A 20-year record (1998–2017) of permafrost, active layer and meteorological conditions at a high Arctic permafrost research site (Bayelva, Spitsbergen)

Most permafrost is located in the Arctic, where frozen organic carbon makes it an important component of the global climate system. Despite the fact that the Arctic climate changes more rapidly than the rest of the globe, observational data density in the region is low. Permafrost thaw and carbon release to the atmosphere are a positive feedback mechanism that can exacerbate global warming. This positive feedback functions via changing land–atmosphere energy and mass exchanges. There is thus a great need to understand links between the energy balance, which can vary rapidly over hourly to annual timescales, and permafrost, which changes slowly over long time periods. This understanding thus mandates long-term observational data sets. Such a data set is available from the Bayelva site at Ny-Ålesund, Svalbard, where meteorology, energy balance components and subsurface observations have been made for the last 20 years. Additional data include a high-resolution digital elevation model (DEM) that can be used together with the snow physical information for snowpack modeling and a panchromatic image. This paper presents the data set produced so far, explains instrumentation, calibration, processing and data quality control, as well as the sources for various resulting data sets. The resulting data set is unique in the Arctic and serves as a baseline for future studies. The mean permafrost temperature is −2.8 C, with a zero-amplitude depth at 5.5 m (2009–2017). Since the data provide observations of temporally variable parameters that mitigate energy fluxes between permafrost and atmosphere, such as snow depth and soil moisture content, they are suitable for use in integrating, calibrating and testing permafrost as a component in earth system models. Published by Copernicus Publications. 356 J. Boike et al.: A 20-year record (1998–2017) of Bayelva, Spitsbergen The presented data are available in the Supplement for this paper (time series) and through the PANGAEA and Zenodo data portals: time series (https://doi.org/10.1594/PANGAEA.880120, https://zenodo.org/record/ 1139714) and HRSC-AX data products (https://doi.org/10.1594/PANGAEA.884730, https://zenodo.org/record/ 1145373).


Introduction
Permafrost, which is defined as ground that has remained frozen continuously for 2 years or more, covers large parts of the land surface in the Northern Hemisphere, amounting to about 15 million km 2 (Brown et al., 1998).The range in temperatures and water and ice content of the upper surface layer of seasonally freezing and thawing ground (the active layer) determines the biological and hydrological processes that operate in these areas.Thermal degradation of permafrost over the last few decades has been reported from 10 available circum-Arctic boreholes (Romanovsky et al., 2010) with a recent update by Romanovsky et al. (2017).Warming and thawing of permafrost and an overall reduction in the area that it covers under future climate change scenarios have been predicted in all recent climate models, but at widely varying rates (Koven et al., 2012).Continued observations, not only of permafrost thermal data but also of the multiple other types of data required to understand the changes to permafrost, are therefore of great importance.The data required include information on the upper boundary condition of the soil (specifically on snow cover), on atmospheric conditions and on various subsurface state variables (e.g., volumetric liquid water content).The seasonal snow cover in Arctic permafrost regions insulates the permafrost surface for many months of the year and has an important effect on the thermal regime of permafrost (Langer et al., 2013;López-Moreno et al., 2016).The soil's water content determines not only its hydrologic and thermal properties, but also the amount of latent heat that is either required for the seasonal thaw in spring or produced during fall.In view of these dependencies, the data sets (including snow cover and the thermal state of the soil and permafrost), ideally at the same resolution as any meteorological input data, will be of great value for evaluating permafrost models (or land surface models intended for permafrost regions).In this paper we present data that incorporate subsurface components of heat and mass flux, properties of the snow cover and weather data from the Bayelva high Arctic permafrost site.
The Bayelva research site on the Spitsbergen island in the Svalbard archipelago (78.551 • N, 11.571 • E) has been used for research by the Alfred Wegener Institute (AWI) in collaboration with academic partners since 1998, with the original objective of developing and testing permafrost process models (Ippisch, 2001;Stern, 2017).Major developments in earth system models, e.g., through the European PAGE21 project (http://www.page21.org),Permafrost Carbon Net-work projects (http://www.permafrostcarbon.org), and satellite calibration and validation missions, have subsequently led to sustained interest in the data produced from a wider modeling community.This publication provides information on the site and a full documentation of the data set collected between 1998 and 2017 that may be required for input into earth system models (Chadburn et al., 2017;Ekici et al., 2014Ekici et al., , 2015)).

