Hydrological dynamics of snowmelt induced streamflow in a high mountain catchment of the Pyrenees under contrasting snow accumulation and duration years

Snowmelt drives a large portion of streamflow in many mountain areas of the world. However, the water paths from snowmelt to the arrival of the water in the streams are still largely unknown. This work analyzes for first time the influence of snowmelt on spring streamflow with different snow accumulation and duration, in an alpine catchment of the central Spanish Pyrenees. This study presents the water balance of the main melting months (May and June). Piezometric values, water temperature, electrical conductivity and isotope data (δ18O) allow a better understanding of the hydrological functioning of the basin during these months. Results of the water balance calculations showed that snow represented on average 73% of the water available for streamflow in May and June while precipitation during these months accounted for only 27%. However, rainfall during the melting period was important to determine the shape of the spring hydrographs. On average, 78% of the sum of both the snow water equivalent (SWE) accumulated at the beginning of May and the precipitation in May and June converted into runoff during the May–June melting period. The average evaporation‐sublimation during the 2 months corresponded to 8.4% of the accumulated SWE and rainfall, so that only a small part of the water input was ultimately available for soil and groundwater storage. When snow cover disappeared from the catchment, soil water storage and streamflow showed a sharp decline. Consequently, streamflow electrical conductivity, temperature and δ18O showed a marked tipping point towards higher values. The fast hydrological response of the catchment to snow and meteorological fluctuations, as well as the marked diel fluctuations of streamflow δ18O during the melting period, strongly suggests short meltwater transit times. As a consequence of this hydrological behaviour, independently of the amount of snow accumulated and of melting date, summer streamflow remained always low, with only small runoff peaks driven by rainfall events.


