Climate model performance and change projection for freshwater fluxes: Comparison for irrigated areas in Central and South Asia

Journal of Hydrology: Regional Studies 5 (2016) 48–65

Contents lists available at ScienceDirect

Journal of Hydrology: Regional Studies journal homepage: www.elsevier.com/locate/ejrh

Climate model performance and change projection for freshwater fluxes: Comparison for irrigated areas in Central and South Asia Shilpa M. Asokan a , Peter Rogberg a , Arvid Bring a,b , Jerker Jarsjö a , Georgia Destouni a,∗ a b

Department of Physical Geography, and Bolin Centre for Climate Research, Stockholm University, Sweden Water Systems Analysis Group, Institute for the Study of Earth, Oceans and Space, University of New Hampshire, USA

a r t i c l e

i n f o

Article history: Received 28 April 2015 Received in revised form 15 November 2015 Accepted 20 November 2015 Keywords: CMIP5 global climate models Hydro-climate Freshwater change Central Asia South Asia Monsoon driven seasonality

a b s t r a c t Study region: The large semi-arid Aral Region in Central Asia and the smaller tropical Mahanadi River Basin (MRB) in India. Study focus: Few studies have so far evaluated the performance of the latest generation of global climate models on hydrological basin scales. We here investigate the performance and projections of the global climate models in the Coupled Model Intercomparison Project, Phase 5 (CMIP5) for freshwater fluxes and their changes in two regional hydrological basins, which are both irrigated but of different scale and with different climate. New hydrological insights for the region: For precipitation in both regions, model accuracy relative to observations has remained the same or decreased in successive climate model generations until and including CMIP5. No single climate model out-performs other models across all key freshwater variables in any of the investigated basins. Scale effects are not evident from global model application directly to freshwater assessment for the two basins of widely different size. Overall, model results are less accurate and more uncertain for freshwater fluxes than for temperature, and particularly so for model-implied water storage changes. Also, the monsoon-driven runoff seasonality in MRB is not accurately reproduced. Model projections agree on evapotranspiration increase in both regions until the climatic period 2070–2099. This increase is fed by precipitation increase in MRB and by runoff water (thereby decreasing runoff) in the Aral Region. © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction To study climate-driven change in local-regional freshwater systems, downscaled climate model data are often used, from either statistical or dynamical downscaling methods, and subsequently processed through hydrological modeling. This approach provides a higher-resolved local-regional view of climate and hydrology than direct hydro-climatic output of global climate models. However, both the downscaled climate data and the hydrological model that uses them still depend fundamentally on the climate forcing and boundary conditions provided as output from global climate models. The driving

∗ Corresponding author at: Department of Physical Geography, Stockholm University, SE-106 91 Stockholm, Sweden. E-mail address: [email protected] (G. Destouni). http://dx.doi.org/10.1016/j.ejrh.2015.11.017 2214-5818/© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

49

global climate model may thus have greater impact on hydrological projection uncertainty than the hydrological modeling (Raje and Krishnan, 2012). Furthermore, climate models are now used for multi-decadal predictions of climate change, in addition to the previous main focus on projecting differences between hypothetical future scenarios (Trenberth, 2010). Direct climate model output is also used outside the climate science community, e.g., for understanding of and adaptation to climate-driven freshwater changes (Arnell, 1999; Gleick and Chalecki, 1999; Lettenmaier et al., 1999; Kundzewicz and Stakhiv, 2010; Törnqvist et al., 2014). Global climate models thus influence the downscaled climate forcing used in hydrological modeling, multi-decadal predictions of hydro-climate, and direct freshwater assessments based on global climate model output. It is therefore important to study and inform the hydrological science community and other users about how the global climate models represent observations and agree among them with regard to freshwater conditions and changes. Simulations and projections of Earth’s past and future climate, including hydro-climate, are provided by the Coupled Model Intercomparison Project (CMIP), which is coordinated by the World Climate Research Programme and supports the assessment reports of the Intergovernmental Panel on Climate Change (IPCC). The global climate models in Phase 5 of the project (CMIP5; Taylor et al., 2012) are developed on their predecessors, the CMIP3 models (Meehl et al., 2007), with more complete representations of external forcing and with increased resolution. One of the earliest evaluations of CMIP5 datasets by Knutti and Sedláˇcek (2013) showed similar model spread in CMIP3 and CMIP5 projections on global scale. Furthermore, a recent study by Mueller and Seneviratne (2014) indicated shortcomings in the CMIP5 climate model simulations of evapotranspiration and precipitation on regional scale. However, few studies have so far evaluated CMIP5 model performance on hydrological basin scales. For planning and sustainable management of freshwater resources under both global and local-regional changes, hydrological drainage basins are recommended or even mandated as relevant spatial units (Pahl-Wostl, 2007; UNECE, 2009). Therefore, also climate model results for freshwater conditions and changes, as required in management, planning and adaptation for freshwater security and sustainability, need to be evaluated on hydrological basin scales. Moreover, hydrological basins offer a substantial modeling advantage of water and constituent balance closure by their topographic integration of both water fluxes (Karlsson et al., 2012; Destouni et al., 2013; Van der Velde et al., 2013; Törnqvist et al., 2014; Jaramillo and Destouni, 2014) and waterborne mass fluxes (Jarsjö and Destouni, 2004; Darracq et al., 2005; Shibuo et al., 2006; Destouni and Darracq, 2009; Törnqvist et al., 2011; Visser et al., 2012). Törnqvist et al. (2014) is one recent study that has applied a hydrological basin perspective for observation-based evaluation of CMIP5 performance with regard to freshwater fluxes, their resulting net water balance, and their changes in the Lake Baikal drainage basin. Other CMIP5 performance studies with focus on freshwater changes have not considered the aspect of basin-scale water balance (Deng et al., 2013; Siam et al., 2013), have not accounted for the historic water-use alterations within the basins (Ibrahim et al., 2015; Wambura et al., 2015; Yan et al., 2015), or have mostly discussed anthropogenic influences on the global scale (Alkama et al., 2013). Furthermore, freshwater changes do not only depend on atmospheric climate change but also on direct change drivers in the landscape (Foley et al., 2005; Shibuo et al., 2007; Weiskel et al., 2007; Wisser et al., 2010; Asokan et al., 2010; Destouni et al., 2013). A recent worldwide study shows that landscape drivers are needed to explain observed historic freshwater changes in 74% of investigated hydrological basins over all continents (Jaramillo and Destouni, 2014); only in 26% of the studied basins worldwide can the observed atmospheric climate changes alone explain the observed freshwater changes. Adequate assessment of freshwater changes hence require account of both atmospheric climate change and changes in the landscape (Milly et al., 2002; Seneviratne et al., 2006; Piao et al., 2007; Destouni et al., 2010; Asokan and Destouni, 2014), which poses an even greater quantification challenge than just atmospheric-driven water changes (Milly et al., 2005; Groves et al., 2008; Bengtsson, 2010; Jarsjö et al., 2012). Main human pressures that alter freshwater fluxes across the world include expansion of irrigated and non-irrigated agriculture, deforestation, and other human-driven land-use and water-use changes in the landscape (Gordon et al., 2005; Destouni et al., 2013; Törnqvist et al., 2015). In densely populated areas, such as many regions of Asia, water diversions and extractions for human uses amount to a considerable fraction of the original freshwater flows in hydrological basins (Shibuo et al., 2007; Destouni et al., 2010; Jarsjö et al., 2012; Törnqvist and Jarsjö, 2012; Asokan and Destouni, 2014; Karthe et al., 2015); as such, these diversions and extractions can greatly influence water fluxes and water availability in the landscape, in addition to such influences of atmospheric climate change. In the present study, we investigate the performance of CMIP5 climate models in two Central and South Asian hydrological basins with previously well-investigated and compared freshwater changes, driven by direct human changes in the landscape (primarily irrigation developments in both regions) in addition to atmospheric climate change (Destouni et al., 2013; Asokan and Destouni, 2014). The two basins are: the Aral Sea drainage basin in Central Asia (1,888,810 km2 , including also the terminal Aral Sea itself and referred to as the Aral Region in the following) (Shibuo et al., 2007; Destouni et al., 2010), and the Mahanadi River Basin in India (135,084 km2 , referred to as MRB) (Asokan et al., 2010). In this study, we investigate and compare the CMIP5 model ability to reproduce observed historic conditions and project future changes in freshwater fluxes and their resulting net water balance in these two Asian basins. We further compare the CMIP5 model performance with that of the predecessor CMIP3 model generation (Solomon et al., 2007); the latter has also previously been analyzed for the Aral Region (Jarsjö et al., 2012). For future projections, we evaluate the consistency among individual CMIP5 model implications for future water fluxes and their changes. The two investigated Asian basins are similar with regard to the primary direct human drivers of historic freshwater changes over the last century; irrigation developments in the basins have over this time period driven evapotranspiration

