Assessing the hydrological response from an ensemble of CMIP5 climate projections in the transition zone of the Atlantic region (Bay of Biscay)

1 Assessing the hydrological response from an ensemble of CMIP5 climate 1 projections in the transition zone of the Atlantic region (Bay of Biscay). 2 Maite Meaurio(a), Ane Zabaleta(a), Laurie Boithias(b,c), Ane Miren Epelde(a), Sabine 3 Sauvage(b), Jose-Miguel Sanchez-Perez(b), Raghavan Srinivasan(d), Iñaki 4 Antigüedad(a). 5 (a) Hydrogeology and Environment Group, Science and Technology Faculty, University6 of the Basque Country UPV/EHU, 48940 Leioa, Basque Country, Spain 7 (b) EcoLab, Université de Toulouse, CNRS, INPT, UPS, 31400 Toulouse, France8 (c) Géosciences Environnement Toulouse, Université de Toulouse, CNES, CNRS,9 IRD, UPS, 31400 Toulouse, France 10 (d) Spatial Sciences Laboratory, Texas A&M University (TAMU), 77843 College11 Station, Texas, USA. 12 *Corresponding author information:13 Maite Meaurio 14 e-mail address: [email protected] 15 16 17 18 19 20 21 This is the accepted manuscript of the article that appeared in final form in Journal of Hydrology 548 : 46-62 (2017), which has been published in final form at https://doi.org/10.1016/j.jhydrol.2017.02.029. © 2017 Elsevier under CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) 2 Abstract 22 The climate changes projected for the 21st century will have consequences on 23 the hydrological response of catchments. These changes, and their consequences, are 24 most uncertain in the transition zones. The study area, in the Bay of Biscay, is located 25 in the transition zone of the European Atlantic region where hydrological impact of 26 climate change has been scarcely studied. To assess the hydrological effects of 27 climate change, 16 climate scenarios including 5 General Circulation Models (GCM) 28 from the 5º report of the Coupled Model Intercomparison Project (CMIP5), 2 statistical 29 downscaling methods and 2 Representative Concentration Pathways were considered 30 in a hydrological model (SWAT). Projections for future discharge (2011-2100) were 31 divided into three 30-year horizons (2030s, 2060s and 2090s) and a comparison was 32 made between these horizons and the baseline (1961-2000). The results show that the 33 downscaling method used resulted in a higher source of uncertainty than GCM itself. In 34 addition, the uncertainties inherent to the methods used at all the levels do not affect 35 the results equally along the year. In spite of those uncertainties, general trends for the 36 2090s predict seasonal discharge decreases by around -17% in autumn, -16% in 37 spring, -11% in winter and -7% in summer. These results are in line with those 38 predicted for France and the Iberian Peninsula in the Atlantic region. Trends for 39 extreme flows were also analysed: the most significant trend shows an increase in the 40 duration (days) of low flows. From an environmental point of view, and considering the 41 need to meet the objectives established by the European Water Framework Directive 42 (WFD), this would be a drawback for the future planning on water management. 43 Keywords: CMIP5, hydrological trend, high flow and low flow, SWAT model, 44 Atlantic region, transition zone. 45 Highlights: 46 3 Hydrological impact studies scarcity in transition zone of Atlantic region addressed 47 Downscaling method used resulted in a higher source of uncertainty than GCM itself 48 Results in line with those predicted for Atlantic region (France, Iberian Peninsula) 49 Uncertainties inherent to methods used do not affect results equally along the year 50 Highest decrease in low flows is a drawback for future planning on water management 51 1. Introduction 52 Climate change will have effects on hydrological systems, which in turn will 53 impact ecological, social and economic systems (Bender et al., 1984; Dibike and 54 Coulibaly, 2005; Brauman et al., 2007; Vörösmarty et al., 2010). These effects can be 55 studied both at local and regional levels, providing important information for territorial 56 and sectoral planning (Lahmer et al., 2001). In some areas where water scarcity is not 57 a key aspect of the territorial management, as it is the case of the Basque Country 58 (Bay of Biscay, Cantabrian Sea), few studies have been carried out to evaluate the 59 possible effects of the climate change on catchment hydrology. 60 The most commonly used method for evaluating climate change impact on 61 hydrological systems is to introduce General Circulation Models (GCMs) into 62 hydrological models (Gosling et al., 2011). This provides a good tool for studying the 63 relationship between climate and water resources, considering also the effect of human 64 activities (Jothityangkoon et al., 2001; Leavesley, 1994). However, GCMs usually have 65 little spatial resolution and if they are introduced directly into hydrological models, the 66 performance is poor (Fowler et al., 2007). This is the reason for the need of performing 67 a statistical or dynamical downscaling of general circulation models. 68 Uncertainties related to the impact of climate change appear at the four levels of 69 the sequence: GCMs, Representative Concentration Pathway (RCP) or emission 70 4 scenario, downscaling and hydrologic projections. For example, Chen et al. (2011) 71 assessed the uncertainty of downscaling methods and the results showed that impact 72 studies based on only one downscaling method should be interpreted with caution. 73 Wilby (2005) found that the uncertainty related to hydrological model calibration is 74 comparable with that involved in greenhouse and other pollutant emissions. Wilby and 75 Harris (2006) determined that the greatest uncertainties derive more from the choice of 76 GCM and downscaling methods and less from the hydrological models and emission 77 scenarios. Some research supports the argument that the choice of hydrological model 78 has a relatively minor impact on the results of hydrological simulations based on 79 climate projections (Boyer et al., 2010; Bates et al., 2008; Kay et al., 2006) and major 80 uncertainties come from the GCM structure (Arnell, 1999; Bergstrom et al., 2011; 81 Nijssen et al., 2001; Kay et al., 2009; Chen et al., 2011; Arnell et al., 2011; Teng et al., 82 2012). For this reason, the use of an ensemble of climate models gives a better 83 estimate of uncertainty (e.g. IPCC, 2007, 2013; Johnson and Sharma, 2009; Stahl et 84 al., 2011). Therefore, the inherent uncertainties of the methods used at each of these 85 levels propagate the uncertainties of the previous levels of the sequence, with the 86 result that all single uncertainties are then propagated in the hydrological models 87 (Wilby et al., 2006). 88 Europe is a representative region of global changes due to climate warming 89 (Shorthouse and Arnell, 1999). There is a clear contrast between the north and the 90 south of the continent, hence, an increase in precipitation and, therefore, an increase in 91 water discharge, has been pointed out in the north (e.g. Arnell, 1998; Kiely, 1999; Xu 92 and Halldin, 1997; IPCC, 2007, 2014). By contrast, in the south the trend is reversed: a 93 decrease of precipitation is predicted and, consequently, a decrease in discharge (e.g. 94 Mimikou et al., 2000; Ayala-Carcedo and Iglesias, 2000; Ribalaygua et al., 2013; 95 Lespinas et al., 2014; Touhami et al., 2015; Valverde et al., 2015; IPCC, 2014). Due to 96 its location in the Bay of Biscay (Fig. 1), the arid climatological conditions projected for 97 5 southern Europe do not seem to be representative for the Basque Country. However, 98 neither is it clear that the hydrological changes projected in this area will follow the 99 discharge increasing trend projected for northern Europe. In this sense, Coch and 100 Mediero (2015) analyzed low flows to identify different areas of hydrologic trends of the 101 Iberian Peninsula and Mediero et al. (2015) investigated the flood-prone regions in 102 Europe. Both studies reached the same conclusion: The Basque Country area would 103 be located in the Atlantic region. This area is characterized by Atlantic frontal systems 104 coming from the west, usually from autumn to spring, being summer the dry season. 