| INTRODUCTION
Snowmelt plays a critical role on streamflow generation in coldregions mountain headwaters (Barnhart et al., 2016;Li et al., 2017) and provides large amounts of water for ecosystems and human uses in their surrounding lowlands (Viviroli et al., 2020).During recent last years, intense research has been conducted in order to improve observational and modelling capabilities, and to better understand the physical mechanisms that connects the snow dynamics and streamflow generation in alpine catchments (Gordon et al., 2022).One of the most challenging aspects of this research topic is to determine the timing and routing from the snowmelt onset into the river flow (Ceperley et al., 2020).Routing may involve processes such as water percolation through the snowpack, the portion of snowmelt that quickly reach the streams as surface runoff, and water that infiltrates to aquifers or circulates as subsurface flow (Carroll et al., 2019).The difficulty to analyse such routing dynamics relies partly on the complexity of maintaining hydrological and hydrometrics measurements in snow dominated areas (Ala-aho et al., 2017).
Depending on the dominant hydrological processes, the transit time of melted snow to reach the stream at each catchment will vary and consequently will strongly determine its vulnerability to drought periods and climate change scenarios (Jeelani et al., 2017;Taylor et al., 2013).Some studies suggest that snowmelt dominated catchments show higher runoff coefficients than ephemeral snowpack and rain dominated catchments (Barnhart et al., 2016;Berghuijs et al., 2014;Li et al., 2017;Lone et al., 2023).However, other studies have not found any strong relationship between changes in the snowpack duration and magnitude of the annual runoff (L opez-Moreno, Pomeroy, et al., 2020).The transit time of snowmelt water in a catchment determines to which extent the accumulated snowpack during the precedent winter(s) and spring season(s) will affect the streamflow during summer time.Some studies have identified a clear role of the antecedent snowpack to explain anomalies in summer streamflow (Carroll et al., 2019;Godsey et al., 2014;Rebetez & Reinhard, 2008).
For example, summer low flows in Czechia are driven by seasonal precipitation and evapotranspiration but also by previous winter snowpack dynamics (Jenicek & Ledvinka, 2020).Conversely, the analysis of 380 Swiss catchments revealed that snow water equivalent and winter precipitation plays a minor role in the magnitude and timing of the warm season low flows (Floriancic et al., 2020).
The comparison between streamflow diel cycles and snow depletion time series also provides useful information about the snowmelt contribution to the total streamflow and their transit time (Holko et al., 2021;Jin et al., 2012;Kirchner et al., 2020;L opez-Moreno et al., 2023;Miller et al., 2020).During the melting season, rain provides a large streamflow contribution, and the meltwater contribution is often difficult to infer.Stable water isotopes (generally δ 2 H and δ 18 O) have helped to better understand the contribution of snowmelt to streamflow and the residence time of melting water in the catchments (Leuthold et al., 2021;McGill et al., 2021;Penna et al., 2017), thanks to the more depleted isotopic values of the melted snow compared to streamflow (McGill et al., 2021;Vystavna et al., 2021).However, a full separation of the contribution of each component is difficult to obtain, since it requires a very intense spatially and temporally isotopic sampling of each component.Further, at the catchment scale there is still a high spatial, as well as temporal (inter-and intra-annual) variability of the isotopic signal of the snowpack, precipitation (liquid and solid) and streamflow water (Wenninger et al., 2011).For this reason, the available literature often uses the evolution of the isotopic composition of water to perform qualitative rather than quantitative analyses, in combination with other source of data such as water characteristics (i.e., water temperature or electrical conductivity, geochemistry) and piezometric levels (Woelber et al., 2018).
In line with this, we analysed the streamflow response of a snow dominated basin in the central Spanish Pyrenees, in combination with water table data, streamflow, isotopic composition of precipitation, and additional information of water temperature and electrical conductivity.The general objective was to better understand the hydrological dynamics induced by snowmelt in this experimental catchment (Izas catchment), which is representative of large subalpine sectors in the Pyrenees.The results of this study are important to better predict the future hydrological response of similar catchments in the Pyrenees when snow duration and accumulation will decrease as a consequence of temperature scenarios for the next decades (L opez-Moreno et al., 2017;L opez-Moreno, Pomeroy, et al., 2013).The specific objectives of this work were: i.To estimate the water balance of the catchment during the main melting period (May and June).
ii.To improve the knowledge on the time in which snowmelt is converted into runoff.iii.To determine the possible influence of the cumulative winter snowpack on the observed hydrological behaviour during spring and early summer.iv.To assess the extent to which the annual hydrologic balance and hydrograph might change in a likely future with less snow.

| STUDY AREA
The Izas Research Experimental Catchment (42 44 0 N, 0 25 0 W) is located in the headwaters of the Gállego River in the central Spanish Pyrenees (Figure 1).The catchment has an area of 0.33 km 2 with altitudes ranging from 2075 m a.s.l.(gauging station) to 2325 m a.s.l.Landcover is dominated by subalpine meadows with some rocky outcrops.Slopes are gentle all over the catchment (mean slope ≈ 16 ) (L opez-Moreno, Fassnacht, et al., 2013).Grasslands are mostly composed by Festuca eskia, Nardus stricta, Trifolium alpinum, Plantago alpine and Carex sempervirens (Revuelto et al., 2017).Soils are generally well developed presenting an approximate depth of 1 m, with some areas accumulating deeper soils where some ephemeral springs are active during spring and early summer.Saturated soils during the melting period produce numerous solifluidal forms such as active lobes and terracets (García-Ruiz et al., 2021).
Snow covers most of the catchment for long periods, with the onset of snow cover being generally observed along November and melting starting in April or early May.Snow cover depletion is normally completed by the end of May or early June, even if snow patches occasionally last until late June.Snow depth shows a large interannual variability and also strong spatial variability, the later mostly driven by a combination of wind transport and the influence of elevation and topography on shortwave radiation (Revuelto et al., 2014).More than half of the annual precipitation (2000 mm year À1 ) falls as snow (Anderton et al., 2004).The catchment benefits from a transitional climate from Atlantic to Mediterranean, where winter and spring are the most humid seasons while summer is the driest, when precipitation is mostly the result of convective thunderstorms (del Barrio et al., 1997).
Annual mean temperature is +3 C, with mean daily temperature below 0 C for an average of 130 days per year (Revuelto et al., 2017).

