source report

Journal of Hydrology 479 (2013) 146–158

Contents lists available at SciVerse ScienceDirect

Journal of Hydrology

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / j h y d r o l

Impact of climate change on water resources status: A case study for Crete Island, Greece

Aristeidis G. Koutroulis a, Ioannis K. Tsanis b,a,⇑, Ioannis N. Daliakopoulos a, Daniela Jacob c a Department of Environmental Engineering, Technical University of Crete, Greece b Department of Civil Engineering, McMaster University, Hamilton, Canada1 c Climate Service Center, Max Planck Institute for Meteorology, Hamburg, Germany

a r t i c l e i n f o

Article history: Received 12 May 2012 Received in revised form 29 October 2012 Accepted 22 November 2012 Available online 11 December 2012 This manuscript was handled by Konstantine P. Georgakakos, Editor-in-Chief, with the assistance of Aiguo Dai, Associate Editor

Keywords: Climate change impacts Crete Water resources Hydrological changes

0022-1694/$ - see front matter � 2012 Elsevier B.V. A http://dx.doi.org/10.1016/j.jhydrol.2012.11.055

⇑ Corresponding author at: Department of Environm University of Crete, Greece. Tel.: +30 2821037799; fa

E-mail address: [email protected] (I.K. Tsanis). 1 On leave.

s u m m a r y

An assessment of the impact of global climate change on the water resources status of the island of Crete, for a range of 24 different scenarios of projected hydro-climatological regime is presented. Three ‘‘state of the art’’ Global Climate Models (GCMs) and an ensemble of Regional Climate Models (RCMs) under emis- sion scenarios B1, A2 and A1B provide future precipitation (P) and temperature (T) estimates that are bias adjusted against observations. The ensemble of RCMs for the A1B scenario project a higher P reduction compared to GCMs projections under A2 and B1 scenarios. Among GCMs model results, the ECHAM model projects a higher P reduction compared to IPSL and CNCM. Water availability for the whole island at basin scale until 2100 is estimated using the SAC-SMA rainfall–runoff model And a set of demand and infra- structure scenarios are adopted to simulate potential water use. While predicted reduction of water availability under the B1 emission scenario can be handled with water demand stabilized at present val- ues and full implementation of planned infrastructure, other scenarios require additional measures and a robust signal of water insufficiency is projected. Despite inherent uncertainties, the quantitative impact of the projected changes on water availability indicates that climate change plays an important role to water use and management in controlling future water status in a Mediterranean island like Crete. The results of the study reinforce the necessity to improve and update local water management planning and adaptation strategies in order to attain future water security.

� 2012 Elsevier B.V. All rights reserved.

1. Introduction

Climate change is expected to affect precipitation and evapo- transpiration patterns (Tsanis et al., 2011), and consequently vari- ables such as local water availability, river discharge, and the seasonal availability of water supply (Arnell et al., 2011). As de- mand for freshwater on a global scale rises due to a variety of fac- tors including population growth, water pollution and economic progress, land use and climate change render its availability into the future uncertain (Davies and Simonovic, 2011). Social and envi- ronmental aspects such as agriculture, tourism and biodiversity conservation are connected to water resources quality and avail- ability, and therefore adaptation measures for water will be strongly bound with policies in a wide spectrum of disciplines (Iglesias et al., 2011).

The latest review on the current state of the art on climate change research for the Mediterranean region by Ludwig et al.

ll rights reserved.

ental Engineering, Technical x: +30 2821037849.

(2011) shows that recently observed trends and projections from climate model ensembles indicate a strong susceptibility to change in hydrological regimes, an increasing general shortage of water re- sources and consequent threats to water availability and manage- ment. These projections enhance the necessity for more robust water management, pricing and recycling policies, in order to se- cure adequate future water supply and prevent tensions among users (García-Ruiz et al., 2011).

Despite the increasing research efforts, there are still consider- able uncertainties in the future climate drivers and in how global hydrological systems will respond to their behaviour (Harding et al., 2011). A number of studies (e.g. Akhtar et al., 2008; Alcamo et al., 2007; Barnett and Pierce, 2009; Charlton and Arnell, 2011; Christensen and Lettenmaier, 2007; Fujihara et al., 2008; Georgakakos et al., 2012) have described the impacts of the expected climate change on global and regional water resources with respect to the various inherent uncertainties. The increasing availability of climatic outputs from general and regional circula- tion models provide the potential of exploring model uncertainties in predicting future climate through the use of ensembles, at global (Manning et al., 2009) but more importantly at regional scales where forcing data is often less accessible but more accurate.

A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158 147

Indicative of the instability in the Mediterranean is the island of Crete, where analysis of climate models data depicts that precipi- tation on average is likely to be less frequent but more intense and droughts are likely to become more frequent and severe in some regions (Koutroulis et al., 2010, 2011; Tsanis et al., 2011). Shorter rainy periods and seasonality shifts could seriously affect water resources by significant reduction of water availability with wide ranging consequences for local societies and ecosystems. It is indicative that, during the last decade, the island of Crete has faced an increased number of droughts (Koutroulis et al., 2011). More- over, the rapid development of Crete in the last 30 years has ex- erted strong pressures on many natural resources. Urbanization and growth of agriculture and tourism industry have strong impact on the water resources of the Island by substantially increasing water demand. Water use in Crete increased following the expan- sion of irrigated land by over than 55% during the period 1985– 2000 (Donta et al., 2005). Regarding future water demands of the island, recent estimates forecast total uses for the year 2015 in the order of 550 Mm3/year, which represents 7% of the mean an- nual precipitation. This highlights that any arising water stress is- sue will be due to poor extraction or retention technology rather than actual availability. It is therefore considered essential to tackle the increasingly severe water problems that the island will face via strategic policies adopting to climate change by using inte- grated water management systems.

This paper assesses the implications of global climate change on the water resources status for the island of Crete for a range of 24 different scenarios from a combination of projected hydro-climato- logical regimes, demand and supply potentials. For this purpose, ‘‘state of the art’’ climate model results within WATCH FP6 (Harding et al., 2011) modelling framework under A2 and B1 emis- sion scenarios and FP6 ENSEMBLES (van der Linden and Mitchell, 2009) regional climate model results under A1B emission scenario are compared, to explore the water resources availability during the 21st century. A two-step bias correction procedure is adopted to adjust precipitation on the observed long-term frequency and intensity distribution based on the period 1970–2000 (Ines and Hansen, 2006; Law and Kelton, 1982; Wood et al., 2002). The cor- rection involves truncating the RCM rainfall distribution and then mapping it onto a gamma distribution fitted to the observed inten- sity distribution. The SAC-SMA hydrological model (Burnash, 1995) is used to estimate the evolution of hydrological variables and water availability on the island of Crete. The outcome of the anal- ysis is useful for the comprehension of the role and consequently the priority of certain water resources related infrastructure development.

2. Methodology

2.1. Framing the problem

The present water resources assessment includes various water management issues regarding actions up to present as well as fu- ture perspectives for policy, management and hydrological (cli- mate) regime, resulting in the modelling framework presented in Fig. 1. These future scenarios include the augmentation of agricul- tural practices, improvement and extension of the already estab- lished irrigation network as well as tourism, permanent population and demand trends. For the purposes of this study, it’s considered that demand can exceed supply, when for example scheduled irrigation is not fully satisfied. In this context, a poorly formed irrigation policy can lead to irrigation infrastructure always inferior to the demand for water resources. Furthermore, in this study water availability is defined as the sum of abstracted or potentially abstracted groundwater and available surface freshwa-

ter runoff that is or can be potentially exploited. For convenience, the reference periods were split to a historic hydrologic regime (1970–2000) and two future periods (2000–2050 and 2050– 2100) and represent the spatial average of all the grid cells that cover Crete in the WATCH and ENSEMBLES domain. The observa- tion period (1970–2000) was used for weighting of Ensembles RCMs, bias correction of Ensembles and WATCH model results and interpretation for the two future (2000–2050 and 2050– 2100) periods.