50

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

Fig. 1. Regional location map of Mahanadi River Basin (MRB) and Aral Region (Source: Google Map, edited to illustrate basin boundaries, the major rivers – Mahanadi River in MRB and Amu Darya and Syr Darya Rivers in Aral Region – and irrigated areas within basins; the latter are shown in green color). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

increase and associated runoff decrease. Main differences between the two basins are their spatial scale and their climate. These differences facilitate distinction of the possible effects of basin size (scale effects) and prevailing climate conditions on the CMIP5 performance with regard to freshwater changes. Not least the scale aspect is important, as climate model output is considered to be more reliable for larger areas (Flato et al., 2013); a main question is then if such scale effects are evident in the CMIP5 model performance for freshwater, so that direct application of the output of global climate models is more accurate for the large Aral Region than for the much smaller MRB. Furthermore, the prevailing climate conditions imply quite different hydrological seasonality between the two basins; in particular the monsoon-driven high hydrological seasonality in MRB may be essential for freshwater resource management and the CMIP5 performance in reproducing this high seasonality is specifically investigated for the monsoon-dependent MRB. 2. Materials and methods 2.1. Study areas Fig. 1 shows the locations and extents of the Aral Region in Central Asia and the MRB in India. The Aral Region extends between the geographical co-ordinates of 54◦ 30 –78◦ 30 E longitude and 34◦ 30 –52◦ 30 N latitude. MRB is located between 80◦ 30 –86◦ 50 E longitude and 19◦ 20 –23◦ 35 N latitude. The total basin area of MRB is about seven percent of the total area of the Aral Region. Besides scale, the other main difference between these two regions lies in their hydro-climate and its seasonality, from the continental semi-arid conditions of the Aral Region, to the tropical monsoon-driven conditions of the MRB. The two basin systems differ also in that the Aral Region is endorheic, discharging into a terminal inland water body (the Aral Sea), while the MRB is exorheic, discharging into the ocean (the Bay of Bengal). However, both the Aral Region and the MRB have undergone similar major expansion of irrigated agriculture during the second half of the twentieth century. Due to their large hydro-climatic differences, in particular regarding average annual runoff (R), the irrigation impacts on average water availability have in comparative terms been relatively small in the tropical

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

51

MRB but severe in the semi-arid Aral Region, where more than 80% of the pre-irrigation runoff R from the drainage basin to the Aral Sea was diverted for irrigation (Destouni et al., 2010; Asokan and Destouni, 2014). The Aral Sea desiccation is a direct consequence of this large water diversion relative to the original R and constitutes one of the worst environmental disasters of our time (Micklin, 2007; Törnqvist et al., 2011; Gaybullaev et al., 2012; Bengtsson, 2014). During the twentieth century, average annual precipitation (P) was 260 mm/year in the Aral Region, which is five times lower than that of about 1300 mm/year in the MRB, while average annual temperature (T) was about 25 ◦ C in the MRB and 8 ◦ C in the Aral Region (Asokan and Destouni, 2014). The Aral Region is primarily seasonal with respect to T, where average seasonal T was 15 ◦ C in the growing season (March–October) and 0.5 ◦ C in the non-growing season (November–February) during the twentieth century. Average seasonal P varies only slightly in the Aral Region and was on average 120 mm over the growing season and 112 mm over the non-growing season during the twentieth century. In contrast, seasonality is much more pronounced for P than for T in the MRB: during the twentieth century, average seasonal P was here 1207 mm over the wet season (June–November) and 97 mm/season over the dry season (December–May), while average seasonal T was quite steady over the year, at 26 ◦ C and 25 ◦ C for the wet and dry seasons, respectively. The areas currently under irrigation within each basin are shown in green color in Fig. 1. An average annual irrigation water amount of 11 km3 /year is re-distributed from Mahanadi River to the irrigated areas in the MRB (Asokan et al., 2010), while a similar amount of about 14 km3 /year is re-distributed from the Amu Darya and Syr Darya rivers to the irrigated areas in the Aral Region (Shibuo et al., 2007, Destouni et al., 2010; Asokan and Destouni, 2014). As a direct result of this water diversion to irrigated areas, evapotranspiration (ET) is estimated to have increased by 38 mm/year in the MRB and by 15 mm/year in the Aral Region (Asokan and Destouni, 2014). In absolute ET terms, the loss of water by the irrigation-driven increase of ET from the basin to the atmosphere is thus smaller in the Aral Region than in the MRB. However, the average annual runoff R from the drainage basin to the Aral Sea is drastically reduced by this loss, to being only 6 mm/year by the end of the twentieth century (from 38 mm/year in the first half of the century). In contrast, the average annual R from the MRB to the Bay of Bengal is still relatively high at 515 mm/year by the end of the twentieth century (Asokan and Destouni, 2014). This comparison of twentieth-century changes between the two regions clarifies that widely different changes in freshwater resource availability and security may be related to more or less similar water-related changes in ET and in associated latent heat flux in the atmospheric climate system (Asokan and Destouni, 2014). Consequently, model focus on only atmospheric climate change may not sufficiently distinguish changes that are of primary importance for freshwater resources in the landscape. Furthermore, improvements of atmospheric climate modeling in successive model generations may or may not constitute improvements from a freshwater change perspective. These different possibilities are a main motivation for the present study of CMIP5 model performance regarding freshwater conditions and changes in the two regional basins. 2.2. Observation and model data sources 2.2.1. Observation-based hydro-climatic data Observation data for T and P over the twentieth century is obtained from the CRU TS 2.1 database (Mitchell and Jones, 2005). The monthly mean values of T and P in that database have been constructed by interpolating station data as a function of latitude, longitude and elevation and were assessed by cross-validation and comparison with other climatologies (New et al., 1999). The T and P data are here assessed over the whole drainage basin area for the Aral Region and the MRB, as delineated by Destouni et al. (2010) and Asokan et al. (2010), respectively, and summarized in Appendix A, Table A1. River discharge and associated runoff data for the Aral Region is derived for most of the twentieth century from Mamatov (2003), Jarsjö and Destouni (2004), Shibuo et al. (2007), Destouni et al. (2010) and Jarsjö et al. (2012). For the MRB, river discharge data is only available (through the Central Water Commission, New Delhi) for a relatively short time period, from 1990 to 2002, as compiled and reported by Asokan and Dutta (2008). Discharge station information for both sites is specified in Appendix A, Table A2. From the available historic observation data on P and R, ET is estimated by honoring and accounting for the water flux constraints implied by basin-scale closure of long-term average annual water balance as ET ≈ P–R. This estimate assumes that average annual water storage change S = P–R–ET is close to zero over long (climatic) averaging periods, as also assumed and tested in several previous studies (Destouni et al., 2013; Van der Velde et al., 2013; Törnqvist et al., 2014; Jaramillo and Destouni, 2014). 2.2.2. CMIP3 and CMIP5 model data For the present two regional study basins, we investigate the hydrological flux results provided directly as output from 22 global climate models contributing to CMIP5 (Taylor et al., 2012). Such direct use and interpretation of climate model output has also been carried out and reported for other hydrological basins across the world (Bring and Destouni, 2011; Jarsjö et al., 2012; Törnqvist et al., 2014; Bring et al., 2015). For comparison of different climate model generations, we also compare the hydrological output of 14 climate models contributing to CMIP3 (Meehl et al., 2007) with that of the CMIP5 models and corresponding regional observation data. Previous results have shown that resolution biases of global climate models should be small for hydrological basins of medium to large size (106 –107 km2 ) and monthly to annual time-scales, thus justifying a direct use of projection results in hydrological interpretation and modeling under such conditions (Milly et al., 2002; Wood et al., 2004). Direct investigation

