, assessment for the location of interest or regionally or globally, assessment againstthe large scale drivers/indices of local/regional precipitation and the relationship between them). Based on the limited considerationand assessment here, and for the broad scale projections presented here, it is probably best to use projections from all the available

Fig. 10. Projected changes in future mean annual runoff (2046–2075 relative to 1976–2005 for RCP8.5) from H08 modelling for Brahmani-Baitarani and Koshi grid cells with future climate inputs informed by simulations from the best five, best 10, best 15, best 20 and all 42 GCMs. Foreach box-plot, the dot shows the median, the boxes show the 25th and 75th percentiles and the dotted lines show the 10th and 90th percentiles.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

106

GCMs to provide an indication of the full range of uncertainty.

5. Summary and conclusion

The paper presents future climate and runoff projections, estimated using a consistent method, across South Asia. The futurerunoff is modelled using the Budyko relationship and H08 global hydrological model, with future climate inputs informed by pro-jections from the 42 CMIP5 GCMs. The modelling results indicate that future runoff will increase throughout most of the region(Indian sub-continent) except in the far north-east (southern part of the Tibetan Plateau) and the arid far north-west region. Themedian projection shows increases in mean annual runoff of 20–30% in the Indian sub-continent for 2046–2075 relative to1976–2005 for RCP8.5 greenhouse gas concentration pathway. However, there is considerable uncertainty in the projections, with a10th and 90th percentile range of more than 50%.

The uncertainty in the future runoff projections come mainly from the uncertainty in the future precipitation projections, bothwithin and between GCMs. Practically all the 42 GCMs project a wetter future throughout the Indian sub-continent (particularly inthe south), but there is little agreement between GCMs in the direction of precipitation change in the far north-east and far north-westof the South Asia region. The percentage change in precipitation is generally amplified as a 1.5–2.0% (for wet areas) to> 2% (for dryareas) change in runoff. The change in runoff by 2046–2075 is driven mainly by the change in precipitation, moderated (where runoffincrease is projected) or intensified (where runoff decrease is projected) by higher temperature and potential evaporation. Thedifferent hydrological modelling approaches (H08 hydrological model, Budyko relationship, runoff outputs from GCMs) show similarresults, providing confidence in the broad scale runoff projections presented here.

The differences in modelled future runoffs using future climate inputs derived from different empirical scaling methods can besignificant, but are relatively smaller than the uncertainty resulting from the large range in future precipitation projections. Thechoice of scaling method to reflect changes in future precipitation (and climate) characteristics that drive runoff can therefore beimportant in hydrological modelling studies to inform planning and adaptation, particularly where simulations of river flow char-acteristics other than just the long-term averages are needed. However, the methods used (including downscaling), particularlywhere significant modelling effort is required, must account for uncertainty from the range of GCMs and supported by scientificunderstanding of atmospheric-oceanic circulation processes in a warmer world.

There is little difference in the projections from using only the ‘better’ performing GCMs versus using all 42 GCMs in the modellinghere. Therefore, for the broad scale projections presented here, it is best to use projections from all the GCMs to provide an indicationof the full range of uncertainty.

Conflict of interest

None.

Acknowledgements

This work is funded by the Australian Government through the South-Asia Development Programme and the SustainableDevelopment Investment Portfolio (SDIP) run by the Department of Foreign Affairs and Trade (DFAT).

References

Abell, R., Thieme, M.L., Revenga, C., Bryer, M., Kottelat, M., Bogutskaya, N., Coad, B., Mandrak, N., Balderas, S.C., Bussing, W., Stiassny, M.L.J., Skelton, P., Allen,G.R., Unmack, P., Naseka, A., Ng, R., Sindorf, N., Robertson, J., Armijo, E., Higgins, J.V., Heibel, T.J., Wikramanayake, E., Olson, D., López, H.L., Reis, R.E.,Lundberg, J.G., Pérez, M.H.S., Petry, P., 2008. Freshwater ecoregions of the world: a new map of biogeographic units for freshwater biodiversity conservation.BioScience 58 (5), 403–414. http://dx.doi.org/10.1641/B580507.

Akhtar, M., Ahmad, N., Booij, M.J., 2008. The impact of climate change on the water resources of Hindukush–Karakorum–Himalaya region under different glaciercoverage scenarios. J. Hydrol. 355 (1–4), 148–163. http://dx.doi.org/10.1016/j.jhydrol.2008.03.015.