Site description
The Bayelva site is located in the European high Arctic, on western Spitsbergen in the Svalbard archipelago.It is situated about 3 km from the small village of Ny-Ålesund, which serves as a permanent hub for researchers and their logistical support.Ny-Ålesund was a coal mining area from 1916 to 1962, since then the village has been gradually transformed into a research community.Long-term meteorological monitoring was initiated in 1969 by the Norwegian Sverdrup Station but research topics now also include terrestrial ecosystems, glacier monitoring, detailed atmospheric research and the Kongsfjorden ocean system, coordinated by the Ny-Ålesund Science Managers Committee (http://nysmac.npolar.no/).
The West Spitsbergen Current, a branch of the North Atlantic, warms this area to an average air temperature of between −17.0 and −3.8 • C in January and from 4.6 to 6.9 • C in July (18-year period: 1 August 1993 to 31 July 2011; Maturilli et al., 2013).It also provides about 400 mm of precipitation annually, which falls mostly as snow between September and May.Significant air temperature rise has been detected since 1960, which has generally been attributed to changes in the radiation budget and atmospheric circulation patterns (Førland et al., 2011;Hanssen-Bauer and Førland, 1998).This rise is also reflected in the permafrost temperatures, as recorded from deep boreholes in the mountains at Janssonhaugen (Isaksen et al., 2007a(Isaksen et al., , b, 2001)).For the 6-year period from 1999 to 2005 the borehole temperatures have increased: at 25 m by 0.26 • C, at 30 m by 0.19 • C and at 40 m by between 0.07 and 0.09 • C (Romanovsky et al., 2017).Continuous permafrost underlies the unglaciated coastal areas to a depth of about 100 m and the active layer thickness (ALT) at the end of summer ranges between 1 and 2 m (Humlum, 2005).ALT data are available from the Global Terrestrial Network for Permafrost (GTNP) database (http://gtnpdatabase.org/) for several locations on Svalbard but for different time periods and distances between 25 and 115 km away from Bayelva.The site with the longest record, Janssonhaugen, is located 115 km to the southeast and had an average thaw depth of 175 cm for 1998-2013.The region has experienced increases in cloudiness, precipitation, and the number and intensity of cyclones in recent years, especially during the winter months (Hanssen-Bauer and Førland, 1998;Sepp and Jaagus, 2011).The increase in cloudiness (Maturilli and Kayser, 2016) has led to an increase in incoming long-wavelength radiation, resulting in a major change to the winter radiation budget for this region, as measured at the German-French (AWIPEV) research station, which is located within the village of Ny-Ålesund (Maturilli et al., 2014).This research station carries out long-term monitoring of radiation (Baseline Surface Radiation Network, BSRN; http://bsrn.awi.de/) and basic meteorological data (http://www.awipev.eu/).Additional meteorological data (e.g., air temperature at 10 m, wind speed and direction at 2 and 10 m) are available in Maturilli et al. (2013).
The data presented herein were collected from the high Arctic Bayelva River catchment area (Fig. 1), about 3 km away from Ny-Ålesund.A legacy from past mining activities is the physical disturbance of the ground in and around Ny-Ålesund (e.g., compaction and reworking of the soil).Traffic in the village (people, cars, snow mobiles) also affects the surface conditions, especially in winter.Thus, we chose a location that was sufficiently away to be unaffected by these disturbances.The Bayelva catchment area lies between two mountains (Zeppelinfjellet and Scheteligfjellet), with the glacial Bayelva River originating from the two branches of the Brøggerbreen glacier.The terrain flattens out to the north of the Bayelva site and the Bayelva River flows into the Kongsfjorden fjord and the Arctic Ocean about 1 km from the site (Fig. 1).Over the past three decades the Bayelva catchment area has been the focus of intensive investigations into fluvial hydrology (sediment transfer and geochemistry; Hodson et al., 2002), soil and permafrost conditions (Boike et al., 2008b;Roth and Boike, 2001;Westermann, 2010;Westermann et al., 2011b), the surface energy balance (Boike et al., 2003b;Westermann et al., 2009), and the micrometeorological processes controlling the surface gas and energy exchanges (Lloyd, 2001b;Lloyd et al., 2001;Lüers et al., 2014).A permafrost maximum, minimum and average temperature with depth (trumpet curve) was shown for data for 1 year (August 2009-August 2010) by Boike et al. (2012).The data from this Bayelva site have also been used in earth system modeling (Ekici et al., 2014(Ekici et al., , 2015)).Nearby investiga-tions by Japanese and Italian researchers include vegetation analysis with respect to periglacial and glacial landforms and topography (Cannone et al., 2004;Ohtsuka et al., 2006), as well as investigations into the plant-and plot-scale dependence of CO 2 emissions on biotic and abiotic factors at the start and end of the growing season (Cannone et al., 2016;Uchida et al., 2006).
The Bayelva site is located on top of the Leirhaugen hill (25 m a.s.l.), on permafrost patterned ground (Fig. 2).The hill consists mainly of rock but is partly covered by a mixture of sediments that consist of glacial till, together with fine grained glaciofluvial sediments and clays from the last glacial advance, likely deposited by the Kongsfjorden glacier (J.Tolgensbakk, personal communication, 2007).According to the geological map, the site is part of the Brøggerbreen formation (lithography sandstone, shale, conglomerate), but the gray color of the sediments suggests that the material was deposited by the Kongsfjorden glacier and not the adjacent Brøggerbreen glacier, which deposits redder material.Thus, the site is part of the Kongsfjorden Formation, with the main lithography sandstone, shale and coal (Hjelle et al., 1999).Since the hill has a maximum altitude of about 38 m, which is below the upper marine limit, these sediments may also include marine deposits.
The vegetation cover in the vicinity of the Bayelva site (i.e., the area of the hill) has been estimated to be approximately 50-60 %, with the remainder being bare soil with a small proportion of stones (cobbles and gravel) (Lloyd et al., 2001).Within the rest of the Bayelva River catchment area, sparse vegetation alternates with bare soil and sand, or with rock fields.
The area of the Bayelva site is covered with non-sorted circles, also known as mudboils (Fig. 2), which are only present on the hill itself, with other patterned ground phenomena (such as sorted circles and stripes) occurring on and in the area surrounding Leirhaugen hill.These non-sorted circles formed under localized favorable conditions following the last glacial period.The bare soil centers of these circles are about 1 m in diameter, surrounded by a vegetated rim that consists of a mixture of low vascular plants (including various species of grass and sedge such as Carex spp., Deschampsia spp., Eriophorum spp., Festuca spp.and Luzula spp.), catchfly, saxifrage, willow, various other locally common species (Dryas octopetala, Oxyria digyna, Polygonum viviparum), and some unclassified species of moss and lichen (Ohtsuka et al., 2006;Uchida et al., 2006).
The soils on the hill generally range from silty loam to silty clay with a few large stones.The silt content decreases from more than 50 % at the top of the soil profile to less than 30 % at the bottom of the soil profile at about 1 m depth, while the clay content increases to more than 50 %.Concentrations of total organic carbon, total nitrogen and total sulphur are highest at the bottom of the soil profile, peaking below the centers of the non-sorted circles.The organic carbon concentration is high (> 6 % weight) at the bottom of the soil profile.Nitro-gen is elevated beneath the vegetated rim of the non-sorted circles (Boike et al., 2008a).
The Bayelva site can be temporarily shaded from direct solar radiation by the mountains on either side.Snow is present for up to 9 months of the year and strong winds can encourage drifting and snow redistribution.The snow depth varies within the Bayelva catchment area, as do its physical characteristics, such as the snow water equivalent and the internal stratigraphy and density of the snowpack.A 15 m × 10 m fence surrounds the Bayelva site to prevent reindeer from damaging the equipment, and a small wooden container (about 0.90 m × 1.40 m, and 2.3 m high) houses the data acquisition and transfer system (Fig. 2).Warm weather events during the winter can interrupt the build-up of snow cover resulting in the formation of either internal ice lenses or basal ice, depending on the heat content of the snowpack.Table 1 summarizes the characteristics of the site retrieved from both previous publications and data included in this paper.

