Received: 25 April 2019 | Revised: 3 July 2019 | Accepted: 9 July 2019 DOI: 10.1002/ece3.5506 ORIGINAL RESEARCH Future hydrological constraints of the Montseny brook newt (Calotriton arnoldi) under changing climate and vegetation cover José L. J. Ledesma1,2 | Albert Montori3 | Vicent Altava‐Ortiz4 | Antonio Barrera‐Escoda5 | Jordi Cunillera5 | Anna Àvila6 1 Center for Advanced Studies of Blanes, Spanish National Research Council, Abstract Blanes, Spain The Montseny brook newt (Calotriton arnoldi) is a critically endangered amphibian 2 Department of Aquatic Sciences and species which inhabits a small 20 km2 holm oak and beech forest area in NE Spain. Assessment, Swedish University of Agricultural Sciences, Uppsala, Sweden Calotriton arnoldi strictly lives in running waters and might be highly vulnerable to 3 GRENP (Grup de Recerca de l'Escola de la hydrological perturbations expected to occur under climate and vegetation cover Natura de Parets del Vallès), Life‐Tritó del Montseny, Diputació de Barcelona, Parets changes. Knowledge about the potential response of the species habitat to environ‐ del Vallès, Spain mental changes can help assessing the actions needed for its conservation. Based on 4 Department of Applied Research and knowledge of the species and supported by observations, we proposed daily low and Modelling, Meteorological Service of Catalonia, Barcelona, Spain high streamflow event thresholds for the viability of C. arnoldi. We used the rainfall– 5 Department of Climatology, Meteorological runoff model PERSiST to simulate changes in the frequency and duration of these Service of Catalonia, Barcelona, Spain 6 events, which were predicted under two climate and four vegetation cover scenarios CREAF, Campus de Bellaterra (UAB), Cerdanyola del Vallès, Spain for near‐future (2031–2050) and far‐future (2081–2100) periods in a reference catch‐ ment. All future scenarios projected a significant decrease in annual streamflow (from Correspondence José L. J. Ledesma, Center for Advanced 21% to as much as 67%) with respect to the reference period. The frequency and Studies of Blanes, Spanish National Research length of low streamflow events will dramatically increase. In contrast, the risk of Council (CEAB‐CSIC), Accés a la Cala Sant Francesc 14, 17300 Blanes, Spain. catastrophic drift linked to high streamflow events was predicted to decrease. The Email: jose.ledesma@ceab.csic.es potential change in vegetation toward an expansion of holm oak forests will be more Funding information important than climate changes in determining threshold low flow conditions. We Svenska Forskningsrådet Formas, Grant/ thus demonstrated that consideration of potential changes in vegetation and not only Award Number: 2015-1518; Spanish Government through a Juan de la Cierva changes in climate variables is essential in simulating future streamflows. This study grant, Grant/Award Number: FJCI-2017- shows that future low streamflow conditions will pose a severe threat for the survival 32111; Spanish Government through a MICINN project, Grant/Award Number: of C. arnoldi and may help taking management actions, including limiting the expan‐ CGL2017-84687-C2-2-R; LIFE‐Tritó sion of holm oak forest, for ameliorating the species habitat and help its conservation. Montseny project from the European Commission, Grant/Award Number: LIFE15 KEYWORDS NAT/ES/000757 amphibians, Calotriton arnoldi conservation, catchment management, endangered species, environmental change, Mediterranean climate, PERSiST model, statistical downscaling This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2019 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd. 9736 | www.ecolevol.org  Ecology and Evolution. 2019;9:9736–9747. LEDESMA et al. | 9737 1 | I NTRO D U C TI O N Prey abundance in these refuge microsites may be very low, po‐ tentially impairing an appropriate development of the individuals The Montseny brook newt (Calotriton arnoldi; Figure 1) was first de‐ (Clergue‐Gazeau, 1969, 1971). Other hydrological threats concern scribed in 2005 as a new amphibian taxon in the Montseny moun‐ just the opposite, the risks associated with very high flows. Newts tains, 40 km NE of Barcelona (Spain) (Carranza & Amat, 2005). and other organisms inhabiting running waters are highly vulnera‐ Phylogenetic analysis based on 12S and 16S DNA sequences indi‐ ble to high streamwater velocity and flooding (Walls, Barichivich, cated that the newt populations from Montseny had evolved sepa‐ & Brown, 2013). These events have been reported to cause cata‐ rately from the Pyrenean brook newt (Calotriton asper) populations strophic drift in C. asper (Colomer, Montori, García, & Fondevilla, for more than one million years and now they constitute two clearly 2014; Montori et al., 2012). differentiated species (Valbuena‐Ureña, Amat, & Carranza, 2013). Consequently, C. arnoldi might be highly vulnerable to hydro‐ Calotriton arnoldi is found in only seven locations distributed in a logical perturbations, either linked to low flows during dry peri‐ small 20‐km2 area, and catchment divides appear to constitute a ods or linked to high hydraulic discharge associated with extreme natural barrier that hampers interbreeding between populations precipitation events. Both situations may be more frequent in a (Carranza & Amat, 2005). Because of this lack of connectivity world undergoing climate change. Precipitation has not shown a and the low population densities (an estimated total of ca. 2,000 clear trend in recent decades in NE Iberian Peninsula (González‐ adults), C. arnoldi is highly endangered and it is included in the Hidalgo, López‐Bustins, Štepánek, Martín‐Vide, & Luisa, 2009; del IUCN Red List of Threatened Species. The Diputació de Barcelona Río, Herrero, Fraile, & Penas, 2011), but air temperature is rising was awarded a LIFE project (LIFE‐Tritó Montseny) in 2016 devised (Vicente‐Serrano et al., 2014) and, particularly, the Montseny mas‐ to protect C. arnoldi. Among others, project objectives included in‐ sif has experienced an increase of around 0.3°C per decade in the creasing the scientific knowledge of the ecological requirements of second half of the 20th century (Peñuelas & Boada, 2003). A tem‐ the species and assessing possible threats that may affect it. There perature increase can lead to enhanced evapotranspiration (Wang, is still little knowledge about the species ecological requirements Dickinson, & Liang, 2012), which eventually will result in reduced and its habitat potential response to environmental changes. This streamflow. Besides climate, landscape cover type has also an in‐ information is of paramount importance to guide efforts for its fluence on water resources. Several catchment studies have shown conservation. that an expansion of forest cover leads to a decrease in water yield Calotriton arnoldi is found in oligotrophic, cold, fast running (Bosch & Hewlett, 1982; Gallart & Llorens, 2004). In Montseny, the waters draining typical holm oak (Quercus ilex) and beech (Fagus holm oak forest appears to be expanding with altitude at the ex‐ sylvatica) forests in Montseny and it appears to completely depend pense of beech forests and heathlands (Peñuelas & Boada, 2003), on aquatic resources, as neither larvae nor adults have been found which may further contribute to reduce streamflow by increas‐ to inhabit the terrestrial environment. Its remarkable morpholog‐ ing evapotranspiration. Thus, both climate and vegetation cover ical traits, including reduced lung volume that may reduce buoy‐ changes may threaten C. arnoldi populations. However, so far no ancy, and its thin skin provide further evidence of its adaptation to studies have quantified the magnitude, duration, and frequency of the aquatic environment (Valbuena‐Ureña et al., 2013). Therefore, future hydrological extremes in the region that can affect the spe‐ drying out of headwater streams will pose a serious risk for the cies survival. Hydrological consequences of climate and vegetation species, either directly because of the unfavorable habitat con‐ changes are better explored at the catchment scale using rainfall– ditions, or indirectly due to the need of the individuals to move runoff models, which can be calibrated against observed stream‐ to wetted refuge areas in streambanks or to the hyporheic zone. flow (Lupon, Ledesma, & Bernal, 2018; Oni et al., 2016; Seibert & McDonnell, 2010). Here, we used the rainfall–runoff model PERSiST (Precipitation, Evapotranspiration and Runoff Simulator for Solute Transport; Futter et al., 2014) to investigate how changes in herein proposed daily low and high hydrological thresholds might threat popula‐ tions of C. arnoldi. Streamflow was simulated at two future inter‐ vals (near‐future 2031–2050 and far‐future 2081–2100) under two climate scenarios (RCP4.5 and RCP8.5) in a reference catchment in Montseny. The hydrological effects of four plausible vegetation cover scenarios were also considered: (i) vegetation as present, (ii) displacement of holm oak and beech into heathland, (iii) expansion of holm oak over the beech but similar heathland cover, and (iv) holm oak covering the whole catchment. The relative role of climate versus vegetation cover in future hydrological extremes threaten‐ F I G U R E 1 Adult individual of Montseny brook newt (Calotriton ing C. arnoldi was also examined. Finally, management implications arnoldi) in its natural habitat were discussed. 9738 | LEDESMA et al. F I G U R E 2 Reference study catchment site including vegetation covers, gridded climate data locations (black squares), and streamflow measurement location (inverted black triangle). The location of the catchment within Spain is shown in the inset Higher altitudes are dominated by heathlands primarily composed 2 | M ATE R I A L S A N D M E TH O DS of heather (Calluna vulgaris and gramineae) and comprise 32.7% of the catchment. Holm oak (Quercus ilex) forest dominates the lower 2.1 | Catchment selection and characterization half (52.2%) of the catchment, whereas beech (Fagus sylvatica) for‐ Because C. arnoldi is catalogued as critically endangered in the IUCN est occupies a 200 m stripe on the north‐facing slope between the Red List of Threatened Species, providing details on specific loca‐ heathlands and the holm oak forest and covers 15.1% of the catch‐ tions where populations can be found is not advised in the context ment. Apart from low‐intensity grazing in the heathland‐dominated of the Life‐Tritó Montseny project. The Diputació de Barcelona is areas, there is currently no relevant human‐related activity in the developing strict plans to ensure the survival of C. arnoldi, and this catchment. information is considered sensitive. For this reason, here we used real data from an actual catchment at Montseny but refer to it as 2.2 | Streamflow and climate data “reference catchment” (RC0 hereafter), and avoided providing geo‐ graphical details of its location and on whether it harbors C. arnoldi Streamflow was measured at the outlet of RC0 between January or not. Because of the high similarity in climate, lithology, vegetation, 2011 and December 2015 in a stilling pond using a 120° V‐notch and topography of RC0 with other neighboring catchments (which weir equipped with a limnimetric scale. A pressure sensor system, might or might not harbor populations of C. arnoldi), we consider that consisting of a submerged Schlumberger Mini‐Diver sensor and a any results and conclusions obtained by studying this catchment can Schlumberger Baro‐Diver above the water surface, was used to be extrapolated to other sites at Montseny where C. arnoldi inhabits. record water levels every 15 min. Sensor water‐level data were RC0 is a 2 km2 Mediterranean forest headwater catchment lo‐ tested against manual water‐level readings from the limnimetric cated in the Montseny Natural Park, NE Spain (Figure 2). The climate scale, showing a strong correlation (R 2 = .98; p < .0001; N = 96). here is subhumid Mediterranean, with mild winters, wet springs Subsequently, high‐frequency measurements were converted to and autumns, and dry summers. Mean annual air temperature is streamflow through the specific stage‐flow rating curve for the 11.5 ± 0.5°C (average ± SD) and total annual precipitation averages location, and aggregated into daily data. A total of 1,618 daily 983 ± 247 mm, less of 3% falling as snow (gridded data from the streamflow observations were available, that is, 89% of the days Meteorological Service of Catalonia for the period 1971–2005). within the 5‐year observed period. Streamflow will be denoted Gentle slopes are present in the upper parts of the catchment, and here by areally normalized units of mm/day (or mm/year where ap‐ steeper slopes are found in mid‐ and downstream valleys (Figure 2). propriate) for easy comparison to other sites. The proposed daily LEDESMA et al. | 9739 TA B L E 1 Summary of vegetation cover proportion scenarios that routed the water inputs from precipitation to the soil, a shallow “soil” layer for fast runoff responses, and a deeper “groundwater” Holm oak Beech Heathland Scenario forest (%) forest (%) (%) layer for base flow regulation. The average of the gridded mean daily temperature and total daily precipitation from the two grid points (i) Vegetation as present 52.2 15.1 32.7 located within the catchment generated at the SMC was used as (ii) Beech increased 59.8 23.9 16.4 input data. Gridded climate data products are a reliable alternative (iii) Heathland protected 59.8 7.6 32.7 to instrumental measurements as inputs to rainfall–runoff models (iv) Holm oak whole 100 0 0 (Ledesma & Futter, 2017). The model was calibrated for the 5‐year period with streamflow low and high streamflow event thresholds will be denoted by units observations (2011–2015) using a combination of manual and au‐ of L/s, which are more intuitive for specific events in an ecological tomatic calibration techniques. Manual calibration has been proven context. as a robust method for obtaining acceptable simulations within the Gridded climate data were computed at the Meteorological Integrated Catchment (INCA) family of models (Cremona, Vilbaste, Service of Catalonia (SMC). Instrumental measurements of daily Couture, Nõges, & Nõges, 2017; Futter et al., 2014; Ledesma, Köhler, temperature (maximum and minimum) and precipitation were inter‐ & Futter, 2012), of which PERSiST is the common hydrological polated in space using a linear multiregression method (Altava‐Ortiz, model. An initial manual calibration employing sensible values and Bladé, & Llasat Botija, 2010). The process considered altitude, lati‐ thresholds based on expert judgement was used to approximate the tude, longitude, and sea distance within different climatic clusters dynamics and magnitude of the simulated streamflow to those of the from