2.1 Study area

Two partly glacierized meso-scale catchments in the Swiss Alps, in the headwater of the Rhine River, were examined in this study: the Hinterrhein catchment and the larger Schwarze Lütschine catchment (Fig. 1, Table 1). Based on the dataset by Freudiger et al. (2018), used in this study, around the year 1900 glacier coverage was approximately 32 % of the Hinterrhein catchment area and around 25 % of the Schwarze Lütschine catchment area. Glaciers in both catchments retreated considerably during the 20th century. The Hinterrhein catchment is characterized by small, scattered glaciers, which by 1973 lost around half their area, leading to a glacier coverage of only 7 % in 2010 (Table 1). In the Schwarze Lütschine catchment losses in relative glacier area have been smaller. This difference in glacier coverage is related to elevation with considerably higher maximum elevations in the Schwarze Lütschine catchment compared to the Hinterrhein catchment (Table 1).

2.2 Data and data preparation

The application of bias correction algorithms to climate model outputs is generally based on three datasets: historical observations as reference (also called “target”) data, historical climate model simulations, and the corresponding climate model projections. In the present study the historical reference data for the study catchments were derived from an observation-based interpolation product, i.e., the 1 km × 1 km gridded daily air temperature and precipitation datasets from the HYRAS product (Rauthe et al., 2013; Frick et al., 2014). Area-weighted mean values of precipitation and air temperature were extracted for the study catchments. The extracted catchment mean precipitation time series were corrected for undercatch based on the method by Sevruk (1989) and were then further adjusted through validation with long-term annual mean precipitation sums resulting from a water balance approach (for details see Stahl et al., 2017). The resulting time series of catchment mean precipitation and air temperature were used as input for the calibration of the glacio-hydrological model and as historically observed climate data (HOCD) for the bias correction.

Table 2GCM–RCM combinations from the EURO-CORDEX initiative used in this study.

GCM institutions: ¹ CNRM-CERFACS (Centre National de Recherches Météorologiques-Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique); note that a warning concerning an inconsistency in the historical run of CNRM-CM5 has been issued on the CORDEX errata page (https://www.hzg.de/ms/euro-cordex/078730/index.php.en, last access: 4 March 2019) after data had been downloaded and selected for this study. ² EC-Earth consortium. ³ IPSL (Institut Pierre-Simon Laplace). ⁴ MPI-M (Max Planck Institute for Meteorology); ⁵ CORDEX errata page (https://www.hzg.de/ms/euro-cordex/078730/index.php.en, last access: 4 March 2019) notes snow accumulation issues for these GCM–RCM runs.

Download Print Version | Download XLSX

The climate model datasets were obtained from the Coordinated Regional Climate Downscaling Experiment (CORDEX, http://www.cordex.org/, last access: 4 March 2019) via the Earth System Grid Federation (ESGF) archive (http://www.cordex.org/data-access/esgf/, last access: 4 March 2019). CORDEX is a collaborative effort within the climate modelling community where general circulation models (GCMs) are downscaled using regional climate models (RCMs). Since all catchments in this study are located in Switzerland, GCM–RCM simulations were selected from the European domain of the CORDEX project (EURO-CORDEX, http://www.euro-cordex.net/, last access: 4 March 2019). EURO-CORDEX provides simulations at 0.11^∘ (∼12.5 km horizontal resolution) and 0.44^∘ (∼50 km horizontal resolution). Given that the catchments used in this study are situated in the Alpine domain, only the higher-resolution 0.11^∘ simulations were used. Two Representative Concentration Pathways (RCPs) were selected for this study: RCP4.5 represents an intermediate mitigation scenario, where greenhouse gas (GHG) emissions will peak around 2040 and then steadily decrease, and RCP8.5 represents a more pessimistic scenario, which assumes that GHG emissions will continue to increase throughout the 21st century (Meinshausen et al., 2011).

Precipitation (P) and air temperature (T_a) data were provided by the 10 GCM–RCM combinations shown in Table 2 for the time period 1970–2099. For each catchment, raw GCM–RCM data were extracted using an area-weighted method as shown in Hakala et al. (2018). Based on the areal fraction of an RCM grid cell overlying a particular catchment, five RCM grid cells contribute to each catchment. All GCM–RCM combinations used in this study utilize a Gregorian calendar.