Data description
The Bayelva site was powered by a solar panel and wind generator during the period from 1998 to 2013 and data were retrieved manually through site visits every 4 to 6 weeks, when visual inspections were made of the sensors.
Since 2013 the site has been connected to the main electricity supply and the Ny-Ålesund data network, and data have since then been transferred automatically resulting in much improved data collection with almost no data gaps.The data gaps prior to 2013 resulted mainly from problems with the energy supply for the site, such as problems with the solar and wind charge regulator.The current online data transfer and real time data visualization enables real time data checking and identification of possible sensor failure, as well as remote programming.Major problems that have affected the continuous data collection or the data quality include reindeer disturbance (e.g., reindeer pulling cables out of the ground or destroying sensors by rubbing against the stations, as occurred in 2003) and the gradual uplift of sensors as a result of freeze-thaw processes, which affected soil temperature measurements.
Details of the sensors are provided in the following sections, as well as descriptions of the data quality and cleaning routine (Sect.4).The instruments are divided into aboveground (meteorological) and below-ground sensors (e.g., soil sensors).Further detailed information on the sensors can be found in Table 2, which summarizes all of the parameters and instruments (Table 2), as well as in Appendices C and D (metadata and description of instruments).Figure 3 presents time series of parameters between 1998 and 2017.The average total phytomass in this subzone is lower than 3 t ha −1 and the net annual production is lower than 0.3 t ha −1 .

Weather station data
Meteorological data have been recorded at the AWIPEV research base in Ny-Ålesund since 1992, including radiation component data for the worldwide Baseline Surface Radiation Network.The AWIPEV meteorological data are described by Maturilli et al. (2013) and can be downloaded from PANGAEA (http://bsrn.awi.de/data/data-retrieval-via-pangaea/).Climate records covering a longer period of time are available from Svalbard since 1898, with the establishment of a permanent weather station in 1911 (Førland et al., 2011).
Since the Bayelva site is located close to the leading edge of the Brøggerbreen glacier, a different local micro-climate can be expected from Ny-Ålesund.A climate station was therefore installed in 1998, within the fenced area.The temporal resolution of the meteorological measurements was increased in October 2009 from 60 to 30 min.

Air temperature, relative humidity and snow temperature
Air temperature and relative humidity (hourly averages, since 2009 half-hourly averages) are measured at a height of 2 m above the ground surface using PT100 platinum resistance temperature sensors and Rotronic MP103A and Vaisala HMP45 capacity sensors, protected by unventilated shields.The heights of the shields are not adjusted during periods of snow cover accumulation or ablation.The uncertainty in the temperature measurements ranges between 0.03 and 0.5 • C, depending on the sensors used; the uncertainty in the relative humidity measurements ranges between 2 and 3 %.The PT100 temperature sensors were calibrated in ice-slush water for at least 12 h to determine the absolute accuracy at 0 • C. The precision of the temperature sensors was calculated from the variation in values recorded during the calibration period, at constant temperatures in ice-slush water, and can be found in Table 2.The change of sensors for relative humidity in 2009 resulted in a 7 percentage points lower average relative humidity between 2009 and 2017 as compared to the period before, independent of the season.PT100 temperatures sensors were installed at the weather station in 1998, close to the ground surface.When covered by snow these sensors record the snow temperature, while during the snow-free period they record (unshielded) air temperatures.The temperature cables were originally attached to a small (vertical) bamboo stick, with sensors sticking out a few centimeters from the stick (Appendix C, Fig. C5).Visual inspections during late spring (May 2016) revealed that melting occurred around the bamboo stick, leaving an air gap around the stick.The temperature sensors were nevertheless well away from the stick (in a horizontal direction) and remained embedded in the snow.The height of the sensors was changed on a number of occasions following damage by rein-deer.Sensors were originally mounted at heights of 0.35 and 0.48 m to record snow temperatures, but these heights were changed during subsequent years (Table 2).Information concerning the changes in sensor height is included in the data description that accompanies the data (Appendix C, Fig. C5).

Wind speed and direction
The wind speed and direction were measured with an R.M.

Radiation
Net radiation was measured between 1998 and 2009 using a CSI (Campbell Scientific Ltd.) Q7 net radiometer or an NR Lite net radiometer (Kipp & Zonen) and since 2009 with an NR01 four-component sensor (Hukseflux).Changes in the sensor setup were partly due to destruction by reindeer which prompted an improvement of the fence in 2017.A wind correction was performed according to the manufacturer's suggestion for the Q7 and NR Lite net radiometers (Campbell Scientific Ltd.; details in Appendices C and D).An SP 1110 (Skye Instruments) pyranometer was also installed in 1998 and a CG1 (Kipp and Zonen, Appendix C3) downwardlooking pyrgeometer (to detect reflected longwave radiation) in 2003.Regular site visits were made during which the pyranometers were checked for condensation, dirt, physical damage or hoar frost.Nevertheless, because the instruments were largely unattended most of the time, we estimate the field accuracy to have been between ±10 % (for the Q7 sensor) and ±20 % (for the NR Lite).Details on the various net radiometer setups and correction are provided in Appendix D.

Rainfall
An unheated and unshielded tipping bucket rain gauge (R M Young, 52203) was installed on a pole in August 1999, with the top of the bucket about 1.6 m above ground level.The instrument measures only liquid precipitation (rainfall) and not winter snowfall.The tipping bucket was regularly checked in summer by pouring a known volume of water into the bucket and by frequent visual inspections for dirt or snow during each site visit.At wind speeds above 12.5 m s −1 (hourly or half-hourly average) the rain gauge was found to record precipitation even when there was none, due to vigorous shaking of the pole; these data were flagged accordingly within the data series.Until 2010, precipitation was only recorded as hourly totals, but the provided data set uses 30 min temporal resolution (half-hours are presented as "NaN", since the hourly total is given at full hour).Total daily precipitation figures from a gauge in Ny-Ålesund are available through the Norwegian Meteorological Institute (formerly www.eklim.noand www.met.en).

Snow depth
The snow depth around the station has been continuously monitored since 1998 with an SR50 sonic ranging sensor (Campbell Scientific Ltd.) and with an additional SR50 sensor at a nearby (about 85 m away; Fig. 2) eddy covariance station since 2006.The sensor recorded raw distance data from the sensor to the object (in this case the ground or snow surface).The data obtained from the sonic sensor were corrected using the speed of sound at 0 • C and the air temperature measured at the Bayelva site, according to the formula provided by the manufacturer (Campbell Scientific Ltd.; Appendix D, Sect.D4).To obtain the snow depth the distance of the sensor from the ground surface was recorded annually and subtracted from the corrected distance data.Snow depths are provided in the data supplement for this paper.
Due to seasonal thawing of the ground, the surface can subside over the season (Overduin and Kane, 2006); thus, negative distance rates of a few centimeters are computed (and not set to zero).The data are not removed from the series since the opposite process, frost heave, takes place during fall freeze-back and thus the surface presumably bounces back to the original height (prior to ground thaw and subsidence).Since the ground surface has little or no vegetation cover the signal is returned with a high degree of accuracy (Table 2).An additional snow depth laser distance sensor (Jenoptik SHM30) was installed at the Bayelva site climate tower in August 2013.The data from the three different snow depth sensors were combined to obtain a continuous record and fill any gaps left by the failure of individual instruments (level 2 data containing all original snow data from the three sites merged into one data product can be found in Sect.4).Any differences in snow depth between these sensors may therefore be due to microtopographic and microclimatic variations between these two locations.