Arnell, N.W., 2011. Incorporating climate change into water resources planning in England and Wales. J. Am. Water Resour. Assoc. 47 (3), 541–549.Arora, V.K., 2002. The use of the aridity index to assess climate change effect on annual runoff. J. Hydrol. 265, 164–177.Ashfaq, M., Shi, Y., Tung, W.-w., Trapp, R.J., Gao, X., Pal, J.S., Diffenbaugh, N.S., 2009. Suppression of south Asian summer monsoon precipitation in the 21st century.

Geophys. Res. Lett. 36, L01704. http://dx.doi.org/10.1029/2008GL036500.Brekke, L.D., Dettinger, M.D., Maurer, E.P., Anderson, M., 2008. Significance of model credibility in estimating climate projection distributions for regional hydro-

climatological risk assessments. Clim. Change 89 (3–4), 371–394. http://dx.doi.org/10.1007/s10584-007-9388-3.Budyko, M.I., 1974. Climate and Life. Academic, Orlando, Fla.Chaturvedi, R.K., Joshi, J., Jayaraman, M., Bala, G., Ravindranath, N.H., 2012. Multi-model climate change projections for India under representative concentration

pathways. Curr. Sci. India 103 (7), 791–802.Chen, J., Brissette, F.P., Leconte, R., 2011. Uncertainty of downscaling method in quantifying the impact of climate change on hydrology. J. Hydrol. 401, 190–202.

http://dx.doi.org/10.1016/j.jhydrol.2011.02.020.Chiew, F.H.S., 2006. Estimation of rainfall elasticity of streamflow in Australia. Hydrol. Sci. J. 51 (4), 613–625.Chiew, F.H., McMahon, T.A., 1991. The applicability of Morton’s and Penman’s evapotranspirationestimates in rainfall‐runoff modeling. J. Am. Water Resour. Assoc.

27, 611–620. http://dx.doi.org/10.1111/j.1752-1688.1991.tb01462.x.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 Press, Wallingford, United Kingdom, pp. 256–262Publication Number 308.

Chiew, F.H.S., Teng, J., Vaze, J., Post, D.A., Perraud, J.-M., Kirono, D.G.C., Viney, N.R., 2009a. Estimating climate change impact on runoff across south-east Australia:method, results and implications of modelling method. Water Resour. Res. 45, W10414. http://dx.doi.org/10.1029/2008WR007338.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

107

Chiew, F.H.S., Teng, J., Vaze, J., Kirono, D.G.C., 2009b. Influence of global climate model selection on runoff impact assessment. J. Hydrol. 379, 172–180. http://dx.doi.org/10.1016/j.jhydrol.2009.10.004.

Chiew, F.H.S., Kirono, D.G.C., Kent, D.M., Frost, A.J., Charles, S.P., Timbal, B., Nguyen, K.C., Fu, G., 2010. Comparison of runoff modelled using rainfall from differentdownscaling methods for historical and future climates. J. Hydrol. 387, 10–23. http://dx.doi.org/10.1016/j.jhydrol.2010.03.025.

Dalton, J.: 1802, ‘Experimental Essays on the Constitution of Mixed Gases: On the Force of Steam or Vapor from Water or Other Liquids in Different Temperatures,Both in a Torricelli Vacuum and in Air; on Evaporation; and on Expansion of Gases by Heat’, Manchester Literary Philosophical Society Mem. Proceedings 5,536–602.

Devkota, L.P., Gyawali, D.R., 2015. Impacts of climate change on hydrological regime and water resources management of the Koshi River Basin, Nepal. J. Hydrol.:Regional Stud. 4 (Part B). http://dx.doi.org/10.1016/j.ejrh.2015.06.023.

Dooge, J.C.I., Bruen, M., Parmentier, B., 1999. A simple model for estimating the sensitivity of runoff to long-term changes in precipitation without a change invegetation. Adv. Water Resour. 23, 153–163. http://dx.doi.org/10.1016/S0309-1708(99)00019-6.

Ekström, M., Grose, M.R., Whetton, P.H., 2015. An appraisal of downscaling methods used in climate change research. WIREs Clim. Change 6, 301–319. http://dx.doi.org/10.1002/wcc.339.

Evans, J., Ekström, M., Ji, F., 2011. Evaluating the performance of a WRF physics ensemble over south-east Australia. Clim. Dynam. http://dx.doi.org/10.1007/s00382-011-1244-5.

Fowler, H.J., Blenkinsop, S., Tebaldi, C., 2007. Linking climate change modelling to impacts studies: recent advances in downscaling techniques for hydrologicalmodelling. Int. J. Climatol. 27, 1547–1578. http://dx.doi.org/10.1002/joc.1556.