105 Furthermore, the 4th report of the International Panel on Climate Change (IPCC, 2007) 106 provides a vulnerability map of Europe for the XXI century, where the Atlantic region 107 encompasses the northern Iberian Peninsula, western France, the Netherlands, 108 Belgium, northern part of Germany, western Denmark and the UK. As a general trend, 109 the report predicts an increase in the future winter storms and flooding for this region. 110 Research works conducted since 2000 in the Atlantic region to evaluate the 111 impacts of climate change in the water resources are summarized in Table 1. This 112 table was done to identify possible future hydrological trends in this region, thus, 113 research works that were published in high impact journals and some reports were 114 collected. In addition, to have a homogeneous view, the selection was made only with 115 works that presented their results in mean discharge difference (%) with respect to their 116 baseline. When necessary, mean difference values and ranges were calculated (for 117 example to obtain seasonal values from monthly ones). Despite some seasonal trends 118 in Table 1 are difficult to interpret, general trends of mean discharge evolution can be 119 derived considering those more clearly observed. A significant decrease of discharge is 120 observed in all the studies for summer and spring seasons. This decrease is even 121 more important towards the end of the century. In winter trends are not so clear. In the 122 UK (b, in Table 1) and in the Iberian Peninsula (d, Table 1), both increase and 123 decrease discharge can be expected depending on the study, whereas in France (c, 124 6 Table 1) the observed trend is decreasing. Trends in spring are similar to those in 125 winter though decrease of discharge prevails. Annual trends are determined by trends 126 in winter and spring. 127 From the aforementioned research works (Table 1), it can be deduced the 128 general idea that a transition zone exists between northern and southern Atlantic 129 region, where expected trends in discharge (in winter and spring) change from 130 increasing in the north to decreasing in the south. It is not an easy task to locate this 131 zone, due to the low spatial resolution of climate models; according to Habets et al. 132 (2013) this zone is located in northern France and following IPCC (2007) and 133 Goubanova and Li (2007) it would be in the north of the Iberian Peninsula, that includes 134 the study region. The uncertainties involved, precisely, in the climate predictions for 135 that transition zone are large and difficult to identify. Therefore, it should be an in-depth 136 studied area. However, compared with other European regions, the number of 137 research works in this zone is low (Table 1); this evidences the strength of the current 138 work. 139 In this context, the aim of this study is to assess the possible future effects of 140 climate change on the hydrology of a catchment located in the Atlantic region of the 141 Iberian Peninsula. For this purpose, five GCMs of the 5º report of the Coupled Model 142 Intercomparison Project (CMIP5), two downscaling methods and two RCPs were 143 considered, using a total of 16 climate projections (Table 2). 144 The projected climate variables were introduced in the Soil and Water 145 Assessment Tool or SWAT model (Arnold et al., 1998) to evaluate the hydrological 146 impact of climate change, focusing on the following partial objectives: 147 1) To evaluate the baselines of considered downscaled projections of climate 148 variables with respect to the observed data (1961-2000). 149 7 2) To study the average hydrological impact of future climate projections in three 150 horizons: 2030s for 2011-2040, 2060s for 2041-2070, 2090s for 2071-2100 151 (annual, seasonal and monthly). 152 3) To assess possible trends in extreme daily discharges (2011-2100) and 153 compare the observed figures for the reference period (1961-2000). 154 4) To identify the differences between climate projections and identify the greatest 155 uncertainties in the different steps of the followed methodology. 156 2. Methodology 157 2.1. Description of the study area 158 The study area is the catchment of the Upper Nerbioi River (185 km2), which is 159 located in the centre-west of the Basque Country (Bay of Biscay), at an average 160 latitude of 43º and longitude of 3º (Fig. 1). Its direction South-North, is very common in 161 catchments of the Basque Country, as well as its geology and land use. The catchment 162 is located at the interface between the Atlantic climate in the north and the 163 Mediterranean climate in the south. In addition, this is one of the catchments with 164 longest records of discharge series in this zone. 165 Average annual rainfall is about 1,000 mm and it is distributed quite evenly 166 throughout the year: close to 300 mm in winter and autumn; 230 mm in spring and 30 167 mm in summer, as an average (1961-2014). The mean annual temperature is around 168 12 °C, being the seasonal averages for winter and summer 8 ºC and 20 ºC (1961-169 2014), respectively. 170 The mean elevation of the catchment is around 200 m above sea level (m.a.s.l.) 171 (Fig. 1). The lithology is dominated by siltstones, clays and sandstones with medium-172 low permeability (Geographical Database of the Basque Government, 173 www.geoeuskadi.net). In the southeast part, at an average altitude of 1,100 m.a.s.l., 174 8 there is a highly permeable late Cretaceous limestone platform. The main soil types are 175 Cambisols, Rankers and Gleysols (FAO, 1977), which are characterized by high clay 176 and silt contents. Land use in the catchment is divided into native forests, exotic 177 plantations and pasturelands. The areas with the highest slopes (>35%) are covered 178 by forest and tree plantations, while the flatter areas (7-15%) host pasturelands. The 179 average slope in the catchment is around 17%. It should be noted that possible future 180 changes in land use have not been taken into consideration in this work. 181 Mean annual discharge at the outlet of the catchment is 3 m3 s-1. The mean in 182 spring and autumn is around 3.5 m3 s-1; in winter 5 m3 s-1 and in summer 0.8 m3 s-1 183 (1996-2013). Discharge data from a gauging station (Gardea; http://www.bizkaia.eus) 184 located at the outlet of the catchment (Fig. 1) were used in this work to calibrate and 185 validate the hydrological model. Discharge data have been recorded at this gauging 186 station since 1995. This station was designed to precisely measure mean and high 187 flows. 188 2.2. Description of the hydrological model: SWAT 189 The SWAT model is a basin-scale continuous in time and semi-distributed 190 model operating on a daily time step. It was developed to evaluate the impact of 191 management practices on water, sediment and agricultural chemical yields in 192 ungauged basins (Arnold et al., 1998). It can be used in a broad range of conditions 193 and it has already been widely used to study the impacts of environmental and climate 194 change (e.g. Bouraoui et al., 2002; Li et al., 2009; Abbaspour et al., 2009; Bekele and 195 Knapp, 2010; Zhang et al., 2014). 