more than 1,250 rain gauges and 660 temperature stations observed streamflow. This step also helped identifying credible pa‐ distributed over Catalonia for the 1970–2015 period. Climatic clus‐ rameter ranges for a subsequent automated calibration procedure, ters were defined as a result of the maximization of monthly data which consisted on a Monte Carlo scheme that optimized the values spatial correlation in the entire region. Residual values (i.e., interpo‐ of four calibration criteria (i.e., performance metrics): (i) the Nash– lated minus observed values) were geographically corrected using Sutcliffe (NS) efficiency index (important to fit high flows; Nash & a kriging interpolation method, which took into account the spatial Sutcliffe, 1970), (ii) the log(NS) (important to fit low flows), (iii) the autocorrelation of the involved data. The interpolation was done for relative volume differences (RVD) of simulated versus observed 1‐km2 grid cells. Two of the data points from the grid laid within the streamflow (important to ensure that cumulative modeled and ob‐ RC0 catchment area, one in the upper part of the catchment and the served streamflows are as similar as possible), and (iv) the ratio of ob‐ second one in the valley near the outlet (Figure 2). Daily averages of served to simulated variance (VAR) in streamflow values (important these two points (for both mean temperature and total precipitation) because automated calibration strategies tend to produce modeled were calculated for the 5‐year period 2011–2015 and assumed to series that are less variable than observed data) (Oni et al., 2016). be representative of the whole catchment. In any case, there were The Monte Carlo process was carried out in 100 iterations of 1,000 only minor differences between the data series from these two grid runs each, and a single best‐performing parameter set was selected points (data not explicitly shown). for final manual tuning. Model parameter sensitivity was assessed using the 100 best‐ performing parameter sets obtained in the automated Monte Carlo 2.3 | Rainfall–runoff modeling: characterization and calibration. For each parameter, the ensemble of values from the 100 present‐day calibration parameter sets was compared to a rectangular distribution using a The rainfall–runoff model PERSiST (Futter et al., 2014) was used to Bonferroni‐corrected Kolmogorov–Smirnov (KS) test. A significant simulate streamflow at the outlet of the RC0 catchment (inverted KS statistic (adjusted p < .01) implied that the posterior distribution black triangle in Figure 2). PERSiST is a conceptual, semi‐distributed was not rectangular and thus that streamflow simulations were sen‐ tool for simulating water fluxes, including streamflow, through phys‐ sitive to the specific parameter (Futter et al., 2014). ically based routing schemes at daily time steps. It involves a flexible framework based on a number of interconnected buckets within a 2.4 | Selection of critical low and high streamflow mosaic of landscape units, which are specified by the modeler based thresholds for Calotriton arnoldi on the conceptual perception of the runoff generation process. The model requires daily time series of air temperature and precipitation Calotriton arnoldi has been very recently discovered, and at present, as input data. Parameter descriptions and other model details can be no direct observations exist that can provide critical threshold values found elsewhere (Futter et al., 2014; Ledesma & Futter, 2017; Lupon for deleterious low or high streamflows for the species. Furthermore, et al., 2018). an exhaustive bibliographic exploration did not return relevant re‐ The application of PERSiST to RC0 consisted of three landscape sults concerning the definition of threshold values for any amphibian units, which corresponded to the catchment vegetation covers de‐ species, which may have helped to constrain and compare our pro‐ scribed above (holm oak forest, beech forest, and heathland) in ap‐ posed low and high flow thresholds. Thus, here we propose thresh‐ propriate proportions, and three soil buckets including a “quick” box old values based on knowledge of C. arnoldi biological cycle (larval 9740 | LEDESMA et al. and adult phases), feeding habits and survival strategy, and observa‐ Four landscape scenarios were also considered to be combined tion of the RC0 hydrological behavior. A daily low flow threshold of with the climate scenarios in order to account for potential changes Qlow = 0.33 L/s was selected. At this streamflow value, downstream in vegetation cover proportions linked to the changes in climate. reaches near the RC0 catchment outlet show no superficial water Based on the results presented by Peñuelas and Boada (2003), the flow because they have low slopes and are covered by sand, gravels, following scenarios were considered: (i) Vegetation proportions are and pebbles (personal observation). Therefore, Qlow has been consid‐ maintained as present; (ii) holm oak forest takes over 50% of the ered as an adequate working threshold for dry conditions. beech forest area and, in its turn, beech forest takes over 50% of No observation or model results exist for C. arnoldi to determine the heathland area; (iii) holm oak forest takes over 50% of the beech the risk of catastrophic drift for high flow thresholds. Studies inves‐ forest area and the heathland area does not change; and (iv) holm tigating the related C. asper species have shown that intense precip‐ oak forest takes over the whole catchment area (Table 1). The single itation events (>50 mm/day) cause catastrophic drift of larvae and best‐performing parameter set obtained during model calibration adults, leading to reduced annual population numbers (Montori et was used to simulate future streamflows for the 16 future scenarios al., 2012). In the RC0 and other catchments at Montseny, high flow (2 periods × 2 climates × 4 vegetation covers) and for the reference peaks strongly depend on the antecedent conditions of humidity in scenario. the catchment soils (Àvila, Piñol, Rodà, & Neal, 1992). Precipitation events of 50 mm/day will result in very different streamflow re‐ 2.6 | Statistical analyses sponses depending on whether they occur in the dry (April to September) or wet season (October to March). Flow increase rate Analysis of variance (ANOVA) was applied to estimate the relative is thus the relevant variable to assess risk in the present case. To contribution of period (near future and far future), climate scenario depict this, we generated the variable “flow difference in consec‐ (RCP4.5 and RCP8.5), and vegetation cover scenario (Table 1) to the utive days” (Qdiff ), which was defined as the difference in observed total variation in the number of days with daily simulated streamflow streamflow between two consecutive days. Threshold values were not reaching/exceeding the specified thresholds: Q low = 0.33 L/s for defined as the 99th percentile of the Qdiff values for the whole year low flows, and Qdiff‐all = 97 L/s (whole year) and Qdiff‐dry = 42 L/s (dry (Qdiff‐all = 97 L/s; N = 1,612) and for the dry season (Qdiff‐dry = 42 L/s; season) for flow increase events in consecutive days. This was done N = 776). All days above Qdiff‐all (or Qdiff‐dry) were considered as sin‐ to disentangle the relative importance of climate versus vegetation gle‐day events independently on whether they were followed