Introduction of a Superconducting Gravimeter as Novel Hydrological Sensor for the Alpine Research Catchment Zugspitze

The Zugspitze Geodynamic Observatory Germany has been set up with a worldwide unique installation of a superconducting gravimeter at the summit of Mount Zugspitze. With regard to hydrology, this karstic high-alpine site is largely dominated by high precipitation amounts and a long seasonal snow cover period with significant importance for water supply to its forelands, while it shows a high sensitivity to climate change. However, regarding the majority of alpine regions worldwide there is only weak knowledge on temporal water storage variations due to only sparsely distributed hydrological 20 and meteorological point sensors and the large variability and complexity of alpine signals. This underlines the importance of well-equipped areas such as Mount Zugspitze serving as natural test laboratories for an improved monitoring, understanding and prediction of alpine hydrological processes. The observatory superconducting gravimeter OSG 052 supplements the existing sensor network as a novel hydrological sensor system for the direct observation of the integral gravity effect of total water storage variations in the alpine research catchment Zugspitze. Besides the experimental setup and the available datasets, 25 the required gravimetric prerequisites are presented such as calibration, tidal analysis and signal separation of the superconducting gravimeter observations from the first 2 years. The snowpack is identified as primary contributor to seasonal water storage variations and thus to the gravity residuals with a signal range of up to 750 nm/s2 corresponding to 1957 mm snow water equivalent measured at a representative station at the end of May 2019. First hydro-gravimetric sensitivity analysis are based on simplified assumptions of the snowpack distribution within the area around Mount Zugspitze. These reveal a 30 snow-gravimetric footprint of up to 4 km distance around the gravimeter with a dominant gravity contribution from the snowpack in the Partnach spring catchment. This study already shows that the hydro-gravimetric approach can deliver important and representative integral insights into this high-alpine site. This work is regarded as a concept study showing preliminary gravimetric results and sensitivity analysis for upcoming long-term hydro-gravimetric research projects. https://doi.org/10.5194/hess-2021-78 Preprint. Discussion started: 8 March 2021 c © Author(s) 2021. CC BY 4.0 License.

LiDAR observation techniques are not able to measure the hydrologically relevant SWE value directly, but rather rely on additional snow density data from local snow pit or snow weight measurements.
The last years of terrestrial gravimetric research have seen a transformation of SG installations from low noise sites for the 70 analysis of global geophysical phenomena to specific sites of interest for the monitoring of near-surface mass transport processes. These include the development of SGs as hydrological sensors for the direct, integral and non-invasive monitoring of total water storage variations in a radius of up to 4 km around the SG in a minimized field enclosure (Güntner et al., 2017).
Just recently, Chaffaut et al. (2020) reported about an SG installation in a lower mountain catchment in the French Vosges for the analysis of water storage dynamics. In this catchment, however, seasonal snow cover plays a minor role. In addition, SGs 75 are applied in connection with absolute gravimeters (AGs) and relative spring gravimeters (RGs) in hybrid approaches for hydro-gravimetry (Naujoks et al., 2010) and even for volcano monitoring (Carbone et al., 2019) and geothermal mass movements (Schäfer et al., 2020). These hybrid approaches exploit the advantages of the various types of gravimeters with the AGs providing long-term gravity changes, SGs for the continuous high-precision temporal gravity changes at the measurement sites and RGs for additional spatiotemporal variations in the area of interest. 80 Any available catchment data can be either used directly to analyse the dynamics of individual components of the catchment water balance, or used in combination with snow-hydrological models where it can provide information on the initial and boundary conditions and also information relevant for model parameters, as well as important data to calibrate the model.
While snow cover and snow height data are able to condition the model behaviour to some degree (Weber et al., 2020a), it has 85 recently been shown that integral data from satellite (Bahrami et al., 2020) and terrestrial gravimetry (Güntner et al., 2017) that provide footprint averaged time series of terrestrial water storage anomaly (TWSA) can greatly improve the identification of water balance components and relevant hydrological processes in the catchment.
Gravimetric methods are already applied in the RCZ. Episodic AG observations have been carried out since 2004 with a FG5 90 absolute gravity meter by the Leibniz University Hannover (LUH) for the analysis of long-term gravity changes at Mount Zugspitze and for a long-range gravimeter calibration base (Timmen et al., 2006;Peters et al., 2009). Timmen et al. (2021) estimate a geophysical trend of -20 nm/s²/yr. with an uncertainty of 3 nm/s²/yr. (single standard deviation 1σ) from AG observations between 2004 and 2019 as a consequence of Alpine mountain uplift and hydrological mass loss. Monthly RG observations have been done with a transportable spring gravimeter since 2014 by the Technical University of Munich (TUM) 95 for the analysis of periodic permafrost changes and cavity detection in a tunnel (Kammstollen) of Mount Zugspitze (Scandroglio et al., 2019). The most important gap in the hybrid gravimetric approach has now been closed by the installation of the SG at the summit of Mount Zugspitze enabling the separation of short-term, seasonal, interannual and long-term gravity changes. The overall research question to be discussed in this study is to what extent the OSG 052 can contribute to a better understanding of hydrological processes in high-alpine catchments. This addresses the benefits and improvements of the hydrogravimetric approach but also its limitations. After a presentation of the available data sets from gravimetry to hydrology and meteorology in the RCZ (section 2), the applied gravimetric methods for the separation of the hydrological signal from the gravity observations are shown (section 3). Section 4 contains the hydro-gravimetric results and sensitivity analysis with regard 105 to various water storage components. Section 5 summarizes the main results and provides an outlook on future hydrogravimetric projects.