2.2. Scenarios and storylines

The IPCC Third Assessment Report (TAR) has published a set of emissions scenarios, called the Special Report on Emissions Scenar- ios (SRES). Four scenario storylines, labelled A1, A2, B1 and B2, were the result of analyzing different possible future development pathways of the main demographic, economic and technological drivers of future greenhouse gas and sulphur emissions (Parry et al., 2004; Nakićenović et al., 2000) and a basis of a set of 40 sce- narios. For the purposes of this study, three of these scenarios were chosen based on the hydrologic simulation of the WATCH and ENSEMBLES climate model input data through continuous rain- fall–runoff modelling. In brief, the B1 storyline and scenario family describes a confluent world that promotes sustainable develop- ment on a global scale, rapidly shifting towards a service and information economy based on clean and resource-efficient tech- nologies. On the contrary, the A2 storyline and scenario family, which is part of Scenario A or Business-as-Usual (BAU), describes a weakly globalized world with continuously increasing popula- tion. Economic development is primarily regionally oriented and per capita economic growth and technological change more frag- mented and slower. Lying somewhere between the above, the A1B, a subgroup of the A1 scenario family, describes a future world of very rapid economic growth, sustainable global population, and the rapid introduction of new and more efficient technologies lead- ing to a balance across fossil intensive and non-fossil energy sources.

The developed demand and infrastructure scenarios are based on the water practices of the local water management authority (Papagrigoriou et al., 2001) and have two basic components: D, the level of demand of water for the various uses, and S the pro- jected water supply potential derived from the combination of the technical infrastructure and the projected water availability (dams, barrages, groundwater abstractions, etc.). For each compo- nent, two alternative storylines were established focusing on the analysis periods (2000–2050 and 2050–2100) and the different cli- mate models projections for the various emission scenarios.

Regarding the demand component, the Existing situation of water demand (D1) storyline includes the current level of demand (2000), based on data that was assembled from irrigation and water supply authorities in the Island. The estimation of irrigation demand is based on a detailed analysis of the irrigated areas per municipality and the potential of each basins for groundwater abstractions. The Future water demand (D2) storyline depicts a real- istic future water demand as it is outlined by the local irrigation facilities development programs, the future irrigation extend plans, the estimate of future population trends and a projected increase of tourism activities.

In the supply component, the Business as usual (S1) storyline in- cludes all the existing works of exploitation of water resources, as they have been recorded in the frame of the ‘‘Integrated water re- sources management of Crete study’’ (Papagrigoriou et al., 2001) including dams, barrages, groundwater abstractions, etc. Finally, the Future technical infrastructure (S2) storyline includes the future technical infrastructure based on the proposed technical work of exploitation of water resources that has been evaluated as mature

Fig. 1. Water resources assessment framework.

148 A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158

enough for construction according to the local water authorities. The future infrastructure action also includes new groundwater abstraction wells in certain regions (very limited in number) where certain quantities can be withdrawn without endangerment of quality of groundwater resources.

2.3. Bias correction of climate model data

Statistical bias corrected hydrological variables significantly im- prove the ability of streamflow simulation when compared to raw outputs (Sharma et al., 2010). The method to adjust biases used in the present study, corrects the frequency and intensity of modelled precipitation relative to a target station (Ines and Hansen, 2006) in a two-step procedure for each one of the 12 months of the year. A calibrated threshold ~x0 is used to discard rainfall values when mod- elled precipitation frequency Fbmodel is greater than the observed frequency Fobs. The correction of intensity involves the mapping of modelled precipitation intensity distribution to the observed for values above the calibrated threshold using a predefined distri- bution for both datasets. The adjustment procedure of frequency truncates the empirical distribution of the raw model precipitation above a ~xi threshold dataset, so that mean frequency of precipita- tion above this threshold matches observed precipitation fre- quency. This threshold ~xi for each calendar month I is estimated according to the following equation:

~xi ¼ F�1bmodelðFobsð~x0ÞÞ ð1Þ

where F(�) and F�1(�) are the cumulative distribution function (CDF) and its inverse for either modelled or observed values. In the final step corrected modelled precipitation x0 of day i of month I is esti- mated by substituting the fitted CDFs into the following equation:

x0i ¼ F�1I;obsðFI;bmodelðxiÞÞ; xi P ~xi 0; xi < ~xi

( ð2Þ

This same method was used to adjust modelled temperature biases but without the use of frequency distribution correction and no truncation to the modelled temperatures.

2.4. Hydrological modelling

The SAC-SMA Sacramento model is a lumped continuous rain- fall–runoff model that can estimate stream flow from precipitation P and potential evapotranspiration ET0 records (Podger, 2004; Tsanis and Apostolaki, 2009). The water balance simulation of a watershed is based on soil moisture accounting. The soil moisture storage increases by rainfall P and reduces by actual evapotranspi- ration ETa, direct runoff from the impervious area, surface runoff Qsur from the pervious area and baseflow that add up to total flow Q and infiltration I between the upper and lower zone that dis- charges at a slower rate (slow runoff). The size and relative wet- ness of the storage determines the depth of rainfall absorbed, actual evapotranspiration, and the amount of water moving verti- cally or laterally out of the store. These processes are described by 16 parameters (Vrugt et al., 2006) that need to be determined by the user or an optimization process using a suitable objective func- tion. Because of the number and nature of those parameters, objec- tive function responses can contain multiple optima. In order to apply the SAC-SMA hydrological model at basin scale, the input variables of P and ET0 are first interpolated using the Inverse Dis- tance Weighting (IDW) method (Wei and McGuinness, 1973) and then averaged over the area of each respective basin.

In order to eliminate subjective judgment of an expert user in model parameter selection, calibration was based on an applica- tion of Genetic Algorithms (GAs). GAs are adaptive random search algorithms that mimic the principal of selection and evolution of the fittest in a natural system. Given a defined search space, GAs have a globally oriented searching approach and are thus poten- tially useful in solving complex optimization problems (Wang, 1998) with one or more objective functions towards a Pareto solu- tion, as discussed by van Werkhoven et al. (2009). While resulting hydrological model parameter estimates may lack direct physical sense, they often represent ‘‘effective’’ watershed properties that compensate for lamped model inadequacies (Vrugt et al., 2006). In this study, a set of three objective functions was used: (a) the Nash–Sutcliffe (Nash and Sutcliffe, 1970) model accuracy statistic calculated on flows (NSEQ) given by:

A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158 149

NSEQ ¼ 1 � PT

t¼1 Q t � Q tm

� �2 PT

t¼1ðQ t � QÞ2

ð3Þ

where Qt is the observed and Q tm the modelled discharge at time t and Q the average discharge of the dataset, (b) the Nash–Sutcliffe logarithmic transformed flows NSElnQ which shows better efficiency for low flows (Pushpalatha et al., 2012) and (c) the R2 efficiency cri- terion, given by:

R2 ¼ N P

Q t Q tm � P

Q t � � P

Q tm � �

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi N P ðQ tÞ2 �

P ðQ tÞ2

q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi N P ðQ tmÞ

2 � P ðQ tmÞ

2 q

0 B@

1 CA

2

ð4Þ

representing the percentage of the initial uncertainty explained by the model. NSEQ and NSElnQ range from �1 to 1 while R2 ranges from 0 to 1. For all above criteria, the perfect fit between observed and calculated values, which is unlikely to occur, is the unity.

River basin scale simulation studies require long-term continu- ous streamflow records, which are unavailable in ungauged basins. As a rule, the calibrated parameter values of a specific catchment should not be used in other catchments, unless the reliability of this transfer can be assessed (Burnash, 1995). Several methods have been used in order to regionalize model parameters including applications of spatial proximity (Vandewiele and Elias, 1995; Merz and Blöschl, 2004; Parajka et al., 2005), flow duration curves (Yu and Yang, 2000), basin similarity (Parajka et al., 2005), neural networks (Heuvelmans et al., 2006), and regional calibration meth- ods (Fernandez et al., 2000; Hundecha and Bárdossy, 2004). In this study, the global parameter mean (e.g. Kim and Kaluarachchi, 2008; Jin et al., 2009) for all basins is used to make estimations on all 110 Cretan basins.