Time lapse photography of snow cover
In order to monitor the timing and pattern of snow cover changes an automated camera system (K1; Umwelt-und Ingenieurtechnik, Dresden) was set up in August 2013 to photograph the land surface in those areas in which the instruments were located (Fig. C2).The camera recorded one image every hour; during the polar night a light was switched on automatically while each image was being recorded.

Dielectric number of snow
The dielectric number (also referred to as relative permittivity) of the snow was measured using a vertically installed time-domain reflectometry (TDR) probe.The length of the TDR probe was 0.5 m between 1998 and 2007 and 0.33 m from 2009 onwards.Dielectric numbers are provided in the data set for both the snow-covered and the snow-free periods (recording only air dielectric numbers).When the snow cover depth exceeds the length of the TDR probe the dielectric number data can be used to infer dry snow densities using empirical relationships (Schneebeli et al., 1998).In wet snow, the measured snow dielectric number can be used for the calculation of the snow liquid water content using the composite approach with ice added as an additional phase (e.g., Roth et al., 1990) and measured snow densities (Appendix E, Table E1).

Vertical profiling of the physical properties of snow
Snow depth and vertical profiles for temperature, density and dielectric numbers were collected from the site and surrounding areas during a number of end-of-snow (spring) seasons.These profiles, which were collected at irregular intervals between 2000 and 2016, provide additional information on the snowpack.They typically reveal the presence of a number of ice layers (including a basal ice layer) and snow densities of between 220 and 440 kg m −3 .The typical end-of-spring average snow density before snow melts is 350 kg m −3 .The end-of-spring season snow water equivalent varies between the years due to variations in snow depth (Fig. 3).Details of these profiles, including some stratigraphic information, are provided in Appendix E (Table E1).

Subsurface data on permafrost and the active layer
Data have been collected from a variety of installations that were changed over time because of the deterioration of sensors or disturbance by reindeer.As a result, there are a number of data series from different profiles rather than one continuous data series.Some measurements were also discontinued (e.g., soil heat flux).

Instrument installation and soil sampling
In order to take into account any possible effects of small variations in vegetation and microtopography at the site (e.g., due to the presence of non-sorted circles), instruments were installed in a number of different positions within a profile.Deterioration of data quality and the failure of some sensors also led to the installation of new sensors in new profiles.

Instrument installation and soil sampling 1998
The first set of instruments was installed in 1998, within one of the non-sorted circles (Appendix C, Fig. C7), for which a trench 2 m wide and 1.4 m deep was excavated across the circular feature (Appendix C, Fig. C8).The surface was carefully cut and the excavated soil stockpiled separately according to depth and soil horizon, in order to be able to restore the original profile following installation of the instruments.2. This plot (divided in three parts for better visibility) is also provided in Appendix H.
www.earth-syst-sci-data.net/10/355/2018/ Earth Syst.Sci.Data, 10, 355-390, 2018 J. Boike et al.: A 20-year record (1998Boike et al.: A 20-year record ( -2017) ) of Bayelva, Spitsbergen entire profile was quite heterogeneous but consisted mainly of silty clay, with a stone content of less than 10 %.The stone content increased to about 50 % in a slightly inclined layer approximately 20 cm thick that occurred at a depth of 50 cm; the stones were quite large (including cobbles in excess of 6 cm in diameter) but showed no preferred orientation.Below 85 cm depth the soil was coal-rich and contained massive coal lenses.Sensors were installed in the undisturbed wall of the soil pit (Appendix C, Fig. C8b) to obtain an instrumented cross section of the pedon.The pit was then backfilled.Soil samples were collected during the installation so that physical parameters could be analyzed.The stratigraphic and textural information from the soil profile, including soil organic carbon (C), nitrogen (N) and sulfur (S) concentrations, can be found in Appendix F.

Soil sampling 2007
A number of additional soil profiles in the vicinity of the Bayelva site were sampled in 2007 (Fig. 2).Data for the soil texture; bulk density; and carbon, nitrogen and sulphur (C, N, S) content of these profiles is also included in Appendix F (Table F1).

Instrument installation and soil sampling 2009
A new set of measurement profiles was established within a second non-sorted circle close to the 1998 profiles (Appendix C, Fig. C9).The site was chosen for its similarity to the 1998 site, in order to provide an analogous continuation of the long-term subsurface data from the 1998 profiles.Data on the soil texture; its bulk density; and carbon, nitrogen and sulphur content can be found in Appendix F.
All profiles show cryoturbation features at the surface through patterned ground (non-sorted circles) and also within the soil profile, e.g., roots that are transported downwards through cryoturbation (Figs.C8 and C9).

Soil profile sensor installation 1998
Pairs of soil temperature and soil water content sensors were installed over a cross section in 1998; their positions are shown in Appendix C (Fig. C8).The probes automatically recorded temperatures from that time until January 2012.The temperatures were measured using CSI (Campbell Scientific Ltd.) thermistors connected to a CR10X data logger with an AM416 multiplexer.The voltage measured for each sensor was stored in a database.Temperatures were calculated from the raw data using the Steinhart-Hart equation (Steinhart and Hart, 1968) and sensor calibration at 0 • C (Table D2).The average accuracy determined through calibration was 0.01 • C with a precision better than 0.002 • C.However, the max deviation at 0 • C was 0.1 • C.After installa-tion, the sensors cannot be recalibrated, certainly not without a lasting physical disturbance of the profile.However, temperatures during spring thaw and fall refreezing are stable at the phase change temperature (the zero-curtain effect in permafrost soils).Assuming that freezing point depression (i.e., soil type and soil water composition) does not change significantly from year to year, these periods can be used to evaluate sensor stability.During the first 2 years (1998)(1999) the temperatures during these periods were stable to within 0.1 • C. Subsequently, temperature readings from 9 of the 32 sensors started to drift and showed a positive offset relative to the ∼ 0 • C zero-curtain period.The temperature sensor at, for example, 19 cm depth showed correct temperatures during the fall freeze-back of 1998 and 1999, but in the fall of 2000 the measured soil temperatures shifted to positive values by 0.5 • C (offset from the fall zero-curtain).This offset increased further to about 6 • C in 2009.All sensors with temperature offsets of 0.5 • C or more were flagged accordingly within the data set.The positive temperature offset was most likely due to additional resistance, either in the cables due to leakage currents or in the multiplex due to corrosion.In addition to the temperature offset, the data quality of the years 2002 to 2007 was strongly affected by irregular spikes caused by the multiplexer.In the worst year, 2002, these spikes occurred for 5 to 7 % of all measured values in all soil temperature series of the 2-D profile.Spikes were flagged automatically and manually within the data set.The external charge controller regulating the solar and wind energy supply was changed in 2002 following a power failure, resulting in increased noise levels in the measured temperatures between 2002 and 2012 and the subsequent discontinuation of this data series.