by an‐ cover in future hydrological extremes that might affect C. arnoldi. other day above Qdiff‐all (or Qdiff‐dry) or not. The fraction of total variation ascribed to each ANOVA component Besides optimizing the performance metrics described above (i.e., period, climate scenario, or vegetation cover scenario) was equal (“hard calibration” criteria), the PERSiST calibration also aimed at to the sum of squares for that component divided by the total sum of simulating a number of days below Q low as similar as possible to squares from the ANOVA, following Futter et al. (2011) and Ledesma the observed number of days below Q low (“soft calibration” criteria) et al. (2013). (Lupon et al., 2018). Analogously, the calibration aimed at simulating In order to investigate how the results would have changed if a number of days above Qdiff‐all and Qdiff‐dry as similar as possible to a different value of Q low would have been selected, we obtained the observed number of days above Qdiff‐all and Qdiff‐dry. the number of simulated days below Q low for each scenario (refer‐ ence + 16 future scenarios) using five other values of Q low (namely 2, 1, 0.5, 0.25, and 0.2 L/s). We considered that the conclusions of 2.5 | Future climate and hydrological projections the study would hold if the simulated days below Q low estimated for Two emission scenarios from the Representative Concentration the reference and the future scenarios using different Q low thresh‐ Pathways (RCP) were considered: the RCP4.5 (moderate carbon di‐ olds were linearly correlated with the analogous values using the oxide [CO2] emission scenario) and the RCP8.5 (extreme‐high CO2 selected Q low of 0.33 L/s. emission scenario), to obtain near‐future (2031–2050) and far‐fu‐ ture (2081–2100) daily time series of temperature and precipitation 3 | R E S U LT S at the same grid points used for generating inputs for streamflow calibration. Historical CO2 concentration values were used to obtain 3.1 | Streamflow model calibration and sensitivity the reference period 1981–2000. Reference and future projected analysis data were based on a statistical downscaling technique performed at the SMC (Altava‐Ortiz et al., 2010) and driven by a global set of For the observation period (2011–2015), streamflow was dominated daily simulations from the Max‐Planck Institute Earth System Model by low base flow. RC0 showed fast responses to some (but not (MPI‐ESM; Giorgetta et al., 2013), which were available from the all) precipitation events, which in the large storm cases generated World Data Centre for Climate Portal CERA (https​://cera-www. as much as 52 mm/day of streamflow. A cross‐correlation analysis dkrz.de). Daily precipitation and temperature data from the two grid showed that the time lag that maximized the correlation between points used for calibration were again averaged prior using them as daily precipitation and daily observed streamflow was 1 day (Figure inputs in PERSiST to simulate future streamflows. S1). PERSiST captured remarkably well this pattern including both LEDESMA et al. | 9741 Precipitation (mm) 100 80 60 40 20 0 60 PERSiST simulated streamflow 50 Measured streamflow 40 Streamflow (mm) 30 20 10 0 May-11 May-12 May-13 May-14 May-15 Mar-12 Mar-13 Nov-15 Mar-14 Jan-11 Jul-11 Jan-12 Jul-12 Sep-12 Jan-13 Jul-13 Sep-13 Jan-14 Jul-14 Jan-15 Jul-15 Sep-11 Sep-14 Sep-15 Mar-11 Nov-11 Nov-12 Nov-13 Nov-14 Mar-15 F I G U R E 3 Daily observed and simulated streamflow (lower panel) and corresponding precipitation (upper panel) time series in the reference catchment the dynamics and the magnitude of streamflow and even the ex‐ water in the “soil” and “groundwater” layers of the holm oak forest tremely high flow events, except for an intermediate event at the and heathland vegetation covers, with lower values producing better end of 2015 (Figure 3). The ratio of observed to simulated variance model performances. The other sensitive parameter was the so‐called was 0.97, indicating just a slightly higher variance in the simulated “evapotranspiration adjustment” in the holm oak forest, which is used flows. Cumulative observed streamflow was only slightly underes‐ to limit evapotranspiration under dry conditions. Higher values of the timated by the cumulative simulated streamflow (3%). The NS ef‐ evapotranspiration adjustment generated better model performances. ficiency index was 0.89, indicating a very good fit of high flows. Finally, the log(NS) was 0.89, indicating a very good fit of low flows. 3.2 | Projected future climate PERSiST was also able to simulate a very similar amount of days below and above the proposed daily low and high streamflow According to both RCP4.5 and RCP8.5 scenarios, mean annual tem‐ threshold values compared with the observed amount of days. The perature in the near future (2031–2050) is projected to increase number of days in which streamflow was below Q low in the obser‐ 1.3°C with respect to the mean annual temperature (11.6°C) of the vation record (N = 1,618) was 44. For the same record, the number reference period (1981–2000). In the far future (2081–2100), tem‐ of daily simulated streamflows below this threshold was 42. There perature will increase 1.8°C according to the RCP4.5 scenario and as were 16 observed single‐day events above Qdiff‐all, whereas PERSiST much as 3.9°C according to the RCP8.5 scenario. In the near future, simulated 14 single‐day events. For the Qdiff‐dry threshold, there the increase will be more moderate in winter and spring months, were eight observed versus six simulated single‐day events. whereas it will be more pronounced in summer and, especially, au‐ Simulated streamflow was sensitive to four out of the 28 model tumn (Figure 4a). The increase in temperature in the far future will parameters tested using the Bonferroni‐corrected KS test (Table S1). be more evenly distributed throughout the year, although slightly Three of the sensitive parameters related to the residence time of less pronounced in winter and slightly more pronounced in summer. 9742 | LEDESMA et al. Average annual precipitation will decrease with respect to the streamflow events between 10 and 30 mm/day will be less frequent reference period value (941 mm/year) in all future scenarios. In the in both the near‐ and, especially, the far future. Intermediate and near future, RCP4.5 projected an 11% decrease, whereas RCP8.5 base flow conditions will yield lower stream water discharge than projected a 16% decrease (down to 805 mm/year). Similarly, RCP4.5 reference conditions according to all scenarios, that is, for any given projected a 17% decrease in the far future, whereas RCP8.5 pro‐ exceedance time above 10%, streamflow magnitude will be lower in jected a 28% decrease for this period (i.e., down to 714 mm/year). all cases (Figure 5b). In all cases, the decrease will be relatively more pronounced in the Consequently, all scenarios projected a significant overall de‐ already dry summer months, whereas winter precipitation will be crease in average annual streamflow with respect to the reference relatively similar to reference period precipitation (Figure 4b). period (346 mm/year), from a 21% decrease (near‐future RCP4.5 with vegetation as present) to as much as a 67% decrease (far‐fu‐ ture RCP8.5 with whole holm oak forest cover; Table 2). In parallel, 3.3 | Projected future streamflow the number of days with simulated streamflow lower than