| DATA AND METHODS
Table 1 shows the different measurements collected during the period between water years 2017 and 2020 and used in this study to assess the hydrological response of the Izas catchment during the melting period and the subsequent weeks.The data gaps correspond to sensor failures or changes of the measurement technology.

| Meteorological and snow data
We used temperature, precipitation (Geonor T-200B with wind shield) and snow depth (ultrasonic sensor) data from the automatic weather station located in the catchment (Revuelto et al., 2017).Data were recorded at 10 min interval and were aggregated into daily values.
F I G U R E 1 Map of the Izas catchment showing the location of the automatic weather station (yellow triangle), gauging station (blue square) and piezometers (red circles).Red line outlines the drainage area of the gauging station.Pink areas in the small map indicates the surface covered by the time lapse camera.
T A B L E 1 Data available for this study for the temporal period analysed.Water year spans from October to September.Time series from 1st April to 1st July were extracted to characterize the main melting period of the catchment.
Daily information of snow cover area was retrieved from a digital camera (Campbell CC640 digital camera) with a resolution of 640 Â 480 pixels that allows to cover around 75% of the drainage catchment (Figure 1).Daily photos were orthorectified and binarized into snow cover maps (presence/absence of snow).Data affected by low clouds were not included in the analysis.The daily photos were used to create series summarizing the snow cover area that authors consider fully representative of the conditions over the entire catchment (Revuelto et al., 2020).In addition, periodic field surveys were performed to derive distributed information on snow depth.In 2017 and 2018, snow depth maps were made using a terrestrial laser scanner (TLS) at dates close to the maximum snow accumulation (Revuelto et al., 2014).Normally, snow depth maps of the catchment were made by merging point clouds from two different scan positions (Figure 1).Meteorological and snow information were used to characterize the meteorological conditions and the magnitude and persistence of snow in the catchment during spring over the 2017-2020 period.

| Hydrological data and water sampling for isotopic analyses
Water level was measured every 5 min at the gauging station (Vshape weir), using a CT2X Seametrics probe (Seametrics, USA).Water level data were corrected from barometric pressure fluctuations using the software Aqua4plus 2.2 and converted into runoff (l s À1 ).The water level information for 2018 was missing due to a sensor malfunctioning.Water temperature and temperature-corrected electrical conductivity were also measured with the same probe.Electrical conductivity data were only available in 2019 and 2020 years when a proper calibration provided reliable values.
Nine piezometers were drilled at different locations within the catchment (see Figure 1) to monitor the fluctuations of the water 2019 and 2020, we collected a total of 22 water samples from the laminar flow flowing directly from the edges of the snow patches, with the aim of having a reference for the isotopic value of the water produced by the snowmelt.We also collected bulk precipitation fallen between two sampling days using a water collector designed to prevent evaporation (Gröning et al., 2012).For all samples, 15 mL of water were conserved in narrow neck propylene tubes and stored in an isothermal bag with cold ice packs to avoid evaporation during the 4 h of transport to the laboratory facilities.In the laboratory, the water tubes were kept in a fridge at +6 C. A Picarro L2130-i isotope analyser was used to measure δ 18 O isotopes in streamwater and precipitation samples.The isotopic values were determined from the eight replicates of the same sample to minimize sample carryover effects (Penna et al., 2012).A total of 553 samples of streamflow and 19 for precipitation were analysed for this study.Thus, the Penman-based unsaturated evapotranspiration routine of Granger and Gray (1989) was used for evaporation and the energybudget snowmelt model (EBSM) developed by Gray and Landine (1988) was used for snowmelt.EBSM (Dornes et al., 2008).In this study, we only used the sublimation and evaporation results from the CRHM model, as we consider that the quantification of SWE peak from our direct observation is much more accurate than simulated values, especially at a site where the undercatch strongly influences the measurement of solid precipitation (Buisán et al., 2017).