Observations
Mount Zugspitze is Germany's highest mountain with an altitude of 2962 m. There are half a million tourists every year and correspondingly high media attention. The worldwide unique setup of the Zugspitze Geodynamic Observatory Germany at the 110 summit of Mount Zugspitze (Figure 1) is explained including all relevant sensors and datasets. In addition, the available hydrological and meteorological datasets from the alpine research catchment Zugspitze are introduced.

The Zugspitze Geodynamic Observatory Germany
The Zugspitze Geodynamic Observatory Germany (ZUGOG) has been set up by the German Research Centre for Geosciences (GFZ) at the summit of Mount Zugspitze in a former laboratory of the Max Planck Institute for Extraterrestrial Physics (MPE), 115 which was built in 1963 for the observation of cosmic rays ( Figure 2). The 10 m high tower-like aluminium structure of the lab with an area of approx. 50 m² has a very steep roof to keep it free from snow. In addition, the position at the summit prevents hydrological mass variations above the sensor and simultaneously increases the hydro-gravimetric footprint. Inside the lab, there is a measuring room with two concrete piers on the ground floor. While the first concrete pier is occupied by the SG, the second one is intended for absolute and relative gravimeters as well as other instruments. On the first floor are basic 120 sleeping facilities for overnight stays if necessary. A ventilation has been installed in order to reduce the heat produced by the compressor of the SG. This is necessary as the lab itself heats up considerably during sunny days. In addition, a thermally insulated box has been built around the SG including heaters to keep the sensor at a stable ambient temperature of around 25°C. Temperature and humidity sensors have been installed in the lab. The power supplies of all electronics and compressor are secured by uninterruptible power supplies (UPS). The lab is accessible all year round with cable cars from Germany and 125 Austria. The UFS provides personnel and technical support as well as infrastructure during maintenance trips.  In September 2017, the OSG 052 was warmed up to room temperature at its former location in Sutherland, South Africa, and sent to the manufacturer GWR Instruments, Inc. in San Diego for refurbishment after observing in parallel with the dual sphere OSG D-037 between 2008 and 2017 . The refurbishment of the 10 years old SG at GWR included the 135 thermal levellers, an upgrade of the electronics from version GEP-2 to GEP-3, a modification of the dewar to enable cooling from room temperature down to 4 K with the accompanying refrigeration system, a replacement of the GPS antenna and a barometer specifically calibrated for a working altitude of 3000 m as well as an Intel mini Personal Computer for the operation of the UIPC software under Windows 7.  After returning to GFZ, the OSG 052 has been moved to ZUGOG in September 2018 by truck, cogwheel train and helicopter at operating temperatures of 4 K. The first weeks of gravity observations with the OSG 052 showed an instrumental malfunction with a very large negative drift of about -50 nm/s²/day and several small offsets. This led to the decision at the 155 end of October 2018 to warm up and re-cool the SG in order to eliminate the abnormal drift. However, the cooling process stopped at 160 K, and the dewar had to be pumped out to recreate a proper vacuum. After that, the temperatures started to decrease again and, finally, the levitation of the sphere could be completed and the SG has been in nominal operation since 29 December 2018. ZUGOG is part of the International Geodynamics and Earth Service (IGETS; Boy et al., 2020)  Level 1 raw gravity and atmospheric pressure data with sampling rates of 1 s and 1 minute (Voigt et al., 2019) on a regular 160 basis to the publicly accessible IGETS data base hosted by GFZ . In addition, the continuous Global Navigation Satellite Systems (GNSS) station ZUGG (Ramatschi et al., 2019) has been installed at 9 Sep 2018 nearby the lab for monitoring of deformations (Figure 2 top left) and is recording since then.
For the monitoring of local hydrological and meteorological variations, several environmental sensors have been installed. A 165 snow scale and three laser-based snow height sensors have been installed in front of the lab in order to quantify the accumulated snow masses on this horizontal plane during the winter months. After the experiences from the first winter 2018/2019, the pole with the snow height sensors had to be extended from 2.5 to 4 m ( Figure 2 bottom left). Another laser-based snow height sensor has been installed with a view to the slope directly below the SG. Laser-based sensors have been preferred instead of the widely used ultrasonic sensors because the snow cover is not horizontal. A small meteorological station outside the lab 170 observes temperature and humidity as well as wind speed and direction. All data sets are parts of a remotely controlled monitoring system.