Q low will Daily extreme high flow events (Q > 30 mm/day) at RC0 will be as dramatically increase with respect to the reference period (Table 2; frequent in the near future as in the reference period (ca. 0.5% of Figure 5b). Most scenarios simulated streamflow below Qlow during the time) according to all scenarios (Figure 5a). On the contrary, all 10%–20% of the time, with a maximum of 30% for the far‐future far‐future scenarios projected less frequency of these events. Daily RCP8.5 scenario with whole holm oak forest cover, whereas for the reference period the same estimate was 2.6%. Far‐future and 30 RCP8.5 scenarios will tend to be drier than their corresponding near‐ 1981–2000 (a) future and RCP4.5 scenarios, respectively. Whole holm oak forest 2031–2050 RCP4.5 cover scenarios projected remarkably dry conditions, whereas sce‐ 25 2031–2050 RCP8.5 narios with an increased beech forest cover or a protected heathland 2081–2100 RCP4.5 2081–2100 RCP8.5 cover projected similar conditions, which were slightly drier than the 20 corresponding vegetation as present scenarios. Temperature (oC) In the reference period, most events with consecutive days 15 below Qlow were shorter than a week and only two events were longer than a month (Figure 6). However, the number and length of these events will dramatically increase in the future, with all scenar‐ 10 ios predicting at least one event with consecutive days below Q low longer than 3 months (Table 2; Figure 6). The number and length 5 of these events was especially magnified during the whole holm oak forest cover scenarios. In the near‐future scenarios, days below 0 Q low will be approximately homogeneously distributed over winter, 300 summer, and autumn months, with lower proportions during spring (b) (Table S2). In the far future, as in the reference period, days below 250 Q low will be more predominant during summer. During the reference period, maximum Qdiff was ca. 35 mm and the single‐day events larger than Qdiff‐all occurred mainly through Precipitation (mm) 200 November to March, typically the wettest and coldest months in Montseny. In all future scenarios, especially the RCP8.5 scenarios, 150 the number of single‐day events above Qdiff‐all will decrease (Table 2) and concentrate in March and December. A number of single‐day 100 events above Qdiff‐dry were slightly less than half of those above Qdiff‐ all in all cases. Furthermore, Qdiff had no clear relationship with the amount of precipitation during the single‐day event for any of the 50 scenarios or the observations, that is, the same precipitation amount (including high rainfall events) generated a wide range of Qdiff values, 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec especially during the dry season (Figure 7). The ANOVA (Table S3) revealed that vegetation cover scenarios F I G U R E 4 Monthly average temperature (a) and precipitation explained more of the variance in the number of days with simulated (b) for the reference period (1981–2000) and for Representative streamflow below Q low (50%) than the climate scenarios (11%) and Concentration Pathway (RCP) scenarios 4.5 and 8.5 for near‐ the period considered (21%) (Table 3). In contrast, climate scenarios (2031–2050) and far future (2081–2100). Error bars in (a) are standard deviations. Dotted light lines in (b) are 90th (above) and explained most of the variance in the number of simulated single‐day 10th (below) percentiles events above Qdiff‐all (47%) and Qdiff‐dry (53%), whereas vegetation LEDESMA et al. | 9743 F I G U R E 5 Flow duration curves 60 1981-2000 10.00 for 0%–10% (a) and 10%–100% (b) time (a) 2031-2050 RCP4.5 i (b) 2031-2050 RCP8.5 i exceedance for each of the 16 scenarios 2031-2050 RCP4.5 ii 50 2031-2050 RCP8.5 ii considered (2 future periods × 2 climate 2031-2050 RCP4.5 iii 2031-2050 RCP8.5 iii scenarios × 4 vegetation cover scenarios) 2031-2050 RCP4.5 iv 1.00 and for the reference period. Note the 40 2031-2050 RCP8.5 iv 2081-2100 RCP4.5 i logarithmic scale in the x‐axis in (a) and 2081-2100 RCP8.5 i Streamflow (mm) 2081-2100 RCP4.5 ii the y‐axis in (b). Dotted line corresponds 2081-2100 RCP8.5 ii to the daily low flow threshold 30 2081-2100 RCP4.5 iii 2081-2100 RCP8.5 iii 0.10 Q low = 0.33 L/s 2081-2100 RCP4.5 iv 2081-2100 RCP8.5 iv 20 0.01 10 0 0.00 0.0% 0.1% 1.0% 10.0% 10% 25% 40% 55% 70% 85% 100% Percentage of time flow equalled or exceeded Percentage of time flow equalled or exceeded TA B L E 2 Number of days with daily simulated streamflow not reaching/exceeding the specified thresholds: Q low = 0.33 L/s for low flows, and Qdiff‐all = 97 L/s (whole year) and Qdiff‐dry = 42 L/s (dry season) for flow increase events in consecutive days Period Climate scenario Vegetation scenario Q low (days) Q low (events) Qdiff‐all (days) Qdiff‐dry (days) Q (mm/year) 1981–2000 Reference i 187 11 74 33 346 2031–2050 RCP4.5 i 758 31 63 25 272 2031–2050 RCP8.5 i 722 24 49 22 238 2031–2050 RCP4.5 ii 853 32 52 20 240 2031–2050 RCP8.5 ii 802 23 46 17 208 2031–2050 RCP4.5 iii 760 33 61 25 270 2031–2050 RCP8.5 iii 721 24 50 20 236 2031–2050 RCP4.5 iv 1,527 58 49 17 196 2031–2050 RCP8.5 iv 1,376 39 40 15 166 2081–2100 RCP4.5 i 854 31 54 26 238 2081–2100 RCP8.5 i 1,369 62 45 18 165 2081–2100 RCP4.5 ii 936 32 48 22 208 2081–2100 RCP8.5 ii 1,533 63 36 15 143 2081–2100 RCP4.5 iii 852 30 54 26 235 2081–2100 RCP8.5 iii 1,369 62 43 17 164 2081–2100 RCP4.5 iv 1,393 48 46 27 166 2081–2100 RCP8.5 iv 2,152 68 34 14 114 Note: Data are for each of the 16 scenarios considered (2 future periods × 2 climate scenarios × 4 vegetation cover scenarios) and for the reference period. Number of events with consecutive days below Qlow and annual average streamflow are also shown. All simulated days above Qdiff‐all or Qdiff‐ dry were considered independent (i.e., single‐day events), and thus, the number of days reported corresponds to the number of events. cover scenarios explained a moderate amount of variance. The pe‐ 4 | D I S CU S S I O N riod considered explained 17% of the variance in the number of simulated single‐day events above Qdiff‐all and a negligible variance 4.1 | Overall streamflow model performance and of above simulated Qdiff‐dry single‐day events (Table 3). Simulated catchment hydrological behavior days below Q low using a range of Q low thresholds were linearly cor‐ related with the simulated days below Q low from the selected Qlow of PERSiST reproduced remarkably well the flow magnitude and 0.33 L/s, as obtained from the reference and the 16 future scenarios dynamics at RC0, as well as the number of days below Q low and (Figure S2). number of single‐day events above Qdiff‐all and Qdiff‐dry. Combining 9744 | LEDESMA et al. 20 F I G U R E 6 Duration and number of 1981–2000 2031–2050 RCP4.5 i events with consecutive days below the 18 Number of events with consecutive days below Qlow 2031–2050 RCP8.5 i daily low flow threshold Q low = 0.33 L/s 2031–2050 RCP4.5 ii 2031–2050 RCP8.5 ii for each of the 16 scenarios considered (2 16 2031–2050 RCP4.5 iii future periods × 2 climate scenarios × 4 2031–2050 RCP8.5 iii 14 2031–2050 RCP4.5 iv vegetation cover scenarios) and for the 2031–2050 RCP8.5 iv reference period 2081–2100 RCP4.5 i 12 2081–2100 RCP8.5 i 2081–2100 RCP4.5 ii 2081–2100 RCP8.5 ii 10 2081–2100 RCP4.5 iii 2081–2100 RCP8.5 iii 2081–2100 RCP4.5 iv 8 2081–2100 RCP8.5 iv 6 4 2 0 <7 days 7 days - 14 days 15 days - 30 days 31 days - 60 days 61 days - 90 days >90 days Duration of the event specific streamflow characteristics (here the specific thresholds here would have yielded the same results in relative terms (Figure S2), Q low, Qdiff‐all, and Qdiff‐dry) with commonly used model performance and therefore, the conclusions discussed below apply to a range of metrics (here NS, log(NS), RVD, and VAR) during calibration proce‐ low streamflow values that could have been selected as thresholds. dures is recommended to achieve accurate streamflow simulations Our simulations indicate that projected climate and vegetation for practical applications (Pool, Vis, Knight, & Seibert, 2017; Vis, cover changes will severely threaten C. arnoldi survival by increasing Knight, Pool, Wolfe, & Seibert, 2015). An accurate model simula‐ the frequency and duration of very low streamflows, which may lead tion was indeed achieved in the present application, suggesting to periods of prolonged lack of surficial flow. Previous studies with that a robust parameterization was used for future simulations of C. asper reported that climate change severely reduced the species streamflow. distribution range because of the reduced possibility of dispersal, Observed and simulated patterns of streamflow at RC0 indicate leading to a loss of genetic diversity (de Pous, Montori, Amat, & that the catchment is very flashy, that is runoff generation is no‐ Sanuy, 2016). Our results suggest that C. arnoldi will need to migrate tably fast, and the sensitivity analysis suggested that streamflow is in search of suitable habitats during periods of drought and the lack mostly sensitive to residence time of soil water and groundwater. of suitable sites may lead to important losses both in population indi‐ This suggests that short water residence times related to topogra‐ viduals and genetic diversity. Furthermore, in case of dwindling pop‐ phy and seasonality drive runoff generation. Seasonality refers to ulations of C. arnoldi, low genetic diversity existing in the remaining the different hydrological behavior during wet versus dry conditions populations constitutes an additional bottleneck that may lead to as, for example, high flow peaks strongly depend on the antecedent extinction (Valbuena‐Ureña et al., 2013). soil wetness (Àvila et al., 1992). Here, streamflow was also sensitive Critical low streamflow projections were not very different con‐ to the evapotranspiration adjustment parameter used to limit evapo‐ sidering the moderate (RCP4.5) or the extreme (RCP8.5) scenarios, transpiration under dry conditions, which indicates that during these whereas vegetation cover had an important role in determining the periods there is low hydrological connectivity between soils and the low streamflow hydrology of the catchment. Scenarios in which the stream and precipitation inputs contribute to increase catchment open heathlands are maintained irrespective of the tree species cov‐ water storage rather than to streamflow generation or evapotrans‐ ering the rest of the catchment present considerably lower number piration. Consistently, precipitation was only weakly related to Qdiff of days below Q low than scenarios where holm oak forest covered during the observation period and this behavior was well captured the whole catchment. Holm oak forests have water demands that are by the future simulations, especially during the dry season (Figure 7). significantly larger than those of open heathlands and can transpire even during dry periods (Ladekarl, Rasmussen, Christensen, Jensen, & Hansen, 2005; Paço et al., 2009), and so evapotranspiration will 4.2 | Vulnerability of Calotriton arnoldi to future low make up a significantly larger proportion of the water balance under streamflows whole holm oak forest cover scenarios. Our simulations thus suggest Establishing critical streamflow thresholds for amphibian species is that the potential change in vegetation cover toward an expansion of difficult, especially for endangered species because observations and holm oak forests at the expense of beech forests and heathlands will data are scare, but we consider our proposed Qlow value an adequate be more important than climate change itself in determining critical working threshold. Nevertheless, small changes in the Qlow proposed low streamflow conditions in the Montseny mountains. LEDESMA et al. | 9745 75 TA B L E 3 Relative contribution of period (near‐ and far‐future), (a) climate scenario (RCP4.5 and RCP8.5), and vegetation cover Observed 60 scenario (Table 1) to the total variation in the number of days with Reference daily simulated streamflow not reaching/exceeding the specified 45 RCP 4.5 thresholds: Q low = 0.33 L/s for low flows, and Qdiff‐all = 97 L/s (whole year) and Qdiff‐dry = 42 L/s (dry season) for flow increase events in 30 RCP 8.5 consecutive days Qdiff (mm) 15 Component Q low (%) Qdiff‐all (%) Qdiff‐dry (%) 0 Period 21 17 0.3 Climate scenario 11 47 53 –15 Vegetation cover 50 33 22 scenario –30 Model 82 97 76 –45 Residual 18 3 24 40 (b) 30 during spring may have deleterious consequences because this is a sensitive period in the C. arnoldi life cycle (reproduction occurs 20 at this time), and drift risk is larger for eggs and larvae than for adults. Immature individuals of the related C. asper species can Qdiff (mm) 10 inhabit the terrestrial environment, and this represents a refuge that might help recovering population sizes after extreme stream‐ 0 flow increase events (Colomer et al., 2014; Montori et al., 2012). By contrast, immature individuals of C. arnoldi do not inhabit the terrestrial surroundings and remain in the aquatic environment –10 until sexual maturity, which makes this species more vulnerable to floods and high water velocities. Nevertheless, our simulations –20 0 25 50 75 100 125 show that the number of single‐day events above Qdiff‐dry will de‐ Precipitation (mm) crease in the future and so the potential risk of high streamflow events for C. arnoldi during its sensitive life cycle period will be F I G U R E 7 Scatter plots of “flow difference in consecutive reduced, as well as the potential risk of catastrophic drift overall. days” (Qdiff ), versus daily precipitation for (a) the whole year and (b) the dry season (April to September) for future Representative In contrast to the importance of vegetation cover scenarios to Concentration Pathway (RCP) scenarios 4.5 and 8.5 and for the explain the number of days below Qlow, climate scenarios appear to reference period. Dashed lines correspond to the threshold values explain most of the variance in the number of single‐day events above defined as the 99th percentile of the Qdiff values for (a) the whole Qdiff‐all and, especially, above Qdiff‐dry. RCP8.5 scenarios are extreme‐ year (Qdiff‐all = 97 L/s), and for (b) the dry season (Qdiff‐dry = 42 L/s) high CO2 emission scenarios in which temperatures are predicted to using the observed data, which are also shown be higher and large rainfall events more sporadic than the correspond‐ ing RCP4.5 scenarios. These conditions will limit the opportunities for generation of high streamflow increase events due to the combination 4.3 | Vulnerability of Calotriton arnoldi to future high of increased evapotranspiration and lower probability of large rainfall streamflows events occurring under pre‐existent wet conditions. Risk of catastrophic drift in C. asper was related to precipitation events