Soil moisture is a key state variable of the Earth's environmental system, controlling various processes at various scales: the exchange of water and energy between the land surface and the atmosphere, runoff generation and groundwater recharge, vegetation development and growth in natural and managed systems, or the release of greenhouse gases from soils.

But while soil moisture is a much desired quantity in research and applications, its observation remains a challenge . Numerous techniques exist for measuring soil moisture at specific points in space, including vertical profiles, such as time domain reflectometry (TDR), frequency domain reflectometry (FDR), and soil sampling. While these techniques all have specific uncertainties and limitations – including even the benchmark standard technique of thermogravimetry – they share one fundamental limitation: the uncertain spatial representativeness. This shortcoming results from the small support volume (footprint of the measurement in the order of centimeters) in combination with the strong and potentially abrupt variation of soil moisture in space . This issue can be overcome by very dense in situ sensor networks see, e.g.,. At the other end of the scale continuum, remote sensing techniques allow for a higher spatial coverage and a volume- or area-integrated measurement approach. and provide an overview of techniques based on optical, thermal, passive microwave, and active microwave measurements. Yet, apart from the fact that these are indirect observations and are thus inherently uncertain, they typically have limited (and often imprecisely characterized) penetration depths that cannot be assumed to be representative of the root zone. Over densely vegetated terrain the uncertainty of these remote sensing techniques increases. Furthermore, both airborne and spaceborne remote sensing of soil moisture often come along with infrequent measurement intervals due to a limited overpass frequency.

Cosmic-ray neutron sensing (CRNS) is a modern measurement technique to close the scale gap in soil moisture observation. According to simulations by , the sensor can provide volume-integrated soil moisture estimates with an exponentially shaped horizontal footprint of hundreds of meters (the plot scale) and a vertical footprint of tens of centimeters, corresponding to the upper root zone. and showed that the temporal resolution of standard detectors is in the range of 3 to 12 h and strongly depends on the detector technology used. The principles of the CRNS method were introduced about 12 years ago by , followed by , who presented the first results using mobile measurements and a conversion procedure to obtain soil moisture. presented a complete overview over the method and the then newly established CRNS network in the USA. published a short synthesis of recent advancements in CRNS at that time, but soon afterwards important updates with respect to snow detection and road effects on roving CRNS measurements were published.

The CRNS sensor is sensitive to the ambient density of neutrons in the near-surface atmosphere. The main source of cosmic-ray-induced neutrons exhibits energies between 105 and 109 eV . Their interaction in the ground leads to neutrons in the energy range of 10-2 to 107 eV, of which epithermal neutrons (1–105 eV) are most sensitive to hydrogen and water . Hence, standard CRNS sensors are equipped with a polyethylene shield that reduces the thermal neutron (< 1 eV) fraction and slows down faster neutrons to detectable energies . Some CRNS sensors employ an additional “bare” counter which is sensitive to thermal neutrons only.

To date, a wide range of neutron detectors optimized for soil moisture monitoring exist, and these use various gases or coating materials for neutron detection. An overview of detectors used in our field campaign is provided in Sect. . For a detailed description of the technical components that make up standard CRNS units, including detector, telemetry, and additional atmospheric sensors, the reader is referred to and .

The intensity of detected neutrons is mainly controlled by the interaction with hydrogen pools in the sensor footprint, of which soil moisture is typically the most important, though not the only one. A standard approach to estimate the gravimetric (θgrv) or volumetric soil moisture (θvol, often referred to as θ) from epithermal neutron count rates N is to use the transfer function proposed by : 1θgrv=a0N/N0-a1-a2,θvol=ϱbϱw⋅θgrv.