Hydrological and meteorological datasets in the research catchment Zugspitze
ZUGOG is connected to the UFS as the home base of a large research consortium operating a dense hydrological and meteorological sensor network for more than 20 years. Long-term meteorological datasets are available on hourly to yearly 175 basis from the Climate Data Center (CDC) of DWD for the station at the summit of Mount Zugspitze (Station ID 5792) including relative humidity, air temperature, precipitation height and form, wind speed and direction and air pressure. LWD provides several hydrological and meteorological datasets from a station at the Zugspitzplatt at an altitude of 2420 m including the SWE of the snowpack recorded by a snow scale. Moreover, in the last few years three further meteorological stations were set up at the Zugspitzplatt as well as on two mountain ridges in the frame of the Virtual Alpine Observatory project (VAO). 180 Gauge stations have been set up for the observation of the discharge at Partnach spring and Bockhütte (Figure 1), while another gauge station is planned for the Hammersbach catchment. The Partnach spring catchment (RCZ) covers an area of about 11 km² located in the Northern Limestone Alps. Its main characteristics are a mean annual precipitation sum of 2080 mm and an average temperature of -4.5°C regarding the climatic reference period 1981 to 2010 (Weber et al., 2016). The altitudes vary 185 between 2962 m (summit of Mount Zugspitze) and 1430 m at Partnach spring where the catchment is exclusively drained due to the special geological karst situation being therefore regarded as a natural lysimeter. Germany's highest glacier remainsthe Northern and Southern Schneeferner -are also located in the RCZ as well as permafrost in the rock walls of Mount Zugspitze (Krautblatter et al., 2010). The RCZ is part of the Bockhütte catchment with an area of 25 km². The Hammersbach catchment covers an area of 14 km² in northeastern direction from Mount Zugspitze and includes the Höllentalferner glacier. 190 Karst hydrological characteristics can be found in Lauber and Goldscheider (2014). All datasets are compiled in the Alpine Environmental Data Analysis Center (AlpEnDAC) as part of the VAO. Overall, this high-alpine region has one of the highest densities of meteorological stations worldwide and serves therefore as an ideal reference for testing new measurement and modelling approaches. 195

Gravimetric methodology
The essential prerequisite for the application of the OSG 052 as a hydrological sensor is the separation of the hydrological signal from the raw gravity observations. The required steps are explained on the basis of the continuous time series from 29 Dec 2018 to 31 Dec 2020 (2 years of observation).