2.5. Limitations

The complexity of climate change impact studies is enhanced from the uncertainty in climate change modelling and the long planning horizons. Capturing the inter-annual weather variability and extremes rather than the statistical properties of climate still presents a challenge for models (e.g. Wilks and Wilby, 1999; Sri- kanthan and McMahon, 2001; Kyselý and Dubrovský, 2005) thus hindering the accuracy of a water availability estimation. A step further, the main shortcoming of statistical bias correction meth- ods is the assumption of stationarity implying that the statistical properties of a time-series remain constant through time. As a re- sult, the transfer functions estimated using the historical climate conditions is assumed to remain valid for correcting biases in fu- ture precipitation or temperature time-series (Rojas et al., 2011). The same limitation holds for hydrologic modelling of bias cor- rected future climate conditions using a model calibrated with his- torical time-series as, for example, future parameters and states may shift unpredictably due to temperature increase. Hydrological model parameter uncertainty resulting from both calibration and regionalization should also be carefully investigated and studied further. Moreover, demand and supply estimates of existing condi- tions for the present study are based on the latest large scale cen- sus in the frame of Papagrigoriou et al. (2001), by request from the Water Authority of Crete. As such, the estimates presented herein have inherent limitations, enhanced by the lack of information regarding technological advancement that may directly affect water treatment and therefore change the water balance. Finally, the modelling approach used herein ignores the dynamic interac- tions between elements of social–economic–environmental sys- tem and future is treated in a deterministic way.

3. Case study

The island of Crete occupies the southern part of the country of Greece (Fig. 2). With an area of 8265 km2, Crete covers almost 6.3% of the area of Greece. The mean elevation is 482 m ranging from sea level to 2450 m and the average slope 228 m/km with the topography fracturing into small catchments with ephemeral streams and karst geology. Crete has a typical Mediterranean is- land environment with about 53% of the annual precipitation occurring in the winter, 23% during autumn and 20% during spring while there is negligible rainfall during summer (Koutroulis and Tsanis, 2010; Naoum and Tsanis, 2004). The average precipitation for a normal year in the island of Crete is approximately 934 mm or 7697 Mm3 (Tsanis and Naoum, 2003). This in addition to non- uniform precipitation distribution in the island (a reduction of al- most 300 mm from the west to the east part of the island and a strong orographic effect) makes the water availability a very small but crucial portion of the total supply (Tsanis et al., 2011).

In order to calibrate the SAC-SMA hydrological model at basin scale input variables were collected from 53 rainfall and 15 tem- perature stations for the period 1970–1999 at monthly time step. While P could be used directly from measurements, values of ET0 were estimated from temperature using the Blaney–Criddle equa- tion (Blaney and Criddle, 1962) in order to have better comparison with future climatic data. Variables were spatially interpolated using the Inverse Distance Weighting (IDW) method (Wei and McGuinness, 1973) and then averaged over the area of each respective basin. The optimization process based on genetic algo- rithms was applied to 17 gauged basins (Fig. 2) where surface run- off Qsur measurements were available at a monthly time step. The calibration yielded satisfactory results for 15 out of the 17 basins (Table 1) with the R2, NSEQ and ranging from 0.52 to 0.90, 0.51 to 0.90 respectively. The NSElnQ of the calibration ranged from 0.20 to 0.81 showing that the model was somewhat less efficient in low flows for the calibrated basins. In two basins, Bramianos and Agios Vasillios, the calibrations results were not satisfactory and they were rejected from the rest of the methodology. A global mean of the SAC-SMA model parameters yielded a total R2 of 0.81 in the comparison of observed and modelled surface flow and was therefore considered adequate to make estimations for the entire 110 basins on the Island.

Following the standardization with the respective basin areas, the annual water balance breaks down to 68–76% evapotranspira- tion, 14–17% infiltration and 10–15% runoff (Table 2). Total water uses in the region in 2000 amounted to 420 Mm3 (Papagrigoriou et al., 2001), approximately 5.5% of the precipitation of a normal year and 16% of the total water potential. An average of 65% of the total water use is supplied by groundwater exploitation while the remaining 35% is obtained from winter spring and stream dis- charges. Of this, 16% is used for domestic, tourist, and industrial uses, 3% for livestock and a vast 81% for irrigated agriculture on less than 30% of the total cultivated land, using mainly ground water in drip irrigation methods. Irrigation and tourism create a marked seasonal pattern in water demand with heavy summer loads, as the annual volume of water abstracted exceeds 50% of the average annual runoff and 35% of the groundwater potential.

In the frame of the EU FP6 project ‘‘WATCH’’ (WATer and global CHange’’), ran from 2007 to 2011, a range of datasets were gener- ated which are publicly available (http://www.eu-watch.org/ data_availability) including 20th Century WATCH Forcing Data (WFD), the 21st Century WATCH Driving Data (WDD). Six datasets available for the period 2001–2100 based on three different GCMs, ECHAM5 (Jungclaus et al., 2006; Roeckner et al., 2003), CNRM (Deque et al., 1994; Deque and Piedelievre, 1995), and IPSL (Hourdin et al., 2006) and two emission scenarios (A2 and B1),

Fig. 2. Location of Crete Island, delineated watersheds and the mesh of the ENSEMBLES RCMs and WATCH climate models data (WFD). Red areas represent gauged watersheds at the outlets. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1 Selected hydrologic characteristics and SAC-SMA calibration results for 15 gauged basins on the island of Crete.

Area (km2) Average annual P (mm) Average annual ET0 (mm) Average runoff Qsur (mm) R 2 NSEQ NSElnQ

Patelis 60.5 804.0 1460.0 102.3 0.87 0.87 0.73 Kalamafkianos 35.2 733.1 1516.8 176.9 0.78 0.80 0.75 Myrtos 97.8 759.6 1501.1 63.9 0.63 0.63 0.59 Arvi 26.1 723.2 1471.7 240.7 0.79 0.79 0.64 Aposelemis 201.0 922.2 1383.1 31.0 0.69 0.69 0.53 Anapodaris 506.1 670.3 1399.7 11.8 0.89 0.89 0.69 Giofyros 158.4 802.8 1414.4 37.2 0.80 0.79 0.75 Gazanos 186.7 884.5 1399.1 31.3 0.71 0.70 0.50 Koutsoulidis 121.0 924.8 1427.8 48.3 0.86 0.86 0.80 Geropotamos 396.5 680.9 1423.6 14.8 0.90 0.89 0.21 Platys 203.2 925.9 1450.6 28.9 0.90 0.90 0.69 Prassanos 88.2 1100.6 1451.3 67.0 0.64 0.64 0.80 Kakodikianos 77.4 1292.8 1482.4 76.8 0.52 0.51 0.48 Sebroniotis 27.5 1285.3 1441.9 217.2 0.83 0.82 0.82 Roumatianos 22.1 1318.0 1432.9 271.9 0.66 0.65 0.70

Table 2 Observed and simulated annual inputs and outputs in the hydrological budget of the island of Crete during normal, humid and dry years using Sacramento for the period 1970– 1999.

Annual P Annual ET0 Runoff Qsur Infiltration I

Observed average (mm) 934 635–710 93–140 131–159 Observed fraction of precipitation (%) 68–76% 10–15% 14–17% Observed average (1000 Mm3) 7.70 0.77–1.56 5.23–5.85 1.08–1.30 Simulated normal year (1978) (1000 Mm3) 7.95 5.84 (73%) 0.87 (11%) 1.25 (16%) Simulated wet year (1981) (1000 Mm3) 8.88 5.82 (66%) 1.39 (16%) 1.67 (19%) Simulated dry year (1985) (1000 Mm3) 5.63 3.96 (70%) 0.73 (13%) 0.94 (17%)

Table 3 List of ensembles regional climate models (RCMs).