Soil profile sensor installation 1999
Temperature sensors were installed in 1999 in one additional vertical profile, 2 m apart, within the fenced-off area, below vegetation (Appendix C, Fig. C6b).A small-diameter (vertical) hole was drilled by a hand drill into the ground in the fall of 1999.The PT100 temperature sensors were calibrated in ice-slush water for at least 12 h to determine the offset at 0 • C. The max deviation at 0 • C was 0.1 • C. Similar to the 107 probes, the average accuracy determined through calibration was 0.01 • C with a precision better than 0.002 • C. Wooden sticks to which sensors had been attached were then inserted into the holes down to 1.50 m into the permafrost beneath the active layer in such a way that the sensors formed a tight fit in the hole and no air was able to pass.The temperature sensors were attached at regular intervals to bamboo sticks.The upper sensors were affected on 11 September 2003 by reindeer disturbance.Furthermore, the upper two sensors moved upwards by about 10 cm between 2003 and 2016 due to frost heave and they were removed from the data series.

Soil profile sensor installation 2009
Pairs of soil temperature (CS107) and water content (TDR) probes were installed during August 2009, close to the first profile that was installed in 1998 (Appendix C, Fig. C9).Soil temperatures were measured at 11 depths in a profile from the surface down to a depth of 1.4 m.The data quality from the 2009 system improved considerably compared to the 1998 system, probably due to the data logging and multiplexer system used (CR10X from 1998 to 2011; CR1000 since 2009).Data collection is ongoing.

Soil dielectric number, volumetric liquid water content and bulk electrical conductivity
Soil profile sensor installations 1998 and 2009 Time-domain reflectometry probes were installed horizontally adjacent to the temperature probes installed in soil profiles in 1998 and 2009 (see above and Appendix C, Table C5).These TDR probes automatically recorded hourly measurements of bulk electrical conductivity (BEC) and the dielectric number (also referred to as relative permittivity), obtained by measuring the amplitude at very long times and the L a /L ratio (ratio of the apparent to real probe length, corresponding to the square root of the dielectric number).From 1998 to 2008 the measurements were obtained using a Tektronix 1502B cable tester connected to a CR10X data logger and custom-made triple-wire 24 cm probes.Since 2009 a CSI TDR100 reflectometer has been used, connected to a CR1000 data logger and 30 cm TDR CS605 probes (Campbell Scientific Ltd.).All TDR probes were calibrated for probe offset following the method described in Heimovaara and de Water (1993) and the CSI TDR 100 manual (http: //www.campbellsci.de/tdr100).The dielectric number data and computed volumetric liquid water in frozen and unfrozen soil are provided with the data set.The calculation for volumetric liquid water content takes into account four phases of the soil medium (air, water, ice, mineral) and uses the mixing model from Roth et al. (1990;Appendix D, Sect. D1).
Time-domain reflectometry can also be used to measure the impedance Z ( ) of the bulk soil, which is related to the soil's bulk electrical conductivity.These data were used to infer the electrical conductivity of soil water and solute transport in the non-sorted circle over a period of 1 year (Boike et al., 2008a).The impedance can be determined from the attenuation of the electromagnetic wave traveling along the TDR probe after all multiple reflections have ceased and the signal stabilized.The bulk electronic conductivities were recorded hourly using the TDR setup described in this section.The equipment installed in 1998 was calibrated in pure water, in air and in NaCl solutions with various known conductivities and is corrected for temperature (25 • C), following the method by Heimovaara et al. (1995).The TDR equipment installed in 2009 was not calibrated for BEC and a probe con-stant of 1 was used for BEC waveform retrieval; Campbell Scientific Ltd. suggests a probe constant (K p ) for the CS605 probes of 1.74.The lack of calibration explains the differences of the data sets recorded between the time 1998-2009 and 2009 to present (Fig. 3i).Measurements of electric conductivity and the dielectric number were affected by irregular spikes and possibly by sensor drift similarly to the soil temperature measurements.Spikes were flagged automatically and manually, while drifts could not be flagged due to the lack of reference data.Overall, the quality of soil data instrumented in 1998 is of reduced quality since July 2005.Values for dielectric number, computed volumetric liquid water content and bulk soil electrical conductivity can be found in the data set.

Soil heat flux
Two heat flux plates (Hukseflux HFP01), one installed at 0.18 m depth below the unvegetated center of the non-sorted circle and the second at 0.24 m depth below the vegetated rim of the non-sorted circle, recorded ground heat flux between September 1998 and September 2009 (Fig. 3).The manufacturer's calibration values were used to record heat flux in W m −2 .

Permafrost and active layer temperature
A 9.3 m borehole was drilled using a rotary drill rig on 30 March 2009 and cased with PVC.No stratigraphic information or drill core was recovered.A thermistor string was installed, with one sensor above the ground surface and nine from the surface down to 9 m depth (at 0, 0.5, 1.0, 1.5, 2.5, 3.5, 5.5, 7.5 and 9 m below ground surface, Appendix C, Fig. C10).The casing was left open (not refilled) so that the thermistor chain could be retrieved or replaced.Because of instrument failure, thermometers were retrieved and replaced several times.From August 2009, a Geoprecision M-Log5W data logger and sensor chain were used.Temperatures were recorded at hourly intervals, with no averaging; no data was recorded between 30 January and 29 March 2011, due to a low battery voltage.Geoprecision claims an accuracy of ±0.05 • C (at 0 • C) and a resolution of 0.01 • C, which is suitable for measurements in the range from −50 to +120 • C.However, comparison measurements using a PT100 thermometer at the same depths in the borehole showed a deviation of up to 0.2 • C. In 2014, a wooden shield was installed over the casing to prevent warming due to radiation and to facilitate natural ventilation (Appendix C, Fig. C10).Caution is warranted when using the borehole temperatures in the uppermost 1-2 m, since they may be affected by air movement in the borehole.These data are flagged in the data series.