Pre-processing and calibration 200
Raw gravity observations are voltage variations in 1 s sampling along with the observed barometric pressure variations and stored in daily files of the Tsoft format (Van Camp et al., 2005). These are compiled into monthly files and converted into the GGP format of IGETS. Short data gaps up to 10 s are interpolated linearly on the full signal. For the decimation from 1 s to 1 min sampling, a double precision Chebyshev filter "g1s1m" with a filter length of 1009 s is applied (Crossley, 2010). For the reduction of the gravity data and the removal of spikes and offsets as well as the filling of longer gaps, the programs DETIDE 205 and DESPIKE of ETERNA 3.4 (Wenzel, 1997) are used. Besides some steps on 7 Jan 2019 during the final centering of the sphere, the time series has shown only two additional steps on 24 Oct 2019 during a simulated power failure for UPS testing and on 7 Dec 2020 after exchanging the CMOS battery on the GEP remote card, which were eliminated manually with the program Tsoft. Overall, the SG shows a very stable performance. Finally, the monthly files in 1 min sampling are further decimated to one long time series in 1 h sampling by the program DECIMATE. 210 For the transition from voltage to gravity variations, the amplitude factor of the OSG 052 has been determined on the basis of two absolute gravimeters and one calibrated spring gravimeter (Table 1). The first estimation was done in 2011 at Sutherland with FG5-301 by the German Federal Agency for Cartography and Geodesy (BKG). In order to validate this result after repeated transport of the SG and refurbishment at GWR, the second estimation was done in Sep 2018 at ZUGOG with FG5X-215 220 by LUH (Timmen et al., 2021), however, with a reduced accuracy due to the malfunction of the SG at this time (see section 2.1). Hence, a third estimation was carried out in Sep and Oct 2019 at ZUGOG on the basis of the relative spring gravimeter CG6-69 of GFZ calibrated in the gravimeter calibration system Hannover (Timmen et al., 2020). Within a least-squares adjustment, the amplitude factor of OSG 052 and a best-fitting polynomial reflecting the irregular drift of the CG6 were determined. The full period of 4 weeks of co-located measurements was divided into blocks from 2 to 4 days with and without 220 overlap and polynomial degrees of 2 and 3. The best fitting solution (smallest standard deviation for the amplitude factor) was found to be for blocks of 3 days, polynomials of degree 3 and 50 % overlap. The final amplitude factor is -749.59 nm/s²/V (1σ=0.22 nm/s²/V) as a weighted mean from calibrations 1-3. The achieved accuracy should be sufficient with regard to the https://doi.org/10.5194/hess-2021-78 Preprint. Discussion started: 8 March 2021 c Author(s) 2021. CC BY 4.0 License.
hydro-gravimetric analysis. Amplitude factor deviations of 1 nm/s²/V correspond to 1.3 x 10 -3 relative and 1 nm/s² absolute gravity deviations for maximum gravity residuals of 800 nm/s² which can be used as measure for the accuracy of the gravity 225 observations at ZUGOG.
The time delay of the OSG 052 was determined within a step response experiment developed by GWR on 1 March 2019 at ZUGOG. 16 introduced step voltages were analysed with the program ETSTEP of ETERNA 3.4 (Wenzel, 1997) and the time delay was estimated to 10.53 s (1σ=0.03 s). 230 The instrumental drift of the OSG 052 has not been determined yet. The first absolute measurements cannot be used as reference value for the drift estimation, as the SG had to be warmed up and cooled again for re-initialisation at the end of 235 December 2018 (section 2.1). Hence, there is no connection to the current continuous SG time series and the second absolute measurements at 26-27 Sep 2019 (Timmen et al., 2021). For the preliminary analysis shown in this study, the drift is assumed to be zero and no trend is subtracted from the SG time series. Further absolute gravity measurements are planned with the FG5X-220 for 2021, which is required for the adequate determination of the rather small drift of some nm/s²/yr.

Tidal analysis 240
In order to reduce the gravity effects from solid Earth and ocean tides, a local tidal model was computed based on 2 years of observations (29 Dec 2018 -31 Dec 2020) with the program ANALYZE of ETERNA 3.4 (Wenzel, 1997) for the analysis of monthly, diurnal, semidiurnal and shorter tidal waves. The estimated amplitude factors and phase leads according to the ETERNA wave grouping for a 1-year gravity time series are displayed in Table 2. The numerical high-pass filtering and a Hann window usually applied for the analysis of diurnal and semi-diurnal waves were deactivated for the simultaneous analysis 245 of the monthly waves resulting in higher standard deviations for the shorter waves. Instead, a Chebyshev polynomial of degree 2 was applied in order to eliminate any long-term instrumental trend signal or long-term variations in gravity. Longer tidal waves (half year periods and longer) are considered by nominal values, i.e. amplitude factors of 1.16 and phases 0°. It should be noted that these results could be reproduced by ET34-ANA-V80 (Schueller, 2020) with very small differences. However, modifications with regard to wave group modelling including higher tidal potential degrees and additional tidal constituents 250 have not been carried out yet but are planned for the future on the basis of longer time series. Along with the tidal waves, the https://doi.org/10.5194/hess-2021-78 Preprint. Discussion started: 8 March 2021 c Author(s) 2021. CC BY 4.0 License. single admittance factor between gravity and barometric pressure was determined with -3.6506 nm/s²/hPa (1σ=0.0393 nm/s²/hPa). In order to reduce the large seasonal gravity signal, a second admittance factor was determined between gravity and the snow water equivalent (SWE) with 0.2965 nm/s²/mm (1σ=0.0007 nm/s²/mm; cf. section 4.2). The gravity variations induced by solid Earth and ocean tides are predicted with the local tidal parameters from Table 2 and shown for the analysed 255 period in Figure 3 (top right).