No. Institute RCM Driving GCM References

1 ETH CLM HadCM Jaeger et al. (2008) 2 ICTP RegCM ECHAM5-r3 Giorgi and Mearns (1999) 3 KNMI RACMO2 ECHAM5-r3 van Meijgaard et al. (2008) 4 METOHC HadRM3Q0 HadCM3Q0 Collins et al. (2010) 5 METOHC HadRM3Q3 HadCM3Q3 Collins et al. (2010) 6 METOHC HadRM3Q16 HadCM3Q16 Collins et al. (2010) 7 C4I RCA3 HadCM3Q16 Kjellström et al. (2005) 8 MPI REMO ECHAM5-r3 Jacob (2001) 9 SMHI RCA BCM Kjellström et al. (2005)

10 DMI HIRHAM ARPEGE Christensen et al. (2006)

150 A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158

interpolated at a spatial resolution of 0.5� (Fig. 2), were used for analysis. For each GCM, three datasets were also available from the control period 1960–2000.

Furthermore, results of an ensemble of RCMs are used focusing on the Island of Crete (Tsanis et al., 2011). Simulations from 10 RCMs, here named ENSEMBLES (Jacob et al., 2007, 2008; Roeckner et al., 2003; van der Linden and Mitchell, 2009) are performed over the European continent at a horizontal resolution of about 25 km (Fig. 2). The RCMs’ lateral boundary conditions are provided by se- ven GCMs for the period 1951–2100 (Table 3). Simulations are forced using observed GHG greenhouse gas and aerosol concentra- tions until 2000 and SRES A1B concentrations scenario until 2100.

A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158 151

The RCMs were chosen based on their spatial and temporal extent as well as their ability to simulate the present climate. RCM spe- cific weights are then calculated in order to construct the optimal ensemble output for precipitation and temperature at a monthly time step and at a watershed level. Weights are calculated accord- ing to the combination of two metrics for both precipitation and temperature time-series. The first metric is a composition of five functions related to probability density distribution match at specific percentiles of the cumulative distribution functions (Christensen et al., 2010). Each of the functions takes into account different aspects of the behaviour of the model parameter (P, T), and thus, their combination gives a complete picture of the model skill. The second metric is related to the ability of the model to rep- resent the annual cycle, that is believed as a good indication of the quality of atmospheric processes description affecting the overall model performance. This metric is based on the so called Taylor diagram (Taylor, 2001) which is used for presenting the data in terms of RMSE (Root Mean Square Error), standard deviation and correlation. Monthly weights per model were constructed from the combination of the above metrics for P and T. The overall weighting effect on historic skill for ENSEMBLES dataset is illus- trated in Tsanis et al. (2011).

Climate models output (precipitation and temperature) is then bias corrected against daily time-series data obtained from 53 rain- fall and 15 temperature stations for the period 1970–2000 and interpolated at basin scale. For precipitation bias correction, the threshold of observed precipitation amount ~x of model daily pre- cipitation is set to 0.1 mm. For the correction of precipitation intensity the gamma distribution is applied to fit the truncated dai- ly modelled and observed precipitation data after Ines and Hansen (2006). Normal instead of gamma distribution is used to map tem- perature distribution (Grillakis et al., 2011). Detailed schematic representations of normal PDFs of the bias adjusted results of all WATCH models and Ensembles members for the two time slices

Fig. 3. Normal PDFs representing the average bias adjusted WATCH models and weigh precipitation (left panels) and (b) average annual temperature (right panels).

under the three emission scenarios, for annual precipitation and temperature, as well as the seasonality shift of these hydro-cli- matic variables are presented in Figs. 3 and 4, respectively. Eventu- ally, monthly aggregated time-series of bias adjusted parameters used to drive the projections of the hydrological model.

For the (D1) storyline, the current total demand is estimated at 535.7 Mm3/year, from which 458.4 Mm3 are consumed in irriga- tion and 77.3 Mm3 in general water supply consisting of household supply and other minor uses (industrial, livestock-farming, etc.). The total existing water demand in storyline D2 is estimated at 775.8 Mm3/year, from which 670.8 Mm3 come from irrigation due to increased cultivated areas and the remaining 105 Mm3 refer to general water supply. Compared to the D1 storyline, the increase of irrigation demand is 44.8%, while the increase of demand of gen- eral water supply is 36%. For the above two conditions, the theoret- ical water demand also represents a ‘‘desirable’’ level of irrigation, taking irrigation system losses into account. In the Business as usual (S1) storyline, a total of 302.0 Mm3 are supplied for irrigation and 69.7 Mm3 for general water supply. An additional 42.8 Mm3, orig- inating from groundwater abstractions, are supplied to the system, shaping the total supply to 414.6 Mm3 (under normal hydrological conditions). The total inflow in the system of water resources for this scenario amounts to 857.0 Mm3/year, including groundwater abstractions, spring discharges and basin outflows under current exploitation conditions or included in planed exploitation. For the S2 storyline, a total of 398.3 Mm3 are supplied for irrigation (59.4% of the demand), 100.4 Mm3 for general water supply, and an additional 35.4 Mm3 from groundwater abstractions are sup- plied to the system, shaping the total supply to 534.2 Mm3 (under normal hydrological conditions). The total inflow in the system of water resources for this scenario amounts in 902.5 Mm3/year as a result of increase of certain groundwater withdrawals (308.0 Mm3/ year) and better exploitation of certain surface flows. The quantity from the annual winter surplus from spring and stream discharges

ted bias-adjusted results of all ENSEMBLES members per time slice for (a) annual

Fig. 4. Seasonality shift of monthly (a) monthly precipitation (left panels) and (b) monthly temperature (right panels).

Table 4 Projected (2000–2050) climate models and hydrological model results for annual precipitation, temperature and availability for the island of Crete, under three emission scenarios.

Comparison to control climate (1970–2000)

IPSL ECHAM CNCM ENSEMBLES IPSL ECHAM CNCM ENSEMBLES

Precipitation (mm) OBS (1970–2000) 934 2000–2050 B1 961 943 931 103% 101% 100%

A2 911 939 908 98% 101% 97% A1B 807 86%

Temperature (�C) OBS (1970–2000) 16.3 2000–2050 B1 18.3 17.4 17.6 2.0 1.1 1.3

A2 18.0 17.3 17.7 1.7 1.0 1.4 A1B 18.0 1.7

Availability (mm) OBS (1970–2000) 299 2000–2050 B1 329 333 310 110% 111% 104%

A2 294 298 296 98% 100% 99% A1B 214 72%

152 A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158

that are not exploited and runoff to the sea amounts to 434.0 Mm3. Finally, a direct connection of the decrease (%) of projected water availability (from Tables 4 and 6) to the supply formulated from the two technical infrastructure scenarios (S1 and S2) depicts the estimation of the impact of climate change to the projected supply potential. Table 5 includes results for the projected annual supply potential according to hydrological modelling for the island of Crete, under the 2 infrastructure (S1 and S2) and the three emis- sion scenarios.

4. Results and discussion

4.1. Historic hydrologic regime (1970–1999)

Continuous hydrological modelling for the 1970–1999 period resulted to the hydrologic regime of the reference period. Stan- dardizing the balances with the respective basin areas, runoff is

10–15% of the approximately 934 mm or 7697 Mm3 of long term annual precipitation that falls on the island of Crete, almost equal with the fraction of water that infiltrates towards slow runoff (14–17%), whereas 68–76% of precipitation goes toward evapo- transpiration (Table 2). The monthly estimated actual evapotrans- piration, runoff and infiltration per basin for the entire island for the period 1970–1999 are shown in Fig. 5. As expected, the esti- mated runoff and infiltration is high during the winter while ac- tual evapotranspiration increases during spring. The hydrological balance of the Island of Crete under normal, wet and dry condi- tions are included in Table 2. Actual evapotranspiration on a dry year consumes a significant amount of precipitation but not as much as on a wet year when there is more water to be evap- orated and plants are able to readily transpire. Also, on a dry year, the total volume of water that discharges is less (about half than on a wet year) but represents a larger fraction of the hydrological balance.