The application of the hydrological model requires catchment mean time series of P and T_a. These were subjected to bias correction. Further data used as model input and for model calibration were not directly bias corrected. Daily potential evapotranspiration was calculated with an air-temperature-based approach provided by Oudin et al. (2005). Catchment-specific air temperature lapse rates were determined based on daily values from the HYRAS product. Based on the reference period from 1976–2006 a mean for each day of the year was calculated and smoothed using an 11-day moving average. A mean precipitation gradient (in percentage per 100 m a.s.l.) was determined from the corrected HYRAS data and applied as constant values in all simulations.

Daily streamflow data for model calibration were provided by the Swiss Federal Office for the Environment (FOEN) and the “Amt für Wasser und Abfall des Kantons Bern”. The available streamflow record for the station Gündlischwand (operated by the Cantone of Berne) at the outlet of the Schwarze Lütschine study catchment covered only the period 1992–1999. By using the record of a downstream station of the Lütschine River (station Gsteig) and subtracting the streamflow of its other major headwater tributary (record from the station Zweilütschinen of the Weisse Lütschine) the streamflow for the Schwarze Lütschine study catchment could be reconstructed for the entire simulation period. This reconstructed streamflow time series was validated with the available streamflow data from the station Gündlischwand for the sub-period 1992–1999 and then used for model calibration. Snow water equivalent (SWE) and snow cover data were derived from a snow map (interpolated grid) product by the OSHD-SLF (2013). The glacier area was assessed based on glacier inventory data by Müller et al. (1976) and Maisch et al. (2000) for the state in the year 1973, by Paul et al. (2011) for the state in 2003, and by Fischer et al. (2014) for the year 2010 (see Table 1). Estimates of glacier volume were derived based on gridded ice thickness data available for the years 1973 and 2010, which were computed using the approach by Huss and Farinotti (2012) and provided by Matthias Huss. Glacier volume for the year 2003 was estimated based on the glacier cover according to Paul et al. (2011) and glacier volume–area scaling. The glacier volume estimate for 1973 was used for model initialization. The estimate for 2003 was incorporated in the model calibration for the period 1976–2006. The estimate for 2010 was not directly used in the calibration but served the validation of model simulations beyond the year 2006.

3.1 Bias correction of climate data

Depending on the GCM–RCM combination, raw climate variables (noBC) of the control period (1976–2006) differ from the reference data (HOCD). To correct these biases, two different bias correction methods were applied to each climate model's T_a and P series: a univariate quantile mapping technique – Quantile Delta Mapping (QDM) – and a multivariate bias correction approach (MBCn). Quantile mapping is based on a transfer function that transforms the cumulative distribution(s) of the modelled data to match the distribution(s) of the observed series. The obtained transfer function is then applied to all climate model data, historical and projected. Thus it corrects systematic distributional biases relative to historical observations and preserves model-projected relative changes. QDM is a variant of quantile mapping by Cannon et al. (2015) that was designed to avoid artificial deterioration of trends arising as a statistical artefact of standard quantile mapping. QDM corrects systematic distributional biases relative to historical observations and preserves model-projected changes in quantiles in the projection period. For a given time slice, the climate model's change signal (Δ) – relative change for precipitation and absolute change for air temperature – is removed from all projected future quantiles in a first step. Quantile mapping is then applied before the projected changes in quantiles are reintroduced to the bias-corrected model output.

The MBCn algorithm by Cannon (2018a) is based on the N-dimensional probability density function transform. This approach was originally developed for image processing (Pitié et al., 2007) but has been converted for post-processing climate model data. MBCn combines QDM and random orthogonal rotations to match the multivariate distributions of climate model data and observed data. In the MBCn approach, a random orthogonal rotation of the data points is applied before QDM. This exposes QDM to a linear combination of the original variables, which is then used to correct the marginal distributions of the rotated data. The QDM-corrected dataset is then rotated back and convergence to the observed multivariate distribution is checked. These steps are conducted iteratively until the multivariate distributions of bias-corrected climate model data and observed climate data match. In this study, 100 iterations were conducted. Both QDM and MBCn were applied in a seasonally dependent fashion. Specifically, bias corrections were applied over 30-year sliding windows. This involved replacing the central 10 years and sliding forward 10 years for each 30-year window, until the end of the projection period was reached. Within each window – to ensure an unbiased seasonal cycle – bias corrections were applied separately for each calendar month. The combination of change-preservation by QDM, which is also a core component of MBCn, with sliding windows ensures that projected trends from the underlying climate model are largely preserved. This follows the general approach and recommendation of Hempel et al. (2013) concerning trend preservation of post-processed climate model output for impact modelling.