In that form, the transfer function requires the calibration of parameter N0 once, in order to take the specific measurement conditions into account. The value not only depends on the local terrain features, soil moisture heterogeneity, and instrumental sensitivity , but also on the local prevalence of other hydrogen pools, such as vegetation, litter, soil organic carbon, and lattice water . The parameters a0, a1, and a2 have proved to be robust in their original formulation across multiple CRNS sites among others, while few locations apparently work better with different values . Soil bulk density ϱb (kg m-3) needs to be measured or estimated locally, while the density of water, ϱw, can be assumed to be 1000 kg m-3.

The calibration of N0 requires the availability of soil moisture observations within the horizontal and vertical footprint of the sensor. have synthesized the state-of-the-art methodology on how to make use of distributed point-scale soil moisture measurements for sensor calibration and proposed adequate sampling designs.

Yet, calibration can only account for the effect of static hydrogen pools at a specific point in time, while hydrogen pools such as vegetation are typically dynamic. In addition, the collection and processing of soil samples for measuring thermogravimetric soil moisture are particularly labor intensive.

One of the challenges in using CRNS for soil moisture observation is thus to separate the part of the signal that is related to soil moisture from those parts affected by other hydrogen pools such as vegetation and soil organic carbon . Further prerequisites include the correction of atmospheric effects such as air pressure, air humidity, and fluctuations of incoming neutron radiation , as well as a better, physically based understanding of the spatial sensitivity of the neutron sensors, both vertically and horizontally, and how it is influenced by dynamic environmental conditions in the footprint .

While the horizontal footprint of a single CRNS sensor already exceeds the spatial support of conventional techniques, further attempts have been made to enhance the spatial coverage such as roving CRNS and networks of CRNS sensors. The term roving CRNS stands for the utilization of mobile CRNS sensors to increase the spatial extent of soil moisture measurements. It has been recognized as a promising approach to increasing the spatial scale see, e.g.,. Within a CRNS roving campaign, the sensor unit is moved (e.g., by car) within the area of interest in order to explore the catchment-scale wetness conditions and related spatial patterns . Under suitable conditions, up to hundreds of square kilometers may be covered in a single day. This coverage makes CRNS roving an ideal method for closing the critical scale gap in soil moisture monitoring toward the scale of small to medium catchments. However, roving relies on campaign-based measurements and thus produces snapshots only. As roving requires neutron measurements with much higher temporal resolution (in the order of minutes), a CRNS rover usually consists of several larger detectors, resulting in an increase in achievable neutron count rates compared to stationary applications. In this context, an increase in detector sensitivity would constitute a promising perspective: obviously, the potential to increase the signal-to-noise ratio by integrating neutron counts over time is limited if the sensor is mobile .

In contrast, networks of CRNS sensors operate multiple stationary CRNS sensors. Some initiatives have established national CRNS monitoring networks with the aim of supporting environmental monitoring at larger scales. These networks are being implemented under the acronym of COSMOS (the COsmic-ray Soil Moisture Observing System). The first network was established in the USA by the University of Arizona and has already deployed more than 60 CRNS sensors at various locations across the USA . The Australian network was supported by the CSIRO research institute and consists of nine sensors distributed across the continent under different environmental conditions . A network has also been established in the United Kingdom by the UK Centre for Ecology and Hydrology (CEH) . Similar initiatives have been started in Kenya and India . Yet, these networks aim at distributing single CRNS sensors at a national scale, similar to the idea of climate stations, providing localized measurements in a sparse network that spans large scales.

To better cover the spatial and temporal scales, the combination of both stationary and roving approaches has been explored : a limited number of stationary sensors have been used to detect the temporal dynamics while roving surveys have provided the means for extrapolation and interpolation beyond the stationary sensors. Despite the promising results, many challenges have been identified due to the heterogeneity of land surface conditions and the validity of assumptions that govern the integration of roving CRNS data, which are sparse in time, with stationary CRNS data, which are sparse in space.