| Meteorological and snow conditions
The four analysed spring seasons (April-June) exhibited strong contrasts in terms of meteorological (temperature, precipitation) and snow conditions (snow depth and snow covered area) (Figure 2).April 2017 started with more than 1 m of snow depth at the ultrasonic sensor of the AWS, but it melted fast in the subsequent weeks.In early May, a significant portion of the catchment (40%) was snow free and average snow depth was only 0.63 m, with very few spots where snowpack exceeded 2 m (Figure 3).

| Water balance and streamwater response
A visual comparison between the hydrological (streamflow, water temperature and electrical conductivity of the streamwater, Figure 4) and the snow and meteorological conditions (Figures 2 and 3) revealed that the spring streamflow was determined by a mixed influence of snowmelt and precipitation (mostly as rain) events.The 3 years with available data suggest that the streamflow was mostly controlled by rain events after mid-June, independently of the snowpack accumulation and its duration in spring.
Water temperature also allowed to assess the relevance of snow In combination with streamflow data, the depth of the water table (and local snow depth) observed at nine locations in the catchment (2019 and 2020) yielded some additional information about the snowmelt influence on the dynamics of spring streamflow in the Izas catchment (Figure 6).The depth of the water table showed different responses among piezometers.However, several common dynamics may be observed.Before the start of the main melting season (April and May), the water table was low (i.e., deeper than 0.6-0.9m) at all location.However, when melt starts rapid water table fluctuations were observed in most locations, leading to several short periods close to saturation.Water table reaching the surface only was observed for short periods during the main melting events.Overall, the 2019 year showed shallower water table, associated to a deeper and longer-lasting snowpack than in 2020.In 2020, the water table dropped significantly in most of the piezometers, by the end of May, when snow cover was almost depleted in the catchment.On the contrary, in 2019, during the same period, water table was still close to the surface in several piezometers coinciding with a much larger snow-covered area in the catchment (Figure 2).Despite this general pattern, higher water levels were observed in some piezometers that had greater snow depth in their surroundings in-2020 than in 2019.

| DISCUSSION
This work combined several data sources, including snow measurements, meteorological information as well as hydrological records in order to assess the water balance and better understand the hydrological response of the Izas catchment during the snow melting period (May and June).

| Spring water balance in Izas catchment
Results showed that snowmelt of accumulated snow until early May is the main water input to the catchment (in average 73% SWE compared to 27% of precipitation in May and June) and drives much of the soil water fluctuations and streamflow during the melting period (May and June), which is in accordance with results found in other cold mountain sectors (Barnhart et al., 2016;Feng et al., 2022;Schreiner-McGraw & Ajami, 2022).Rainfall enhances the streamflow peaks controlled by melt, and keep the peak flows high once snow cover is almost depleted over the catchment (Gordon et al., 2022).
The importance of liquid precipitation for the spring hydrological response of snow-dominated catchments was also highlighted for alpine sites in the Dolomites (Penna et al., 2016).Runoff coefficients during the snow melting period were high.On average 78% of the sum of both SWE peak and precipitation in May and June was converted into streamwater over the May-June period.Using the CRHM model, we estimated that on average 8.4% of the input water evaporated or sublimated, leaving only 13.6% available for soil and groundwater recharge.These figures should be viewed with caution given the obvious uncertainties involved in estimating the individual components of the water balance, but can be considered a good approximation and are within the range of other water balances conducted in alpine and snow dominated catchments (Kraller et al., 2012;Krogh & Pomeroy, 2019;Zappa et al., 2003).