larger than 50 mm/day (Colomer et al., 2014; Montori et al., 4.4 | Implications for conservation and management 2012). In these studies, no indication of associated streamflows was provided as the investigated catchments were ungauged. As discussed by O'Driscoll et al. (2018), climate impact assessment At RC0, the threshold based on actual streamflow increase rate studies, such as the one presented here, can be interpreted in quali‐ better describes the potential impact of high flows on C. arnoldi tative terms based on alternative scenarios of internally consistent or other species, as the relationship between precipitation and narratives, or “storylines,” which are based on a restricted set of pos‐ streamflow is quite loose (Figure 7). During the dry season (April sible sources of uncertainty and thus are representations of a sub‐ to September), Qdiff values were noticeably lower than those in set of all possible futures (Cremona et al., 2017; van Vuuren, Vries, the wet season and single‐day events above Qdiff‐dry concentrated Beusen, & Heuberger, 2008). In our case, combining an educated in April and May, both in the reference and future scenarios (data selection of critical streamflow thresholds based on hydrological not explicitly shown). The occurrence of these flashy events observations and C. arnoldi ecological requirements, with moderate 9746 | LEDESMA et al. (e.g., RCP4.5) and extreme (RCP8.5) climate scenarios, and four DATA AVA I L A B I L I T Y S TAT E M E N T plausible vegetation cover scenarios ensured that a wide range of All data associated with the paper can be accessed through figshare possible futures was covered. These relevant storylines on possible at https​://figsh​are.com/artic​les/Ledes​ma_et_al_2019_ECE_Datas​ futures offer a starting point for decision‐making regarding C. arnoldi et_xlsx/8869598 (https​://doi.org/10.6084/m9.figsh​are.8869598). management in the context of the LIFE‐Tritó Montseny project that can guide future actions and research. Our study shows that predicted future low streamflow condi‐ ORCID tions will pose a severe threat for the survival of existing C. arnoldi José L. J. Ledesma https://orcid.org/0000-0002-4181-5498 populations, or the viability of those planned to be reintroduced. This thread will be mostly driven by the hydrological consequences of the expected expansion of holm oak forests. In contrast, the risk of cat‐ REFERENCES astrophic drift due to high streamflow events will be reduced, espe‐ cially under high RCP emission scenarios. Therefore, we recommend Altava‐Ortiz, V., Bladé, I., & Llasat Botija, M. D. C. (2010). Caracterització i monitoratge de les sequeres a Catalunya i nord del País Valencià càlcul consideration of future potential changes in landscape (e.g., changes in d'escenaris climàtics per al segle XXI. Doctoral dissertation, Universitat vegetation cover) and not only predicted changes in climate variables de Barcelona. Departament d'Astronomia i Meteorologia. Retrieved in simulations of future streamflows using rainfall‐runoff models. from http://world​c at.org/z-wcorg/​database This work may help to take management actions (e.g., limit the Àvila, A., Piñol, J., Rodà, F., & Neal, C. (1992). Storm solute behavior in a montane Mediterranean forested catchment. Journal of Hydrology, expansion of the holm oak forest) for ameliorating the C. arnoldi 140(1–4), 143–161. https​://doi.org/10.1016/0022-1694(92)90238-q newt habitat. This is relevant because very few case studies have up Bosch, J. M., & Hewlett, J. D. (1982). A review of catchment experiments to now examined the effects of hydrological regimes on amphibian to determine the effect of vegetation changes on water yield and populations (Cayuela, Besnard, Béchet, Devictor, & Olivier, 2012) or evapotranspiration. Journal of Hydrology, 55(1–4), 3–23. https​://doi. org/10.1016/0022-1694(82)90117-2 the efficacy of landscape management practices devised to help the Carranza, S., & Amat, F. (2005). Taxonomy, biogeography and evolution conservation of such populations (Hartel, Băncilă, & Cogălniceanu, of Euproctus (Amphibia: Salamandridae), with the resurrection of the 2011; Shoo et al., 2011). We used a holistic approach that brings genus Calotriton and the description of a new endemic species from together different scientific disciplines (hydrological modeling, cli‐ the Iberian Peninsula. Zoological Journal of the Linnean Society, 145(4), 555–582. https​://doi.org/10.1111/j.1096-3642.2005.00197.x matology, biology, and ecology) for providing a basis for C. arnoldi Cayuela, H., Besnard, A., Béchet, A., Devictor, V., & Olivier, A. (2012). management and conservation strategies. This type of approach is Reproductive dynamics of three amphibian species in Mediterranean encouraged for formulating science‐based, sustainable management wetlands: The role of local precipitation and hydrological regimes. of both amphibian species and ecosystems. Freshwater Biology, 57(12), 2629–2640. https​://doi.org/10.1111/ fwb.12034​ Clergue‐Gazeau, M. (1969). Alimentation de l'Euprocte dans les grottes. AC K N OW L E D G M E N T S Annales de Spéleologie, 24, 361–364. Clergue‐Gazeau, M. (1971). L'Euprocte pyrénéen. Conséquence de la vie JLJL was funded by the Swedish Research Council Formas (grant cavernicole sur son développement et sa reproduction. Annales de 2015‐1518) and the Spanish Government through a Juan de la Cierva Spéleologie, 25, 825–960. Colomer, M. A., Montori, A., García, E., & Fondevilla, C. (2014). Using grant (FJCI‐2017‐32111). AA acknowledges funding from the Spanish a bioinspired model to determine the extinction risk of Calotriton Government through a MICINN project (CGL2017‐84687‐C2‐2‐R) asper populations as a result of an increase in extreme rainfall in a and the LIFE‐Tritó Montseny project from the European Commission scenario of climatic change. Ecological Modelling, 281, 1–14. https​:// (LIFE15 NAT/ES/000757). We would like to thank Martyn Futter and doi.org/10.1016/j.ecolm​odel.2014.02.018 Cremona, F., Vilbaste, S., Couture, R. M., Nõges, P., & Nõges, T. (2017). Jan Seibert for valuable discussions on the rainfall‐runoff modeling Is the future of large shallow lakes blue‐green? Comparing the and Irene Fraile for field work at Montseny. response of a catchment‐lake model chain to climate predic‐ tions. Climatic Change, 141(2), 347–361. https​://doi.org/10.1007/ s10584-016-1894-8 C O N FL I C T O F I N T E R E S T de Pous, P., Montori, A., Amat, F., & Sanuy, D. (2016). Range con‐ traction and loss of genetic variation of the Pyrenean en‐ None declared. demic newt Calotriton asper due to climate change. Regional Environmental Change, 16(4), 995–1009. https​://doi.org/10.1007/ s10113-015-0804-3 AU T H O R S ' C O N T R I B U T I O N S del Río, S., Herrero, L., Fraile, R., & Penas, A. (2011). Spatial distribution of recent rainfall trends in Spain (1961–2006). International Journal of JLJL and AA conceived the idea, designed the methodology, pro‐ Climatology, 31(5), 656–667. https​://doi.org/10.1002/joc.2111 posed the structure, and wrote the paper. JLJL performed the hydro‐ Futter, M. N., Erlandsson, M. A., Butterfield, D., Whitehead, P. G., Oni, logical modeling and analyzed the data. AM provided support on the S. K., & Wade, A. J. (2014). PERSiST: A flexible rainfall‐runoff mod‐ elling toolkit for use with the INCA family of models. Hydrology and Calotriton arnoldi ecology and contributed