196 SWAT divides the catchment into sub-basins which are subdivided into 197 Hydrological Response Units (HRUs) with homogeneous land use, soil characteristics 198 and slope gradient. Two methods are used to simulate surface runoff in the SWAT 199 9 model: the modified SCS curve number (USDA Soil Conservation Service, 1972) and 200 the Green-Ampt infiltration method (Green and Ampt, 1911), which requires a sub-daily 201 precipitation time step. In this case the observed meteorological data used to calibrate 202 and validate the model were daily and therefore the SCS method was used. The model 203 calculates the peak runoff rate with a modified rational method (Chow et al., 1988). The 204 lateral subsurface flow in the soil profile is determined for each soil layer, using the 205 kinematic storage routing model (Sloan and Moore, 1984), which is calculated 206 simultaneously with percolation. The groundwater flow contribution to the total 207 streamflow is simulated by creating shallow aquifer storage (Arnold and Allen, 1996) 208 where percolation from the bottom of the root zone is considered as recharge to the 209 shallow aquifer. The potential evapotranspiration can be estimated using the 210 Hargreaves (Hargreaves and Samani, 1985), Priestley-Taylor (Priestley and Taylor, 211 1972) and Penman-Monteith (Monteith, 1965) methods. This study uses Hargreaves, 212 which only requires precipitation and maximum and minimum temperature, since these 213 were the only meteorological data available. The flow is routed through the channel 214 using either the variable storage coefficient method (Williams, 1969) or the Muskingum 215 routing method (Overton, 1966). In this study the former method was used because it 216 better suited the observed discharge. 217 2.3. Hydrological model input and data source. 218 SWAT requires topographic, land use/cover, soil and meteorological data. The 219 source for the Digital Elevation Model (LIDAR 2008, 5x5m), land use classification 220 (2005, 1: 10,000) and part of the soil map (1: 25,000) is the Basque Government’s 221 Geographical Database (www.geoeuskadi.net). The remainder of the soil map was 222 obtained from the soil map of Araba province (1:200,000) (Iñiguez et al., 1980). Soil 223 properties were obtained from these two sources and the plant growth properties for 224 each land cover were directly obtained from the SWAT database. 225 16 the test underestimates the probability of detecting trends. This serial correlation may 382 therefore influence the results of the test (Douglas et al., 2000). To avoid these 383 possible effects, before using the Mann-Kendal test the trend free pre-whitening 384 approach developed by Yue et al. (2002) was applied to the serial data. With the Mann-385 Kendall test it is possible to identify increasing and decreasing trends and the 386 probability of occurrence (P) of those trends. The value of P can vary between 0 and 1, 387 where 0 indicates that there is no probability of occurrence in the trend and 1 indicates 388 maximum probability. The criteria suggested by the IPCC (Mastrandrea et al., 2010) in 389 its 5th report were used to evaluate P. In this document, likelihood refers to a 390 probabilistic assessment of some well-defined past or future outcomes. The categories 391 defined and used in this research are: P>0.99, virtually certain trend; P>0.95, extremely 392 probable trend; P>0.90, very probable trend and P>0.66, probable trends. Values of P 393 below 0.66 are considered to represent non-probable trends in this work. 394 3. Results and discussion 395 3.1. SWAT calibration (1996-2006), validation (2007-2013) and simulation 396 uncertainty 397 The parameters changed in the calibration process of SWAT for the Upper 398 Nerbioi, their value-range used in the autocalibration and the final values are shown in 399 Table 3. These are some of the most common parameter changes usually performed in 400 SWAT to calibrate the model (Arnold et al., 2012) and all of them were changed taking 401 into account the catchment characteristics. The results of the calibration (1996-2006) 402 and the validation (2007-2013) are displayed in a daily hydrograph with the observed 403 and simulated discharge (Fig. 2). The calibration can be seen to fit the observed data 404 well, although the peak magnitude is underestimated in some high flows. In previous 405 works carried out with the SWAT model (daily time step) in the Basque Country, the 406 underestimation of the peak magnitude is usual (Zabaleta et al., 2014; Peraza et al., 407 17 2015; Epelde et al., 2015; Meaurio et al., 2015). This inaccuracy may be related to the 408 inability of the model to properly consider precipitation intensity and spatial-temporal 409 distribution when simulating rapid hydrological responses at the daily time step (Qiu et 410 al., 2012). 411 In general, simulated discharge peaks fit observed data better during the 412 validation (Fig. 2). The set of statistical indices calculated for daily discharge (Table 4) 413 shows that the model performs satisfactorily during both, calibration and validation 414 (Moriasi et al., 2007). 415 Analyzing years with a low and high aridity index (AI), it is possible to assess 416 whether the simulation is good enough to make long-term hydrologic projections and 417 evaluate when the greatest uncertainties are found (low or high AI). The set of 418 statistical indices (Table 4) shows that simulation performance for years with high and 419 low AI is at least “good”, although some parameters are slightly poorer for years with 420 high AI (more uncertainty). In addition, the set of statistical indices were also applied for 421 the entire period (1996-2013) on a seasonal scale. It is thus possible to evaluate the 422 model performance considering low (summer), intermediate (spring and autumn) and 423 high (winter) flows and determining where the largest uncertainties are. Winter and 424 spring present "good" statistical results, autumn is "at least satisfactory" and the 425 statistical indices show that although the hydrograph seems not to fit properly in 426 summer (low r2, NSE and RSR), the water yield is simulated correctly (low PBIAS). 427 Note that summer discharges, being the lowest, are more vulnerable to measurement 428 errors. Therefore, although the simulation does perform well, summer is the season 429 associated to the highest modeling uncertainty. However, according to Moriasi et al. 430 (2007) the values of most of the statistical indices shown in Table 4 were "good" or 431 "very good" at monthly time step. Since these analyses were made with daily values 432 they are considered to be at least "good". Additionally, the p-factor and r-factor 433 18 obtained with the SWAT-CUP program (the range of the parameters is shown in Table 434 3) for calibration and validation are 0.81 and 0.41 respectively, which is considered 435 "good" (Abbaspour et al., 2015). As a consequence, it can be said that the 436 performance of the model is good enough for carrying out future hydrological 437 projections with a certain degree of confidence. 438 3.2. Assessment of the baseline hydrological projections 439 The bias-corrected precipitation and maximum and minimum temperature for 440 the baseline of each downscaled GCM was introduced in the calibrated and validated 441 SWAT project. The first step was to assess how the hydrological simulations obtained 442 for baseline (1961-2000) adjust to the ones performed using observed meteorological 443 data (OBS_SIM) for the same period. The mean monthly discharges (m3 s-1) obtained 444 are shown in Fig. 4. This figure shows that as in the case of precipitation, hydrological 445 simulations obtained using the baseline climate projections downscaled with the SDSM 446 method fit much better to OBS-SIM than those downscaled with the AN method. In fact, 447 the adjustment for discharge series obtained using SDSM downscaling to OBS_SIM is 448 really good in autumn (-9%) and winter (-2%) whereas in spring and summer the 449 discharge is underestimated by about -22% and -71%, respectively. However, those 450 differences are higher in all seasons for the discharge series obtained using the AN 451 method; around -22% in autumn, -11% in winter, -46% in spring and -83% in summer. 