| Transit times from snowmelt to runoff
The measurement of piezometric levels during spring in 2019 and 2020 revealed that the storage of infiltrated water from snowmelt in the catchment was very variable among different points of the basin.Meltwater infiltration is probably controlled by the soil types and the terrain slope (Woelber et al., 2018).In most cases, the water levels fluctuations were very fast.Water level increases when melting starts, even if the entire catchment was snow covered.Saturation conditions (when water table reaches the surface) only happened during short periods.Saturation was often associated to the snow depletion period at each specific point.Afterwards, water levels declined considerably, and saturation conditions were not reached even in periods of heavy rain.Therefore, results suggest that under rainy conditions, the overland flow controls the hydrological response in the catchment.
When less than half of the catchment area was covered with snow, a sudden increase in water temperature, electrical conductivity alpine catchments with relatively shallow soils (Ceperley et al., 2020;Segura, 2021).Such behaviour contrasts with other alpine and subalpine catchments, where thick soils or sedimentary deposits favour the existence of alpine aquifers (Cochand et al., 2019;Hayashi, 2020;L opez-Moreno et al., 2023) and intense subsurface flow (Ceperley et al., 2020;Jin et al., 2012;Tague & Grant, 2009) that favour longer transit times and a slow hydrological response independently regardless the short term climatic fluctuations.
Isotopic values of stream water showed marked diel cycles of water isotopy during the melting periods, with a statistically significant correlation between the magnitude of the daily isotopic contrast and the mean daily temperature (as a proxy of snowmelt).This is a clear indication of the low transit time of water in the Izas catchment during melting period.However, groundwater storage and interflow processes can be present in the catchment.Direct in-situ observations suggest that that most of the tributaries to the main stream are completely dry during the driest period of the summer, but there is always some runoff at the gauge station thanks to lateral flow and few small perennial springs.(Floriancic et al., 2020).The lack of relation between antecedent snowpack and summer streamflow contrast to other mountain snow-dominated sectors, where snowmelt drives the streamflow anomalies several months after the snow depletion (Godsey et al., 2014;Staudinger et al., 2017).
Thus, streamflow reacted immediately to the onset of melt events, but also declined quickly when new snowfalls or cold periods occur.
After these interruptions, streamflow raised quickly when conditions that favour melting returned, or rain events occurred (Figure 2).
This work confirmed the relevance of monitoring stable isotopes for a better understanding of the catchment streamflow in relation to snowpack evolution.The δ 18 O magnitude and spatial variability across the catchment increases while the snow disappears from values very close to the isotopic signal of water from snowmelt to more enriched values, which is consistent with previous results (Dietermann & Weiler, 2013;Feng et al., 2002;Holko et al., 2013).Isotopic values from water flowing from springs are highly determined by snowmelt and they follow a similar enrichment speed to the one observed for streamwaters.The spring located on the north facing slope exhibited systematically lower values than the ones on the south facing slope, In the former, snowpack is significantly thicker and last for longer time (Revuelto et al., 2014), the isotopic values of water was very close to the values of sampled melt water.On the south facing slope, the spring is activated earlier by the snowmelt and also dries out much earlier than on the north facing slope, which explains the earlier enrichment of the water isotopes.The isotopic values of streamwater and water flowing from springs were significantly lower in 2018 than in other years.Unfortunately, we have not snowmelt samples of that year, and we can only hypothesise that this could be due to some heavy snowfall event, or the very heavy rain-on-snow event that affected the Pyrenees in April (Baladima et al., 2022) which could have very depleted isotopic values and influenced the isotopic signal of snowpack that year.