52

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

and use of climate model output is also relevant considering recent findings, which indicate that advantageous effects of downscaling approaches may be limited for such hydrological conditions (Bring et al., 2015). Downscaling may be a necessary bias-reducing step for relatively small catchments and/or extreme events (e.g., Kay et al., 2009), reducing the spread in model output by altering the signals from global climate models (Weiland et al., 2010). However, such dampening of output signal spreading may also increase ambiguity in comparative model-observation studies, such as the present one. Importantly, both downscaling and direct use of climate model output for hydrological modeling and interpretation depend on the hydrological realism of the climate model output, which is what the present study aims at investigating. The investigated climate models are specified in Appendix A, Table A3. The CMIP5 models are chosen based on their provision of explicit output for all hydro-climatic variables T, P, R and ET (including both evaporation and sublimation) that are analyzed and compared with observation data in this study. The available fewer CMIP3 models with relevant simulations are used for direct comparison of model output regarding the atmospheric drivers T and P in the historic time period 1961–1990. The CMIP5 model performance is further analyzed for both this historic time period and for projections of possible future water resource changes for all hydro-climatic variables T, P, R and ET. In analogy with previous regional assessments we consider here the SRES A2 scenario of the CMIP3 models, which reflects a fragmented and regionally oriented economic development in the Asian region (IPCC, 2007). As the climate modeling progressed from CMIP3 to CMIP5, the scenarios have been re-designed in the latter to four representative concentration pathways (RCPs) RCP2.6, RCP4.5, RCP6.0 and RCP8.5. Each of these scenarios corresponds to a specific radiative forcing value in the year 2100 relative to pre-industrial values (+2.6, +4.5, +6.0 and +8.5 W/m2 respectively) (Moss et al., 2010). In this study we consider scenarios RCP2.6 and RCP8.5 so as to cover a maximum range of greenhouse gas concentration trajectories (PCDMI, 2013). We process the monthly average values of the climate model outputs of T, P, R and ET to obtain annual area-averaged values for the Aral Region and the MRB. In order to compare the different model results with observations and among models, all model data are re-gridded onto a uniform 0.5 × 0.5◦ grid. Our assessment of climate model performance (both CMIP3 and CMIP5) is carried out for a historic time period of 1961–1990, for which observed data are available for both T and P in both regions. Furthermore, the ability of climate models to represent the net annual water balance in the regions is quantified by considering and comparing their results for long-term average annual change in water storage, S = P–R–ET, with P, R and ET being the outputs obtained from the CMIP5 global climate models. The model outputs of P, R and ET are also directly compared with the available P and R observation data and the observation-based water-balance constrained estimates of ET ≈ P–R . Furthermore, for the MRB, which has particularly high, monsoon-driven hydrological seasonality in water fluxes, an analysis of monthly P, R and ET is also carried out, in order to evaluate the ability of climate models to accurately represent this high seasonality on basin-scale; this ability is of primary importance for water resource planning, adaptation and management in tropical regions.

3. Results and discussion 3.1. CMIP5 and CMIP3 comparison Fig. 2 shows the comparison between CMIP3 and CMIP5 model results for T and P for the historic (1961–1990) period and for projections for the near future (2010–2039) period, along with observations in the historic (1961–1990) period. Overall, the CMIP5 results of absolute T and P deviate more from observations than the CMIP3 results for both regions. With regard to P, the CMIP3 models over-estimate P in the Aral Region whereas they under-estimate it in MRB. The CMIP5 model results follow the same pattern, although with greater deviation from observations (Fig. 2). Previously reported results for the Aral Region (Jarsjö et al., 2012) on the performance of CMIP3 compared with the earlier CMIP1 generation of general circulation models show that, for temperature T, the CMIP3 results deviated less from observations than the CMIP1 results. For precipitation P, however, the CMIP3 results deviated more from observations than the CMIP1 results. This indicates progressively less accurate P results (increased overestimation of P) from CMIP1 through CMIP3–CMIP5 in comparison with observations. However, for the Aral Region, bias correction of observed P decreases the difference between observations and model results if only undercatch correction is considered (Adam and Lettenmaier, 2003) (Appendix B, Fig. B1a). In contrast to the uncorrected P data, the undercatch-corrected P data thus indicate improved P results (decreased overestimation of P) in CMIP5 for the Aral Region. However, if also orographic P correction is accounted for (Adam et al., 2006), the model bias switches from overestimation to underestimation (Appendix B, Fig. B1a). On average between the directly observed and the bias-corrected P data for the Aral Region, the accuracy level of climate model results for precipitation P may thus have remained essentially the same between model generations. For the MRB, however, undercatch correction further increases the model underestimation of observed P while orographic correction has insignificant effect (Appendix B, Fig. B1b). Regarding the spread of individual model results around the model ensemble mean, the insert panels of Fig. 2 show that the coefficient of variation (CV) of CMIP5 outputs around the ensemble mean has mostly decreased in comparison with the CMIP3 results, except for P in the Aral Region. Hence, the ensemble model precision of CMIP5 has mostly increased whereas the accuracy of CMIP5 models has mostly decreased in comparison with CMIP3. With regard to the tropical MRB, our results agree with earlier findings for Indian summer monsoon rainfall by Shashikanth et al. (2014), but here with a more complete

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

53

Fig. 2. Observations and climate model results for temperature and precipitation. Temperature in (a) the Aral Region and (b) the MRB. Precipitation in (c) the Aral Region and (d) the MRB. Observations are shown by the grey line, with 30 years running average in black line. Model projections include: average of 14CMIP3 models; average of 22CMIP5 models for the RCP 2.6 scenario and the RCP 8.5 scenario. Average model values for the historic period 1961–1990 and for the near future period 2010–2039 are shown as filled triangles (CMIP3) and filled circles (CMIP5). The average value of observations for the 1961–1990 period is shown by the filled square. The coefficient of variation (CV in%) of individual model results around the CMIP3 and CMIP5 ensemble means are shown in the inserts of each panel; the same quantities and units apply for the main and the insert axes.

set of models: they compared five CMIP3 and CMIP5 climate models, also finding the model uncertainty (spreading) in the CMIP5 rainfall projections to be lower than that in the CMIP3 projections. Furthermore, Fig. 2 shows that the model-projected changes in T and P from the historic period 1961–1990 to the nearfuture period 2010–2039 are greater for both CMIP5 scenarios (RCP 2.6 and RCP 8.5) than for the A2 scenario of CMIP3. With respect to change in T for Aral, it is 1.5 ◦ C for the CMIP3 A2 scenario, while the CMIP5 scenarios RCP 2.6 and RCP 8.5 show T increase of 1.7 ◦ C and 1.9 ◦ C, respectively. For the MRB, the increase in T for the CMIP3 A2 scenario is 0.88 ◦ C, which is less than that of the CMIP5 RCP 2.6 (1.05 ◦ C) and RCP 8.5 (1.25 ◦ C) scenarios. With respect to change in P for Aral, it is an increase of 10 mm for the CMIP3 A2 scenario, while the CMIP5 scenarios RCP 2.6 and RCP 8.5 show P increase of 13 mm and 18 mm, respectively. For the MRB, the increase in P for the CMIP3 A2 scenario is negligible, while the CMIP5 scenarios RCP 2.6 and RCP 8.5 show P increase of 15 mm and 27 mm, respectively. Overall, CMIP5 thus projects greater changes than CMIP3 for both T and P in our two study basins. Noting the decreased model accuracy compared with T and P observations, the reliability of these greater change projections by CMIP5 may be questioned; a recent CMIP5 performance assessment by Mueller and Seneviratne (2014) also indicates that systematic bias in T may lead to regional overestimations of ET and P.

3.2. CMIP5 performance relative to historic observations Based on the ability of individual CMIP5 models to reproduce the observation-based data, we arrange the models in decreasing order of performance in Table 1 for the Aral Region and Table 2 for the MRB. Model performance is measured in terms of the absolute value of the deviation of a model result from the corresponding observation-based data. Rank 1 is given to the model with the smallest absolute deviation value; for the Aral Region this is 12 mm/year for P, 3 mm/year for R, 0.6 mm/year for ET, 4 mm/year for P, 0.05 ◦ C for T and 0.02 ◦ C for T; for the MRB, it is 17 mm/year for P, 19 mm/year for R, 4 mm/year for ET, 60 mm/year for P, 0.06 ◦ C for T and 0 ◦ C for T. Considering both regions, the model GFDL-ESM2G comes out as the best performing model on average (overall rank 1). For the Aral Region, the MPI-ESM-MR performs on average equally well as GFDL-ESM2G (both achieve overall rank 1). For the MRB, the second best performing model is GFDL-CM3 (overall rank 2). For both study basins, versions of the GFDL model thus perform on average better relative to observations than the other investigated CMIP5 models.