Climate model data is often simultaneously bias corrected and downscaled as the reference data stems from stations or higher-resolution observations in comparison to the coarse grid resolution of RCMs. Undesirable effects in downscaling to finer scales have been one of the major limitations of current bias correction methods (Maraun, 2013; Ehret et al., 2012; Maraun et al., 2017). Such artefacts can occur in complex terrain in particular and if the scale gap between climate model outputs and impact model data is considerable. In general, bias correction based on spatial resolutions that differ substantially should be avoided or handled with great care. In this study the discrepancy in resolution is assumed to be acceptable as the bias correction was based on spatially aggregated mean climate variables for the meso-scale catchments (54 and 180 km²) with the original resolution of the underlying gridded datasets (GCM–RCM data: 0.11^∘, historical HYRAS data: 1 km) becoming of secondary importance.

Table 3Model performance criteria for the calibration (1 October 1976–30 September 2003) and validation (1 October 2003–31 December 2006) of the hydrological model formulated (see footers) that the ideal value for a perfect fit is 1.0.

Formulation of model performance criteria: ¹ $R_{\mathrm{eff}}=\mathrm{1}-\frac{\sum {\left(Q_{\mathrm{obs}}-Q_{\mathrm{sim}}\right)}^{\mathrm{2}}}{{\left(Q_{\mathrm{obs}}-\stackrel{\mathrm{‾}}{Q_{\mathrm{obs}}}\right)}^{\mathrm{2}}}$ where Q_obs and Q_sim, respectively, are observed and simulated streamflow (mm day⁻¹). ² See Gupta et al. (2009). ³ 1−V with $V=\frac{\sum \left|\left(Q_{\mathrm{obs}}-Q_{\mathrm{sim}}\right)\right|}{\sum \left(Q_{\mathrm{obs}}\right)}$ where Q_obs and Q_sim, respectively, are observed and simulated streamflow (mm day⁻¹). ⁴ $\mathrm{1}-R_{\mathrm{eff}}-\mathrm{0.1}V$ with R_eff¹ and V⁶; see Lindström et al. (1997). ⁵ $\mathrm{1}-\sqrt{\frac{\mathrm{1}}{n}{\left(C_{\mathrm{sim}}-C_{\mathrm{ref}}\right)}^{\mathrm{2}}}$ with C_ref the snow-covered catchment area fraction (C) (–) as per gridded SWE reference data, C_sim the simulated C, and n the number of time steps. ⁶ $\mathrm{1}-\frac{\sum \left|\left(S_{\mathrm{ref}}-S_{\mathrm{sim}}\right)\right|}{\sum S_{\mathrm{obs}}}$ with S (mm) the mean SWE for elevation range below 2500 m a.s.l. where S_ref is derived from SWE reference data and S_sim is simulated. ⁷ $\mathrm{1}-\frac{\left|\mathrm{\Delta }{W}_{\mathrm{sim}}-\mathrm{\Delta }{W}_{\mathrm{obs}}\right|}{\mathrm{\Delta }{W}_{\mathrm{obs}}}$ with ΔW (mm) the change of glacier ice volume in water equivalent between the years 1973 and 2003, where ΔW_obs corresponds to an estimate based on observed glacier area and ΔW_sim is simulated.

Download Print Version | Download XLSX

3.2 Hydrological model simulations

The HBV model (Bergström, 1976; Lindström et al., 1997) is a semi-distributed bucket-type runoff model. Here the software implementation HBV-light (Seibert and Vis, 2012) was used, which recently has been extended to represent coupled glacio-hydrological processes of partly glacierized catchments (Seibert et al., 2018). This version of the HBV model also allows tracking of the different components of streamflow resulting from rainfall (Q_R), snowmelt (Q_S), and glacier ice melt (Q_I) (Weiler et al., 2018; Seibert et al., 2018). The HBV model requires daily precipitation, air temperature, and potential evapotranspiration data as input to simulate daily runoff. In addition, linear gradients of air temperature and precipitation are needed for the interpolation over elevation zones. A general description of the basic model structure and the process conceptualization of the HBV model are found elsewhere (e.g., Lindström et al., 1997; Seibert and Vis, 2012; Seibert et al., 2018). Snow and ice accumulation and melt are based on a widely used air temperature index approach using a threshold air temperature as a model parameter to differentiate between precipitation falling as snow and rain as well as to simulate melt of snow and ice by additionally using a degree-day factor. Differences in the melt of glacier ice compared to snow are represented by another model parameter. The influence of differences in aspect on snow and ice melt was taken into account by distinguishing three aspect classes and applying an additional aspect factor parameter (Hagg et al., 2007; Hottelet et al., 1993). The latest version of the HBV-light software with the implementation of the coupled glacio-hydrological processes and the adjustment of glacier geometry to glacier mass changes based on the Δh parametrization by Huss et al. (2010) is explained in detail in Seibert et al. (2018). It should be noted that with the implementation in HBV-light only one glacier per catchment or subcatchment can be represented. Hence, glacier cover areas in each of the two case study catchments were aggregated and simulated as one “virtual” model glacier.