| Implications of the results to anticipate hydrological changes under a future with les snow and uncertainties in the research
The overall results indicated that snow strongly influences hydrology during the melting period.The expected future with a lower and shorter snow cover and a greater influence of rainfall (L opez-Moreno et al., 2017) may lead to shifts in the occurrence of the maximum peak discharge; and also an earlier rise in water temperature in the rivers, which may affect river ecology (Kamarianakis et al., 2016).However, the fast hydrological response of the catchment, the limited water storage capacity of the ground, and the importance of spring rainfall suggest that the main characteristics of the annual water balance and its hydrograph would not change in a drastic way.These results must be considered as local and explained by the main lithological, edaphic and climatological characteristics of the studied catchment.Mountain regions where most of the precipitation only fall during the coldest months of the year and where melt plays a major role in groundwater recharge will show a major dependence with the amount of timing of snow dynamics (Fayad et al., 2017).
The water balance simulation presented in the study is susceptible to evident uncertainties.These uncertainties can be reduced in future research by obtaining sublimation estimates from eddycovariance towers and employing more sophisticated simulations of the catchment's hydrological cycle that offer a more accurate representation of soil characteristics and groundwater processes.In addition, despite the considerable number of water samples taken in this study, the application of hydrograph separation based on water isotopes is complicated by the lack of detailed control of isotopic variations in individual precipitation events, water stored in soils and groundwater, and distributed samples of snowpack isotopy (Kamensky, 1998;Lee et al., 2010;Leuthold et al., 2021;Schmieder et al., 2016).Such monitoring should also be considered in further research.

| CONCLUSIONS
The combined study of meteorological, snowpack characteristics, pie- However, due to the harsh meteorological conditions of the 2018 snow season, we only obtained information from one scan position for this year.Therefore, the 2018 snow depth map presents large areas affected by topographic shadows.In 2019 and 2020, snow depth maps were created by photogrammetry based on Structure from Motion (SfM) algorithms with photos retrieved from a fix wing unmanned aerial vehicle (Ebee+) following the methodology presented byRevuelto et al. (2021).
Snow depth maps near the maximum snow accumulation (late April or early May, Figure2) were converted to snow water equivalent (SWE) maps by multiplying the snow depth by the manually measured snow density (using a Snowhydro ® SWE tube, L opez-Moreno,Leppänen, et al., 2020) on the same days next to the automatic weather stations.The assumption of a constant spatial distribution of density is an obvious simplification, but previous studies in the catchment have shown that the snowpack at the time of SWE peak is almost isothermal and the variations in density are very small (generally below <10% that is close to the instrumental error, L opez-Moreno, Fassnacht, et al., 2013;L opez-Moreno, Pomeroy, et al., 2013).Most of the precipitation in May and June falls as rain.This is added to existing SWE and considered as total water inputs into the catchment area for May and June.The total runoff of May and June is compared with the water inputs to obtain the runoff coefficient for the period in question.Evaporation and sublimation were calculated using the Cold Regions Hydrological Model-CRHM(Pomeroy et al., 2022) with the same configuration used in L opez-Moreno, Fassnacht, et al. (2013).
The basin was almost free of snow by 1st June.Spring 2018 was the snowiest year.By early May, almost 1.5 m of snow depth was measured in the AWS, and by mid-May, 100% of the catchment was still covered by a snowpack, which, on average, measured 2.6 m in depth.A significant snow cover lasted until mid-June and snow remnants lasted until early July.In spring 2019, snow was also relatively abundant compared to 2017 (1 m in depth as average).May was dominated by intense melt.June started with a 50% of the catchment covered by snow, which mostly disappeared in the next 2 weeks.At the beginning of May 2020, 80% of the catchment was covered by snow, with also an average of 1 m in depth.May and the first days of June where particularly warm, driving a fast melt.
melting in streamflow generation.Water was noticeably cold (normally <3 C, interrupted by some rainfall events and very warm days), and ruled by the snowmelt dynamics during the period encompassing mid-May to mid-June.Only when the snow cover area was less than approximately 10%, did water temperature increase quickly until reaching 15-20 C by the end of June.This was not the case in June 2018, when the deep and longest-lasting snowpack limited the temperature increase of streamflow water even in late June.During 2019 and 2020, electrical conductivity and temperature of the streamflow water showed a very similar temporal evolution.In both cases, a regular increase associated to the melting of snow cover was detected around the snow depletion date.Comparison between 2019 and 2020 years revealed that the shallower snowpack and its F I G U R E 2 Daily values of mean temperature (red line), precipitation (grey bars) and snow depth (blue line) measured at the in Izas catchment automatic meteorological station during the period 1st April to 1st July (2017 to 2020).Dots shows of the snow covered area (SCA in % of the area covered by the camera) obtained from time lapse photography.shorter duration in 2020, led to an earlier and faster increase in conductivity during the melting period.The water balance (Figure5) showed that the SWE is by far the largest potential contributor to catchment stream runoff (on average 73% snow and 27% liquid precipitation), even in low snow years (66% snow and 24% precipitation for 2017).The runoff coefficient averaged 78% (oscillating between 75% and 81%), and evaporationsublimation averaged 8,4% (oscillating between 7,3% and 9,5%) of the F I G U R E 4 Daily values of mean discharge (black line), streamwater temperature (red line) and electrical conductivity (blue line) during the period 1st April to 1st July at the Izas gauging station.F I G U R E 3 Snow depth for different dates with available data closer to the annual maximum snow accumulation.Snow depth records are acquired by using a TLS (2017 and 2018) and a UAV (2019 and 2020).total water input, indicating a very low amount of water available for soil and groundwater recharge.