Table 6 Projected (2050–2100) climate models and hydrological model results for annual precipitation, temperature and availability for the island of Crete, under three emission scenarios.

Comparison to control climate (1970–2000)

IPSL ECHAM CNCM ENSEMBLES IPSL ECHAM CNCM ENSEMBLES

Precipitation (mm) OBS (1970–2000) 934 2050–2100 B1 932 827 935 100% 89% 100%

A2 769 673 764 82% 72% 82% A1B 688 74%

Temperature (�C) OBS (1970–2000) 16.3 2050–2100 B1 20.0 19.3 18.9 3.7 3.0 2.6

A2 21.5 20.2 20.7 5.2 3.9 4.4 A1B 20.9 4.6

Availability (mm) OBS (1970–2000) 299 2050–2100 B1 290 233 286 97% 78% 96%

A2 198 155 189 66% 52% 63% A1B 144 48%

Table 5 Projected annual supply potential according to hydrological modelling results for the island of Crete, under the two infrastructure (S1 and S2) and the three emission scenarios.

Supply S1 (Mm3) Supply S2 (Mm3)

IPSL ECHAM CNCM ENSEMBLES Average IPSL ECHAM CNCM ENSEMBLES Average

2000–2050 B1 456 462 430 449 588 595 554 579 A2 408 413 410 410 525 532 529 529 A1B 297 297 382 382

2050–2100 B1 402 323 397 374 518 416 511 482 A2 275 215 262 251 354 277 338 323 A1B 200 200 257 257

Fig. 5. Monthly estimated (a) actual ET, (b) Runoff and (c) Infiltration for the period 1970–1999. Solid boxes signify values from 1st to 3rd quantile while whiskers extend for the zero to the 4th quantile.

A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158 153

4.2. Projected water resources availability

4.2.1. Period 2000–2050, emission scenario B1 For the period 2000–2050 and according to WATCH modelling

results (Table 4 and Fig. 3) for emission scenario B1, projected average annual precipitation is expected to be similar to that of the control period (1970–2000). The three GCMs (IPLS, ECHAM, CNCM) indicate that average annual precipitation could slightly in- crease (1%) to an average of 945 mm (from 931 mm to 961 mm depending on the model). The current long term average annual temperature (16.3 �C) could increase by an average of 1.5 �C (from 1.1 �C to 2.0 �C). Hydrological modelling shows that during this

period we can expect a decreasing trend of water availability of 2.6 mm per year (Fig. 6a) which nevertheless stabilizes the abrupt change during the control period (1970–2000) leading to an aver- age increase of water availability between 4% and 11% (Fig. 7). Compared the control period, during infrastructure scenarios S1 and S2, an additional 34 Mm3 (15–47 Mm3) and 45 Mm3 (20– 54 Mm3), respectively, are expected to flow into the system (Ta- ble 5). Furthermore, the combination of D1 and S1 results to an additional 35 Mm3 available water resources, thus reducing the deficit from 23% to 16% (Table 7). In the case of D1 and S2 an excess of 43 Mm3 is estimated. In the case of D2 and S1, a severe lack of water resources by 327 Mm3 (�42%) could be observed, compared

Fig. 6. Black lines depict the historical simulated annual availability based on observations. Red dotted lines correspond to historical and projected multi-model mean annual availability. Trend lines represent the average annual availability slopes for the observed period (in black) and for the ensemble projections (in red) under (a) B1, (b) A2 and (c) A1B emission scenarios. Grey areas indicate the amplitude of historical and projected availability of multi-model RCM and GCM ensembles and hydrological modelling. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

154 A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158

to the future demand. A similar status of water stress, with a deficit of 197 Mm3 could affect the region under the assumption of D2 and S2 scenarios.

4.2.2. Period 2000–2050, emission scenario A2 For the period 2000–2050 and according to WATCH modelling

results (Table 4) for emission scenario A2, projected average an- nual precipitation is expected to be similar to that of the control period. The three GCMs indicate that average annual precipitation could slightly decrease (2%) to an average of 919 mm (from 908 mm) (Fig. 3). Average annual temperature could increase by an average of 1.4 �C (from 1.0 �C to 1.7 �C). Hydrological modelling shows that projected average annual supply is expected to have a slight decrease by 1% (Table 4, Fig. 7) or 2.1 mm per year (Fig. 6b) over the control period. For both S1 and S2 scenarios, 5 Mm3 less

water (from 2 Mm3 to 9 Mm3) is expected to flow into the system (Table 5), compared to the control period (1970–2000). For the A2 scenario, the combination of D1 and S1 results to a water status similar to that of the control period (1970–2000), with a deficit of 24% (Table 7). In the case of D1 and S2 a slight deficit of 7 Mm3 could be observed. For the combination of D2 and S1, a se- vere lack of water resources by 366 Mm3 (�47%) could be ob- served. A similar status of severe water stress, with a deficit of 247 Mm3 could affect the region under the assumption of D2 and S2.

4.2.3. Period 2000–2050, emission scenario A1B According to the Ensembles results from 10 RCMs (Table 4), and

based to the control climatology, future projections show an aver- age decrease of 14% in average annual rainfall (807 mm) and an

Fig. 7. Normal PDFs representing the average bias adjusted WATCH models and weighted bias-adjusted results of all ENSEMBLES members per time slice for average annual availability (left panels). Seasonality shift of monthly availability (right panels). Three models for WATCH and 10 for ENSEMBLES used to construct the corresponding PDFs.

A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158 155

average temperature increase of 1.7 �C, leading to a water avail- ability decrease by 28%. For the period 2000–2050 and according to Ensembles modelling results (Table 4) for emission scenario A1B, projected average annual supply is expected to decrease by 28% (Fig. 7), nevertheless without any significant trend (Fig. 6c). For the infrastructure scenarios S1 and S2, a decrease of 118 Mm3 and 152 Mm3, respectively, is expected to flow into the

Table 7 Estimated excess-deficit of the water balance from the combination of all components.

Scenario Period Emission scenario

Hydrologic regime

Demand Infrastructure Est. d (Mm

1 BAU 1970– 2000

Normal D1 S1 536

1 2000– 2050

B1 WATCH D1 S1 536 2 D1 S2 536 3 D2 S1 776 4 D2 S2 776 5 A2 WATCH D1 S1 536 6 D1 S2 536 7 D2 S1 776 8 D2 S2 776 9 A1B ENSEMBLES D1 S1 536

10 D1 S2 536 11 D2 S1 776 12 D2 S2 776 13 2050–

2100 B1 WATCH D1 S1 536

14 D1 S2 536 15 D2 S1 776 16 D2 S2 776 17 A2 WATCH D1 S1 536 18 D1 S2 536 19 D2 S1 776 20 D2 S2 776 21 A1B ENSEMBLES D1 S1 536 22 D1 S2 536 23 D2 S1 776 24 D2 S2 776

system, denoting a more severe situation (Table 5). According to scenario A1B based on the Ensembles modelling results, and in contrast with WATCH climate modelling results, for the period 2000–2050, the combination of D1 and S1 (BAU) indicates an ex- treme water deficit of 239 Mm3 (�45%) (Table 7). In the cases of D1 and S2 the deficit is less than that of the previous case, but still emerging. A deficit of 154 Mm3 (�29%) could be observed. For the

emand 3)

Est. supply (Mm3)

Est. deficit (Mm3)

Est. excess or Deficit (%)

Difference from BAU (Mm3)

414 �122 �23 0

449 �87 �16 35 579 43 8 165 449 �327 �42 �205 579 �197 �25 �75 410 �126 �24 �4 529 �7 �1 115 410 �366 �47 �244 529 �247 �32 �125 297 �239 �45 �117 382 �154 �29 �32 297 �479 �62 �357 382 �394 �51 �272 374 �162 �30 �40 482 �54 �10 68 374 �402 �52 �280 482 �294 �38 �172 251 �285 �53 �163 323 �213 �40 �91 251 �525 �68 �403 323 �453 �58 �331 200 �336 �63 �214 257 �279 �52 �157 200 �576 �74 �454 257 �519 �67 �397

156 A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158

combination of D2 and S1, a severe lack of water resources by 479 Mm3 (�62%) could be observed. A similar status of severe water stress, with a deficit of 394 Mm3 could affect the region un- der the assumption of D2 and S2.