Freychet, N., Hsu H-H, C., Chou, C., Wu, C.H., 2015. Asian summer monsoon in CMIP5 projections: a link between the change in extreme precipitation and monsoondynamics. J. Clim. 28 (4), 1477–1493. http://dx.doi.org/10.1175/JCLI-D-14-00449.1.

Frost, A.J., Charles, S.P., Timbal, B., Chiew, F.H.S., Mehro-tra, R., Nguyen, K.C., Chandler, R.E., McGregor, J.L., Fu, G., Kirono, D.G.C., et al., 2011. A comparison ofmulti-site daily rainfall downscaling techniques under Australian conditions. J. Hydrol 408, 1–18.

Ghimire, S., Choudhary, A., Dimri, A.P., 2015. Assessment of the performance of CORDEX-South Asia experiments for monsoonal precipitation over the himalayanregion during present climate: part I. Climate dynamics. p1-24. http://dx.doi.org/10.1007/s00382-015-2747-2.

Gupta, S., Deshpande, D., 2004. Water for India in 2050: first order assessment of available options. Curr. Sci. India 86, 1216–1224.Haddeland, I., Clark, D.B., Franssen, W., et al., 2011. Multimodel estimate of the global terrestrial water balance: setup and first results. J. Hydrometeorol. 12,

869–884.Hanasaki, N., Kanae, S., Oki, T., Masuda, K., Motoya, K., Shirakawa, N., Shen, Y., Tanaka, K., 2008. An integrated model for the assessment of global water

resources—Part 1: model description and input meteorological forcing. Hydrol. Earth Syst. Sci. 12, 1007–1025. http://dx.doi.org/10.5194/hess-12-1007-2008.Hay, L.E., Clark, M.P., 2003. Use of statistically and dynamically downscaled atmospheric model output for hydrologic simulations in three mountainous basins in the

western United States. J. Hydrol. 282, 56–75.Held, I.M., Soden, B.J., 2006. Robust responses of the hydrological cycle to global warming. J. Clim. 19, 5686–5699.Hijioka, Y., Lin, E., Pereira, J.J., Corlett, R.T., Cui, X., Insarov, G.E., Lasco, R.D., Lindgren, E., Surjan, A., et al., 2014. Asia. In: Barros, V.R. (Ed.), Climate Change 2014:

Impacts, Adaptation, and Vulnerability. Part B: Regional Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panelof Climate Change. Cambridge University Press, pp. 1327–1370.

Immerzeel, W.W., Bierkens, M.F.P., 2012. Asia’s water balance. Nat. Geosci. 5 (12), 841–842.Immerzeel, W.W., van Beek, L.P.H., Bierkens, M.F.P., 2010. Climate change will affect the Asian water towers. Science 328 (5984), 1382–1385. http://dx.doi.org/10.

1126/science.1183188.IPCC, 2013. Climate change 2013: the physical science basis. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V.,

Midgley, P.M. (Eds.), Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge UniversityPress, Cambridge, United Kingdom and New York, NY, USA, pp. 1535.

Jiménez Cisneros, B.E., Oki, T., Arnell, N.W., Benito, G., Cogley, J.G., Döll, P., Jiang, T., Mwakalila, S.S., 2014. Freshwater resources. In: Field, C.B., Barros, V.R.,Dokken, D.J., Mach, K.J., Mastrandrea, M.D., Bilir, T.E., Chatterjee, M., Ebi, K.L., Estrada, Y.O., Genova, R.C., Girma, B., Kissel, E.S., Levy, A.N., MacCracken, S.,Mastrandrea, P.R., White, L.L. (Eds.), Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of WorkingGroup II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and NewYork, NY, USA, pp. 229–269.

Jourdain, N., Gupta, A., Taschetto, A., Ummenhofer, C., Moise, A., Ashok, K., 2013. The indo-Australian monsoon and its relationship to ENSO and IOD in reanalysisdata and the CMIP3/CMIP5 simulations. Clim. Dyn. 41 (11–12), 3073–3102. http://dx.doi.org/10.1007/s00382-013-1676-1.

Kumar, P., Wiltshire, A., Mathison, C., Asharaf, S., Ahrens, B., Lucas-Picher, P., Christensen, J.H., Gobiet, A., Saeed, F., Hagemann, S., Jacob, D., 2013. Downscaledclimate change projections with uncertainty assessment over India using a high resolution multi-model approach. Sci. Total Environ. (Supplement S18–S30),468–469. http://dx.doi.org/10.1016/j.scitotenv.2013.01.051.

Lopez, A., Fung, F., New, M., Watts, G., Weston, A., Wilby, R.L., 2009. From climate model ensembles to climate change impacts and adaptation: a case study of waterresource management in the southwest of England. Water Resour. Res. 45, W08419. http://dx.doi.org/10.1029/2008WR007499.