Non-tidal gravity reductions
Besides tidal variations, the gravity observations include significant non-tidal effects shown in Figure 3. For a more detailed compilation of temporal gravity field variations see e.g. Voigt et al. (2016a). The signal admittance factor from the tidal analysis of -3.6506 nm/s²/hPa includes the maximum correlated signal between observed gravity and barometric pressure. For a refined modelling of gravity variations induced by mass redistributions in the atmosphere, the Atmospheric attraction 265 computation service (Atmacs; Klügel and Wziontek, 2009) provides effects from local to global scales with a temporal resolution of 3h based on 3D ECMWF weather data. However, the limited spatial resolution of the weather models of 7 km https://doi.org/10.5194/hess-2021-78 Preprint. for Europe shows that the complex topography around the station cannot be represented sufficiently. Therefore, the local part from Atmacs is not used. Instead, a single admittance factor is estimated between the gravity observations reduced by regional and global atmospheric effects (i.e. the local atmospheric effect) and barometric pressure variations with -2.9190 nm/s²/hPa 270 (1σ=0.0274). This is the local atmospheric effect which is added to the regional and global parts from Atmacs. A future task is to consider the complex alpine topography in the area around the gravimeter either by using a weather model with a higher spatial resolution or by setting up a local model based on an array of available barometers (Riccardi et al., 2007). Additional Hydrological gravity variations can be subdivided into those from local scales (up to several meters around the gravimeter) 290 over alpine catchment scales (from several meters to kilometers) to non-local scales (from several kilometers). Non-local hydrological gravity variations include both attraction effects and surface loading, while for local to catchment scales only the attraction effects from mass redistributions is considered. The non-local hydrology is provided by the EOST Loading Service

315
An essential task is the reduction of local hydrological signals in order to enhance the sensitivity towards the catchment scale.
While the steep roof of the lab and its position above the slope at the summit are very advantageous, most disturbing signals are expected from snow masses on a horizontal plane with an area of 5 m x 10 m directly in front of the lab (Figure 2 bottom left) where a local snow monitoring network has been set up. A first estimation of the maximum signal from snow heights of 4 m and densities of 300 kg/m³ reveals a significant gravity effect of 25 nm/s² with a seasonal character superimposed by 320 somewhat smaller event-like signals during heavy snowfall events. As this is only a fraction of 1:30 with regard to the total gravity residual range and due to initial problems as a consequence of very harsh conditions during winter 2018/19 (frozen snow height sensors, torn cables, too low sensor pole height), a model-based description of the local snowpack situation has not yet been set up completely.

Hydro-gravimetric results and sensitivity analysis 325
The subsequent hydro-gravimetric analysis based on the gravity residuals from Figure 3 (bottom) focus, for the first time, on a high-alpine region largely dominated by seasonal snow cover. The gravity residuals reflect the total water storage variations from local to catchment scales as the balance of precipitation, glacial melt, retention in the karst, runoff and evapotranspiration.
At Mount Zugspitze, this includes a large annual precipitation of approx. 2000 mm with 80 % as snowfall from autumn to late spring above approx. 1800 m, leading to a profound seasonal snow storage. The discharge measured at the gauges, is largely 330 characterized by a high-alpine snowmelt runoff regime. The discharge peaks occur in summer and include besides large portions of melted snow also components of rainfall events, and to a minor extent melted ice from the Northern and Southern Schneeferner as well as the Höllentalferner glaciers (Hagg et al., 2012) and from thawing permafrost below the summit of Mount Zugspitze (Krautblatter et al., 2010). Moreover, the water storage in the vadose and phreatic zones and subsequently also the discharge are influenced by the geological karst situation in this region. Evapotranspiration has a large geographical heterogeneity of the determining parameters on catchment scale (Wetzel 2004). At the surface of the highly karstified limestone of this region, only short distances of overland flow can be observed. Accordingly evapotranspiration is reduced due to lacking surface storage capacities in soils or fine-grained sediments.
The large complexity and variability of the hydrological parameters show the high potential of continuous gravity observations 340 providing the integral signal of all mass redistributions in the vicinity of the gravimeter and serving as constrains for the hydrological modelling on catchment scale. The following hydro-gravimetric analysis should be regarded as a preliminary concept study to demonstrate potentials and limitations of integrating gravimetric signals into analyses of high-alpine hydrological processes.

Water balance 345
The gravity residuals from the hydro-gravimetric approach in Figure 3 Timmen et al. (2021). They suggest that the main contribution is caused by glacier diminishing, and a smaller part is explained by mountain uplift (1 mm causes -2 nm/s²). With a multi-year continuous gravity time series from OSG 052, it is possible to study the evolution of seasonal and -in combination with absolute gravity observations -also longterm water storage variations. 355 However, for the hydrological decomposition of the gravity residuals into individual water storage components, complementary data from meteorological and hydrological techniques are needed. The gravimetric method is known to be most sensitive to local mass variations in vertical direction with a signal attenuations by 1/r² (r being the distance between gravimeter and source mass) and further attenuation towards increasing horizontal directions. Hence, the essential question is 360 how sensitive the gravity residuals are with regard to individual water storage components from local to catchment scales. This question will be addressed in the following sections.