4.2.4. Period 2050–2100, emission scenario B1 According to WATCH modelling results (Table 6) for emission

scenario B1, projected average annual precipitation is expected to decrease slightly (4%, 989 mm) in comparison to the control period (1970–2000). Average annual temperature could increase by an average of 3.1 �C (varying from 2.6 �C to 3.7 �C depending on the model). Hydrological modelling shows that during this period we could expect a stabilized 10% less available water resources, vary- ing from 3% to 22%, depending on the model (Fig. 7) but without significant trends (Fig. 6a). For the period 2000–2050 and accord- ing to WATCH modelling results (Table 5) for emission scenario B1, projected average annual supply is expected to be 10% less. For the infrastructure scenarios S1 and S2, an average decrease of 41 Mm3 (a decrease from 13 Mm3 to 92 Mm3, depending on the model) and 52 Mm3 (from 16 Mm3 to 118 Mm3 less, depending on the model), respectively, is expected to flow into the system, in comparison to the control period (Table 7).

4.2.5. Period 2050–2100, emission scenario A2 WATCH modelling results (Table 6) under emission scenario A2,

show more severe changes to hydro-climatic variables. Projected average annual precipitation is expected to drop to 735 mm (vary- ing from 673 mm to 769 mm depending on the model), 21% less in comparison to the period 1970–2000. Average annual temperature could increase by an average of 4.5 �C (from 3.9 �C to 5.4 �C depending on the model). Under these changes, average water availability could decrease by 40% (from 33% to 48% less, depend- ing on the model) compared to past climate conditions (Table 6). For the period 2000–2050 and according to WATCH modelling re- sults (Table 5) for emission scenario A2, projected average annual supply is expected to decrease by 40%. It is important to note that out of all scenarios, this appears to be the most alarming, showing an ever decreasing water availability trend (Fig. 6b). For the infra- structure scenarios S1 and S2, an average decrease of 164 Mm3

(from 153 Mm3 to 200 Mm3, depending on the model) and 211 Mm3 (from 180 Mm3 to 257 Mm3, depending on the model), respectively, is expected to flow into the system, in comparison to the control period (1970–2000) (Table 7).

4.2.6. Period 2050–2100, emission scenario A1B According to the Ensembles results (Table 6), in comparison to

the control climatology, future projections show an mean decrease of average annual rainfall by 26%, reaching 688 mm, and an aver- age temperature increase of 4.6 �C, leading to a water availability decrease of 48%. For the period 2050–2100 and according to Ensembles modelling results (Table 5) for emission scenario A1B, projected average annual supply is expected to be decrease by 52%. While the trend of availability is not significant (Fig. 6b) hydrological modelling predicts that by the end of the century, an- nual water availability will be as low as 80 mm/year, which is a re- cord value for the studied dataset. For the infrastructure scenarios S1 and S2, the flow into the system is expected to decrease by 215 Mm3 and 277 Mm3, respectively, denoting the most severe sit- uation among the analyzed scenarios (Table 7).

5. Conclusions

Despite limitations and uncertainties, this study presents a wide range of draft estimates and results, providing water re- sources management community with a glimpse into a very plau-

sible future where the quantitative impact of climate change on water availability can be substantial, especially in a Mediterranean island like Crete. RCM model results (Ensembles) project increased precipitation reduction and consequently show increased water insufficiency compared to GCM projections (WATCH). Among WATCH GCM results, ECHAM model projects higher water re- sources reduction compared to IPSL and CNCM results. Overall, a robust signal of water insufficiency is projected for all the combi- nations of emission, demand and infrastructure scenarios, with the estimated deficit ranging from 10% to 74%. Assuming that all climate scenarios are equally probable, average water availability is expected to drop from 93% during 2000–2050 to a devastating 70% of the observed average (Fig. 7), which is already insufficient to cover current demand.

The outcomes of the above analysis are useful for understand- ing the role and consequently the priority of certain water re- sources related infrastructure development. Out of all discussed demand and supply scenarios, the only window for improving the current status is given under scenario B1 for the period 2000–2050, with the assumption that the demand during this per- iod does not increase (an outcome against development strategies) and water resources infrastructure improves. Even as such, assum- ing demand scenarios equally probable or an average rise of de- mand by 20%, it becomes evident that water resources management should consider infrastructure and adaptation strate- gies to mitigate risks of the forecasted deficit.

For the combination of D2 and S2 and for a normal hydrological year, it is observed that the increase of irrigated cultivated areas causes an increase in water demand and despite the increase of allocated quantity for irrigation by roughly 100 Mm3/year, the ser- vice of water demand is lower by 6.5%. This is an important conclu- sion which should enjoy further attention with regard to growth policies as current policy for new water resources infrastructures is very closely related to the growth of new irrigated areas. This leads to a great increase of irrigation demand level, as this has been determined in the present study, precluding the practicability of this scenario. The conclusion is that an alternative policy of devel- opment of new infrastructures should be adapted. This policy should not only give priority to the increase of irrigated areas but also promote a more sustainable irrigation practice for existing and new agricultural land.

The EU Water Framework Directive and the policies of droughts and adaptation to climate change (CEC, 2008, 2009), provides a specific framework of objectives, principals, definitions and mea- sures to adopt, for assessing the impact of climate change on water resources. This enables decision-makers to develop and constantly review water management plans. According to the local water authorities, the overwhelming majority of small scale works (bar- rages, small dams) are projects of purely local character regarding their incorporation in the integrated system of water resources, thus yielding negligible effects at municipal or greater level. There- fore, they should be evaluated only based on technical, economi- cally and social criteria of the local region in which they are sited and which they will serve. Regarding large scale water resources constructions, which have a wider-than-regional character, some of them are presented as additional supply for covering existing demand (e.g. the dam of Valsamiotis under construction at Chania district in Western Crete), while others have a capacity exceeding present demand and are consequently be associated with agricul- tural expansion (e.g. the dams of Roumatianos and Derianos in Chania, the dam of Plakiotissa in Heraklion and Amariou dam in Rethymno).

Despite this growth of infrastructure aiming to store winter and spring stream flows, the inadequacy to control large outflow quan- tities that runoff or infiltrate to the sea remains a problem. Adap- tation is unlikely to be facilitated through the introduction of

A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158 157

new and separate policies, but rather by the revision of existing policies that currently undermine adaptation and the strengthen- ing of policies that currently promote it (Iglesias et al., 2011). Such strategies of adaptation to consider include wastewater recycling and reuse that are estimated to lead to water savings of up to 5% of the total irrigation water of Crete (Tsagarakis et al., 2004; Agraf- ioti and Diamadopoulos, 2012). In view of the new results pre- sented here, local long-term water resources management plans could be updated including the procedure on data analysis and output interpretation. The current financial stress and the continu- ously reduced national investment programmes call for low cost, short and long term water management strategies in order to tackle the climate induced changes in water resources.

Acknowledgments

The financial support of this work has been provided by the European Commission through the WATCH FP6, COMBINE FP7, IM- PACT2C FP7 and ECLISE FP7 projects. The ENSEMBLES data used in this work was funded by the EU FP6 Integrated Project ENSEMBLES (Contract Number 505539) whose support is gratefully acknowl- edged. We also thank the three anonymous reviewers whose com- ments have improved the paper.

References

Agrafioti, E., Diamadopoulos, E., 2012. A strategic plan for reuse of treated municipal wastewater for crop irrigation on the Island of Crete. Agric. Water Manage. 105, 57–64.

Akhtar, M., Ahmad, N., Booij, M.J., 2008. The impact of climate change on the water resources of Hindukush–Karakorum–Himalaya region under different glacier coverage scenarios. J. Hydrol. 355, 14–163.

