5. DISCUSSION
Despite the obvious difficulty of gathering data in high alpine catchments, the complementary information derived from natural and artificial tracer data and from a rainfall-runoff model (i.e. a water balance) allowed a thorough quantification of recharge and discharge components of the catchment of the active rock glacier Innere Ölgrube at present. Recharge components vary seasonally as well as diurnally reflecting varying contributions from rainfall, snowmelt and ice melt. Discharge components are delayed to a certain extent due to the storage capabilities of the shallow aquifer within the rock glacier catchment (Wagner et al., 2020a), depending on the flow paths and saturation status of the groundwater body (Winkler et al., 2016, 2018).
The tracer velocities are in the range of reported values in active layers (Buchli et al., 2013; Krainer & Mostler, 2002; Tenthorey, 1992). The low recovery rate (< 50 %) likely indicates that some of the tracer is lost to the deeper domains of the rock glacier (potentially the unfrozen base layer). The tracer is presumably stored there for a longer period of time, thereby preventing it from reaching the measurement device at a detectable concentration within the observation period. Yet, the fast recovery of at least 40% of the tracer clearly indicates a fast flow component likely related to lateral flow on top of the permafrost table towards the rock glacier spring. This is also reflected in the hydrograph analysis. Concerning the slow flow component related to the unfrozen sediment layer (subpermafrost flow) at the base of the rock glacier and potentially other fine sediments within the catchment (e.g. till deposits and moraines between the cirque glaciers and the rock glacier) EC and isotopic data was analysed.
Figure 8 depicts the seasonal variations in individual “recharge” and discharge component contributions. “Recharge” is inferred from the rainfall-runoff model (Figure 8a) and does not take the loss due to evapotranspiration into account, as this process is happening in the production store (Figure 3) where the different input components are already mixed and evapotranspiration is dependent on the saturation / water level within that store. Nevertheless, evapotranspiration is limited in such an alpine environment. Discharge is inferred from event water analysis (Figure 8b), where higher mineralized water (higher EC) is related to longer stored water within the groundwater body and low mineralized water is related to event water derived from snowmelt, ice melt or rainfall. The actual differentiation of event water into rainfall, snowmelt and ice melt is not possible for that time frame (see Figure 6c,d for a detailed analysis over a short period of time using isotopic data in addition to EC).
Combining the information of Figure 8a and 8b allows for an interesting interpretation:
The groundwater component (longer stored water within the aquifer) is the most important component during periods of little to no recharge (e.g. winter time; Wagner et al., 2020a). Snowmelt and (subordinately) rainfall are the dominant recharge components during early summer, rapidly increasing the event water contribution to ~60 %. In particular during the summer with little to no rain and after snow has melted, ice (glacier or permafrost ice) runoff and groundwater discharge become the relevant components. Separating groundwater and ice melt water due to the diurnal pattern of melting process (related to variations in air temperature) is possible. The pronounced change in discharge pattern in early autumn observable in Figure 6b (reduced discharge and attenuated diurnal variations caused by declining air temperatures in early September) indicates the transition from the event-water dominated period (late April - August) to the groundwater dominated period (September until onset of the snowmelt in April-early May). Some caution is required when interpreting relative discharge component contributions during the onset of snowmelt in April (Figure 8), since the only available EC time series during that month starts on April 25, 2018. Thus, the groundwater contribution during (early) April is likely underestimated in Figure 8.
[Insert Figure 8]
To identify and quantify the main source of the ice melt water potentially from permafrost ice melt or from the upper cirque glaciers is more complex. Natural and artificial tracer test results show that event water travels through the rock glacier within several hours. Ice-derived meltwater can be observed with a time lag in the order of ~ 16 hours relative to air temperature indicated by EC and isotopic data (Figure 6c). This time lag in combination with the fast response due to event water suggests some distance of the melt water source to the spring/rock glacier and favour the cirque glaciers as the main melt water source. This is further substantiated by comparing the time lag to the artificial tracer test results: The linear distance between the lowermost point of the cirque glacier and the gauging station is approximately 1500 m, indicating an actual flow distance ≥ 1500 m. Typically, runoff from glaciers reaches a maximum a few hours after the peak in meltwater production (Cuffey & Paterson, 2010), thus the runoff peak at the glacier precedes the peak at the gauge by ≤ 16 h. Both estimates suggest that the water travels from the front of the glacier to the gauging station at a velocity ≥ 0,026 m/s. This is in good agreement with the artificial tracer test results yielding slightly higher linear velocities. Note also that melting rates at the cirque glaciers reflect diurnal variations in radiation, heat and vapour content of the adjacent air (Cuffey & Paterson, 2010). The coarse grained active layer covering the permafrost ice within the catchment protects it from radiation and induces a damped and retarded variation in temperatures, decoupling it from external weather and climate conditions (Haeberli, 1985; Jones et al., 2019; Vonder Mühll, 1993; Wagner et al., 2019). Accordingly, melting of permafrost ice is expected to follow changes in atmospheric conditions in a more damped and retarded fashion, suggesting cirque glacier melting to account for the observed diurnal variation in discharge and natural tracers.