The model was calibrated for the period from 1976–2003, preceded by a 3-year warm-up period, by optimizing a weighted objective function, giving special attention to streamflow dynamics (50 %), snow simulation (25 %), and glacier volume change (25 %). The Lindström measure (Lindström, 1997) was used for the streamflow's general dynamic and volume errors, while the Nash–Sutcliffe efficiency (Nash and Sutcliffe, 1970) was computed based on logarithmically transformed streamflow. Additionally the Nash–Sutcliffe efficiency was computed for the streamflow only during the summer months, from June to September. To calibrate the snow simulations the snow-covered area fraction of the catchment and the mean SWE of the elevation range < 2500 m a.s.l. were used. Elevations below 2500 m a.s.l. represent the crucial range for the snow line and in this range the gridded SWE interpolation used as reference data is well-founded on station data. Glacier volume was considered in the calibration process using glacier volume estimates for the years 1973 and 2003. The automated multi-criteria calibration was based on a genetic algorithm for parameter optimization (see Seibert, 2000). A 3-year model validation period (1 October 2003–31 December 2006) completed the historical reference period 1977–2006. Resulting performance measures for the calibration and validation period are summarized in Table 3 (see Supplement for additional figures comparing simulated variables and reference data). The retreat of the glaciers required all experiments to be run in a transient mode, i.e., the model was forced with climate model scenario data for the period from October 1976 to September 2099.

Figure 2Annual precipitation sums for days with air temperatures above or below 0 ^∘C.

Download

3.3 Data analysis

Effects of the bias correction approaches on the hydrological simulation were based on comparisons of the simulation results for the historical reference period 1976–2006 using P and T_a time series derived from the HYRAS datasets as input (Sim_HOCD) and simulations forced with P and T_a series from 10 different GCM–RCM outputs for the two different RCP scenarios, each uncorrected (Sim_noBC) and bias corrected based on QDM (Sim_QDM) and on MBCn (Sim_MBCn). In total, this led to 61 hydrological model runs (1 Sim_HOCD, 20 Sim_noBC, 20 Sim_QDM, and 20 Sim_MBCn) per catchment. In a first step (Sect. 4.1), the different P and T_a series were evaluated for the amount of precipitation occurring at air temperatures above and below 0^∘ C due to the importance for the simulation of snow accumulation and melt processes. Furthermore, the simulation results were assessed in terms of SWE, glacier ice volume (V_I) evolution (Sect. 4.2), and eventually streamflow with its three individual components Q_R, Q_S, and Q_I (Sect. 4.3).

4.1 Climate variables bias correction

The two applied bias correction methods led to differences concerning the interdependence of P and T_a. The distribution of annual precipitation sums during air temperatures above and below 0 ^∘C in the entire ensemble is represented in Fig. 2, while results for the individual GCM–RCM output series are provided in the Supplement. Generally, the uncorrected climate model data (noBC) have a wider variability than the reference data (HOCD). Particularly for the Schwarze Lütschine the uncorrected data yielded precipitation amounts remarkably higher than historically observed. However, differences also existed between the correction methods. For both catchments precipitation falling above air temperatures of 0 ^∘C was overestimated with QDM. Accordingly, precipitation falling below air temperatures of 0 ^∘C was underestimated in the univariate bias-corrected data. MBCn appears to have better reproduced the historical reference data in this respect.