Lutz, A.F., Immerzeel, W.W., Shrestha, A.B., Bierkens, M.F.P., 2014. Consistent increase in High Asia’s runoff due to increasing glacier melt and precipitation. NatureClim. Change 4 (7), 587–592. http://dx.doi.org/10.1038/nclimate2237.

Manabe, S., 1969. Climate and the ocean circulation: i. The atmospheric circulation and the hydrology of the earth’s surface. Mon. Wea. Rev. 97, 739–774. http://dx.doi.org/10.1175/1520-0493(1969)097,0739:CATOC.2.3.CO;2.

Maraun, D., 2013. Bias correction, quantile mapping, and downscaling: revisiting the inflation issue. J. Clim. 26 (2013), 2137–2143. http://dx.doi.org/10.1175/JCLI-D-12-00821.1.

McMahon, T.A., Adeloye, A.J., 2005. Water Resources Yield. Water Resources Publications, LLC, pp. 220.McMahon, T.A., Pegram, G.G.S., Vogel, R.M., Peel, M.C., 2007. Revisiting reservoir storage-yield relationships using a global streamflow database. Adv. Water Resour.

30, 1858–1872. http://dx.doi.org/10.1016/j.advwatres.2007.02.003.Mishra, V., Kumar, D., Ganguly, A.R., Sanjay, J., Mujumdar, M., Krishnan, R., Shah, R.D., 2014. Reliability of regional and global models to simulate precipitation

extremes over India. J. Geophy. Res. (Atmos) 119 (15), 9301–9323.Morton, F.I., 1983. Operational estimates of areal evapotranspiration and their significance to the science and practice of hydrology. J. Hydrol. 66, 1–76.Moss, R.H., Edmonds, J.A., Hibbard, K.A., Manning, M.P., Rose, S.K., van Vuuren, D.P., Carter, T.R., Emori, S., Kainuma, M., Kram, T., Meehl, G.A., Mitchell, J.F.B.,

Nakicenovic, N., Riahi, K., Smith, S.J., Stouffer, R.J., Thomson, A.M., Weyant, J.P., Wilbanks, T.J., 2010. The next generation of scenarios for climate changeresearch and assessment. Nature 463 (7282), 747–756.

Mountain Research Initiative EDW Working Group, 2015. Elevation-dependent warming in mountain regions of the world. Nat. Clim. Change 5, 424–430. http://dx.doi.org/10.1038/nclimate2563.

Nepal, S., Shrestha, A.B., 2015. Impact of climate change on the hydrological regime of the indus, ganges and brahmaputra river basins: a review of the literature. Int.J. Water Resour. Dev.t 31 (2), 201–218. http://dx.doi.org/10.1080/07900627.2015.1030494.

Ogata, T., Ueda, H., Inoue, T., Hayasaki, M., Yoshida, A., Watanabe, S., Kira, M., Ooshiro, M., Kumai, A., 2014. Projected future changes in the Asian monsoon: acomparison of CMIP3 and CMIP5 model results. J. Meteorol. Soc. Jpn. Ser. II 92 (3), 207–225. http://dx.doi.org/10.2151/jmsj.2014-302.

Oudin, L., Andréassian, V., Lerat, J., Michel, C., 2008. Has land cover a significant impact on mean annual streamflow? An international assessment using 1508catchments. J. Hydrol. 357, 303–316. http://dx.doi.org/10.1016/j.jhydrol.2008.05.021.

Palazzi, E., von Hardenberg, J., Terzago, S., Provenzale, A., 2015. Precipitation in the karakoram-himalaya: a CMIP5 view. Clim. Dyn. 45 (1-2), 21–45. http://dx.doi.org/10.1007/s00382-014-2341-z.

Penman, H.L., 1948. Natural evaporation from Open Water, Bare soil and grass. Proc. R. Soc. Lond. 193, 120–145.Piani, C., Haerter, J.O., Coppola, E., 2010. Statistical bias correction for daily precipitation in regional climate models over Europe. Theor. Appl. Climatol. 99 (1–2),

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

108

187–192.Saha, A., Ghosh, S., Sahana, A.S., Rao, E.P., 2014. Failure of CMIP5 climate models in simulating post-1950 decreasing trend of Indian monsoon. Geophys. Res. Lett.