452 Therefore, it is clear that, in this case, the choice of the downscaling method is 453 the cause of a higher uncertainty source in the obtained discharge series than the 454 choice of the GCM itself. 455 3.3. Hydrological impact of future climate scenarios: annual and seasonal scales 456 (2011-2100) 457 19 To evaluate the impact of climate change on the hydrology of the Upper Nerbioi 458 catchment area, future hydrological projections divided into three time horizons (2030s, 459 2060s, and 2090s) were compared with their baseline average discharge (1961-2000). 460 The difference in average discharge (in %) is shown in Fig. 5. Focusing on the 461 downscaling method (AN or SDSM), the hydrological projections derived from climate 462 projections that use the AN method always show a smaller discharge decrease than 463 those downscaled with SDSM (with the exception of CMCC_CESM_AN_R85). 464 Considering that climate projections obtained with the SDSM downscaling method fit 465 better to OBS_SIM series data for the baseline period, it seems more reasonable to 466 consider those hydrological projections derived from SDSM downscaled series. 467 It is also important to compare the two different RCPs because, one would 468 expect that the difference between the 2060s and 2090s for the projections with RCP 469 4.5 would be minimal, while for 8.5 the difference would continue to increase. Indeed, if 470 the projections of CMCC-CESM are not considered, the difference in discharge at 471 annual scale between the baseline and the projections with RCP 4.5 is -6% for the 472 2030s, -8% for the 2060s and -9% for the 2090s, while for RCP 8.5 it is -13% for the 473 2030s; -15% for the 2060s and -20% for the 2090s. In the RCP 4.5 scenario, the 474 seasonal decrease of discharge throughout the century is lower. In some seasons, as 475 in summer, a stabilization of the discharge can be observed, and in others (e.g. 476 autumn) the increase in the average flow in the 2090s almost compensates for the 477 decreases observed during the 2060s. This is not the case for the RCP 8.5 scenario 478 where discharge continues to decrease until the end of the century. 479 Undeniably CMCC_CESM_AN_R85 is most noteworthy because it projects 480 higher discharge than the baseline. CMCC_CESM_SDSM_R85 decreases respect to 481 the baseline, but as it happens with CMCC_CESM_AN_R85 the discharge increases 482 throughout the century. The results for this GCM were analysed separately due to 483 20 those different trends shown. Fig. 6 shows the difference (%) between CMCC-484 CESM_AN_R85 and CMCC-CESM_SDSM_R85 future discharges with regard to their 485 baseline on a seasonal and annual scale. In order to compare not only the trends but 486 also the discharge, in terms of illustrative average flow for each period, the average 487 discharge for each time horizon considered is also displayed in Fig. 6. These are the 488 two projections that simulate highest discharge at annual scale as well as at seasonal 489 scale. 490 As discussed above, the downscaling method and the selection of RCP have a strong 491 influence on the results. The projections were therefore classified taking into 492 consideration the downscaling method (AN or SDSM) and the scenarios (RCP 4.5 or 493 RCP 8.5) (Fig. 7), analysed in four different groups (with the exception of the CMCC-494 CESM projections). Thus, the results are an ensemble of projections but it is possible 495 to analyse differences between these factors (downscaling method and RCP). At 496 annual scale, and for the end of the century (2090s), discharge decreases by -9% and -497 20% for RCP 4.5 and RCP 8.5 scenarios, respectively, with little differences between 498 downscaling methods. These results are consistent with most of the studies carried out 499 in the Atlantic region of France and in the North of the Iberian Peninsula (Table 1). 500 However, focusing on the seasonal changes, significant differences can be found 501 depending on the use of the downscaling method. Summer is the season when the 502 greatest differences can be observed; slightly increasing (<5% for 2090s) for climate 503 projections derived from AN downscaling, and clearly decreasing (-15 to -25% for 504 2090s) for SDSM-derived ones. The AN downscaling method simulated considerably 505 less discharge than the SDSM downscaling method. Hence, the difference decreases 506 between baseline and future projections are bigger for the SDSM method, although the 507 projections downscaled with SDSM (independent of RCP) always projected more 508 discharge than those downscaled with the AN method (Fig. 7). In other seasons 509 (autumn, winter and spring), discharge decreased regardless of the method chosen, 510 21 with higher changes when using the AN method in autumn and smaller changes in 511 spring. The results obtained using different downscaling methods are most similar in 512 winter. This is the season that has most weight in the annual discharge, hence its effect 513 can be observed at annual scale. Considering all the models, autumn is the season 514 with the most significant discharge decrease (-17%) followed by spring (-16%), winter (-515 11%), and summer (-7%) for 2090s. These results are slightly different from previous 516 studies undertaken in the Atlantic region of France and the north of the Iberian 517 Peninsula (Table 1). In most of the research works carried out in these areas the 518 highest discharge decreases occur in summer, and depending on the study are 519 followed by autumn or spring (Table 1). In the present study, considering the average 520 value, summer is not the most affected season in percentage. This could be explained 521 by the influence of the projections downscaled with the AN method (Fig. 7). 522 Fig. 8 is an ensemble showing a combination of the 16 hydrological projections 523 analysed (average of the mean monthly discharge (m3 s-1) represented in a 524 hydrological year) and their evolution over time measured at the 3 time horizons 525 (2030s, 2060s and 2090s). In order to consider all the discharge predictions obtained, 526 the projections were not divided based on the downscaling method or the RCP. The 527 highest discharge values represent the maximum value of the mean monthly discharge 528 of all of the projections, while the lowest values represent the minimum ones. The 529 possible discharge range is the highest in winter and autumn. In these seasons the 530 possible mean discharge may vary by 4 m3 s-1. In spring the discharge may range 531 between 1 and 2.1 m3 s-1, while the range in summer may be the lowest: between 0.1 532 and 0.3 m3 s-1. In spring, summer and the beginning of autumn, the projected discharge 533 is always lower than the OBS_SIM. However, the results for spring and summer have 534 to be considered with special care because, as discussed previously, the baseline of 535 the hydrological projections are underestimated and it is therefore probable that a 536 similar phenomenon happens in the case of future projections. With regard to the 537 22 evolution of the discharge over the century, the projected lowest discharge decreases 538 from the 2030s to the 2090s. The projected highest discharges show the same trend in 539 spring and summer, whereas they may even increase in autumn and winter. 