54

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

Table 1 Ranking of climate model performance for the Aral Region. Performance is assessed in terms of model ability to reproduce observation-based mean values of precipitation P, runoff R, evapotranspiration ET and temperature T for the period 1961–1990 and change in the mean values of P and T between 1961–1980 and 1986–2005. The ranking is based on the absolute value of the deviation of model results from observation-based data, with rank 1 given to the model with the smallest absolute value of the deviation. P GFDL-ESM2G MPI-ESM-MR CSIRO-Mk3.6.0 MPI-ESM-LR CanESM2 NorESM1-ME bcc-csm1-1-m MRI-CGCM3 FGOALS-g2 IPSL-CM5A-MR FIO-ESM GFDL-CM3 CNRM-CM5 CCSM4 GISS-E2-R bcc-csm1-1 MIROC5 IPSL-CM5A-LR MIROC-ESM-CHEM GISS-E2-H MIROC-ESM BNU-ESM

1 6 4 3 5 2 10 9 16 8 17 13 7 11 22 15 12 14 19 21 18 20

R

ET

T

P

T

Average rank value

Overall rank

12 9 4 7 10 2 15 19 22 13 6 18 20 5 1 21 11 17 16 3 14 8

13 5 8 1 18 11 2 7 9 10 16 6 14 15 22 3 12 4 19 21 20 17

4 17 3 18 8 11 6 2 15 5 16 12 14 13 7 1 20 9 21 10 22 19

1 6 15 12 7 14 13 5 2 22 8 18 10 20 4 17 11 19 3 9 21 16

13 1 15 14 9 18 12 16 2 8 6 5 7 10 19 20 11 17 4 22 3 21

7.3 7.3 8.2 9.2 9.5 9.7 9.7 9.7 11.0 11.0 11.5 12.0 12.0 12.3 12.5 12.8 12.8 13.3 13.7 14.3 16.3 16.8

1 1 3 4 5 6 6 6 9 9 11 12 13 14 15 16 16 18 19 20 21 22

Table 2 Ranking of climate model performance for the Mahanadi River Basin (MRB). Performance is assessed in terms of model ability to reproduce observationbased mean values of precipitation P, runoff R, evapotranspiration ET and temperature T for the period 1961–1990 and change in the mean values of P and T between 1961–1980 and 1986–2005. The ranking is based on the absolute value of the deviation of model results from observation-based data, with rank 1 given to the model with the smallest absolute value of the deviation.

GFDL-ESM2G GFDL-CM3 CanESM2 NorESM1-ME BNU-ESM MIROC-ESM-CHEM MPI-ESM-LR MIROC-ESM IPSL-CM5A-MR MIROC5 CCSM4 MPI-ESM-MR FIO-ESM IPSL-CM5A-LR FGOALS-g2 CNRM-CM5 MRI-CGCM3 bcc-csm1-1-m bcc-csm1-1 CSIRO-Mk3.6.0 GISS-E2-H GISS-E2-R

P

R

ET

T

P

1 8 6 5 4 3 10 2 11 18 7 12 9 14 16 15 20 13 19 17 21 22

3 7 1 9 4 6 13 5 12 22 2 16 18 17 10 15 19 11 8 14 20 21

10 5 11 12 6 15 1 13 2 8 18 3 9 4 16 7 20 14 19 17 21 22

1 9 2 7 11 17 5 16 4 8 13 3 12 15 6 10 14 19 20 18 21 22

1 7 6 9 22 11 16 3 19 4 17 15 8 13 14 18 2 20 21 10 5 12

T 17 3 22 7 6 4 11 21 14 2 8 20 15 13 18 16 9 10 1 19 12 5

Average rank value

Overall rank

5.5 6.5 8.0 8.2 8.8 9.3 9.3 10.0 10.3 10.3 10.8 11.5 11.8 12.7 13.3 13.5 14.0 14.5 14.7 15.8 16.7 17.3

1 2 3 4 5 6 6 8 9 9 11 12 13 14 15 16 17 18 19 20 21 22

The actual deviations from observation-based data of the two overall best performing models for each region are summarized in Table 3. For the main water fluxes, the relative model deviation from observations is greater for runoff R than for precipitation P and evapotranspiration ET, and particularly so for the large Aral Region, for which the two best performing models overestimate R by 5–6 times. Overall, with regard to possible scale effects on climate model performance for water fluxes, the two best performing models for each study basin do not, in relative terms, perform better for the large Aral Region than for the small MRB (relative results in parenthesis, Table 3). With regard to water flux changes, due to insufficiently long observation time series for R and thereby also for ET ≈ P–R, the only modeled such change that can be directly compared with observation data is the precipitation change P. For P, the modeled change to the near future (increase of 13–18 mm for the Aral Region and 15–27 mm for the MRB; see previous section and Fig. 2) is of the same order or smaller than the model deviation from observations for the two overall

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

55

Table 3 Deviation of model results from observation-based data for models with overall rank 1–2 in Tables 1–2, with regard to mean values of precipitation P, runoff R, evapotranspiration ET and temperature T for the period 1961–1990 and change in the mean values of P and T between 1961–1980 and 1986–2005. Positive and negative deviation values imply model overestimation and underestimation, respectively, in relation to the observation-based data. For the main water fluxes, the relative deviation of model results from observations is also given in parenthesis. Deviation of model result from observation-based data Water fluxes (mm/year) ◦

Changes

Rank 1–2 models

Overall rank

T ( C)

P (mm/year) [%]

R (mm/year) [%]

ET (mm/year) [%]

P (mm/year)

T (◦ C)

Aral Region GFDL-ESM2G MPI-ESM-MR

1 1

−0.10 2.16

11.6 [4] 42.5 [17]

73.3 [645] 57.5 [505]

−46.6 [−19] 11.3 [5]

3.6 −14.3

0.27 0.02

MRB GFDL-ESM2G GFDL-CM3

1 2

0.06 −0.86

−16.6 [−1] 208.5 [16]

132 [26] 268 [52]

−111 [−14] −48 [−6]

60 90

0.43 −0.02

Fig. 3. CMIP5 model results compared with observation-based data for 1961–1990. Results are shown for average annual precipitation (P), runoff (R) and evapotranspiration (ET) in (a) the Aral Region and (b) the MRB. For both basins, panel c illustrates the model-implied average annual water storage change (S). Climate model results are shown for the ensemble mean of 22CMIP5 models, and for the individual and the mean results of the two best-performing models on average for each basin. Error bars show the model standard deviation around the ensemble mean.

best-performing models (absolute value range of 3.6–14.3 mm for the Aral Region and 60–90 mm for the MRB; Table 3). The relatively small deviation for the larger Aral Region might be indicative of a scale effect for the climate model results of P. However, similarly for both study basins, the modeled temperature change T to the near future (increase of 1.7–1.9 ◦ C for the Aral Region and 1.05–1.25 ◦ C for the MRB; previous section and Fig. 2) is greater than the deviation from observations of the two overall best-performing models (absolute value range of 0.02–0.27 ◦ C for the Aral Region and 0.02–0.43 ◦ C for the MRB; Table 3). For T, there is thus no clear scale effect indication in the comparison between model results and observations. Furthermore, if priority is, for example, given to model performance for runoff R through the landscape, the best performing models are for both regions mostly others than the overall best-performing ones. More generally, no single climate model consistently outperforms other models for all hydro-climatic variables and their changes. This general finding supports earlier results of Gleckler et al. (2008) and emphasizes the relevance of considering results from multiple climate models rather than using only one selected model for freshwater change assessments (Lambert and Boer, 2001; Tebaldi and Knutti, 2007; Jarsjö et al., 2012). Fig. 3a and b summarize the observation-based water fluxes within the Aral Region and the MRB, the ensemble mean of the 22 model projections of CMIP5, the projections of the two overall best-performing models (GFDL-ESM2G and MPI-ESM-MR for the Aral Region and GFDL-ESM2G and GFDL-CM3 for the MRB; for further model specification, see Appendix A, Table A3), and the mean model outputs of the latter for the climatic period 1961–1990. Appendix C (Fig. C1 for the Aral Region and Fig. C2 for the MRB) shows the performance of individual climate models relative to observations. With regard to P, the mean of the two overall best-performing models is in agreement with observed P for both regions. With respect to R, however, neither the ensemble mean of 22 models nor the 2-best-model mean accurately reproduces the small observed R in the Aral Region, which is currently close to negligible after the impacts of irrigation in this region. For the MRB, the mean value of R for the two overall best-performing models overestimates the observed R, while the model ensemble mean underestimates