to the writing. VA, AB, and Earth System Sciences, 18(2), 855–873. https​://doi.org/10.5194/ JC carried out the climate modeling. All authors contributed critically hess-18-855-2014 to the drafts and gave final approval for publication. LEDESMA et al. | 9747 Futter, M. N., Löfgren, S., Köhler, S. J., Lundin, L., Moldan, F., & evergreen oak savannah: The role of trees and pasture. Journal Bringmark, L. (2011). Simulating dissolved organic carbon dynamics of Hydrology, 369(1–2), 98–106. https​://doi.org/10.1016/j.jhydr​ at the Swedish integrated monitoring sites with the integrated catch‐ ol.2009.02.011 ments model for carbon, INCA‐C. Ambio, 40(8), 906–919. https​://doi. Peñuelas, J., & Boada, M. (2003). A global change‐induced biome shift org/10.1007/s13280-011-0203-z in the Montseny mountains (NE Spain). Global Change Biology, 9(2), Gallart, F., & Llorens, P. (2004). Observations on land cover changes and 131–140. https​://doi.org/10.1046/j.1365-2486.2003.00566.x water resources in the headwaters of the Ebro catchment, Iberian Pool, S., Vis, M. J. P., Knight, R. R., & Seibert, J. (2017). Streamflow char‐ Peninsula. Physics and Chemistry of the Earth, 29(11–12), 769–773. acteristics from modeled runoff time series importance of calibra‐ https​://doi.org/10.1016/j.pce.2004.05.004 tion criteria selection. Hydrology and Earth System Sciences, 21(11), Giorgetta, M. A., Jungclaus, J., Reick, C. H., Legutke, S., Bader, J., 5443–5457. https​://doi.org/10.5194/hess-21-5443-2017 Böttinger, M., … Stevens, B. (2013). Climate and carbon cycle changes Seibert, J., & McDonnell, J. J. (2010). Land‐cover impacts on streamflow: from 1850 to 2100 in MPI‐ESM simulations for the Coupled Model A change‐detection modelling approach that incorporates param‐ Intercomparison Project phase 5. Journal of Advances in Modeling eter uncertainty. Hydrological Sciences Journal‐Journal Des Sciences Earth Systems, 5(3), 572–597. https​://doi.org/10.1002/jame.20038​ Hydrologiques, 55(3), 316–332. https​://doi.org/10.1080/02626​ González‐Hidalgo, J. C., López‐Bustins, J. A., Štepánek, P., Martín‐ 66100​3683264 Vide, J., & de Luisa, M. (2009). Monthly precipitation trends on the Shoo, L. P., Olson, D. H., McMenamin, S. K., Murray, K. A., Van Sluys, Mediterranean fringe of the Iberian Peninsula during the second‐ M., Donnelly, M. A., … Hero, J.‐M. (2011). Engineering a future for half of the twentieth century (1951–2000). International Journal of amphibians under climate change. Journal of Applied Ecology, 48(2), Climatology, 29(10), 1415–1429. https​://doi.org/10.1002/joc.1780 487–492. https​://doi.org/10.1111/j.1365-2664.2010.01942.x Hartel, T., Băncilă, R., & Cogălniceanu, D. (2011). Spatial and temporal Valbuena‐Ureña, E., Amat, F., & Carranza, S. (2013). Integrative phylo‐ variability of aquatic habitat use by amphibians in a hydrologically geography of Calotriton newts (Amphibia, Salamandridae), with modified landscape. Freshwater Biology, 56(11), 2288–2298. https​:// special remarks on the conservation of the endangered Montseny doi.org/10.1111/j.1365-2427.2011.02655.x brook newt (Calotriton arnoldi). PLoS ONE, 8(6), e62542. https​://doi. Ladekarl, U. L., Rasmussen, K. R., Christensen, S., Jensen, K. H., & org/10.1371/journ​al.pone.0062542 Hansen, B. (2005). Groundwater recharge and evapotranspiration Van Vuuren, D. P., de Vries, B., Beusen, A., & Heuberger, P. S. C. (2008). for two natural ecosystems covered with oak and heather. Journal Conditional probabilistic estimates of 21st century greenhouse gas of Hydrology, 300(1–4), 76–99. https​://doi.org/10.1016/j.jhydr​ emissions based on the storylines of the IPCC‐SRES scenarios. Global ol.2004.05.003 Environmental Change‐Human and Policy Dimensions, 18(4), 635–654. Ledesma, J. L. J., & Futter, M. N. (2017). Gridded climate data products https​://doi.org/10.1016/j.gloen​vcha.2008.06.001 are an alternative to instrumental measurements as inputs to rain‐ Vicente‐Serrano, S. M., Lopez‐Moreno, J.‐I., Beguería, S., Lorenzo‐ fall‐runoff models. Hydrological Processes, 31(18), 3283–3293. https​ Lacruz, J., Sanchez‐Lorenzo, A., García‐Ruiz, J. M., … Espejo, F. (2014). ://doi.org/10.1002/hyp.11269​ Evidence of increasing drought severity caused by temperature rise Ledesma, J. L. J., Grabs, T., Futter, M. N., Bishop, K. H., Laudon, H., & in southern Europe. Environmental Research Letters, 9(4), 9. https​:// Köhler, S. J. (2013). Riparian zone control on base cation concentra‐ doi.org/10.1088/1748-9326/9/4/044001 tion in boreal streams. Biogeosciences, 10(6), 3849–3868. https​://doi. Vis, M., Knight, R., Pool, S., Wolfe, W., & Seibert, J. (2015). Model cali‐ org/10.5194/bg-10-3849-2013 bration criteria for estimating ecological flow characteristics. Water, Ledesma, J. L. J., Köhler, S. J., & Futter, M. N. (2012). Long‐term dynamics 7(5), 2358–2381. https​://doi.org/10.3390/w7052358 of dissolved organic carbon: Implications for drinking water supply. Walls, S., Barichivich, W., & Brown, M. (2013). Drought, deluge and Science of the Total Environment, 432, 1–11. https​://doi.org/10.1016/j. declines: The impact of precipitation extremes on amphibians in a scito​tenv.2012.05.071 changing climate. Biology, 2(1), 399–418. https​://doi.org/10.3390/ Lupon, A., Ledesma, J. L. J., & Bernal, S. (2018). Riparian evapotranspira‐ biolo​g y201​0399 tion is essential to simulate streamflow dynamics and water budgets Wang, K. C., Dickinson, R. E., & Liang, S. L. (2012). Global atmospheric in a Mediterranean catchment. Hydrology and Earth System Sciences, evaporative demand over and from 1973 to 2008. Journal of Climate, 22(7), 4033–4045. https​://doi.org/10.5194/hess-22-4033-2018 25(23), 8353–8361. https​://doi.org/10.1175/jcli-d-11-00492.1 Montori, A., Richter‐Boix, A., Franch, M., Santos, X., Garriga, N., & Llorente, G. A. (2012). Natural fluctuations in a stream dwelling newt as a result of extreme rainfall: A 21‐year survey of a Calotriton asper population. Basic and Applied Herpetology, 26, 43–56. https​://doi. S U P P O R T I N G I N FO R M AT I O N org/10.11160/​bah.12001​ Nash, J. E., & Sutcliffe, V. (1970). River flow forecasting through concep‐ Additional supporting information may be found online in the tual models part I ‐ A discussion of principles. Journal of Hydrology, 10, Supporting Information section at the end of the article. 282–290. https​://doi.org/10.1016/0022-1694(70)90255-6 O'Driscoll, C., Ledesma, J. L. J., Coll, J., Murnane, J. G., Nolan, P., Mockler, E. M., … Xiao, L. W. (2018). Minimal climate change impacts on How to cite this article: Ledesma JLJ, Montori A, Altava‐Ortiz natural organic matter forecasted for a potable water supply in Ireland. Science of the Total Environment, 630, 869–877. https​://doi. V, Barrera‐Escoda A, Cunillera J, Àvila A. Future hydrological org/10.1016/j.scito​tenv.2018.02.248 constraints of the Montseny brook newt (Calotriton arnoldi) Oni, S., Futter, M., Ledesma, J. L. J., Teutschbein, C., Buttle, J., & Laudon, under changing climate and vegetation cover. Ecol Evol. H. (2016). Using dry and wet year hydroclimatic extremes to guide 2019;9:9736–9747. https​://doi.org/10.1002/ece3.5506 future hydrologic projections. Hydrology and Earth System Sciences, 20(7), 2811–2825. https​://doi.org/10.5194/hess-20-2811-2016 Paço, T. A., David, T. S., Henriques, M. O., Pereira, J. S., Valente, F., Banza, J., … David, J. S. (2009). Evapotranspiration from a Mediterranean