Data quality control and availability
The primary purpose of quality control of observational data is to detect missing data, errors in the data, and possibly to correct errors in order to ensure the highest possible standard of accuracy and optimal use of the data by the broadest range of possible users.We differentiate between different data levels, from online data (http://www.awi.de/im-fokus/permafrost/direkter-draht-in-den-permafrost.html) to level 2 data (Table 3).Data supplied with this paper are level 1 and level 2 data.Level 1 data have undergone an extensive quality-control and are flagged with regards to maintenance periods, physical plausibility, spike and constant value detection and sensor drifts (Table 3).Level 2 data are compiled for special purposes such as a combination of data series from multiple sensors and gap-filling of data (documentation of source data is supplied in the PANGAEA data archive).Examples in this paper of level 2 data are the snow cover, soil temperature and soil volumetric water content data series that have been combined from various stations into a single data series, in order to obtain a long-term picture.There are eight quality control types (Table 3).These flags include informa-tion on the system's maintenance; no data availability; system error; and on consistency checks based on physical limits, gradients and plausibility.Due to the failure of some sensors that cannot be retrieved for repair or recalibration (e.g., sensors installed in the ground), the initial accuracy and precision of the sensors is not maintained.
In the case of soil temperature, the accuracy can be estimated by analysis of temperatures relative to the fall zerocurtain effect assuming soil water composition is similar from year to year.As an example, our temperature data have been checked through the fall zero-curtain effect and thus provide information on any reduction in accuracy (see Table 3, Flag 7).These checks are mandatory when small warming trends are to be estimated and interpreted.The selection of flagged data depends on its application and the required accuracies.
The data from this Bayelva site have been widely used for the development and evaluation of land surface models (e.g., Chadburn et al., 2017;Ekici et al., 2014Ekici et al., , 2015)).The newly collated data set will allow multi-year model runs, with improved quality control and checks.The data set presented herein is freely available, either as a download from PANGAEA (https://doi.org/10.1594/PANGAEA.880120) or from the Supplement of this paper.The data are provided as ASCII files.Other available data sets include a flux data set in the European Fluxnet database (Lüers and Boike, 2013) and aerial imagery of the snow cover (Westermann et al., 2015).Volumetric soil liquid water content and snow cover data are transferred in real time to the NASA SMAP mission (https://smap.jpl.nasa.gov/)and are uploaded to the PAN-GAEA archives on an annual basis.

Outlook
Permafrost around the Arctic is warming and thawing (Christiansen et al., 2010;Romanovsky et al., 2008), and this is true for Svalbard (Isaksen et al., 2007b) and the Bayelva site as well.The record from 2009 to 2017 of permafrost temperature measurements shows a zero-amplitude depth of 5.5 m at a mean temperature of −2.8 • C. Mean annual, summer and winter soil temperature data at all depths have been rising over the period of record (Fig. 4, three depths are shown) as well as in the deeper permafrost (Fig. 3h).Interannual to subdecadal variability is evident in the data and results mostly from differences during the winter months.Future analysis is required to detangle the relationship between potential meteorological drivers and permafrost degradation at this site.
The data set described and distributed in this paper provides a basis for analyzing this relationship at one site and a means of calibrating earth system modeling efforts over a long observational period.Developing predictive capacity for permafrost warming will be key to understanding the role of the permafrost feedback in the global climate system.

Appendix B: Description of HRSC-AX images
HRSC is a multisensor pushbroom instrument with nine charge-coupled device (CCD) line sensors mounted in parallel (Fig. B1) that has been in orbit around Mars since January 2004 on ESA's Mars Express spacecraft (Gwinner et al., 2016).It simultaneously obtains high-resolution stereo, multicolor and multiphase images.Digital photogrammetric techniques are used to reconstruct the topography on the basis of five stereo channels, which provide five different views of the ground.The four color channels (blue, green, red and nearinfrared; Fig. B2) are used to make true orthophotos in color and false color.The particular value of HRSC is the stereo capability, which allows the systematic production of highresolution digital elevation models (DEMs) with grid sizes between 50 and 100 m (Wewel et al., 2000;Scholten and Gwinner, 2004;Scholten et al., 2005;Gwinner et al., 2005Gwinner et al., , 2010)).
Since 1997 different airborne versions of HRSC have been developed, one of which (HRSC-AX) was used to acquire stereo color images over Svalbard.The principles of HRSC-AX data processing are described by Gwinner et al. (2006).Data on the camera orientation are reconstructed from a global positioning system inertial navigation system (GPS INS).HRSC-AX has been used in diverse technical and scientific applications (e.g., Gwinner et al., 1999Gwinner et al., , 2000)).The aerial survey covering the Brøgger peninsula took place on 17 July 2008 at around noon, acquiring data over most of the northern part of Brøggerhalvøya (Fig. B3).A Dornier Do 228 aircraft from the German Aerospace Center (DLR) was used for the survey, flying at an altitude of ∼ 2800 m.A comprehensive description of the initial results is given by Hauber et al. (2011a, b).
Data processing from the raw images to the final data products, including digital photogrammetric processing, was performed with the VICAR (Video Image Communication and Retrieval; http://www.mipl.jpl.nasa.gov/external/vicar.html) software developed at the JPL (Jet Propulsion Laboratory, Pasadena, USA) and the DLR.We provide HRSC-AX data in the form of a digital elevation model and as individual channels (panchromatic nadir channel; red, green, and blue color channels; and CIR, color-infrared, false-color channels).The CIR channels are computed by merging a false-color image (an RGB image where R, G and B correspond to the original infrared, red and green channels, respectively) with the nadir channel.Table B1 lists important key properties of the individual HRSC-AX image files.Metadata for HRSC-AX data are contained in "image labels", which we provide as two XML files (one for the panchromatic and (false) color images, and one for the DEM).The label entries consist of keyword-value pairs; essential keywords are defined below in Table B2.The elevations recorded in the DEM are ellipsoid heights; i.e., they are not computed with respect to a geoid but to a mathematically defined reference surface, which is a rotational ellipsoid with the equatorial A and B axes both having a radius of 6378.14 km and the polar C axis having a radius of 6356.75 km.This results in an offset of about 36.5 m with respect to geoid heights; i.e., sea level in the HRSC-AX DEM is not at 0 m, but at ∼ 36.5 m.The A_AXIS_RADIUS element provides the value of the semimajor axis of the ellipsoid that defines the approximate shape of a target body."A" is usually in the equatorial plane.
km real MAP