540 3.4. Evaluation of trends in duration and severity of extreme flows 541 The results obtained from trend analysis carried out for the duration of extreme 542 flows are shown in Table 5 and Fig. 9. This analysis has been made for the reference 543 period 1961-2000 (Table 5) and the future periods 2011-2040, 2011-2070 and 2011-544 2100 (Fig. 9). However, the most significant trends appear in the longest period (2011-545 2100) and hence, these are the results considered in this study. 546 Analysing the duration (in days) of low flows (<Q20) from 1961 to 2000 (Table 547 5), the discharge simulated for the reference period (OBS_SIM) does not show any 548 significant annual trend. At seasonal scale, a significant trend is only detected in spring 549 when the low flow duration shows a "probable" upward trend. However, some of the 550 discharge series simulated using the nine climate baselines, show significant trends; an 551 increase at annual scale and, depending on the GCM considered, an increase or 552 decrease in spring, summer and autumn (Table 5). 553 In the evaluation of future low flow duration (2011-2100; Fig. 9), a general 554 increasing trend can be observed, although, there are a few decreasing trends. In 555 spring and autumn upward trends predominate. In spring the number of projections 556 with significant increasing trends is higher under RCP 8.5 than in RCP 4.5. On the 557 contrary, in autumn this number is lower. Summer present random significant trends 558 mostly under RCP 8.5, that in general tend to be positive. There are few projections 559 with significant trends in winter, therefore, is not possible to obtain clear conclusions for 560 this season. 561 23 OBS_SIM displays decreasing annual and seasonal trends for high flow 562 duration (above Q80). Annually, the decreasing trend is "extremely probable", on 563 summer is “very probable” and in autumn, winter and spring is “probable” (Table 5). 564 The high flow durations obtained using climate baselines, in general do not show 565 significant trends. However, the most significant trends are for the model BNU-ESM 566 showing “very probable” to “virtually certain” decreasing trends annually and for 567 autumn. 568 The high flow duration (Q80) for future projections (2011-2100) show significant 569 trends annually and in autumn, however, there are as many increasing as decreasing 570 trends (Fig. 9). In spring there are few projections with significant trends being most of 571 them increasing. In winter under RCP 8.5 a general decreasing trend predominates. In 572 summer a change in general trend can be observed form RCP 4.5 to RCP 8.5: under 573 RCP 4.5 decreasing trends prevail, while under RCP 8.5 are mostly increasing. 574 Zabaleta et al. (2012) identified hydrological signs in the catchments of the 575 Basque Country. They used observed daily discharge values in different periods and 576 catchments (regional context). The longest analysed period in their research is 34 577 years (1973-2007). Although this period does not coincide in time with the reference 578 period used in this work (1961-2000), similarity in the increasing trends of duration of 579 low flows (Q20) can be observed. The authors attributed these trends to hydrological 580 signs of climate change. 581 Understanding the lower part of the projected hydrographs is essential for an 582 assessment of impact on freshwater ecosystems. Therefore, besides knowing the trend 583 of the number of days with high (above Q80) and low (below Q20) flows, it is also 584 important to know the volumetric deficit (severity) trend, especially for low flows. The 585 OBS_SIM (1961-2000) low flow deficit does not display any significant trend. Most of 586 24 the baselines do not show significant trends, nevertheless, the few of them where trend 587 is detected, predict an increase in severity. 588 For 2011-2100 there are significant trends in most of the projected simulations 589 for severity. However, these trends are opposite from each other and cannot be related 590 to the use of given GCMs, downscaling methods or RCP scenarios. As a consequence, 591 severity showed very high uncertainty in future hydrologic projection in the Upper 592 Nerbioi catchment. 593 4. Conclusions 594 In this study, to assess future climate change effects (up to year 2100) on the 595 hydrological response of the Upper Nerbioi catchment,16 climate projections combining 596 five GCMs (ACCESS1-0, BNU-ESM, MPI-ESM-RL, MPI-ESM-MR, CMCC-CESM), two 597 downscaling methods (AEMET analogues -ANand Statistical Downscaling Method -598 SDSM-) and two Representative Concentration Pathways (RCP 4.5 and RCP 8.5) were 599 considered. Hydrological simulation was performed using the SWAT model achieving 600 satisfactory results for the calibration and validation periods (1996-2013). 601 Different sources of uncertainties are involved in the hydrologic projections (GCM, 602 downscaling method, RCP, hydrological model). Some conclusions can be drawn from 603 the obtained results even if the quantification of the uncertainties lies outside the scope 604 of this study. The considerable difference between the baselines of the climate models 605 (1961-2000) and the observed meteorological data (the models generally simulate less 606 rainfall, especially in spring and summer) evidences the uncertainty involved for the 607 studied area in the results of the GCMs. 608 However, downscaling method used resulted in a higher source of uncertainty than 609 GCM itself. When simulated discharges for the baselines of all climate projections were 610 compared with the discharge obtained from a simulation made with the observed 611 25 meteorological data (OBS_SIM; 1961-2000), the comparison shows that the GCMs 612 downscaled with the SDSM method achieve a much better adjustment than those 613 downscaled with the AEMET analogues (AN). Nevertheless, they all underestimate the 614 discharge amount. Those uncertainties inherent to the methods used at all the levels 615 do not affect the results equally along the year. The seasons with most variable results, 616 and so, the ones for which it is the most difficult to draw a clear conclusion, are spring 617 and especially summer, for which future discharges could either increase or decrease 618 depending on the downscaling method. 619 From results obtained from four of the analysed GCMs, ACCESS1-0, BNU-ESM, 620 MPI-ESM-RL and MPI-ESM-MR, it can be said that discharge would decrease with 621 respect to the baseline at annual scale. This conclusion is consistent with the trends 622 obtained in the Atlantic region, mostly in France and the Iberian Peninsula (c, d; Table 623 1). The spatially most homogeneous result from Table 1 is the decrease in summer 624 projections for the entire Atlantic region. However, summer is the season that most 625 discrepancies show in the study area; the projections downscaled with the AN method 626 projected around 5% more discharge for the 2090s than for the baseline whereas 627 SDSM projects -15 to -25% less discharge (Fig 7). In fact, the seasons that predict the 628 largest decrease in discharge in the study area are autumn and spring (around -16 % 629 for 2090s). These downwards trends are also detected in the Atlantic region of France 630 and the Iberian Peninsula, although, they are not so strong. The lowest decrease as it 631 happens in other zones of the Atlantic region, is projected for winter. 