56

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

R. For ET, both the model ensemble mean and the mean of the 2 best-performing models are on average closer to the observation-based results than they are for R. Overall, the relative standard deviation of the 22-model output is somewhat greater for the relatively small MRB than for the large Aral Region, thus indicating a possible scale effect on climate model precision (model standard deviation around the model ensemble mean). For the Aral Region, the model precision (standard deviation) has also decreased in CMIP5 relative to CMIP3. However, the two best models for the Aral Region have average rank values of 7 across all hydro-climatic variables (Table 1), while the two best models for the MRB both have lower average rank values of 5.5–6.5 across all variables (Table 2). Scale effects of direct application of global climate models to assessment of freshwater fluxes and their changes are thus not clearly evident from the present results. Particular caution may rather be needed in climate model application to the large but semi-arid, and thereby more water-sensitive, Aral Region than to the smaller MRB. Fig. 3c further shows the model-implied net average annual water balance over the 1961–1990 period, which quantifies the average annual storage change S = P–R–ET over that period; values for the right-hand side variables are then all given from climate model output. Where S is positive (negative) an average annual water addition (depletion) is implied for water storage within the basin, systematically over the whole 1961–1990 period. In the Aral Region, this model-implied water storage change is 7 mm/year for the ensemble mean model result. This model result thus implies an average 7 mm of water per year being added to the region-wide water storage, which in turn implies a cumulative 0.21 m water level increase over the whole 30-year period and the whole region. These model implications are inconsistent with the observed decrease of the Aral Sea level and associated decrease of surrounding groundwater level (Jarsjö and Destouni, 2004; Shibuo et al., 2006; Alekseeva et al., 2009). For comparison, the actual Aral Sea level decrease (21 m decrease over 40 years) distributed over the total Aral Region area of 1,888,810 km2 implies an average annual decrease in water storage of 9 mm/year and a cumulative decrease of 0.27 m over the whole 30-year period. The results of the two best performing climate models thus agree with the observed direction of water storage change in the Aral Region, in contrast to the ensemble mean result (Fig. 3c). However, in terms of the change magnitude, the two best models yield a mean decrease of around 20 mm/year, which overestimates the observation-based average annual decrease in water storage of 9 mm/year. For the MRB, both the ensemble mean and the two overall best-performing climate models imply an average annual decrease in water storage (Fig. 3c). The 2-model mean implies a decrease of about 0.75 m in the water stored within the basin over the whole 30-year period. Assuming, for instance, a soil porosity of 30% on average in the basin, this cumulative decrease corresponds to a basin-wide average drop in groundwater and surface water levels of 2.5 m, which is substantial and could in principle be checked against observations. However, region-specific data is not available, or at least not accessible to us for testing this model result. Overall, the model uncertainty represented by the standard deviation of the 22-model ensemble output is greater, in both relative and direction terms, for the average annual water storage change than for the other water variables P, R and ET, as well as for temperature T.

3.3. CMIP5 projections for the future Fig. 4 shows projections of the CMIP5 models for future changes in T, P, R and ET from the 1961–1990 reference conditions. For both regions and for both scenarios RCP 2.6 and 8.5, the near future (2010–2039) increase in T is projected to be between 1.5 and 2 ◦ C. For the more distant future (2070–2099), the scenario RCP 2.6 shows a T increase of 2 ◦ C, whereas the scenario RCP 8.5 shows a much greater T increase of about 6 ◦ C for both Aral Region. For the MRB, the scenario RCP 2.6 shows a T increase of 1.4 ◦ C, whereas the scenario RCP 8.5 shows a much greater T increase of about 5 ◦ C for the more distant future (2070–2099). Such T increases may then also contribute to increased ET and P (DelGenfo et al., 1991; Trenberth, 1999; Held and Soden, 2000; Huntington, 2006). The projected near-future changes in P are relatively similar and mostly indicating P increase, however with some models also projecting P decrease, for both regions. For the more distant future, a considerably greater P increase is projected for the MRB than for the Aral Region, however with relatively large standard deviation among individual model results (Fig. 4). The freshwater runoff R is mostly projected to decrease in the Aral Region, with overall negative values of R and relatively small standard deviation among models (Fig. 4). For both the Aral Region and the MRB, model projections of future change in ET are highly uncertain for the near-future period (2010–2039), while mostly indicating increase for the more distant future (2070–2099). Overall, Fig. 4 shows that the climate model projections for the future are far more uncertain for freshwater fluxes (see the large standard deviation bars in the six lower panels for P, R and ET in Fig. 4) and thereby also for freshwater availability in the landscape than for temperature (see the relatively small standard deviation bars in the two upper panels for T in Fig. 4). Even change directions are uncertain for some water fluxes in both the near and the more distant future (see the standard deviation bars extending far into both the positive and the negative change directions in the six lower water flux panels of Fig. 4). However, model projections agree relatively well on that the freshwater loss to the atmosphere by ET should be expected to increase in both regional basins until 2070–2099, regardless of model scenario (see results for ET to 2070–2099, and its standard deviation extending primarily in the positive direction, in the two lower panels of Fig. 4). For the Aral Region, with no clear projection of P increase until this period, models thus agree on that the increased loss of freshwater by ET will lead to decreased runoff R through the landscape. For the MRB, with relative agreement in the model

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

57

Fig. 4. Projected future change () in average annual temperature (T), average annual precipitation (P), runoff (R), evapotranspiration (ET) values of CMIP5 ensemble mean (mean of 22 models is shown in purple, and mean of 2 best performing models is shown in grey color) from 1961–1990 to 2010–2039 and 2070–2099. The model standard deviation is shown as blue error bars on the respective columns. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

ensemble on P increase until 2070–2099, this P increase is expected to largely feed into the ET increase, so that the change in R may be relatively small, even though highly uncertain, also for the more distant future. 3.4. CMIP5 representation of high hydrological seasonality Planning for adaptation to and management of future water resource changes may also crucially depend on changes to the seasonality and timing of P and R fluxes and not just on the change of their average values. In this context, we focus here on assessing the ability of CMIP5 to reproduce the extreme, monsoon-driven seasonal variability and timing of freshwater fluxes in the MRB. In the Aral Region, hydrological seasonality is relatively small even though the basin has glaciated parts and the temperature-driven snow and glacial dynamics also drive hydrological seasonality. However, the glaciated parts of this basin have experienced the smallest historic temperature increases over the basin and the temperature change seasonality in the basin reflects primarily the cooling effect of the historic irrigation-driven ET increase (Destouni et al., 2010). For the MRB, Fig. 5 illustrates the monsoon-driven seasonality of monthly freshwater fluxes P, R, ET and storage change S for the historic time period with available monthly observation or observation-derived data (1990–2002). It also shows the differences of the RCP 2.6 and RCP 8.5 scenarios from the CMIP5 historical average for the near future (2010–2039) and the more distant future (2070–2099). As also noted above, the model ensemble mean overall underestimates P compared with the historic observations. Underestimation of the observed peak in monthly P, in August, is then particularly prominent (Fig. 5a). Peak underestimation is even more pronounced for R, which is thereby modeled to have smaller seasonality than exhibited by observations (Fig. 5b). Furthermore, the modeled monthly ET (Fig. 5c) yields a net water balance P–ET–R = S that

58

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

Fig. 5. Seasonal hydro-climatic cycles in the MRB. Results are shown for (a) precipitation (available observations and mean of model ensemble), (b) runoff (available observations and mean of model ensemble), (c) evapotranspiration (mean of model ensemble) and (d) water storage change (mean of model ensemble), for the historic time period 1990–2002 (left, available observations and models), and for the future time periods (models) 2010–2039 (middle) and 2070–2099 (right). The error bars show the standard deviation of individual climate models around the ensemble mean for 1990–2002 and for future periods in scenarios RCP 2.6 and RCP 8.5.

implies large increase of water storage during high flow conditions (Fig. 5d); at maximum, the modeled S of 100 mm/month is even greater than the maximum high-flow R at 69 mm/month. In contrast, for the rest of the year, the climate model ensemble implies systematic water storage decrease on the order of 30–50 mm/month, which in turn leads to the previously discussed model-implied annual decrease in water storage for the MRB (Fig. 3). The MRB contains constructed water reservoirs, such as the Hirakud reservoir, which may contribute to human-driven changes in intra-annual hydro-climatic variability, such as decrease in the short-term variability of R relative to pre-reservoir conditions (Destouni et al., 2013). However, the climate model results do not reproduce the actually observed post-reservoir seasonal variability of R (Fig. 5b). Furthermore, the long-term water storage decrease implied by the climate models corresponds to a cumulative (subsurface and surface) water level drop of 2.2 m on average over the whole basin; since the models are unable to reproduce the post-reservoir seasonality of R, such a large water-level drop may not be consistent with realistic changes in intra-annual water storage variability. In particular, the CMIP5 model ensemble does not reproduce the observed high peak in R during June–August, but shows instead a smaller peak in September, a month later than indicated by observations; this September peak also appears in the model projections of future scenarios. The phase shift in the R peak relative to observations indicates a likely model overestimation of the temporary increase of water storage during high-flow conditions. The model dampening of seasonal R variability found here for the MRB does not resemble the model bias results reported by Siam et al. (2013), who evaluated CMIP5 historical simulations for the tropical Congo basin. Their results indicate instead overestimation of the seasonal amplitude in water fluxes, particularly for runoff in that basin (Siam et al., 2013; their Fig. 6). As the Congo basin is less influenced by regulation and irrigation than the MRB, the difference in climate model performance relative to observations for these basins with different human change drivers points to a need for further basin-scale studies