Snowpack
Representative observations of the snow water equivalent (SWE) for the Zugspitzplatt are available from the LWD station regression factor between gravity residuals and SWE of 0.298 nm/s²/mm (1σ = 0.003 nm/s²/mm). The high correlation of 0.969 between the gravity residuals and SWE is clearly visible and both are following similar seasonal patterns.
In general, the winter seasons of 2018/19 and 2019/20 were very different. The first winter season is characterized by a sharp 370 increase in SWE in mid-January as well as in the second half of May due to massive snowfall. The maximum SWE is extraordinary high, with a value of 1957 mm measured at the LWD station, compared to an annual mean of the maximum SWE at approx. 1350 mm since the installation of the snow scale in 2014 until now. Contrary, the second winter season with a maximum SWE of 1147 mm at the LWD station reflects more a winter with a rather small to normal amount of snow. While in 2019 the seasonal gravity and SWE maxima coincide at 29 May, there is a difference of more than one month between 375 gravity and SWE maxima in 2020, i.e. 14 Mar and 19 Apr, respectively. However, in 2020 there was no distinctive SWE peak rather a longer period with maximal SWE values between these two dates. Higher temperatures during April 2020 led to an early onset of snowmelt in the lower part of the catchment with beginning recharge of the karst water body and increasing runoff at Partnach spring.

Figure 4: Gravity residuals (solid blue), snow water equivalent (solid red) from LWD station at Zugspitzplatt and SWE multiplied with estimated regression factor of 0.298 nm/s²/mm between gravity residuals and snow water equivalent (dashed blue line)
The essential question is how representative the locally observed gravity residuals are with regard to the snowpack of the negligible. The remaining 23 % contribution to the total gravity signal from snow masses comes mainly from the area northwest from ZUGOG but should be much less in reality as there should be less snow in this area due to very steep slopes and prevailing westerly winds.
As the sensitivity diminishes with increasing distances from the gravimeter site, the individual contribution of the snow covered 425 areas in the RCZ are analysed being coloured from yellow to deep red (>10 -3 nm/s²) in Figure 5. The left chart of Figure 6 shows a majority of 58 % from the deep red zone around ZUGOG including two prisms directly below the gravimeter which contribute with 308 nm/s² reduced by 60 nm/s² masking out the area below the ZUGOG building. These effects in the close vicinity should be much smaller in reality compared to our very simple assumption of snow distribution, as the maximum snowpack at the summit is certainly less than the values from the LWD station due to usually strong winds at the summit ridge, 430 which are neglected up to this point. For the setup of more detailed snowpack description in the direct vicinity, a high resolution DTM with a grid spacing of 1 m x 1 m will be used in the future in combination with a detailed 3D surveying of the buildings at the summit (sections 2.1 and 3.3). The right chart of Figure 6 without the two prisms below the gravimeter shows a far more uniformly distributed contribution. This should be considered as target distribution for a thorough gravimetric evaluation of the spatiotemporal snowpack modelling especially in the RCZ in the future. 435 Based on this concept study, the description of the snowpack distribution will be refined in future studies for the entire Zugspitze region and the three catchments. The aim is to set up snowpack models such as SNOWPACK/Alpine3D (Lehning et al., 2006) or use cold region hydrological model frames like the Canadian Hydrological Model (CHM, Marsh et al., 2020) at this location, and, in addition, to statistically describe the main snowpack distribution via LiDAR measurements, similar as 440 presented in Grünewald et al. (2013). With these approaches, we will improve the descriptions of the spatial snowpack distribution in this very complex high-alpine terrain by including detailed descriptions of the snowpack itself, the effect of energy balance on the snowpack, potential wind redistributions as well as further meteorological and gravimetrical influences on the snow cover regarding elevation, aspect and slope -and of course by using the gravity residuals as boundary conditions.