4.2 Hydrological model simulations – cryosphere

Application of the climate scenarios clearly revealed a decreasing role of snow for both study catchments. Figure 3 illustrates a distinctly smaller snow accumulation in the course of a year simulated for the period 2070–2099 (compared to the historical reference period 1977–2006) and a more complete melt during the summer. This extended the snow-free period during the summer in the Hinterrhein catchment. The spread between the simulations diverged for the simulations of future conditions. In the Schwarze Lütschine catchment with its higher maximum elevations all effects were comparable, yet a permanent snow cover remained still present based on most scenarios. As expected, simulations based on the RCP4.5 scenario (not shown) led to a clear but less severe decrease in mean SWE than for the RCP8.5 scenario.

Figure 3Mean annual SWE regime, calculated using the 11-day moving average of daily simulated SWE (catchment mean) during the historical reference period (a and c) and at the end of the scenario period based on the RCP8.5 scenario (b and d).

Download

The differences in the interdependence of precipitation and air temperature resulting from the application of QDM versus MBCn to the GCM–RCM data can be seen in the simulated SWE (Fig. 3). The state of precipitation defined by the calibrated threshold air temperature parameter TT (Schwarze Lütschine TT = −0.29 ^∘C; Hinterrhein TT = −0.73 ^∘C) influenced the snow accumulation and therefore led to differences in the annual SWE regime (Fig. 3). As MBCn-corrected GCM–RCM data caused more precipitation to fall as snow, the accumulated catchment mean SWE in spring was simulated to be up to around 100–200 mm higher in the historical reference period compared to simulations based on QDM-corrected forcing data. Simulated SWE based on the two different bias correction methods differed notably. Comparing the results with the reference simulation (Fig. 3) indicates that MBCn performed better. The systematic difference in simulated SWE resulting from the bias correction methods was a bit less clear for the Schwarze Lütschine catchment in the scenario period, yet overall the differing tendencies between QDM- and MBCn-corrected data were considerable.

Figure 4Simulated glacier ice volume from 1977 to 2099 using the RCP8.5 scenario forcing in the two catchments (a, b). In the lower part of the graphs the boxes in the left figure and the dots in both figures indicate the simulated years of the complete glacier ice melt. For the Schwarze Lütschine only 5 (3) out of the 10 Sim_QDM (Sim_MBCn) simulations led to complete glacier melt by the end of 2099, not allowing any box plots to be shown. Filled black circles are glacier volume estimates based on observed glacier area data in 2003 and 2010.

Download

For the period 1976 to 2099 the glacier volume was simulated to decrease in both catchments. In the Hinterrhein catchment, glaciers diminished continuously from the beginning of the simulation period and were simulated to have disappeared between 2028 and 2055 under the RCP 8.5 scenario depending on the GCM–RCMs and the applied bias correction method (Fig. 4). In the Schwarze Lütschine catchment, data from a few GCM–RCMs resulted in an increase in simulated glacier volume in the 1970s and 1980s, which is in line with the historical reference simulation (Sim_HOCD). In the following years, glacier volume decreased continuously. In contrast to the Hinterrhein catchment, glaciers were not simulated to have disappeared by the end of 2099 based on the RCP4.5 scenario (not shown). However, in the simulations the glacier volume diminished to on average roughly a third of its initial size at the beginning of the simulation period. The RCP8.5 scenario from a few certain GCM–RCM combinations even led to complete glacier disappearance in the Schwarze Lütschine catchment within the 21st century.

Figure 5Observed total streamflow and simulated streamflow components for the historical reference period and for the different simulations under the RCP8.5 scenario. Stacked bar plots show mean values over the historical reference period (a, e) and for the period 2070–2099 (d, h), stacked bar plots for Sim_QDM and Sim_MBCn show ensemble mean with ensemble spread (error bars). Simulation results over the scenario period 2006–2099 (b, c, f, g) are shown as semi-transparent polygons for each GCM–RCM combination.

Download

Focusing on systematic differences between simulations using data corrected based on QDM and MBCn, the simulations of glacier volume showed similar tendencies to those found for SWE. For both catchments, but again more clearly for the Hinterrhein catchment, MBCn-corrected GCM–RCM data resulted in a slower decline in glacier volume in comparison to simulations based on QDM-corrected data. All projections led to complete glacier disappearance in the Hinterrhein catchment by about the year 2050, with a clear tendency towards earlier dates for QDM-based simulations (2028–2041, mean: 2036) compared to MBCn-based simulations (2040–2055, mean: 2047). For the Schwarze Lütschine catchment the range of QDM- and MBCn-based glacier volume simulations overlapped largely as simulations in general diverged considerably. However, for each individual GCM–RCM dataset, glacier melt was simulated to be faster using the QDM-corrected data compared to the MBCn-corrected data. The less intense decline in glacier volumes resulting from MBCn-corrected forcing data appeared to correspond better with the reference simulation (Sim_HOCD) in the initial phase of the historical period and with the observation-based glacier volume estimates for the year 2003 (and also for the year 2010 in case of the Hinterrhein catchment). MBCn thus led to more realistic results for the historical reference period.