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

59

of human-driven effects on runoff seasonality, as well as of the representations of this seasonality and its change drivers in the land surface schemes of climate models. 4. Conclusions We have investigated and compared CMIP5 model performance in reproducing the observed historic and modeling the future evolution of basin-scale hydro-climate, water resource availability and their changes in two extensively irrigated Asian regions: the large semi-arid Aral Region and the smaller tropical MRB. For precipitation, the accuracy of climate model results compared with observations has remained essentially the same or progressively decreased from the CMIP1 through the CMIP3 to the CMIP5 generation of climate models; this applies when considering un-corrected as well as bias-corrected observation data for P. However, the model precision has mostly improved (the CV has decreased) for both temperature and precipitation in comparison with CMIP3. No clear scale effects of direct application of global climate models to freshwater assessment are evident from the present result comparison between the large Aral Region and the relatively small MRB. Rather, the long-term average water storage increase implied by the model ensemble output for the large Aral Region directly contradicts the water storage decrease occurring in this region, as evident in the dramatic Aral Sea desiccation (Jarsjö et al., 2004; Shibuo et al., 2006; Alekseeva et al., 2009). In comparison, the model-implied decrease of long-term average water storage in the smaller MRB may be qualitatively realistic but the change magnitude is highly uncertain, in particular since the CMIP5 model ensemble does not accurately reproduce the seasonal variability in runoff, including the timing and magnitude of the runoff peak. The present scale-effect findings may apply only to the specific two investigated study basins, and similar analysis for other regional basins, with downscaled hydro-climatic variables, may lead to different results. Further investigation across basins of different sizes and in different parts of the world is needed to arrive at more general scale-effect conclusions. For both of the investigated basins in this study, versions of the GFDL model (NOAA Geophysical Fluid Dynamics Laboratory; Table A3, Appendix A) are found to perform on average better relative to observations than other CMIP5 models. However, neither this nor any other single climate model consistently outperforms all of the other models with respect to all key freshwater variables. Overall, climate model projections are considerably more uncertain for freshwater fluxes and their changes than for temperature T and its change, with both inaccuracy and uncertainty being particularly high for the model-implied water storage changes. For future freshwater changes, the CMIP5 models largely agree on that evapotranspiration ET should be expected to increase in both regions until 2070–2099. They also agree on that this ET increase in the Aral Region will mostly be fed by runoff water, thereby leading to decreased R and thus decreased annually renewable freshwater in this Central Asian region. In contrast, for the tropical MRB, the climate models largely agree on that the ET increase here will mostly be fed by an increase in precipitation P, implying relatively small, even though highly uncertain, runoff change in this South Asian region. Conflict of interest The authors declare no competing financial interests or other conflict of interest. Acknowledgments We acknowledge the World Climate Research Programme’s Working Group on Coupled Modeling, which is responsible for CMIP, and we thank the U.S. Department of Energy’s Program for Climate Model Diagnosis and Intercomparison for coordinating support and development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. The research has been funded by the strategic environmental research project Ekoklim at Stockholm University. Appendix A

Table A1 Investigated hydrological regions and total aggregated drainage basin areas. Region

Basin area (km2 )

Reference with basin description

Aral Mahanadi

1.89 × 106 0.14 × 106

Destouni et al. (2010) Asokan et al. (2010)

60

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

Table A2 List of discharge stations and data sources. Region

River

Station coordinates or name

Basin areaa (km2 )

Data source

Amu Darya Syr Darya

Kziljar Karateren

0.68 × 106 0.71 × 106

Mamatov (2003) Mamatov (2003)

Mahanadi

Tikarpara

1.35 × 105

Asokan and Dutta (2008)

Aralb

Mahanadi a

Refers to the area upstream of the corresponding monitoring station. b For the Aral Region, all climate data are calculated for the total drainage area, including the combined individual station catchments and the unmonitored areas draining into the Aral Sea. Table A3 List of investigated climate models in the CMIP3 and CMIP5 model generations. These are referred to as Global Circulation Models (GCMs) for the CMIP3 model generation and as Earth System Models (ESMs) for the CMIP5 model generation. CMIP3 GCMs

CMIP5 ESMs

Institution

CSIRO:Mk3.0

CSIRO-Mk3.6.0

CSIRO in collaboration with Queensland Climate Change Centre of Excellence

ECHAM5/MPI-OM

MPI-ESM-LR MPI-ESM-MR

Max Planck Institute for Meteorology

GFDL:CM2.0; GFDL:CM2.1

GFDL-CM3 GFDL-ESM2G

NOAA Geophysical Fluid Dynamics Laboratory

MIROC3.2

MIROC5

Atmosphere and Ocean Research Institute (The University of Tokyo), National Institute for Environmental Studies, and Japan Agency for Marine-Earth Science and Technology Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The University of Tokyo), and National Institute for Environmental Studies Centre National de RecherchesMeteorologiques/Centre Europeen de Recherche et Formation AvanceesenCalculScientifique NASA Goddard Institute for Space Studies

MIROC-ESM MIROC-ESM-CHEM CNRM-CM3

CNRM-CM5

GISS-ER

GISS-E2-R GISS-E2-H

IPSL-CM4

IPSL-CM5A-LR IPSL-CM5A-MR

Institut Pierre-Simon Laplace

MRI-CGCM2.3.2

MRI-CGCM3

Meteorological Research Institute

CCSM3

CCSM4

National Center for Atmospheric Research

NCPCM

National Center for Atmospheric Research

HadCM3

Hadley Centre for Climate Prediction and Research, Met Office

ECHO-G

Meteorological Institute of the University of Bonn (Germany), Institute of KMA (Korea) and Model and Data Group

HADGEM

Hadley Centre for Climate Prediction and Research, Met Office

INCM3 BNU-ESM CanESM2 FGOALS-g2 FIO-ESM NorESM1-ME bcc-csm1-1 bcc-csm1-1-m

Institute of Numerical Mathematics, Russian Academy of Science College of Global Change and Earth System Science, Beijing Normal University Canadian Centre for Climate Modelling and Analysis LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences; and CESS, Tsinghua University The First Institute of Oceanography, SOA, China Norwegian Climate Centre Beijing Climate Center, China Meteorological Administration

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

61

Appendix B

Fig. B1. Climate model performance for annual average precipitation during 1961–1990 in: (a) the Aral Region and (b) the MRB. Observed precipitation is also shown as corrected for undercatch and orographic factors. Appendix A (Table A3) provides further climate model specifications.

62

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

Appendix C

Fig. C1. Performance of individual CMIP5 models for the Aral Region. Difference of model output from corresponding observations (or observation implications for ET) for the period 1961–1990 and the hydro-climatic variables: (a) temperature T, (b) precipitation P, (c) runoff R and (d) evapotranspiration ET. Appendix A (Table A3) provides further climate model specifications.

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

63

Fig. C2. Performance of individual CMIP5 models for the MRB. Difference of model output from corresponding observations (or observation implications for ET) for the period 1961–1990 and the hydro-climatic variables: (a) temperature T, (b) precipitation P, (c) runoff R and (d) evapotranspiration ET. Appendix A (Table A3) provides further climate model specifications.