41, 7323–7330.Sankarasubramanian, A., Vogel, R.M., 2001. Climate elasticity of streamflow in the United States. Water Resour. Res. 37, 1771–1781.Schaake, J.C., 1990. From climate to flow, chap. 8. In: Waggoner, P.E. (Ed.), Climate Change and U.S. Water Resources. John Wiley, New York, pp. 177–206.Sheffield, J., Goteti, G., Wood, E.F., 2006. Development of a 50-year high-resolution global dataset of meteorological forcings for land surface modeling. J. Clim. 19,

3088–3111. http://dx.doi.org/10.1175/JCLI3790.1.Sivakumar, M.V.K., Stefanski, R., 2011. Climate change in South Asia. In: Lal, R., Sivakumar, M.V.K., Faiz, M.A., Mustafizur Rahman, A.H.M., Islam, K.R. (Eds.),

Climate Change and Food Security in South Asia. Springer Science and Business Media B.V., pp. 13–30. http://dx.doi.org/10.1007/978-90-481-9516-9_2.Sivapalan, M., Blöschl, G., Zhang, L., Vertessy, R., 2003. Downward approach to hydrological prediction. Hydrol. Processes 17, 2101–2111. http://dx.doi.org/10.

1002/hyp.1425.Stainforth, D.A., Aina, T., Christensen, C., Collins, M., Faull, N., Frame, D.J., Kettleborough, J.A., Knight, S., Martin, A., Murphy, J.M., Piani, C., Sexton, D., Smith, L.A.,

Spicer, R.A., Thorpe, A.J., Allen, M.R., 2005. Uncertainty in predictions of the climate response to rising levels of greenhouse gases. Nature 433, 403–406.Stainforth, D.A., Allen, M.R., Tredger, E.R., Smith, L.A., 2007. Confidence, uncertainty and decision-support relevance in climate predictions. Philosophical Trans. R.

Soc. A 365 (1857), 2145–2161.Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2012. An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc. 93, 485–498. http://dx.doi.org/10.1175/

BAMS-D-11-00094.1.Teng, J., Vaze, J., Chiew, F.H.S., Wang, B., Perraud, J.M., 2012a. Estimating the relative uncertainties sourced from GCMs and hydrological models in modeling

climate change impact on runoff. J. Hydromet. 13, 122–139. http://dx.doi.org/10.1175/JHM-D-11-058.1.Teng, J., Chiew, F.H.S., Vaze, J., Marvanek, S., Kirono, D.G.C., 2012b. Estimation of climate change impact on mean annual runoff across continental Australia using

budyko and Fu equations and hydrological models. J. Hydrometeor 13, 1094–1106. http://dx.doi.org/10.1175/JHM-D-11-097.1.United Nations, Department of Economic and Social Affairs, Population Division, 2015. World population prospects: the 2015 revision, key findings and advance

tables. Working Paper No. ESA/P/WP.241.Vetter, T., Reinhardt, J., Flörke, M., et al., 2016. Evaluation of sources of uncertainty in projected hydrological changes under climate change in 12 large-scale river

basins. Clim. Change p1–15. http://dx.doi.org/10.1007/s10584-016-1794-y.Warszawski, L., Frieler, K., Veronika, H., Franziska, P., Olivia, S., Jacob, S., 2014. The inter-sectoral impact model intercomparison project (ISI-MIP): project fra-

mework. P Natl. Acad. Sci. 111, 3228–3232. http://dx.doi.org/10.1073/pnas.1312330110.World Bank, 2016. High and Dry: Climate Change, Water, and the Economy. World Bank, Washington, DC License: Creative Commons Attribution CC BY 3.0 IGO.Zhang, L., Hickel, K., Dawes, W.R., Chiew, F.H.S., Western, A.W., Briggs, P.R., 2004. A rational function approach for estimating mean annual evapotranspiration.

Water Resour. Res. 40, W02502. http://dx.doi.org/10.1029/2003wr002710.Zhang, Y., Zheng, H., Chiew, F.H.S., Jorge, P.A., Zhou, X.Y., 2016. Evaluating regional and global hydrological models against streamflow and evapotranspiration

measurements. J. J. Hydrometeorol. 17, 995–1010. http://dx.doi.org/10.1175/JHM-D-15-0107.1.Zheng, H., Zhang, L., Zhu, R., Liu, C., Sato, Y., Fukushima, Y., 2009. Responses of streamflow to climate and land surface change in the headwaters of the yellow River

Basin. Water Resour. Res. 45http://dx.doi.org/10.1029/2007WR006665. W00A19.Zheng, H., Chiew, F.H.S., Charles, S.P., 2015. Database of CMIP5 climate change projections across south Asia. CSIRO Sustainable Development Investment Portfolio

Project. CSIRO Land and Water, Australia.

H. Zheng et al. Journal of Hydrology: Regional Studies 18 (2018) 92–109

109