Karst groundwater
Besides snow distribution and snow water equivalent, liquid water balance in the karstified RCZ is influencing the SG signal.
Throughout the year, a typical course of runoff with four characteristic periods can be observed (Figure 7). From the end of October to April no recharge of the karst system takes place and the Partnach spring is falling dry. With rising temperatures in April, melting processes are beginning in the lower parts of the catchment and first meltwater pulses can be observed at the 460 Partnach River gauge. Melting period in the upper part of RCZ starts later in May lasting until the beginning of July. The karst system of RCZ is mainly fed by meltwater and discharge at the Partnach spring is continuously high. Liquid precipitation leads to pronounced runoff peaks on top of the increased basal discharge level. During this period of time, runoff at the Partnach spring is a mixture of meltwater from areas with increasing elevations and liquid precipitation. After melting ends, long lasting rainfall and storm precipitation are dominating runoff characteristics with steep rising and falling limbs. The well-developed 465 karst system of RCZ with conduit flow causes these rapid runoff reactions of the Partnach spring. With lowering temperatures during autumn, snow accumulation starts in higher elevations of RCZ and recharge of karst groundwater is reduced because liquid precipitation is more seldom. Sometimes daily melting cycles of the glacier remains (Northern and Southern Schneeferner) can be observed during this usually dry period. At the end of autumn, low temperatures and snowfall in higher elevations are terminating recharge and karst groundwater head is falling step by step beneath the level of the Partnach spring. 470 For the hydrological interpretation of gravity signals from ZUGOG, it is crucial to quantify the water volume stored in the vadose karst zone of RCZ. Thereto pseudo-continuous depletion curves were constructed by splicing short falling hydrograph intervals together (e.g. Lamb and Beven, 1997). The recession constant "α" was calculated for several years, because recession behaviour of the Partnach spring varies from year to year due to unknown processes in the karst system (Figure 8). Based on 485 a mean recession constant "α", a water storage model for the vadose karst zone was developed by an addition of daily discharge volumes during the depletion period. As Figure 8 shows, storage volume in the vadose zone varies between 1.6 and 3.38 x 10 6 m³. Under the assumption of a homogeneous layer of a 6 km² wide groundwater body (Figure 1 corresponding gravity values between 12 and 24 nm/s², respectively. With an uncertainty of a few nm/s², the gravimetric 490 approach should be able to distinguish interannual groundwater height variations. Besides seasonal runoff and corresponding karst groundwater variations, rainfall events on time scales from hours to days produce significant peak-like signals not only in the runoff but also in the gravimetric time series. A homogeneous layer of 1 mm precipitation height on top of the digital terrain model applied (section 4.2) would result in a gravity increase of 0.9 nm/s². 495 This shows that the gravity variations can be used as reference for the estimation of the total sum of precipitation in this alpine terrain with large variability in precipitation instead of using point measurements with precipitation collectors. The same hydro-gravimetric approach might be applied to the estimation of daily evapotranspiration rates during dry days in late summer (August and September), when the seasonal runoff has mainly finished. However, maximum evapotranspiration rates do not exceed 2-3 mm/day inducing a gravity effect of 1.8 to 2.7 nm/s² at the ZUGOG gravimeter site which is at the limit of what 500 can be observed by the gravimeter.
In addition, glacier melting and permafrost degradation also contribute to groundwater and runoff with predominant climatedriven long-term signals but also significant interannual variations e.g. as a consequence of very dry and hot summers (Scandroglio et al., 2019). In addition, cavities inside Mount Zugspitze filled with water through permafrost degradation could 505 be problematic depending on the distance and direction to the gravimeter and their sizes. However, ZUGOG is located above the southern slope of Mount Zugspitze, while permafrost is mostly situated at the northern part. These additional signals will be best captured by the combination of absolute, superconducting and relative gravimetry and other geodetic techniques in a future hybrid approach. Thus, the whole summit area could be covered by a gravity network, constrained by the SG at ZUGOG, and observed regularly within episodic campaigns.