632 For the ensemble of the 16 hydrological projections analysed in the three horizons, 633 the widest range between the monthly highest and lowest discharge values would 634 occur in winter and autumn (around 3-5 m3 s-1) followed by spring (between 0.1-2 m3 s635 1) while the narrowest one would be in summer (between 0.1-0.3 m3 s-1). For spring, 636 summer and the beginning of autumn, simulated discharge is always below the 637 32 Chow, V.T., Maidment, D.R. and Mays, L.W., 1988.Applied Hydrology. McGraw-Hill 773 Inc, New York, USA. 774 Cloke, H. L., Jeffers, C., Wetterhall, F., Byrne, T., Lowe, J., and Pappenberger, F., 775 2010. Climate impacts on river flow: Projections for the medway catchment, UK, with 776 UKCP09 and CATCHMOD. Hydrol. Process. 24 (24), 3476-3489. 777 http://dx.doi.org.10.1002/hyp.7769. 778 Coch, A. and Mediero, L., 2015. Trends in low flows in Spain in the period 1949–2009. 779 Hydrolog. Sci. J. 1-17. http://dx.doi.org//10.1080/02626667.2015.1081202. 780 Da Cunha, L.V., De Oliveira, R.P., Nascimento, J., and Ribeiro, L., 2007. Impacts of 781 climate change on water resources: A case-study for Portugal. Water in Celtic 782 Countries: Quantity, Quality and Climate Variability, 310, 37-48. 783 Diaz-Nieto, J. and Wilby, R.L., 2005. A comparison of statistical downscaling and 784 climate change factor methods: Impacts on low flows in the river Thames, United 785 Kingdom. Climatic Change. 69 (2), 245-268. 786 Dibike, Y.B. and Coulibaly, P., 2005. Hydrologic impact of climate change in the 787 Saguenay watershed: comparison of downscaling methods and hydrologic models. J. 788 Hydrik, 307 (1–4), 145-163. http://doi.org/10.1016/j.jhydrol.2004.10.012. 789 Douglas, E.M., Voguel, R.M. and Kroll, C.N., 2000. Trends in floods and low flows in 790 the United States: Impact of spatial correlation. J. Hydrol. 240, 90–105. 791 http://dx.doi.org/10.1016/S0022-1694(00)00336-X. 792 Ducharne, A., Habets, F., Page, C., Sauquet, E., Viennot, P., Deque, M., Gascoin, S., 793 Hachour, A., Martin, E., Oudin, L., Terray, L. and Thiery, D., 2010. Climate change 794 33 impacts on water resources and hydrological extremes in northern France. 795 Proceedings of the XVIII International Conference on Computational Methods in Water 796 Resources (Cmwr 2010), 243-250. 797 Epelde, A.M., Cerro, I., Sanchez-Perez, J.M., Sauvage, S., Srinivasan, R., and 798 Antigueedad, I., 2015. Application of the SWAT model to assess the impact of changes 799 in agricultural management practices on water quality. Hydrol. Sci. J. 60 (5), 825-843. 800 http://dx.doi.org/10.1080/02626667.2014.967692. 801 FAO, 1977. Guidelines for Soil Profile Description. Rome, Italy. 802 Fowler, H.J. and Kilsby, C.G., 2007. Using regional climate model data to simulate 803 historical and future river flows in northwest England. Climatic Change. 80 (3-4), 337-804 367. http://dx.doi.org/10.1007/s10584-006-9117-3. 805 Fowler, H.J., Blenkinsop, S. and Tebaldi, C., 2007. Linking climate change modelling to 806 impacts studies: recent advances in downscaling techniques for hydrological modelling. 807 Int. J. Climatol. 27 (12), 1547–1578. http://dx.doi.org/10.1002/joc.1556. 808 Geographical data base of the Basque Government, GEOEUSKADI; 809 http://www.geo.euskadi.net 810 Giorgi, F. and Mearns, L.O., 2002. Calculation of average, uncertainty range, and 811 reliability of regional climate changes from AOGCM simulations via the ‘‘reliability 812 ensemble averaging’’ (REA) method. J. Climate. 15, 1141–1158. 813 http://dx.doi.org/10.1175/1520-0442(2003)016%3C0883:COCOAU%3E2.0.CO;2. 814 Görgen, K., Beersma, J., Brahmer, G., Buiteveld, H., Carambia, M., de Keizer, O., 815 Krahe, P., Nilson, E., Lammersen, R., Perrin, C. and Volken, D., 2010. Assessment of 816 34 Climate Change Impacts on Discharge in the Rhine River Basin: Results of the 817 RheinBlick2050 Project, CHR Report, 1–23, 229 pp, Lelystad, ISBN 978-90-70980-35-818 1. 819 Gosling, S.N., Taylor, R.G., Arnell, N.W. and Todd, M.C., 2011. A comparative analysis 820 of projected impacts of climate change on river runoff from global and catchment-scale 821 hydrological models. Hydrol. Earth. Syst. Sc. 15, 279-294. 822 http://dx.doi.org/10.5194/hess-15-279-2011. 823 Goubanova, K. and Li, L., 2007. Extremes in temperature and precipitation around the 824 Mediterranean basin in an ensemble of future climate scenario simulations. Glob. 825 Planet. Chang. 57, 27–42 826 Graham, L.P., Andréasson, J. and Carlsson, B., 2007. Assessing climate change 827 impacts on hydrology from an ensemble of regional climate models, model scales and 828 linking methods – a case study on the Lule River Basin. Climatic Change. 81 (1), 293-829 307. http://dx.doi.org/10.1007/s10584-006-9215-2. 830 Green, W.H. and Ampt, G.A., 1911. Studies on soil physics. The Journal of Agricultural 831 Science. 4 (1), 1-24. http://dx.doi.org/10.1017/S0021859600001441. 832 Gupta, H.V., Sorooshian, S. and Yapo, P.O., 1999. Status of automatic calibration for 833 hydrologic models: Comparison with multilevel expert calibration. J. Hydrol.Eng. 4 (2), 834 135-143. 835 Habets, F., Boe, J., Deque, M., Ducharne, A., Gascoin, S., Hachour, A., Martin, E., 836 Page, C., Sauquet, E., Terray, L., Thiery, D., Oudin, L. and Viennot, P., 2013. Impact of 837 climate change on the hydrogeology of two basins in northern France. Climatic 838 Change. 121 (4), 771-785. http://dx.doi.org/10.1007/s10584-013-0934-x. 839 35 Hargreaves, G. and Samani, Z.A., 1985. Reference crop evapotranspiration from 840 temperature. Appl. Eng. Agric. 1 (2), 96-99. http://dx.doi.org/10.13031/2013.26773. 841 Hewitson, B.C. and Crane, R.G., 1996. Climate downscaling: techniques and 842 application. Clim. Res. 7, 85-95. 843 Hiscock, K., Sparkes, R., and Hodgson, A., 2011. Evaluation of future climate change 844 impacts on European groundwater resources. Climate Change Effects on Groundwater 845 Resources: A Global Synthesis of Findings and Recommendations. 27, 351-365. 846 Hisdal, H., Stahl, K., Tallaksen, L.M. and Demuth, S., 2001. Have streamflow droughts 847 in Europe become more severe or frequent? Int. J. Climatol. 21, 317–333. 848 http://dxdoi.org/10.1002/joc.619. 849 Iñiguez, J., Sánchez‐Carpintero, I., Val, R.M., Romeo, A. and Basconesm, J.C., 1980. 850 Mapa de suelos de Alava. Vitoria‐Gasteiz: Diputación Foral de Alava‐Departamento de 851 Edafología de la Universidad de Navarra. 852 IPCC, 2007. Climate Change 2007: synthesis report. Contribution of working groups I, 853 II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate 854 Change. Core Writing Team, R.K. Pachauri, and A. Reisinger, eds. Geneva: IPCC 855 Secretariat. 856 IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of 857 Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on 858 Climate Change. [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. 859 Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University 860 Press, Cambridge, United Kingdom and New York, NY, USA, 1535pJohnson, F. and 861 Sharma, A., 2009. Measurement of GCM skill in predicting variables relevant for 862 36 hydroclimatological assessments. J. Climate. 22 (16), 4373-4328. 863 http://doi.org/10.1175/2009JCLI2681.1. 864 Jothityangkoon, C., Sivapalan, M. and Farmer, D.L., 2001. Process controls of water 865 balance variability in a large semi-arid catchment: downward approach to hydrological 866 model development. J. Hydro. 254,174–198. http://dx.doi.org/10.1016/S0022-867 1694(01)00496-6. 868 Kay, A.L., Davies, H.N., Bell, V.A. and Jones, R.G., 2009. Comparison of uncertainty 869 sources for climate change impacts: flood frequency in England. Climatic Change. 92, 870 41-63. http://dx.doi.org/10.1007/s10584-008-9471-4. 871 Kay, A.L., Jones, R.G. and Reynard, N.S., 2006. RCM rainfall for UK flood frequency 872 estimation II. Climate change results. J. Hydrol. 318, 163-172. 873 http://dx.doi.org/10.1016/j.jhydrol.2005.06.013. 874 Kendall, M.G., 1975. Rank correlation measures. Charles Griffin, London, UK. 875 Kiely, G., 1999. Climate change in Ireland from precipitation and streamflow 876 observation. Adv. Water. Resour. 