Alcamo, J., Floerke, M., Maerker, M., 2007. Future long-term changes in global water resources driven by socio-economic and climatic changes. Hydrol. Sci. 52 (2), 247–275.

Arnell, N.W., van Vuuren, D.P., Isaac, M., 2011. The implications of climate policy for the impacts of climate change on global water resources. Global Environ. Change 21 (2), 592–603.

Barnett, T.P., Pierce, D.W., 2009. Sustainable water deliveries from the Colorado River in a changing climate. Proc. Natl. Acad. Sci. USA. http://dx.doi.org/ 10.1073/pnas.0812762106.

Blaney, H.F., Criddle, W.D., 1962. Determining consumptive use and irrigation water requirements. US Dept. Agr. Agricultural Research Service Tech Bull, vol. 1275, 59p.

Burnash, R.J.C., 1995. The NWS River Forecast System – catchment modelling. In: Singh, V.P. (Ed.). Computer Models of Watershed Hydrology, pp. 311–366.

CEC, 2008. Commission of the European Communities, Water Scarcity and Droughts Expert Network: Drought Management Plan Report, Technical Report-2008-023 (19.12.08).

CEC, 2009. Commission of the European Communities, White Paper: Adapting to climate change: Towards a European framework for action, COM (2009) 147 final: Climate change and water, coasts and marine issues, SEC(2009) 386, Brussels (01.04.09).

Charlton, M.B., Arnell, N., 2011. Adapting to climate change impacts on water resources in England—an assessment of draft water resources management plans. Global Environ. Change 21 (1), 238–248.

Christensen, N.S., Lettenmaier, D.P., 2007. A multimodel ensemble approach to assessment of climate change impacts on the hydrology and water resources of the Colorado River Basin. Hydrol. Earth Syst. Sci. 11, 1417–1434.

Christensen, O.B., Drews, M., Christensen, J.H., Dethloff, K., Ketelsen, K., Hebestadt, I., Rinke, A., 2006. The HIRHAM Regional Climate Model Version 5 (b). Tech Rep 06-17. DMI, Copenhagen. ISSN 1399-1388

Christensen, J.H., Kjellström, E., Giorgi, F., Lenderink, G., Rummukainen, M., 2010. Weight assignment in regional climate models. Clim. Res. 44, 179–194.

Collins, M., Booth, B.B.B., Bhaskaran, B., Harris, G.R., Murphy, J.M., Sexton, D.M.H., Webb, M.J., 2010. Climate model errors, feedbacks and forcings: a comparison of perturbed physics and multi-model ensembles. Clim. Dyn. 36 (9–10), 1737– 1766.

Davies, E.G.R., Simonovic, S.P., 2011. Global water resources modelling with an integrated model of the social–economic–environmental system. Adv. Water Resource 34 (6), 684–700.

Deque, M., Piedelievre, J.P., 1995. High resolution climate simulation over Europe. Clim. Dyn. 11, 321–339.

Deque, M., Dreveton, C., Braun, A., Cariolle, D., 1994. The ARPEGE/IFS atmosphere model: a contribution to the French community climate modelling. Clim. Dyn. 10, 249–266.

Donta, A., Lange, M.A., Herrmann, A., 2005. Water on Mediterranean Islands: current conditions and prospects for sustainable management. Centre for Environment Research (CER), University of Muenster (Germany), vol. 5, 516p.

Fernandez, W., Vogel, R.M., Sankarasubramanian, A., 2000. Regional calibration of a watershed model. Hydrol. Sci. J. 45 (5), 689–707.

Fujihara, Y., Tanaka, K., Watanabe, T., Nagano, T., Kojiri, T., 2008. Assessing the impacts of climate change on the water resources of the Seyhan River Basin in Turkey: use of dynamically downscaled data for hydrologic simulations. J. Hydrol. 353 (1–2), 33–48.

García-Ruiz, J.M., López-Moreno, J.I., Vicente-Serrano, S.M., Lasanta-Martínez, T., Beguería, S., 2011. Mediterranean water resources in a global change scenario. Earth – Sci. Rev. 105 (3–4), 121–139.

Georgakakos, A.P., Yao, H., Kistenmacher, M., Georgakakos, K.P., Graham, N.E., Cheng, F.-Y., Spencer, C., Shamir, E., 2012. Value of adaptive water resources management in Northern California under climatic variability and change: reservoir management. J. Hydrol. 412–413, 34–46.

Giorgi, F., Mearns, L.O., 1999. Introduction to special section: regional climate modelling revisited. J. Geophys. Res. 104, 6335–6352.

Grillakis, M.G., Koutroulis, A.G., Tsanis, I.K., 2011. Climate change impact on the hydrology of Spencer Creek watershed in Southern Ontario, Canada. J. Hydrol. 409, 1–19.

Harding, R., Best, M., Blyth, E., Hagemann, S., Kabat, P., Tallaksen, L.M., Warnaars, T., Wiberg, D., Weedon, G.P., van Lanen, H., Ludwig, F., Haddeland, I., 2011. Preface to the Water and Global Change (WATCH) special collection: current knowledge of the terrestrial global water cycle. J. Hydrometeorol. 12 (6), 1149–1156.

Heuvelmans, G., Muys, B., Feyen, J., 2006. Regionalisation of the parameters of a hydrological model: comparison of linear regression models with artificial neural nets. J. Hydrol. 319, 245–256.

Hourdin, F., Musat, I., Bony, S., Braconnot, P., Codron, F., Dufresne, J.-L., Fairhead, L., Filiberti, M.-A., Friedlingstein, P., Grandpeix, J.-Y., Krinner, G., LeVan, P., Li, Z.-X., Lott, F., 2006. The LMDZ4 general circulation model: climate performance and sensitivity to parametrized physics with emphasis on tropical convection. Clim. Dyn. 27, 787–813.

Hundecha, Y., Bárdossy, A., 2004. Modelling of the effect of landuse changes on the runoff generation of a river basin through parameter regionalization of a watershed model. J. Hydrol. 300, 230–257.

Iglesias, A., Garrote, L., Diz, A., Schlickenrieder, J., Martin-Carrasco, F., 2011. Re- thinking water policy priorities in the Mediterranean region in view of climate change. Environ. Sci. Policy 14 (7), 744–757.

Ines, A.V.M., Hansen, J.W., 2006. Bias correction of daily GCM rainfall for crop simulation studies. Agric. For. Meteorol. 138 (1–4), 44–53.

Jacob, D., 2001. A note to the simulation of the annual and inter-annual variability of the water budget over the Baltic Sea drainage basin. Meteorol. Atmos. Phys. 77, 61–73.

Jacob, D., Bärring, L., Christensen, O.B., Christensen, J.H., de Castro, M., Deque, M., Giorgi, F., Hagemann, St., Hirschi, M., Jones, R., Kjellström, E., Lenderink, G., Rockel, B., Sanchez, E., Schär, Ch., Seneviratne, S., Somot, S., van Ulden, A., van den Hurk, B., 2007. An inter-comparison of regional climate models for Europe: design of the experiments and model performance. Clim. Change 81 (Supplement 1), 31–52.

Jacob, D., Christensen, O.B., Doblas-Reyes, F.J., Goodess, C., Tank, A.K., Lorenz, P., Roeckner, E., 2008. Information on observations, global and regional modelling data availability and statistical downscaling, Ensembles Technical Reports, ISSN 1752-2854.

Jaeger, E.B., Anders, I., Lüthi, D., Rockel, B., Schär, C., Seneviratne, S., 2008. Analysis of ERA40-driven CLM simulations for Europe. Meteorol. Z. 7, 1–19.

Jin, X., Xu, C., Zhang, Q., Chen, Y.D., 2009. Regionalization study of a conceptual hydrological model in Dongjiang basin, south China. Quatern. Int. 208, 129–137.

Jungclaus, J.H. et al., 2006. Ocean circulation and tropical variability in the coupled model ECHAM5/MPI-OM. J. Clim. 19, 3952–3972.

