Soils are embedded in the environment, coupled to vegetation and atmosphere at the land surface and to groundwater at its lower end. This coupling gives rise to a suite of physical, chemical, and biological dynamics, most of which are highly non-linear and varying in time and space. Soils provide crucial ecosystem functions such as water storage and filtering, food and other biomass production, recycling of carbon and nutrients, biological habitat and gene pool, physical and cultural heritage, source of raw materials, and platforms for human life (United Nations, 2014; Vereecken et al., 2016). Soils are widely distributed on the earth's surface. Flow and transport processes in unsaturated soils occur predominantly in the vertical direction, with the gravity force playing a major role, as abrupt changes in soil properties due to soil horizons and layers are typically more significant than those in the lateral direction and because of the strong coupling between soil, vegetation, and atmosphere. Therefore, despite the relevance of soils for global phenomena, the relevant soil processes are rather local. Here, one aspect that complicates the picture is the heterogeneity of soil properties. Another one is the non-linearity of soil processes. In order to effectively address this complexity, state-of-the-art experimental approaches must be coupled to numerical models for the comprehensive representation of the system properties, states, and fluxes so that the hydrological system can be better understood.

Recently, Vogel (2019) provided a comprehensive discussion about the scales and scaling issues in the context of soil hydrological research and noted the need for looking at small-scale soil properties (i.e. at the pedon scale, at which soil physics is capable of describing states and fluxes with sufficient accuracy) as a necessary step towards understanding and summarizing the processes at larger scales. In this respect, the author stresses the need for a two-step approach based on the accurate description of the soil water dynamics at the pedon scale and accounting for the spatial patterns of functional soil types that constitute the landscape, including the vertical stratification of soil hydraulic properties and structural attributes. However, the author remarks that high-resolution measurements of the relevant states and properties cannot be achieved at the larger scale (i.e. catchment, the typical scale of application of hydrological research). In this context, the intermediate scale of hillslopes is crucial for linking the detailed process understanding to larger-scale dynamics, recognizing hillslopes as key landscape features that organize water availability on land (Fan et al., 2019). In this respect, coupling state-of-the-art hydrological modelling approaches with high-resolution subsurface characterization can lead to an accurate quantification of the soil water dynamics in the vadose zone (Vereecken et al., 2015).

The physical description of the small-scale water movement through the soil's porous structure is typically achieved using the Richards equation. However, a detailed description of the material properties is needed and cannot be fully resolved by direct sampling. Thus, inverse modelling can be a powerful tool for the estimation of the soil hydraulic parameters (e.g. Vrugt et al., 2008), including the recent developments in data assimilation approaches (e.g. Bauser et al., 2016, 2020; Botto et al., 2018). These require dense (in the direction of the dominant flow, typically orthogonal to the soil surface) measurements of soil water content with high temporal resolution and of high quality. Furthermore, in situ sensors can experience all the processes affecting the measured state variables in their natural environment (Wollschläger et al., 2009), which is an important advantage with respect to sample-based determinations from the laboratory.

The performances of hydrological models can be improved by various measured data with high spatial and temporal resolution (Clark et al., 2017). Bronstert (1999) highlighted the importance of linking experimental knowledge to the experience gained from numerical modelling applications as a very valuable synergistic combination. Technological advances in our ability to measure soil hydrological states efficiently at the hillslope scale and beyond are one possible way to gain the much-needed improved understanding of processes that challenge the comprehensive understanding of field-scale hydrology.

In the research framework of the TERENO Harz/Central German Lowland Observatory, the Schäfertal Hillslope represents a benchmark site for developing and testing the integration of state-of-the-art monitoring techniques with advanced modelling approaches. This offers the opportunity to gain a more detailed understanding of processes and to quantify and predict water and matter fluxes at nested spatial scales in the context of climate and land use change. Specifically, the approach followed at the site accounts for the soil spatial variability through detailed soil mapping and is designed to provide in situ data with high-temporal-resolution and dense coverage in the vertical direction about the soil water dynamics in the vadose zone and of its boundary conditions. With this design tailored to the needs of vadose zone modelling, we aim to provide physical models with ideally all the data needed for quantifying and predicting the soil water fluxes at spatial scales ranging from the pedon to the hillslope scale, with important implications in terms of methodological advance and process understanding for catchment-scale processes.