Appendix D. Supplementary data Supplementary data associated with http://dx.doi.org/10.1016/j.ejrh.2015.11.017.

this

article

can

be

found,

in

the

online

version,

at

References Adam, J.C., Lettenmaier, D.P., 2003. Adjustment of global gridded precipitation for systematic bias. J. Geophys. Res. 108 (D9), 1–14. Adam, J.C., Clark, E.A., Lettenmaier, D.P., Wood, E.F., 2006. Correction of global precipitation products for orographic effects. J. Clim. 19 (1), 15–38. Alkama, R., Marchand, L., Ribes, A., Decharme, B., 2013. Detection of global runoff changes: results from observations and CMIP5 experiments. Hydrol. Earth Syst. Sci. 17 (7), 2967–2979, http://dx.doi.org/10.5194/hess-17-2967-2013. Alekseeva, I., Jarsjö, J., Schrum, C., Destouni, G., 2009. Reproducing the Aral Sea water budget and sea–ground water dynamics between 1979 and 1993 using a coupled 3-D sea–ice–ground water model. J. Mar. Syst. 3, 296–309. Arnell, N.W., 1999. Climate change and global water resources. Global Environ. Change 9, S31–S49. Asokan, S.M., Dutta, D., 2008. Analysis of water resources in the Mahanadi River Basin, India under projected climate conditions. Hydrol. Process. 22, 3589–3603, http://dx.doi.org/10.1002/hyp.6962. Asokan, S.M., Jarsjö, J., Destouni, G., 2010. Vapor flux by evapotranspiration: effects of changes in climate, land use and water use. J. Geophys. Res. 115 (D24102), http://dx.doi.org/10.1029/2010JD04417. Asokan, S.M., Destouni, G., 2014. Irrigation effects on hydro-climatic change: basin-wise water balance-constrained quantification and cross-regional comparison. Surv. Geophys. 35, 879–895, http://dx.doi.org/10.1007/s10712-013-9223-5. Bring, A., Destouni, G., 2011. Relevance of hydro-climatic change projection and monitoring for assessment of water cycle changes in the Arctic. Ambio 40, 361–369. ´ J., Prieto, C., Rogberg, P., Destouni, G., 2015. Implications of freshwater flux data from the Bring, A., Asokan, S.M., Jaramillo, F., Jarsjö, J., Levi, L., Pietron, CMIP5 multi-model output across a set of Northern Hemisphere drainage basins. Earth’s Future 3 (6), 206–217, http://dx.doi.org/10.1002/2014EF000296. Bengtsson, L., 2010. The global atmospheric water cycle. Environ. Res. Lett. 5 (025001), http://dx.doi.org/10.1088/1748-9326/5/2/025001.

64

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

Bengtsson, L., et al., 2014. Foreword: international space science institute (ISSI) workshop on the earth’s hydrological cycle. In: Bengtsson, L. (Ed.), The Earth’s Hydrological Cycle, Space Science Series of ISSI. Springer, Dordrecht, Heidelberg, New York London, pp. pp. 1–4, http://dx.doi.org/10.1007/978-94-017-8789-5. Darracq, A., Greffe, F., Hannerz, F., Destouni, G., Cvetkovic, V., 2005. Nutrient transport scenarios in a changing Stockholm and Malaren valley region, Sweden. Water Sci. Technol. 51 (3–4), 31–38. DelGenfo, A.D., Lacis, A.A., Ruedy, R.A., 1991. Simulations of the effect of a warmer climate on atmospheric humidity. Nature 351, 382–385, http://dx.doi.org/10.1038/351382a0. Deng, H., Luo, Y., Yao, Y., Liu, C., 2013. Spring and summer precipitation changes from 1880 to 2011 and the future projections from CMIP5 models in the Yangtze River Basin, China. Quat. Int. 304, 95–106, http://dx.doi.org/10.1016/j.quaint.2013.03.036. Destouni, G., Darracq, A., 2009. Nutrient cycling and N2O emissions in a changing climate: the subsurface water system role. Environ. Res. Lett. 4 (035008), 7pp., http://dx.doi.org/10.1088/1748-9326/4/3/035008. Destouni, G., Asokan, S.M., Jarsjö, J., 2010. Inland hydro–climatic interaction: effects of human water use on regional climate. Geophys. Res. Lett. 37 (L18402), http://dx.doi.org/10.1029/2010gl044153. Destouni, G., Jaramillo, F., Prieto, C., 2013. Hydroclimatic shifts driven by human water use for food and energy production. Nat. Clim. Change 3, 213–217, http://dx.doi.org/10.1038/nclimate1719. Flato, G., et al. 2013. Evaluation of climate models, Climate Change 2013: the physical science basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge, United Kingdom and New York, NY, USA. Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S., Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A., Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K., 2005. Global consequences of land use. Science 309, 570–574, http://dx.doi.org/10.1126/science.1111772. Gaybullaev, B., Chen, S.C., Kuo, Y.M., 2012. Large-scale desiccation of the Aral Sea due to over-exploitation after 1960. J. Mt. Sci. 9, 538–546, http://dx.doi.org/10.1007/s11629-012-2273-1. Gleckler, P.J., Taylor, K.E., Doutriaux, C., 2008. Performance metrics for climate models. J. Geophys. Res. 113, http://dx.doi.org/10.1029/2007jd008972D06104. Gleick, P.H., Chalecki, E.L., 1999. The impacts of climatic changes for water resources of the Colorado and Sacramento-San Joaquin river basins. J. Am. Water Resour. Assoc. 35, 1429–1441, http://dx.doi.org/10.1111/j.1752-1688.1999.tb04227.x. Gordon, L.J., Steffen, W., Jönsson, B.F., Folke, C., Falkenmark, M., Johannessen, A., 2005. Human modification of global water vapor flows from the land surface. Proc. Natl. Acad. Sci. U. S. A. 102, 7612–7617. Groves, D.G., Yates, D., Tebaldi, C., 2008. Developing and applying uncertain global climate change projections for regional water management planning. Water Resour. Res. 44 (W12413), http://dx.doi.org/10.1029/2008wr006964. Huntington, T.G., 2006. Evidence for intensification of the global water cycle: review and synthesis. J. Hydrol. 319, 83–95, http://dx.doi.org/10.1016/j.jhydrol.2005.07.003. Held, I.M., Soden, B.J., 2000. Water vapor feedback and global warming. Annu. Rev. Energy Environ. 25, 441–475, http://dx.doi.org/10.1146/annurev.energy.25.1.441. IPCC, Climate change 2007: synthesis report, contribution of working groups I II and III to the fourth assessment report of the intergovernmental panel on climate change 2007, 104. Ibrahim, N.M., Bomblies, A., Wemple, B.C., 2015. The use of CMIP5 data to simulate climate change impacts on flow regime within the Lake Champlain Basin. J. Hydrol.: Reg. Stud. 3, 160–186, http://dx.doi.org/10.1016/j.ejrh.2015.01.002. Jaramillo, F., Destouni, G., 2014. Developing water change spectra and distinguishing change drivers worldwide. Geophys. Res. Lett. 41, 8377–8386, http://dx.doi.org/10.1002/2014gl061848. Jarsjö, J., Destouni, G., 2004. Groundwater discharge into the Aral Sea after 1960. J. Mar. Syst. 47, 109–120, http://dx.doi.org/10.1016/j.jmarsys.2003.12.013. Jarsjö, J., Asokan, S.M., Prieto, C., Bring, A., Destouni, G., 2012. Hydrological responses to climate change conditioned by historic alterations of land-use and water-use. Hydrol. Earth Syst. Sci. 16, 1335–1347, http://dx.doi.org/10.5194/hess-16-1335-2012. Karlsson, J.M., Lyon, S.W., Destouni, G., 2012. Thermokarst lake, hydrological flow and water balance indicators of permafrost change in Western Siberia. J. Hydrol., 464–465, http://dx.doi.org/10.1016/j.jhydrol.2012.07.037, 459–466. Karthe, D., Chalov, S., Borchardt, D., 2015. Water resources and their management in central Asia in the early twenty first century: status, challenges and future prospects. Environ. Earth Sci. 73 (2), 487–499, http://dx.doi.org/10.1007/s12665-014-3789-1. Knutti, R., Sedláˇcek, J., 2013. Robustness and uncertainties in the new CMIP5 climate model projections. Nat. Clim. Change 3 (4), 369–373, http://dx.doi.org/10.1038/nclimate1716. Kundzewicz, Z.W., Stakhiv, E.Z., 2010. Are climate models ready for prime time in water resources management applications, or is more research needed? Hydrol. Sci. J. 55 (7), 1085–1089, http://dx.doi.org/10.1080/02626667.2010.513211. Lambert, S.J., Boer, G.J., 2001. CMIP1 evaluation and inter comparison of coupled climate models. Clim. Dyn. 17, 83–106. Lettenmaier, D.P., Wood, A.W., Palmer, R.N., Wood, E.F., Stakhiv, E.Z., 1999. Water resources implications of global warming: A U S. regional perspective. Clim. Change 43, 537–579. Mamatov, S., 2003. Study of the groundwater contribution to the Aral Sea region water supply and water quality: strategies for reversibility and pollution control. INTAS project 1014, EU-INTAS Aral Sea Basin Call 2000, Group CR5 Status Report, European Commission, Bruxelles, Belgium. Meehl, G.A., Covey, C., Delworth, T., Latif, M., McAvaney, B., Mitchell, J.F.B., Stouffer, R.J., Taylor, K.E., 2007. The WCRP CMIP3 multi-model dataset: a new era in climate change research. Bulletin of the American Meteorological Society. 88, 1383-1394, 2007. Micklin, P., 2007. The Aral sea disaster. Annu. Rev. Earth Planet Sci. 35, 47–72. Milly, P.C.D., Wetherald, R.T., Dunne, K.A., Delworth, T.L., 2002. Increasing risk of great floods in a changing climate. Nature 415, 514–517. Milly, P.C.D., Dunne, K.A., Vecchia, A.V., 2005. Global pattern of trends in streamflow and water availability in a changing climate. Nature 438, 347–350, http://dx.doi.org/10.1038/nature04312. Mitchell, T.D., Jones, P.D., 2005. An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol. 25, 693–712, http://dx.doi.org/10.1002/joc.1181. Moss, R.H., Edmonds, J.A., Hibbard, K.A., Manning, M.R., 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 change research and assessment. Nature 463, 747–756, http://dx.doi.org/10.1038/nature08823. Mueller, B., Seneviratne, S.I., 2014. Systematic land climate and evapotranspiration biases in CMIP5 simulations. Geophys. Res. Lett. 41 (1), 1–7, http://dx.doi.org/10.1002/2013gl058055. New, M., Hulme, M., Jones, P., 1999. Representing twentieth-century space-time climate variability Part I: development of a 1961–90 mean monthly terrestrial climatology. J. Clim. 12, 829–856. Pahl-Wostl, C., 2007. Transitions towards adaptive management of water facing climate and global change. Water Resour. Manag. 21 (1), 49–62, http://dx.doi.org/10.1007/s11269-006-9040-4. PCDMI, 2013. CMIP Collaboration protocol, Available online at:http://www-pcmdi.llnl.gov/projects/cmip/diagsub.php (accessed 9.9.13). Piao, S., Friedlingstein, P., Ciais, P., de Noblet-Ducoudre, N., Labat, D., Zaehle, S., 2007. Changes in climate and land use have a larger direct impact than rising CO2 on global river runoff trends. Proceedings of the National Academy of Sciences of the United States of America 104, 15242–15247, http://dx.doi.org/10.1073/pnas.0707213104.