B_AXIS_RADIUS
The B_AXIS_RADIUS element provides the value of the intermediate axis of the ellipsoid that defines the approximate shape of a target body."B" is usually in the equatorial plane.
km real MAP

C_AXIS_RADIUS
The C_AXIS_RADIUS element provides the value of the C_AXIS of a solar system body.For triaxial ellipsoidal objects, the C_AXIS is the semiminor axis of the ellipsoid which defines the approximate shape of the body.
km real MAP

BODY_LONG_AXIS_ LONGITUDE
The BODY_LONG_AXIS_LONGITUDE element represents the offset between the longest axis of the triaxial ellipsoid used to model a body and the prime meridian of the body.Its value is the sum of the offset added to the prime meridian.This term is the positive west longitude, measured from the prime meridian.
deg real MAP

CARTESIAN_ AZIMUTH
The CARTESIAN_AZIMUTH element provides the clockwise rotation, in degrees, of the line and sample coordinates with respect to the center of the pixel at the map projection origin (i.e., where LINE_PROJECTION_OFFSET and SAM-PLE_PROJECTION_OFFSET are measured).
deg real MAP

CENTER_LATITUDE
The CENTER_LATITUDE element provides a reference latitude for certain map projections.In many projections, the CEN-TER_LATITUDE along with the CENTER_LONGITUDE defines the point or tangency between the sphere of the planet and the plane of the projection.string MAP

LINE_PROJECTION_ OFFSET
The LINE_PROJECTION_OFFSET element provides the line offset value of the map projection origin position from the center of the pixel at line and sample position 1, 1 (line and sample 1, 1 is considered the upper left corner of the digital array).

SAMPLE_ PROJECTION_ OFFSET
The SAMPLE_PROJECTION_OFFSET element provides the sample offset value of the map projection origin position from the center of the pixel line and sample 1, 1 (line and sample 1, 1 is considered the upper left corner of the digital array).

MAP_SCALE
The MAP_SCALE element identifies the scale of a given map.
The scale is defined as the ratio of the actual distance between two points on the surface of the target body to the distance between the corresponding points on the map.The MAP_SCALE references the scale of a map at a certain reference point or line.
Certain map projections vary in scale throughout the map.In general, the MAP_SCALE usually refers to the scale of the map at the CENTER_LATITUDE and CENTER_LONGITUDE.
The POSITIVE_LONGITUDE_DIRECTION element identifies the direction of longitude (e.g., east, west) for a planet.The IAU definition for direction of positive longitude is adopted.F1.Soil data for texture and C, N and S are listed in Table F1.The apparent dielectric numbers were converted into liquid water content (θ l ) using the semi-empirical mixing model in Roth et al. (1990).Frozen soil is treated as a four-phase porous medium composed of a solid (soil) matrix and interconnected pore spaces filled with water, ice and air.
The TDR method measures the ratio of apparent to physical probe rod length (L a /L) which gives the square root of dielectric number (ε b ).
The bulk dielectric number is then calculated from the volumetric fractions and the dielectric numbers of the four phases (details in Appendix A) using A value of 0.5 was used for α.
It is not possible to distinguish between changes in the liquid water content and changes in the ice content with only one measured parameter (ε b ).Equation (D1) was therefore rewritten in terms of the total water content (θ tot ) and the porosity ( ) as We remark that Eq. (D2) assumes the densities of liquid and frozen water to be the same.This is clearly wrong for free phases and probably also in the pore space of soils.However, the density ratio can be absorbed into the dielectric number ε i , which we do in the following.The resulting fluctuation of ε i is presumed to be small compared to other uncertainties. and to obtain the equation For temperatures above a threshold freezing temperature (T > T f ), all water is assumed to be unfrozen (θ tot equals θ l ).Equation (D5) then reduces to For temperatures equal to or below the threshold freezing temperature (T ≤ T f ) it was assumed that the total water content (θ tot ) remained constant and only the ratio between volumetric liquid water content (θ l ) and volumetric ice content (θ i ) changed.This is a rather bold assumption as freezing can lead to high gradients of matric potential, as well as to moisture redistribution.However, since the dielectric number of ice is much smaller than the dielectric number of liquid water, the error in liquid water content measurements is still acceptable (which is not the case for ice content measurements).Under these assumptions we obtained the following equation for calculating the liquid water content of a fourphase mixture: The error of the volumetric water content measurements using TDR probes was estimated to be between 2 and 5 %, and the precision to be better than 0.5 % (Boike and Roth, 1997).
The availability of reliable temperature data is crucial in this approach.The liquid water content is first calculated for all times when the soil temperature was above the freezing threshold, using Eq.(D5).When the soil temperature was below the freezing threshold the water content immediately prior to the onset of freezing was determined and used as total water content (θ tot ) for calculating the liquid water content during the frozen interval with Eq. (D7).
Since water in a porous medium does not necessarily freeze at 0 • C but at a temperature that depends on the soil type and water content, estimating the threshold temperature is a crucial part of this approach.If the freezing characteristic curve is known for the material then the threshold temperature can be determined from the soil liquid water content.To avoid interpretations of frequent freezing and thawing due to soil temperature measurement errors, short-term temperature fluctuations were smoothed by calculating the mean of a moving window with an adjustable width.The smoothed temperatures were then used to trigger the switch from one equation to the other, rather than using the original temperature time series.

D4 Snow depth correction for air temperature
The raw distance Dsn raw obtained from the SR50 sonic sensor (Campbell Scientific Ltd.) is calculated using the speed of sound at 0 • C and is corrected with the air temperature at 2 m height using the formula provided by the manufacturer:  The pockets/holes are the result of snow removal using a snow density cutter (100 cm 3 ).Snow depth is 47 cm above a basal ice layer at the point of measurement but ranges between 47 and 49 cm.Measurements were taken to the left of the yellow folding rule (left and right profiles in Table E1).Ice layers are present 24-28, 43-47 and 55-56 cm above the basal ice layer, which is 15 cm thick.Further data on temperatures, dielectric numbers and stratigraphy can be found in Table E1.E1).Further data on temperatures, dielectric numbers and stratigraphy can be found in Table E1.Data can be found in Table E1.
Table E1.Snow data obtained from manual probing of snow profiles in spring 2000 (Bayelva), 2006 (Bayelva and Ny-Ålesund) and 2016 (Bayelva).Data include stratigraphy, temperature, dielectric number, density and information on dominant grain shape, size and hand hardness index according to international snow classification (Fierz et al., 2009).Grain shape abbreviations are DH for depth hoar, IFbi for basal ice layer, IFil for horizontal ice layer and PPip for ice pellets.Hand hardness indices are 1 for very soft, 2 for soft, 3 for medium, 4 for hard, 5 for very hard and 6 for ice.Left and right refer to the measurement positions within the profile shown in Fig. E3.The soil organic carbon content (kg m −2 ) has been calculated using the following formula: SOCC = TOC × ρ bulk × z = CD bulk × z.
(F1) The bulk carbon density (CD bulk ) (kg m −3 ) has been calculated using the bulk soil density ρ bulk and the following formula:  2.
Competing interests.The authors declare that they have no conflict of interest.
Special issue statement.This article is part of the special issue "Changing Permafrost in the Arctic and its Global Effects in the 21st Century (PAGE21) (BG/ESSD/GMD/TC inter-journal SI)".It is not associated with a conference.