23, 141–151. 877 Lahmer, W., Pfutzner, B. and Becker, A., 2001. Assessment of land use and climate 878 change impacts on the mesoscale. Physics and Chemistry of the Earth, Part B. Hydrol. 879 Ocean. Atmos., 26 (7–8), 565–575. http://dx.doi.org/10.1016/S1464-1909(01)00051-X. 880 Leavesley, G.H., 1994.Modeling the effects of climate change on water resources – a 881 review. Climatic Change. 28, 159–177. 882 37 Lenderink, G., Buishand, A. and van Deursen, W., 2007. Estimates of future 883 discharges of the river Rhine using two scenario methodologies: direct versus delta 884 approach. Hydrol. Earth Syst. Sc. 11 (3), 1145–1159. http://dx.doi.org/10.5194/hess-885 11-1145-2007. 886 Lespinas, F., Ludwig, W. and Heussner, S., 2014. Hydrological and climatic 887 uncertainties associated with modeling the impact of climate change on water 888 resources of small Mediterranean coastal rivers. J. Hydrol. 511, 403 – 422. 889 http://dx.doi.org/10.1016/j.jhydrol.2014.01.033. 890 Li, Z., Liu, W., Zhang, X. and Zheng, F., 2009. Impacts of land use change and climate 891 variability on hydrology in an agricultural catchment on the Loess Plateau of China. J. 892 Hydrol. 377, 35-42. http://dx.doi.org/10.1016/j.jhydrol.2009.08.007. 893 Mann, H.B., 1945. Non-Parametric tests against trend. The Econometric Society. 13, 894 245–259. http://dx.doi.org/10.2307/1907187. 895 Maraun, D., Wetterhall, F., Ireson, A.M., Chandler, R.E., 2010. Precipitation 896 downscaling under climate change. Recent developments to bridge the gap between 897 dynamical models and the end user. Rev. Geophys. 48 (3), 1944-9208. 898 http://dx.doi.org/10.1029/2009RG000314. 899 Mastrandrea, M.D., Field, C.B., Stocker, T.F., Edenhofer, O., Ebi, K.L., Frame, D.J., 900 Held, H., Kriegler, E., Mach, K.J., Matschoss, P.R., Plattner, G-K., Yohe, G.W. and 901 Zwiers, F.W., 2010. Guidance Note for Lead Authors of the IPCC Fifth Assessment 902 Report on Consistent Treatment of Uncertainties. Intergovernmental Panel on Climate 903 Change (IPCC). Available in http://www.ipcc.ch. Accessed 5 August 2015. 904 38 Meaurio, M., Zabaleta, A., Uriarte, J.A., Srinivasan, R. and Antigüedad, I., 2015. 905 Evaluation of SWAT models performance to simulate streamflow spatial origin. The 906 case of a small forested watershed. J. Hydrol. 525, 326-334. 907 http://dx.doi.org/10.1016/j.jhydrol.2015.03.050. 908 Mediero, L., Kjeldsen, T.R., Macdonald, N., Kohnova, S., Merz, B., Vorogushyn, S., 909 Wilson, D., Alburquerque, T., Bloeschl, G., Bogdanowicz, E., Castellarin, A., Hall, J., 910 Kobold, M., Kriauciuniene, J., Lang, M., Madsen, H., Gul, G.O., Perdigao, R.A.P., 911 Roald, L.A., Salinas, J.L., Toumazis, A.D., Veijalainen, N. and Porarinsson, O., 2015. 912 Identification of coherent flood regions across Europe by using the longest streamflow 913 records. J. Hydrol.. 528, 341-360. http://.dx.doi.org/10.1016/j.jhydrol.2015.06.016. 914 Meinshausen, M., Smith, S.J., Calvin, K., Daniel, J.S., Kainuma, M.L.T., Lamarque, J-915 F., Matsumoto, K., Montzka, S.A., Raper, S.C.B., Riahi, K., Thomson, A., Velders, 916 G.J.M. and van Vuuren, D.P.P., 2011. The RCP greenhouse gas concentrations and 917 their extensions from 1765 to 2300. Climatic Change. 109 (1-2), 213-241. 918 http://dx.doi.org/10.1007/s10584-011-0156-z. 919 Mimikou, M.A., Baltas, E., Varanou, E., Pantazis, K., 2000. Regional impacts of climate 920 change on water resources quantity and quality indicators. J. Hydrol. 234, 95–109. 921 Monteith, J.L., 1965. Evaporation and environment: the state and movement of water in 922 living organisms. Symposia of the Society for Experimental Biology.19, 205–234. 923 Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L., Harmel, R.D. and Veith, 924 T.L., 2007. Model evaluation guidelines for systematic quantification of accuracy in 925 watershed simulations. Watershed Simulations. T. ASABE. 50 (3), 885–900. 926 39 Moss, R. H., Edmonds, J.A., Hibbard, K.A., Manning, M.R., Rose, S.K., van Vuuren, 927 D.P., Carter, T.R., Emori, S., Kainuma, M., Kram, T., Meehl, G.A., Mitchell, J.F.B., 928 Nakicenovic, N., Riahi, K., Smith, S.J., Stouffer, R.J., Thomson, A.M., Weyant, J.P. and 929 Wilbanks, T.J., 2010. The next generation of scenarios for climate change research 930 and assessment. Nature. 463 (7282), 747-756. http://dx.doi.org/10.1038/nature08823. 931 Murphy, J.M., Sexton, D.M.H., Barnett, D.N., Jones, G.S., Webb, M.J. and Stainforth, 932 D.A., 2004. Quantification of modelling uncertainties in a large ensemble of climate 933 change simulations. Nature, 430, 768–772. http://dx.doi.org/10.1038/nature02771. 934 Nash, J.E. and Sutcliffe, J.V., 1970. River flow forecasting through conceptual models: 935 Part 1. A discussion of principles. J. Hydrol.10 (3), 282-290. 936 http://dx.doi.org/10.1016/0022-1694(70)90255-6. 937 Nijssen, B., O’donnell, G.M., Hamlet, A.F. and Lettenmaier, D.P., 2001. Hydrologic 938 sensitivity of global rivers to climate change. Clim. Chang. 50, 143–175. 939 Ouarda, T.B.M.J., Charron, C. and St-Hilaire, A., 2008. Statistical models and the 940 estimation of low flows. Can. Water Resour. J. 33 (2), 195-206. 941 http://dx.doi.org/10.4296/cwrj3302195. 942 Overton, D.E., 1966. Muskingum flood routing of upland streamflow. J. Hydrol. 4, 185-943 200. http://dx.doi.org/10.1016/0022-1694(66)90079-5. 944 Peraza-Castro, M., Ruiz-Romera, E., Montoya-Armenta, L.H., Sanchez-Perez, J.M. 945 and Sauvage, S., 2015. Evaluation of hydrology, suspended sediment and nickel loads 946 in a small watershed in Basque Country (northern Spain) using eco-hydrological SWAT 947 model. Ann. Limnol-Int. J. Lim. 51 (1), 59-70. http://dx.doi.org/10.1051/limn/2015006. 948 40 Perez, J., Menendez, M., Mendez, F. and Losada, I., 2014. Evaluating the performance 949 of CMIP3 and CMIP5 global climate models over the north-east Atlantic region. Clim. 950 Dynam. 43 (9-10), 2663-2680. http://dx.doi.org/10.1007/s00382-014-2078-8. 951 Petisco, S.E. and Martín, J.M., 2006. Escenarios de temperatura y precipitación para la 952 España peninsular y Baleares durante el período 2001-2100 basados en “downscaling” 953 estadístico mediante métodos de análogos. XXIX Jornadas Científicas de la 954 Asociación Meteorológica Española. Pamplona. 955 Priestley, C.H.B. and Taylor, R.J., 1972. On the assessment of surface heat flux and 956 evaporation using large-scale parameters. Monthly Weather Review,100, 81–92. 957 Prudhomme, C., Young, A., Watts, G., Haxton, T., Crooks, S., Williamson, J., Davies, 958 H., Dadson, S. and Allen, S., 2012. The drying up of Britain? A national estimate of 959 changes in seasonal river flows from 11 regional climate model simulations. 960 Hydrological Processes. 26 (7), 1115-1118. http://dx.doi.org/10.1002/hyp.8434. 961 Qiu, L.J., Zheng, F.L. and Yin, R.S., 2012. SWAT-based runoff and sediment 962 simulation in a small watershed, the loessial hilly-gullied region of China: Capabilities 963 and challenges. Int. J. Sediment Res. 27:226–234. http://dx.doi.org/10.1016/S1001-964 6279(12)60030-4. 965 Ribalaygua, J., Pino, M.R., Pórtoles, J., Roldán, E., Gaitán, E., Chinarro, D. and 966 Torres, L., 2013. Climate change scenarios for temperature and precipitation in Aragón 967 (Spain). Sci. Total Environ. 463–464 (0), 1015-1030. 968 http://dx.doi.org/10.1016/j.scitotenv.2013.06.089. 969 Shorthouse, C. and Arnell, N., 1999. The effects of climate variability on spatial 970 characteristics of European river flows. Phys. Chem. Earth 24 (1–2), 7–13. 971 41 Sloan, P.G. and Moore, I.D., 1984. Modeling surface and subsurface stormflow on 972 steeply sloping forested watersheds. Water Resour. Res. 20 (12), 1815-1822. 973 http://dx.doi.org/10.1029/WR020i012p01815. 