The amount of ice melt is in the order of 30% of the annual recharge and will decrease and finally disappear in the future due to glacier melt. Thus, ignoring ice contribution from cirque glacier melt is hypothetical, but interesting at least. This can be assessed by using the calibrated rainfall-runoff model and assuming that ice melt from the cirque glaciers is absent due to potential future complete glacier melt (green line in Figure 9). This was simulated in a simplified manner by actually eliminating the (glacier) ice storage, keeping all the other parameters constant. Increasing runoff during snowmelt periods does not change expectedly; however, the runoff in late summer indicates a certain reduction. Interestingly, base flow during winter months does not change significantly. This is an important finding that further supports that base flow is mainly derived from the unfrozen base layer of the rock glacier (and other fine-grained sediments in the catchment area (e.g. till deposits, moraines); cf. Wagner et al., 2020a). In total, a runoff reduction of almost 30% is expected in this hypothetical scenario due to the disappearance of the cirque glaciers.
Another hypothetical scenario can be constructed by applying the model parameters of a relict rock glacier where the same model was already applied (Table 2; cf. Wagner et al., 2016) for the current setting of the active rock glacier. This trading space-for-time approach accounting for changing watershed behaviour under permafrost-free conditions (Singh et al., 2011) allows to speculate how progressing climate change might influence the discharge pattern, considering the associated uncertainties (purple line in Figure 9).
[Insert Figure 9]
Changes are not great and only a slightly more buffered behavior during storm events or snowmelt periods is visible compared to the scenario where only the ice melt from the glaciers is neglected. Interestingly, slightly slower recessions at the onset of winter periods are observable. This can be explained by a higher routing store x3 for the relict (Schöneben) rock glacier catchment than the active (Innere Ölgrube) rock glacier catchment. This might indicate an increase in storage capacities as permafrost thaw potentially leads to more available pore space in the shallow aquifer system (cf. Rogger et al., 2017). Nevertheless, in the case of the relict Schöneben rock glacier as well as for the active Innere Ölgrube rock glacier, both are known to have a rather fine-grained, unfrozen base layer that is 10-15 meters thick. These similarities might explain their rather similar behavior, although one may speculate about a further increase in storage capacity within the Innere Ölgrube catchment as permafrost thaws eventually. This potential change is depicted by computing the master recession curves of the modelled runoff patterns for the “current” data, the hypothetical scenario of vanished cirque glaciers and the other hypothetical scenario of a relict rock glacier (Figure 10). The comparison to the observed data further indicates the good model fit in addition to the efficiency criteria and visual fit of the hydrographs presented before (and shown again in the inset). The similar master recession curves of the “current” data and the hypothetical one without the glacier ice melt is to be expected, as the actual internal (or model) structure did not change, but only a recharge component was “removed”. The discharge pattern and consequently master recession curve of the hypothetically relict rock glacier catchment (using the parameter set of the relict Schöneben rock glacier; Wagner et al., 2016) indicates a slightly slower recession during winter periods as storage capabilities increased.
[Insert Figure 10]
What remains to be identified is the actual contribution of permafrost ice melt at present and future changes in the runoff pattern due to this additional water in the short term and potential increases in the storage capacity in the long term when the permafrost ice will melt completely. The trading space for time approach presented herein does suggest a slight difference in runoff patterns and supports what was suggested by Rogger et al. (2017): Runoff might be dampened slightly and storage capacities might increase to some extent. However, further research addressing this particular case and ideally monitoring rock glacier catchments without the influence of cirque glaciers within their catchment is needed. Colombo et al. (2018) suggest establishing baselines for future monitoring related to downstream water quality (and quantity) as solute export from deteriorating rock glaciers might provide valuable information.
In alpine catchments, rock glaciers should be seen as hydro(geo)logically (and geomorphologically) conservative systems (Giardino et al., 1992) that become increasingly important as ice glaciers will continue to vanish. For a recent assessment of potential changes in the European mountain cryosphere until the end of the 21st century, see Beniston et al. (2018). There, improvements in understanding changes in re- as well as discharge in high-alpine areas are warranted. The here gathered information and developed understanding of a high-alpine spring catchment that drains an active rock glacier and two cirque glaciers is of importance and should be extended by including downstream observations (cf. Wagner et al., 2016). Another open question remains the actual contribution of permafrost ice melt relative to the glacier ice melt and becomes increasingly important with regard to climate change and a likely further warming trend. A recent review about potential ecosystem shifts in alpine streams highlights the importance of a better understanding of permafrost / rock glacier thaw (Brighenti et al., 2019). Therefore, long-term monitoring of such high-alpine catchments (and streams further downstream) are essential.