S.M. Asokan et al. / Journal of Hydrology: Regional Studies 5 (2016) 48–65

65

Raje, D., Krishnan, R., 2012. Bayesian parameter uncertainty modeling in a macroscale hydrologic model and its impact on Indian river basin hydrology under climate change. Water Resour. Res. 48 (8), http://dx.doi.org/10.1029/2011WR011123, W08522. Seneviratne, S.I., Lüthi, D., Litschi, M., Schär, C., 2006. Land–atmosphere coupling and climate change in Europe. Nature 443 (7108), 205–209, http://dx.doi.org/10.1038/nature05095. Shashikanth, K., Salvi, K., Ghosh, S., Rajendran, K., 2014. Do CMIP5 simulations of Indian summer monsoon rainfall differ from those of CMIP3? Atmos. Sci. Lett. 15, 79–85, http://dx.doi.org/10.1002/asl2.466. Siam, M.S., Demory, M.-E., Eltahir, E.A.B., 2013. Hydrological cycles over the Congo and Upper Blue Nile basins: evaluation of general circulation model simulations and reanalysis products. J. Clim. 26 (22), 8881–8894, http://dx.doi.org/10.1175/jcli-d-12-00404.1. Shibuo, Y., Jarsjö, J., Destouni, G., 2006. Bathymetry-topography effects on saltwater–fresh groundwater interactions around the shrinking Aral Sea. Water Resour. Res. 42 (W11410), http://dx.doi.org/10.1029/2005wr004207. Shibuo, Y., Jarsjö, J., Destouni, G., 2007. Hydrological responses to climate change and irrigation in the Aral Sea drainage basin. Geophys. Res. Lett. 34 (L21406), http://dx.doi.org/10.1029/2007gl031465. Solomon, S., et al., 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge Univ. Press, Cambridge, U. K., and New York, 996 pp. Taylor, K.E., Stouer, R.J., Meehl, G.A., 2012. An overview of CMIP5 and the experiment design. Bull. Am. Meteorol. Soc. 93, 485–498. Tebaldi, C., Knutti, R., 2007. The use of the multi-model ensemble in probabilistic climate projections. Phil. Trans. R. Soc. A 365, 2053–2075, http://dx.doi.org/10.1098/rsta.2007.2076. Törnqvist, R., Jarsjö, J., Karimov, B., 2011. Health risks from large-scale water pollution: trends in Central Asia. Environ. Int. 37 (2), 435–442, http://dx.doi.org/10.1016/j.envint.2010.11.006. Törnqvist, R., Jarsjö, J., 2012. Water savings through improved irrigation techniques: basin-scale quantification in semi-arid environments. Water Resour. Manag. 26, 949–962, http://dx.doi.org/10.1007/s11269-011-9819-9. ´ J., Bring, A., Rogberg, P., Asokan, S.M., Destouni, G., 2014. Evolution of hydro-climate in the Lake Baikal basin. J. Hydrol. 519, Törnqvist, R., Jarsjö, J., Pietron, 1953–1962, http://dx.doi.org/10.1016/j.jhydrol.2014.09.074. Törnqvist, R., Jarsjö, J., Thorslund, J., Rao, P.S.C., Basu, N.B., Destouni, G., 2015. Mechanisms of basin-scale nitrogen load reductions under intensified irrigated agriculture. PLoS ONE 10 (3), e0120015, http://dx.doi.org/10.1371/journal.pone.0120015. Trenberth, K.E., 1999. Conceptual framework for changes of extremes of the hydrological cycle with climate change. Clim. Change 42, 327–339. Trenberth, K., 2010. More knowledge, less certainty. Nat. Rep. Clim. Change 1002, 20–21, http://dx.doi.org/10.1038/climate.2010.06. UNECE, 2009. Guidance on water and adaptation to climate change. Printed at United Nations, Geneva, ISBN 978-92-1-117010-8. Van der Velde, Y., Lyon, S.W., Destouni, G., 2013. Data-driven regionalization of river discharges and emergent land cover-evapotranspiration relationships across Sweden. J. Geophys. Res. Atmos. 118, 2576–2587, http://dx.doi.org/10.1002/jgrd.50224. Visser, A., Kroes, J., Van Vliet, M.T.H., Blenkinsop, S., Fowler, H.J., Broers, H.P., 2012. Climate change impacts on the leaching of a heavy metal contamination in a small lowland catchment. J. Contam. Hydrol. 127, 47–64, http://dx.doi.org/10.1016/j.jconhyd.2011.04.007. Wambura, F.J., Ndomba, P.M., Victor, K., Tumbo, S.D., 2015. Uncertainty of runoff projections under changing climate in Wami River sub-basin. J. Hydrol.: Reg. Stud. 4, 333–348, http://dx.doi.org/10.1016/j.ejrh.2015.05.013. Weiland, F.C.S., van Beek, L.P.H., Kwadijk, J.C.J., Bierkens, M.F.P., 2010. The ability of a GCM-forced hydrological model to reproduce global discharge variability. Hydrol. Earth Syst. Sci. 14 (8), 1595–1621. Weiskel, P.K., Vogel, R.M., Steeves, P.A., Zarriello, P.J., DeSimone, L.A., Ries III, K.G., 2007. Water use regimes: characterizing direct human interaction with hydrologic systems. Water Resour. Res. 43 (W04402), http://dx.doi.org/10.1029/2006wr005062. Wisser, D., Fekete, B.M., Vörösmarty, C.J., Schumann, A.H., 2010. Reconstructing 20th century global hydrography: a contribution to the global terrestrial network-hydrology (GTN-H). Hydrol. Earth Syst. Sci. 14 (1-24), http://dx.doi.org/10.5194/hess-14-1-2010. Wood, A.W., Leung, L.R., Sridhar, V., Lettenmaier, D.P., 2004. Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs. Clim. Change 62, 189–216. Yan, D., Werners, S.E., Ludwig, F., Huang, H.Q., 2015. Hydrological response to climate change: the Pearl River, China under different RCP scenarios. J. Hydrol.: Reg. Stud. 4, 228–245, http://dx.doi.org/10.1016/j.ejrh.2015.06.006.