Kim, U., Kaluarachchi, J.J., 2008. Application of parameter estimation and regionalization methodologies to ungauged basins of the Upper Blue Nile River Basin, Ethiopia. J. Hydrol. 362, 39–56.

Kjellström, E., Bärring, L., Gollvik, S., Hansson, U., Jones, C., Samuelsson, P., Rummukainen, M., Ullerstig, A., Willén, U., Wyser, K., 2005. A 140-year simulation of European climate with the new version of the Rossby Centre regional atmospheric climate model (RCA3). Reports Meteorology and Climatology, 108, SMHI, SE-60176 Norrköping, Sweden, 54p.

Koutroulis, A.G., Tsanis, I.K., 2010. A method for estimating flash flood peak discharge in a poorly gauged basin: case study for the 13–14 January 1994 flood, Giofyros basin, Crete. J. Hydrol. 385, 150–164.

Koutroulis, A.G., Tsanis, I.K., Daliakopoulos, I.N., 2010. Seasonality of floods and their hydrometeorologic characteristics in the island of Crete. J. Hydrol. 394 (1–2), 90–100 (Special issue on Flash Floods).

Koutroulis, A.G., Vrochidou, A., Tsanis, I.K., 2011. Spatial and temporal characteristics of droughts for the island of Crete. J. Hydrometeorol. 12 (2), 206–226.

Kyselý, J., Dubrovský, M., 2005. Simulation of extreme temperature events by a stochastic weather generator: effects of interdiurnal and interannual variability reproduction. Int. J. Climatol. 25, 251–269.

Law, A.M., Kelton, W.D., 1982. Simulation Modelling and Analysis. McGraw-Hill Book Co., USA, p. 400.

Ludwig, R., Roson, R., Zografos, C., Kallis, G., 2011. Towards an inter-disciplinary research agenda on climate change, water and security in Southern Europe and neighboring countries. Environ. Sci. Policy 14 (7), 794–803.

158 A.G. Koutroulis et al. / Journal of Hydrology 479 (2013) 146–158

Manning, L.J., Hall, J.W., Fowler, H.J., Kilsby, C.G., Tebaldi, C., 2009. Using probabilistic climate change information from a multimodel ensemble for water resources assessment. Water Resource Res. 45.

Merz, R., Blöschl, G., 2004. Regionalisation of catchment model parameters. J. Hydrol. 287, 95–123.

Nakićenović, N. et al., 2000. Special Report on Emissions Scenarios A Special Report of Working Group III of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge.

Naoum, S., Tsanis, I.K., 2004. Orographic precipitation modelling with multiple linear regression. J. Hydrol. Eng. 9 (2), 79–102.

Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models. Part I. A discussion of principles. J. Hydrol. 10, 282–290.

Papagrigoriou, S., Kaimaki, S., Perleros, S., Papageorgiou, N., Lazaridis, L., 2001. Integrated Water Resources Management of Crete, 2001 (in Greek).

Parajka, J., Merz, R., Blöschl, G., 2005. A comparison of regionalisation methods for catchment model parameters. Hydrol. Earth Syst. Sci. Discuss. 2, 509–542.

Parry, M.L., Rosenzweig, C., Iglesias, A., Livermore, M., Fischer, G., 2004. Assessing the effects of climate change on global food production under differing socio- economic scenarios. Global Environ. Change 14 (1), 53–67.

Podger, G., 2004. RRL Rainfall–Runoff Library, User Guide. Cooperative Research Centre for Catchment Hydrology, Australia.

Pushpalatha, R., Perrin, C., Le Moine, N., Andréassian, V., 2012. A review of efficiency criteria suitable for evaluating low-flow simulations. J. Hydrol., 171–182, http:// dx.doi.org/10.1016/j.jhydrol.2011.11.055, ISSN 0022-1694.

Roeckner, E., Bauml, G., Bonaventura, L., Brokopf, R., Esch, M., Giorgetta, M., Hagemann, S., Kirchner, I., Kornbluch, L., Manzini, E., Rhodin, A., Schlese, U., Schulzweida, U., Tompkins, A., 2003. The atmospheric general circulation model ECHAM5. Part I: Model description. Max Planck Institute for Meteorology Rep. 349, 127p. <http://www.mpimet.mpg.de/fileadmin/publikationen/Reports/ max_scirep_349.pdf>.

Rojas, R., Feyen, L., Dosio, A., Bavera, D., 2011. Improving pan-European hydrological simulation of extreme events through statistical bias correction of RCM-driven climate simulations. Hydrol. Earth Syst. Sci. 15, 2599–2620. http://dx.doi.org/ 10.5194/hessd-8-3883-2011.

Sharma, M., Coulibaly, P., Dibike, Y., 2010. Assessing the need for downscaling RCM data for hydrologic impact study. ASCE J. Hydrol. Eng. 16 (6), 534–539.

Srikanthan, R., McMahon, T.A., 2001. Stochastic generation of annual, monthly and daily climate data: a review. Hydrol. Earth Syst. Sci. 5, 653–670.

Taylor, K.E., 2001. Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res. 106, 7183–7192.

Tsagarakis, K.P., Dialynas, G.E., Angelakis, A.N., 2004. Water resources management in Crete (Greece) including water recycling and reuse and proposed quality criteria, Mediterranean region. Agric. Water Manage. 66, 35–47.

Tsanis, I.K., Apostolaki, M.G., 2009. Estimating groundwater withdrawal in poorly gauged agricultural basins. Water Resource Manage. 23 (6), 1097–1123.

Tsanis, I.K., Naoum, S., 2003. The effect of spatially distributed meteorological parameters on irrigation water demand assessment. Adv. Water Resource 26, 311–324.

Tsanis, I.K., Koutroulis, A.G., Daliakopoulos, I.N., Jacob, D., 2011. Severe climate- induced water shortage and extremes in Crete. Clim. Change 106 (4), 667–677.

van der Linden, P, Mitchell, J.F.B. (Eds.), 2009. ENSEMBLES: Climate Change and its Impacts: Summary of research and results from the ENSEMBLES project. Met Office Hadley Centre, FitzRoy Road, Exeter EX1 3PB, UK, 160p.

van Meijgaard, E., van Ulft, L.H., van de Berg, W.J., Bosveld, F.C., van den Hurk, B.J.J.M., Lenderink, G., Siebesma, A.P., 2008. The KNMI regional atmospheric climate model RACMO, version 2.1 KNMI-publication TR-302. KNMI, De Bilt, 50p.

van Werkhoven, K., Wagener, T., Reed, P., Tang, Y., 2009. Sensitivity-guided reduction of parametric dimensionality for multi-objective calibration of watershed models. Adv. Water Resource 32, 1154–1169.

Vandewiele, G.L., Elias, A., 1995. Monthly water balance of ungauged catchments obtained by geographical regionalisation. J. Hydrol. 170, 277–291.

Vrugt, J.A., Gupta, H.V., Nualláin, B., Bouten, W., 2006. Real-time data assimilation for operational ensemble streamflow forecasting. J. Hydrometeor. 7, 548–565.

Wang, Q.J., 1998. Using genetic algorithms to optimise model parameters. Environ. Model. Softw. 12 (1), 27–34.

Wei, T.C., McGuinness, J.L., 1973. Reciprocal distance squared method, a computer technique for estimating area precipitation. Technical Report ARS-Nc-8, US Agricultural Research Service, North Central Region, Ohio.

Wilks, D.S., Wilby, R.L., 1999. The weather generation game: a review of stochastic weather models. Prog. Phys. Geograph. 23, 329–357.

Wood, A.W., Maurer, E.P., Kumar, A., Lettenmaier, D.P., 2002. Long range experimental hydrologic forecasting for the eastern United States. J. Geophys. Res. 107 (D20), 4429.

Yu, P., Yang, T., 2000. Using synthetic flow duration curves for rainfall–runoff model calibration at ungauged sites. Hydrol. Process. 14 (1), 117–133.