F
Average daily streamflow δ 18 O values measured at the gauging station (Figure 7) remained low and relatively constant during the snow cover periods.Snowmelt water samples had very similar values between.Years, with a mean value of À10.57‰ (with 10th and 90th percentiles of À10.81‰ and À10.36‰, respectively).When significant snow cover prevailed, streamwater isotopic composition was similar to the snowmelt isotopic signal (ranging from À8‰ to À12‰), but increased significantly towards more enriched δ 18 O values after the snow cover disappeared (during early, late and mid-June in 2017, 2018 and 2019, respectively).Precipitation (mostly in liquid phase) showed more enriched δ 18 O values, ranging from À4‰ to À10‰, with most of the values ranging between À6‰ and À8‰.The year 2018 showed significantly lower isotope values of the streamwater than the years 2019 and 2020, which were below the snowmelt signature until the end of July.Despite this bias, the pattern of increasing values over the season is similar to that of the other 2 years.Comparison between morning and evening streamflow samples (6 AM and 6 PM, respectively) showed higher δ 18 O values in the latter, coinciding with a major influence of snow melting in the evening sample (Figure8a,b).This pattern occurred in the large majority of the days, regardless of rainfall events.The observed daily cycle disappears when snow cover depletes from the catchment (early and end June, for 2017 and 2018, respectively).Figure8c,dshowed the correlation between the daily oscillation of the isotopic morning and evening values of the streamwater and the daily temperature (as a proxy for snowmelt) for days without precipitation and with a snow cover of more than 50%.Both variables are strongly correlated (r 2 values of 0.67 and 0.47 for 2017 and 2018, respectively), indicating a rapid arrival of meltwater in the stream.δ 18 O values from water samples taken weekly across the catchment (Figure 9) showed a parallel evolution to those measured at the gauging station.The lowest and less variable δ 18 O values, which were very similar to the ones of melt water, were recorded when the Time series of mean daily depth to water table (2019 red line and 2020 blue line) observed at 9 locations during the period 1st April to 1st July in the Izas catchment.Red (2019) and blue (2020) dots show the mean snow depth measured with a UAV around (10 m 2 ) each location.F I G U R E 7 Precipitation δ 18 O (average of the 12 antecedent days; triangles) and streamflow δ 18 O signals (average of the morning and evening samples) measured at the Izas gauging station (lines).Blue dashed line and lower and upper limits of the blue band informs respectively of the average and 10th/90th percentiles isotopic values of snowmelt water samples taken in 2017, 2019 and 2020.catchment was snow covered.δ 18 O values progressively increased during the snow free period (July), showing also a higher variability between locations towards the end of the snow season.On most days, the isotope values of water flowing from were within the observed range of the values measured in the streams.The isotopic values of the springs on the north facing slope of the catchment generally had lower values than those on the south facing slope, and they were more often closer to the isotopic signature of snowmelt.