Here, we present the first 21 months of the comprehensive data set measured by the monitoring network STH-net, recently implemented at the Schäfertal Hillslope site, part of an intensive hydrological observatory. The data set includes hourly time series of the meteorological forcing, soil water content measured in situ at different locations and at multiple soil depths along a hillslope transect, and soil physical and physicochemical properties.

The Schäfertal experimental site is a small headwater catchment (1.44 km2) located in the Lower Harz Mountains, in central Germany (51∘39′ N, 11∘3′ E). Environmental research at the Schäfertal catchment was initiated at the end of the 1960s with the implementation of a hydro-meteorological station (Reinstorf et al., 2010), and the infrastructure has continuously been expanded since then. Since 2010, the Schäfertal catchment is one of the highly instrumented intensive research sites within the TERENO Harz/Central German Lowland Observatory (Zacharias et al., 2011, Wollschläger et al., 2017). Due to the geographical settings of the Harz region, the Schäfertal catchment receives only 630 mm of precipitation per year. The average annual air temperature is 6.9 ∘C, with a sub-continental climate (Reinstorf, 2010). The geology of the catchment is dominated by Devonian argillaceous shales and greywackes, covered by periglacial sediments (Borchardt, 1982). Near-surface compacted horizons within the basal layer are known to induce interflow processes in the unsaturated zone (Borchardt, 1982; Gräff et al., 2009). Dominant soil types in the Schäfertal are Gleysols occurring in the valley bottom as well as Luvisols and Cambisols on the loess-covered slopes (Ollesch et al., 2005). The slopes of the catchment are intensively used for agriculture, whilst meadows occupy the valley bottom (Schröter et al., 2015).

Since 2012, a smaller hillslope area named Schäfertal Hillslope site, located downstream of the Schäfertal gauging station, has been instrumented for detailed investigations of the hydrological processes in the unsaturated zone. From 2012 to 2017, the wireless soil moisture monitoring network SoilNet delivered information about the soil water dynamics at three depths within the unsaturated zone with high spatial coverage. In 2018, the SoilNet was disposed, and a new soil monitoring network, named STH-net, was installed, aiming to improve the resolution in the vertical direction at fewer locations selected based on the knowledge about the soil spatial variability and soil water dynamics gained from the previous monitoring experience (see Martini et al., 2015, 2017a, b). The STH-net is described in the following sections of this paper, and its data are now available through the data portal EUDAT (https://doi.org/10.23728/b2share.82818db7be054f5eb921d386a0bcaa74). The Schäfertal Hillslope site includes north- and south-exposed slopes divided by the creek (Schäferbach) in the valley bottom (Fig. 1). In contrast to the slopes upstream, which are primarily covered by cropland, this grassland transect is used as pasture and is not affected by agricultural practices except that the grass is mowed typically once per year. The spatial extent of the hillslope is approximately 250m×80m and presents various topographical and pedological features. The slopes are covered by silty loam Cambisols more evolved towards the foot slope, while loam and silty loam stagnic Gleysols occupy the valley bottom. An extensive description of the soil units mapped at the site is provided in Martini et al. (2015). The STH-net is designed to cover the spatial variability in the soil properties as well as the soil layering with high resolution.

The STH-net comprises eight monitoring stations (named as P1 to P8) arranged along a transect centred within the Schäfertal Hillslope site and aligned along the slope direction (Fig. 1). The stations P1, P2, and P3 are located within the northern (i.e. south-facing) slope and cover the transition between the soil units STU1 and STU2 described in Martini et al. (2015); the stations P4 and P5 fall into the valley bottom, i.e. soil unit STU3; P6, P7, and P8 cover the lower part of the southern (i.e. north-facing) slope, i.e. soil unit STU4. Every station features a soil profile instrumented with time-domain reflectometry (TDR) probes installed every 0.1 m along the vertical direction. A sketch showing the design of a reference monitoring station is presented in Fig. 2. Each of the instrumented soil profiles located on the hillslopes features a minium of seven TDR probes installed at depths of 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, and 0.7 m, whilst an additional probe is installed at P3 at a depth of 0.8 m, and the profiles at P4 and P5 feature additional TDR probes at depths of 0.8, 0.9, 1.0, and 1.1 m in order to cover the deeper soils. In a few cases, the depths of the probes were adjusted to avoid installing the TDR probe at or too close to the boundaries between soil horizons. The exact depth of every TDR probe is reported in the file “STH-net_Soils.txt” and displayed in Fig. 3.

At every station, a well instrumented with a piezometer was installed ca. 2 m to the east of the instrumented soil profiles for monitoring the water level. One station for every topographic unit (i.e. northern slope, valley bottom, and southern slope) was further instrumented with sensors measuring the soil temperature at six depths between 0.05 and 1.0 m.