4.3 Hydrological model simulations – streamflow

Time changes of annual variables and mean monthly hydrological regimes were assessed for streamflow Q and for the individual streamflow components, i.e., the rain component Q_R, the snowmelt component Q_S, and the ice melt component Q_I. Mean annual streamflow of the study catchments showed a small decrease over the entire simulation period from 1976 to 2099 for most simulations, while for some a slight increase was noticed (Fig. 5). However, the simulations based on different GCM–RCM outputs diverged over time. While, on average, the total annual streamflow stayed largely unchanged, its composition clearly changed. The streamflow component from glacier ice melt decreased slowly over time as the glaciers retreated. Likewise, the snowmelt component of streamflow decreased over time. On average, for the RCP4.5 scenario's MBCn-corrected data these decreases were around 14 % in the Hinterrhein and 16 % in the Schwarze Lütschine for the RCP8.5 scenario's QDM-corrected data they were around 53 % in the Hinterrhein and 33 % in the Schwarze Lütschine.

Figure 6Streamflow regimes based on 11-day moving averages of daily streamflow during 30-year periods in the historical reference period and as projected for the period 2070–2099 under the RCP8.5 scenario for the two catchments. Simulation results for each ensemble member are shown as semi-transparent polygons. For the historical reference period the results of the simulations based on the historical reference P and T_a time series are also shown (black lines).

Download

The streamflow simulations reflected the changes from the different bias correction methods found for the cryosphere. Simulations based on QDM-corrected data led to slightly different total streamflow than MBCn-corrected data (Fig. 5a, d and e). These differences were much more pronounced regarding the individual streamflow components. Modelling based on QDM-corrected climate data led to an approximately 10 % higher rain component of streamflow Q_R in comparison to MBCn-corrected simulations. The snowmelt component of streamflow Q_S varies proportionally, being notably smaller when using QDM-corrected GCM–RCM data. Comparing the means of the ice melt components of streamflow Q_I for the 30-year periods at the beginning and at the end of the entire simulation period showed no differences from the bias correction methods for the Hinterrhein catchment and differences in the range of only 1 % for the Schwarze Lütschine catchment.

Simulated streamflow and its components, Q_I, Q_S, and Q_R, also changed seasonally (Fig. 6). In the historical reference period (1977–2006), the two catchments had a nivo-glacial streamflow regime peaking in the summer due to snow and ice melt and with little streamflow during winter. According to the projections the streamflow peak in early summer remained a dominant characteristic until the end of the simulation period. Yet, for the Hinterrhein catchment, the peak's timing was simulated to shift, causing streamflow to concentrate in May and the peak to become much narrower than in the past. For the Schwarze Lütschine catchment the simulations for the RCP8.5 scenario resulted in very variable summer streamflow regimes for 2070–2099 and a tendency towards a lower summer streamflow peak than in the past. In the reference period, the glaciers' influence showed during late summer, where it extended the melt peak into autumn. This effect was simulated to diminish, resulting in decreased total streamflow in late summer. During autumn and winter, simulated streamflow for 2070–2099 was nearly double the level of the historical period mainly due to an increase in the rainfall component of streamflow. Despite similar tendencies of reduced Q_S in the future, differences arising from the different bias correction methods are notable. Q_S was more prominent in all regimes based on MBCn-corrected GCM–RCM outputs, which simulated higher peaks during the snowmelt season and a generally higher fraction during the rest of the year, especially for the future periods. Accordingly, QDM-corrected data supported a larger Q_R component beyond the summer. As a consequence, during low-flow periods in winter, QDM-corrected forcing data overestimated the streamflow in the historical reference period. In contrast, QDM-corrected forced simulations tended to slightly underestimate the streamflow during the spring and summer months, as Q_S was underestimated. Generally, MBCn-corrected data matched more closely with the reference simulations based on observed data.