Summary and conclusions
The superconducting gravimeter OSG 052 has been introduced as a novel hydrological sensor within a worldwide unique installation at the summit of the high-alpine Mount Zugspitze (2962 m). For the first time, a high-quality and publicly available continuous gravity time series of the water storage variations for two years around Mount Zugspitze is provided for an 525 improved understanding of high-alpine catchment processes. As the discharge of the main three high-alpine catchments with their respective gauges at the Partnach spring, Bockhütte and Hammersbach (projected) are largely influenced by seasonal snow cover from autumn to spring, the spatiotemporal variations of the snowpack are the main contributor to the gravity residuals. Sensitivity analysis on the basis of a simplified assumption on snow distribution in this area reveal a snowgravimetric footprint of 3.5 km horizontal and 4 km slant distances around the gravimeter covering an area of 40 km². The 530 large range of gravity residuals up to 750 nm/s² corresponds to the maximum of 1957 mm snow water equivalent on 29th May 2019, measured at the LWD station located on the Zugspitzplatt. This result, together with the low uncertainty of the gravity residuals of a few nm/s², enables various detailed future hydro-gravimetric analysis for this high-alpine site at Mount Zugspitze.
As the snow masses from the Partnach spring catchment (RCZ) contribute to more than 2/3 of the total gravity signal, this catchment will be the main focus for future investigations. 535 During the mainly snow-free season in summer, other water storage components dominate the gravity residuals. Runoff and karst groundwater variations are driven not only by snowmelt and rain but also by glacier melting and permafrost degradation.
While the discharge of the Partnach spring and Bockhütte catchments are well observed by gauge stations, the estimation of catchment-wide total rainfall amounts and evapotranspiration rates might strongly benefit from including the gravity residuals 540 into the analysis. However, the hydro-gravimetric approach shows also a few limitations. The setup of a more detailed and small scale snowpack description, especially in the direct vicinity, is essential in order to increase the sensitivity towards the whole catchment. The sensitivity to the hydrological mass variations depends on the distances and directions from the gravimeter. For the separation 545 of individual water storage components, additional information from hydro-meteorological monitoring and modelling is always required. Nevertheless, the well-equipped high-alpine research catchment Zugspitze is a very good home base for upcoming long-term hydro-gravimetric studies.
The required steps within an upcoming hydro-gravimetric research project can be divided into monitoring, modelling and 550 prediction. The essential prerequisite is to ensure the continuous high-quality operation of the OSG 052 and all associated sensor systems at ZUGOG as well as the hydrological and meteorological sensors in the three relevant catchments with focus on the RCZ. For an enhanced gravimetric monitoring, additional absolute gravity measurements provide the SG drift and support long-term studies. The GNSS station nearby ZUGOG reveals the long-term vertical displacement of the site. It is further intended to install a continuously recording spring gravimeter in a vertical distance of 500 m below OSG 052 inside 555 Mount Zugspitze in a technical room next to the rails of the cogwheel train to quantify the ongoing mass redistributions inside the mountain. In addition, the integration of episodic relative gravity measurements both from the tunnel of Mount Zugspitze and from the RCZ would be highly beneficial in a future hybrid gravimetric approach in order to better capture the spatiotemporal gravity variations on catchment scales -and with this for a more thorough constraining of the hydrological model. 560 For an improved gravimetric modelling, the artificial snow-hydrological mass variations in the direct vicinity of the gravimeter have to be reduced by local monitoring and modelling. With regard to atmospheric gravity effects, the complex alpine topography surrounding ZUGOG has to be taken into account either by using a weather model with a higher spatial resolution or by setting up a local model based on an array of available barometers. Modelling both effects with an uncertainty of a few 565 nm/s² (1σ) leads to an improved signal separation with the gravity residuals being even more representative for the hydrology of the RCZ. As the snowpack signal dominates the gravity residuals, a large focus will be on high-resolution modelling of the spatiotemporal variations of the snowpack amount and distribution as well as the coupling with a hydrological model including the description of the karst zone. In addition to continuous point measurements and in order to allow spatial model validation, a LiDAR installation is planned to track continuously the spatial distribution of snow height in the upper part of the RCZ as 570 well as snow mapping using current remote sensing approaches.
The essential task is to develop an efficient physically based, spatially distributed karst-snow-hydrological model describing relevant physical processes in the RCZ. This will answer the question how robust the estimated gravimetric footprint is with regard to real snowpack distribution conditions. A "forward operator" for the gravimeter should be included which allows observed temporal gravity variations (as well as other model outputs such as runoff, snow pattern etc.) to be used for inverse modelling and conditioning the uncertainties of all predicted processes and states. This will illuminate to what extent the observed gravity variations are able to condition the prediction of snow-and karst-hydrological states and processes in the catchment. Different time scales have to be distinguished, as precipitation events (solid, liquid) and lateral snow distribution induce short-term signal alterations, while evapotranspiration, sublimation, snow melting processes and groundwater recession 580 take place on longer time scales. It will be a challenge to explore whether and to what extend these processes can be disentangled by combining gravimetric and other available information.
The overall aim of such projects is to develop the hydro-gravimetric technique for an improved understanding and a better predictability of alpine mass redistributions on catchment scales which can be transferred to other alpine research catchments 585 worldwide. Finally, an improved knowledge of hydrological variations on catchment scales enhances the resolution of largescale hydrological variations and reduces the spatial and temporal gap to the satellite mission GRACE-FO (Gravity Recovery and Climate Experiment -Follow On), launched in May 2018, which provides gravity variations with a spatial resolution of 300 x 300 km² and a temporal resolution of 1 month.

Data availability 590
Raw gravity and atmospheric pressure data from the OSG 052 at ZUGOG are published (Voigt et al. 2019) and available from the IGETS data base hosted by the Information System and Data Center at GFZ (URL: https://isdc.gfz-potsdam.de/igets-database/). The subsequent gravity residuals and all auxiliary data from ZUGOG can be provided upon request from the author.