A weather station is located in the centre of the hillslope transect next to the creek. All measurement systems comprising the STH-net collect measurements every 10 min, with the only exception of the water level data, which are collected every 2 h.

For the estimation of soil water content using a composite dielectric approach, some physical parameters must be known. These are primarily temperature, porosity, and the dielectric number of the solid matrix (εsoil). Among them, soil temperature plays a major role in determining the global uncertainty. As part of the STH-net, soil temperature is measured in situ at the same time as the TDR waveforms, which enables an accurate temperature correction. The soil porosity was estimated for every sampling point from undisturbed soil cores and introduces an uncertainty. For εsoil we have chosen the value of 4.6, corresponding to the dielectric permittivity of quartz. This value was chosen arbitrarily and hence introduces an uncertainty. For a more extensive discussion about the uncertainty in the soil water content estimation due to the single parameters, we refer to Roth et al. (1990). For the data set presented here, we estimated the uncertainty in the calculated soil water content using the CRIM formula by varying the values of εsoil and porosity between 4 and 6 and between 0.3 and 0.5, respectively (similar to Wollschläger et al., 2010). We obtained values < ± 0.03 m3m-3 as the largest uncertainty in the soil water content estimation. This information is reported in Table 1 along with the measurement range, accuracy, and resolution of the other variables provided within the data set described in this article.

Rain gauges may misestimate the rainfall rate under certain circumstances, especially when rainfall events are associated with strong wind. The experiment described in Basara et al. (2009) shows that a sensor similar to the one installed at the Schäfertal Hillslope site overestimates the rainfall intensity in an urban environment. The rainfall rate data presented in this article were compared to those of several other rain gauges (data from partner research institute, not available here) located ca. 100 m away from the site. The rainfall intensity values measured by our sensor neither underestimate the rainfall rate values nor completely miss rainfall events. With our data set, we make the measured data available to any interested scientists along with all relevant site information and leave them the choice about eventual compensation measures to be applied. The correction function proposed by Richter (1995) is commonly used for studies conducted in central Germany to account for the possible wind-induced underestimation of the rainfall intensity.

Until a few years ago, the Schäfertal catchment used to be affected by significant snowfall, with major snowmelt events occurring between January and April, whose effects on the hydrological processes are described, e.g. in Ollesch et al. (2005). In the last years, however, no significant snowfall events were observed. The last winter period (December 2020 to February 2021), instead, was characterized by exceptionally intense snowfall (with a maximum of ca. 45 cm on 8 February 2021) that accumulated and persisted. Unfortunately, the technical infrastructure currently available at the site does not allow a meaningful estimation of the snow height and distribution during the monitoring period; hence the snowfall events are not recorded by the weather station in use (see Fig. 7). Because of this, the snow contribution to the water balance needs to be derived from the meteorological and soil temperature data available.

Overall, 9.3 % of the soil water content data and 7.6 % of the soil temperature data are missing (particularly until March 2019 for the station North and between April and August 2020 for the station Valley) due to various technical failures.

The STH-net data stored by the three data loggers are accessed and downloaded remotely using the software Loggernet (Campbell Scientific Inc., Logan, UT, USA). The water level data, which are manually downloaded, are the only exception. The data files are regularly quality-checked and uploaded to the EUDAT record STH-net (https://doi.org/10.23728/b2share.82818db7be054f5eb921d386a0bcaa74, Martini et al., 2020), where they remain available for download.

The STH-net data are available under a dynamic identifier (https://doi.org/10.23728/b2share.82818db7be054f5eb921d386a0bcaa74; Martini et al., 2020) at the time of the manuscript submission (from there, all future versions of the archive can be easily accessed) under the Creative Commons Attribution license (CC-BY 4.0).

The STH-net data are archived as separate text files for the different data types: soil water content, soil temperature, water level, and meteorological variables. Furthermore, the geographic coordinates of the measurement locations and the soil information are available for download. The time series data start from 1 January 2019 and continue with hourly time steps until the most recent update. At the time of the manuscript submission, the latest entry refers to 28 February 2021. The water level data are available with a 2 h resolution and cover the time period between 6 March 2020 and 23 February 2021. All the data published in the online archive (https://doi.org/10.23728/b2share.82818db7be054f5eb921d386a0bcaa74, Martini et al., 2020) will be updated on an approximately 3-month basis.