F
I G U R E 8 Daily streamflow δ 18 O values measured at 6 AM and 6 PM at the Izas gauging station during the years 2017 (a) and 2018 (b).δ 18 O values measured once a day (12 AM) during the snow free period later in the season are also shown.Figures C and D show the correlation between the daily oscillation of the streamwater isotopic signal (i.e., difference between the morning and evening δ 18 O values) and the daily temperature (as a proxy for snowmelt).
and δ 18 O values was observed.Further, the piezometric levels and the streamflow showed very low values 2 weeks after the snow depletion, independently of the snowpack magnitude and the duration of the snow season.The fast hydrological response of the catchment during the melting period was also suggested by the rather sudden change in water temperature, electrical conductivity and isotopic composition of streamwater after the snow cover depleted over the catchment.Such fast hydrological and water properties response to dominant climatic conditions is generally characteristic of many small zometric levels, streamflow and streamwater isotopic values and other physical properties have permitted to deepen the understanding on the water balance and the hydrological functioning of the Izas catchment during the melting period in years of contrasted snow duration and thickness.During spring, the hydrological timing and magnitude is mainly controlled by snow melting.On average, 73% of the water available for flow in the catchment in May and June corresponded to snow accumulated until early May whereas 27% corresponded to precipitation (mostly rainfall) in May and June.Over the melting period, 78% of the water available for flow was converted into runoff, the remaining 22% was divided into losses through evapotranspiration and sublimation (8.4%) and water available for storage in the soil and groundwater.Despite its relatively low percentage, liquid precipitation strongly shaped the hydrographs and contributed to the highest spring peak flows.Once snow disappeared from the ground, the piezometric levels quickly decline, and the streamflow values rapidly decrease.The increase of streamwater temperature, electrical conductivity and δ 18 O isotopic values in combination with the strong diel variability of the δ 18 O isotopic values during the snow melt period, which reacts very quickly to the daily temperature, indicated a rapid movement of the meltwater into the streams and a limited role of groundwater recharge, as the water balance also shows.Thus, the interannual variability of the snowpack had a limited role in the summer runoff evolution.Hence, snow rich years might not increase the catchment resilience to summer droughts.The study also highlights the limitations of lacking information on the groundwater isotope end member, which hinders to perform accurate separation of hydrographs.Future sampling strategies and research efforts should aim to address this gap.
Afterearly July, when melting did not drive daily runoff cycles, the sampler only collected one sample per day (12 AM).Water sampler had storage capacity of 24 bottles, making necessary to collect water samples every 12 days.During the days, we collected ISCO samples, 11 water samples from different streams and springs across the basin were also collected following a fixed itinerary that started at 12 AM, aiming to reduce the impact of daily cycles in water isotopy.On eight different days in 2017, table levels with LevelScout sensors (https://www.seametrics.com/product/levelscout/) that were also corrected from barometric pressure fluctuations.In this study, we focused on water table data from 1st April to 1st July to assess water table dynamics from before the onset of the main melting period until some weeks after snow has completely melted in the catchment.An ISCO 3700 automatic water sampler (https://www.teledyneisco.com/en-us/water-and-wastewater/3700-sampler)was used to sample streamwater at the gauging station twice daily in 2017, 2018 and 2019.The sampler malfunctioned in 2020 and also caused data gaps in 2019.The sampler was programmed at 6 AM and at 6 PM in order to capture the diel cycle from inexistent or very low snowmelt conditions (6 AM), and to very high input from snowmelt (6 PM).