Figure 1 .
Figure 1.(a) Location of the Bayelva study site on Spitsbergen, western Svalbard, at 78.92094 • N, 11.83334 • E. (b) The site is located about 3 km from Ny-Ålesund in the Bayelva River catchment area, between two mountains (Zeppelinfjellet and Scheteligfjellet), and in front of the Brøggerbreen glacier.Aerial image captured in August 2008 using the high-resolution HRSC-AX camera (Hauber et al., 2011a, b): data and metadata for the high resolution the HRSC-AX image covering the entire area shown in (b) are provided in Appendix B. (c) The area of the site under summer conditions (August 2008) and (d) under spring conditions (April 2016).

Figure 2 .
Figure 2. Topographical map (a) and overview of the Bayelva study site (b) in Ny-Ålesund, Svalbard.(a) Aerial orthoimage (20 cm px −1 ) with topographic information (2.5 m contour lines) derived from a digital elevation model (DEM) with a cell size of 0.5 m.The locations of the instrumental and sampling sites are marked by colored symbols.The site is located on a small hill; white areas are snow fields remaining late in the season.The orthophoto and DEM were obtained with the HRSC-AX camera in August 2008 (data archived in Pangaea and metadata for the high-resolution DEM covering the entire area shown in Fig. 1b are provided in Appendix B).(b) Most of the instruments are within a fenced-off area to protect them from reindeer damage.Note that the eddy covariance station and permafrost borehole are located outside the fenced area.

Figure 3 .
Figure 3.Time series of Bayelva data provided in this paper.(a)-(f) meteorological data, (g)-(l): soil and subsurface data, (m)-(p): snow data.The data are organized following the structure of Appendix G. Further details on the sensors and periods of operation are given in Table2.This plot (divided in three parts for better visibility) is also provided in Appendix H.

Figure B1 .
Figure B1.Operating principle of the airborne high-resolution stereo camera (HRSC-AX), and viewing geometry of the individual charge-coupled device (CCD) sensors.ND -nadir channel; S1 and S2 -stereo 1 and stereo 2; P1 and P2 -photometry 1 and photometry 2; IR -near-infrared channel; GR -green channel; BL -blue channel; RE -red channel.All nine line sensors have a cross-track field of view of ±6 • .

Figure B3 .
Figure B3.Context map of the Brøgger peninsula, with the thick white outline showing the total coverage of the HRSC-AX survey and with the white square indicating the location of the image tile provided in this publication (base map: hillshade version of ASTER DEM).
element provides a reference longitude for certain map projections.In many projections, the CENTER_LONGITUDE along with the CEN-TER_LATITUDE defines the point or tangency between the sphere of the planet and the plane of the projection.

Figure C5 .
Figure C5.(a) Vertically installed time-domain reflectometry (TDR) air-snow TDR probe (50 cm length) above the ground surface during summer at the Bayelva climate station (1999).During winter (fully snow covered), the dielectric number of snow is recorded.(b) PT100 temperature sensors in air (summer) at 4 and 20 cm above ground surface.(c) PT100 temperature sensors in snow (winter) with air gaps around cables and sticks; the TDR probe is completely covered with snow and not visible from the surface.

Figure C7 .
Figure C7.Soil profile and station location (a) 1998, (b) 1999 and (c) 2009.(b) Upper temperature sensors were affected by reindeer disturbance in September 2003.Furthermore, the temperature sensors were affected by continuous frost heave by about 10 cm between 2003 and 2015.

Figure C8 .
Figure C8.Bayelva soil station, September 1998-January 2012, (78.92094 • N, 11.83334 • E).(a) Soil profile in non-sorted circle (down to 120 cm below ground surface); (b) TDR and temperature probes.Soil texture and C, N and S data are listed in TableF1.

Figure E2 .
Figure E2.Snow profile taken on 9 May 2016, inside the Bayelva fenced area.The pockets/holes are the result of snow removal using a snow density cutter (100 cm 3 ).Snow depth is 47 cm above a basal ice layer at the point of measurement but ranges between 47 and 49 cm.Measurements were taken to the left of the yellow folding rule (left and right profiles in TableE1).Ice layers are present 24-28, 43-47 and 55-56 cm above the basal ice layer, which is 15 cm thick.Further data on temperatures, dielectric numbers and stratigraphy can be found in TableE1.

Figure E3 .
Figure E3.Snow pit, 10 May 2016.The basal ice layer is visible at the bottom of the snow layer.Two profiles were measured: one to the left and one to the right from the yellow meter stick (profile data named left and right in TableE1).Further data on temperatures, dielectric numbers and stratigraphy can be found in TableE1.

Figure E4 .
Figure E4.Measuring snow temperature in profile, 10 May 2016.Data can be found in TableE1.

)Figure H1 .
Figure H1.Time series of Bayelva data provided in this paper (Fig. 3) divided into three plots.(a-f) Meteorological data; (g-l) soil and subsurface data; (m-p) snow data.The data are organized following the structure of Appendix G. Further details on the sensors and periods of operation are given inTable 2.

Table 1 .
Site description parameter for earth system model input.Values have been computed and compiled for the Bayelva site and surrounding areas.

Table 2 .
List of sensors, parameters and instrument characteristics for the automated time series data from the Bayelva site, 1998-2017.Additional (not automated) data on snow cover and soil profiles can be found in Appendices E and F. Positive heights are above surface; negative heights are below surface.

Table 3 .
Description of data level and quality control for data flags.Most flags are run automatically; few are done manually (e.g., three for maintenance, six for plausibility).

Table B1 .
Image size and pixel size of individual HRSC-AX images.

Table B2 .
Description of metadata (keyword-value pairs) contained in the HRSC-AX image headers.

Table B2 .
Description of metadata (keyword-value pairs) contained in the HRSC-AX image headers.