974 Smakhtin, V.Y., 2001. Low flow hydrology: a review. J. Hydrol. 240, 147-186. 975 http://dx.doi.org/10.1016/S0022-1694(00)00340-1. 976 Stahl, K., Tallaksen, L.M., Gudmundsson, L. and Christensen, J.H.,2011. Streamflow 977 data from small basins: a challenging test to high resolution regional climate modeling, 978 J. Hydrometeorol. 12, 900–912. 979 Steele-Dunne, S., Lynch, P., McGrath, R., Semmler, T., Wang, S., Hanafin, J. and 980 Nolan, P., 2008. The impacts of climate change on hydrology in Ireland. J. Hydrol. 356 981 (1-2), 28-45. http://dx.doi.org/10.1016/j.jhydrol.2008.03.025. 982 Szêpszô, G., Lingemann, I., Klein, B., and Kovacs, M., 2014. Impact of climate change 983 on hydrological conditions of Rhine and upper Danube rivers based on the results of 984 regional climate and hydrological models. Nat. Hazards. 72 (1), 241-262. 985 Http://dx.doi.org/10.1007/s11069-013-0987-1. 986 Tavakoli, M., De Smedt, F., Vansteenkiste, T., and Willems, P., 2014. Impact of climate 987 change and urban development on extreme flows in the Grote Nete watershed, 988 Belgium. Nat. Hazards. 71(3), 2127-2142. http://dx.doi.org//10.1007/s11069-013-1001-989 7. 990 Taylor, K. E., Stouffer, R. J. and Meehl, G. A., 2012. An overview of CMIP5 and the 991 experiment design. American Meteorological Society. 93, 485-498. doi:10.1175/BAMS-992 D-11-00094.1. 993 48 Table 3. Most sensitive parameters (ranked from 1 the most sensitive and 13 the less 1091 sensitive) in the Upper Nerbioi River catchment, their description, the range used for 1092 the autocalibration (p-factor 0.81 and r-factor 0.41) and the best value. 1093 Change type Variable name Description Range Best value r CN2 Curve number -0.2-+0.2 -0.07 v ESCO Soil evaporation compensation factor 0.77-0.86 0.83 v GWQMN Depth of water in the shallow aquifer required for return flow to occur 614-655 625.36 r SOL_AWC Available water capacity 0.1-0.5 0.48 v EPCO Plant uptake compensation factor 0.8-0.95 0.87 v REVAPMN Threshold water in shallow aquifer 768-900 892.21 v CH_K2 Main channel conductivity 10-44 38.66 v ALPHA_BF Base flow alpha factor 0.6-0.9 0.77 v SURLAG Surface runnoff lag coefficient 0.5-2.5 1.32 v SMTMP Snow melt base temperature (ºC) 3-9 4.77 v GW_DELAY Delay time for aquifer recharge 1-20 1.4 v SFTMP Snowfall temperature (ºC) 0.39-1.5 0.62 v GW_REVAP Groundwater “revap” coefficient 0.017-0.04 0.026 “v” means the default parameter is replaced by a given value; “r” means the existing parameter value is changed 1094 relatively 1095 1096 49 Table 4. Values obtained for the statistical indices used in the evaluation of the SWAT 1097 model performance at daily time-step. Seasonal statistical values are calculated for the 1098 1996-2013 period. 1099 DISCHARGE Scale NSE r2 slope/int. PBIAS RSR CALIBRATION 1996-2006 0.63 0.68 0.85/0.35 -1.00 0.61 VALIDATION 2007-2013 0.75 0.77 0.91/0.24 0.17 0.50 HIGH AI 2010-2012 0.67 0.76 1.02/0.16 -10.51 0.58 LOW AI 2003-2005 0.74 0.76 1.01/0.16 -5.16 0.51 ALL 1996-2013 1996-2013 0.69 0.72 0.89/0.3 -0.51 0.56 WINTER 0.66 0.68 0.81/0.62 6.26 0.58 SPRING 0.74 0.76 0.87/0.1 17.74 0.51 SUMMER 0.29 0.40 0.61/0.06 12.23 0.84 AUTUMN 0.61 0.72 0.98/0.77 -26.78 0.63 * According to Moriasi et al., (2007) the discharge simulation is satisfactory at monthly time step when the 1100 NSE > 0.5, r2 > 0.5, RSR ≤ 0.7, and PBIAS < 25%. The best value for slope is 1 and 0 for intercept (Arnold 1101 et al., 2012). 1102 1103 1104 50 Table 5. Sign (+ or -) and probability of occurrence (P) of the annual and seasonal 1105 trends detected for the duration (days) of the period below Q20 and above Q80 1106 between 1961 and 2000. Trends with a P higher than 0.66 are represented in bold; 1107 positive values are italicised. 1108 OBS_SIM ACCESS10_AN BNUESM_AN BNUESM_SDSM MPI-ESMRL_AN MPI-ESMRL_SDSM MPI-ESMMR_AN MPI-ESMMR_SDSM CMCCCESM_AN CMCCCESM_SDSM Q20 YEAR 0.39 0.58 0.34 0.91 -0.45 0.73 -0.45 0.50 0.93 0.34 AUTUMN -0.52 0.26 -0.08 0.37 -0.06 0.47 -0.99 0.04 0.65 0.96 WINTER 0.61 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 SPRING 0.86 0.08 0.68 0.16 0.47 0.73 -0.40 -0.81 -0.64 -0.05 SUMMER 0.32 -0.21 0.52 0.85 -0.78 0.45 0.41 0.81 -0.64 -0.41 Q80 YEAR -0.99 -0.17 -0.91 -0.90 0.02 0.02 0.32 0.47 0.37 0.34 AUTUMN -0.75 0.50 -0.98 -1.00 -0.75 -0.50 0.25 0.71 -0.62 -0.52 WINTER -0.68 -0.76 0.49 0.63 0.76 0.62 -0.28 -0.49 -0.62 0.53 SPRING -0.74 -0.64 0.00 -0.48 -0.42 0.37 0.65 0.37 0.77 0.39 SUMMER -0.91 0.00 -0.84 -0.85 0.00 0.09 0.00 0.00 -0.35 0.00 1109 1110 51 1111 Fig. 1. Atlantic region area as described by the IPCC (2007). The location of the 1112 research works summarized in Table 1 is represented in by numbers. The italicised 1113 numbers in bold refer to works carried out in more than two catchments. Location of the 1114 Upper Nerbioi in this context and a map of river catchment with hydro-meteorological 1115 stations settings are also included. 1116 1117 Fig. 2. Daily observed (OBS) and simulated (SIM) discharge for both the calibration 1118 (1996-2006) and the validation (2007-2013) periods, and the precipitation (PCP) 1119 observed in Amurrio station (AEMET 1060). 1120 52 1121 Fig. 3. Difference between observed meteorological parameters; precipitation (PCP) 1122 and average temperature (TMEAN) and climate baselines (1961-2000) before applying 1123 bias correction at annual and seasonal scales. 1124 1125 Fig. 4. Monthly mean discharge (m3 s-1) from 1961 to 2000 obtained from the 1126 hydrological simulation with observed meteorological data (OBS_SIM) and from the 1127 hydrological simulation with the downscaled GCMs baselines. 1128 53 1129 Fig. 5. Annual discharge difference (%) between the 16 hydrological projections and its 1130 respective baseline simulations, divided into three 30-year horizons (2030s, 2060s, 1131 2090s). 1132 1133 Fig. 6. Seasonal discharge difference (%) between CMCC_CESM_AN_R85 and 1134 CMCC_CESM_SDSM_R85 hydrological projections and their respective baselines 1135 divided into three 30-year horizons (2030s, 2060s, 2090s). In addition, the mean 1136 annual and seasonal discharge (m3s-1) is indicated in each bar. 1137 54 1138 Fig. 7. Annual and seasonal discharge difference (%) between hydrological projections 1139 and their respective baselines grouped by downscaling method and RCP. The figure 1140 shows the mean difference between: 1141 - ACCES1-0_AN_R45, BNU-ESM_AN_R45, MPI-ESM-MR_AN_R45 and MPI-1142 ESM-RL_AN_R45, represented as AN_R45. 1143 - BNU-ESM_SDSM_R45, MPI-ESM-MR_SDSM_R45 and MPI-ESM-1144 RL_SDSM_R45, represented as SDSM_R45. 1145 - ACCES1-0_AN_R85, BNU-ESM_AN_R85, MPI-ESM-MR_AN_R85 and MPI-1146 ESM-RL_AN_R85, represented as AN_R85. 1147 - BNU-ESM_SDSM_R85, MPI-ESM-MR_SDSM_R85 and MPI-ESM-1148 RL_SDSM_R85, represented as SDSM_R85. 1149 The results are divided into 3 horizons (2030s, 2060s, 2090s). In addition, the mean 1150 annual and seasonal discharge (m3 s-1) is indicated in each bar. 1151 55 1152 Fig. 8. Mean monthly discharge (m3 s-1) simulated with 16 climate projections. The 1153 highest discharge values represent the maximum value of the mean monthly discharge 1154 of all the projections by month, while the lowest values represent the minimum. The 1155 results are divided into three 30-year horizons (2030s, 2060s, 2090s). The grey colour 1156 represents the range of possible discharge values and the observed mean monthly 1157 discharge (1961-2000) is shown (OBS_SIM). 1158 1159 1160 Fig. 9. Trends for low flow (Q20) duration and high flow (Q80) duration displayed at 1161 annual and seasonal scales for the 2011-2100 period. The projections under 1162 Representative Concentration Pathway 4.5 (RCP 4.5) and 8.5 (RCP 8.5) are displayed 1163 separately. Only values with a probability of occurrence higher than 0.66 are shown. 1164