The Cosmic Sense research unit, funded by the German Research Foundation (DFG), addresses the above challenges in a concerted effort with a consortium of eight German partner institutions. In this context, one component of the Cosmic Sense project is the targeted joint operation of a large number of CRNS sensors in a dense observational network. The scientific aims behind these joint field campaigns (JFCs) are

to systematically explore, at the landscape level, the effect of heterogeneity of soil and vegetation in the CRNS footprint on the neutron signals;

As mentioned in the previous section, networks of a large number of CRNS sensors have been established before; however, as sparse national-scale networks they are not geared towards the scientific subjects elaborated above. Hence, the joint operation of 24 CRNS sensors that took place from early May to late July 2019 is unprecedented in its scope as it features a large number of detectors in an area of just 1 km2, characterized by pronounced soil moisture gradients and heterogeneous land cover. The CRNS monitoring network was complemented by various observational techniques including CRNS roving, thermal imaging from unmanned aerial systems (UASs), multiple FDR sensor networks (SoilNet clusters), and profile probes, as well as manual measurements with FDR probes and thermogravimetry; and, finally, the CRNS network was embedded in an established long-term observation infrastructure TERENO, Terrestrial Environmental Observatories, and multiple synchronous observational campaigns (MOSES, ScaleX, see Sect. ) that focused on water and energy fluxes in the atmospheric boundary layer. The observational period from late spring to midsummer was characterized by pronounced soil moisture dynamics, starting out from fully saturated conditions after persistent heavy rainfall, followed by more than 2 months of drying occasionally interrupted by brief but intense rainfall events.

In this paper, we present this unique data set as a contribution to the CRNS community but also to those researchers interested in exploring the potential of dense volume-integrated soil moisture observations from a hydrological perspective. The data set was split into two subsets and is freely available from EUDAT at https://doi.org/10.23728/b2share.282675586fb94f44ab2fd09da0856883 and https://doi.org/10.23728/b2share.bd89f066c26a4507ad654e994153358b .

We start by providing a detailed overview and justification of the choice of the study area (Sect. ). Section  constitutes the core of this paper: the presentation of the data collected as part of the JFC. Section  gives an overview of the various components, Sect.  documents the underlying data model, and the subsequent sections describe these data sets in more detail. Section  outlines relevant data from third parties which where not collected as part of the JFC and are not part of this data publication. As this paper is about the presentation of the data set, we will not provide any further analysis or interpretation. However, in Sect. , we will illustrate how the neutron count rates of stationary and roving sensors can be used to represent temporal soil moisture dynamics at the catchment scale, including the areal average of soil moisture and its variability in space. Based on that example, Sect.  will conclude by outlining the potential of the presented data set.

This section describes the measurements conducted during the campaign. Other complementary data are described in Sect. .

The core innovation and overall motivation of this data set is the use of a dense network of 24 CRNS sensors in a study area of roughly 1 km2. These instruments of various types and their key observational variables are described in Table .

Around this core data set we carried out various measurements that are required to utilize, study, and evaluate observed neutron counts from the CRNS network for the purpose of soil moisture retrieval: meteorological observations (Sect. ); soil properties and soil moisture measurements for calibration and validation, obtained at selected points and various depths using different techniques (Sect. ); CRNS roving data (Sect. ) and remotely sensed thermal images (Sect. ), both from several campaigns in order to obtain additional information on soil moisture variability in space; and measurements and estimates of aboveground biomass (Sect. ) in order to account for the corresponding hydrogen pools in the CRNS data analysis.

The data presented in this study consist, for large parts, of time series data recorded at well-defined, sparse, and static points in space (e.g., neutron counts, meteorological variables, soil moisture, or permittivity). For such data, we decided to implement a simple, transparent, and easy-to-use data model that is based on standard text tables (character (tabulator) separated values, csv): each sensor (or sensor unit) is characterized in attribute tables, including a unique identifier (ID), location (latitude and longitude in WGS 84 reference system), measurement depth or height (in meters above or below the surface), and a set of sensor-specific attributes that are documented in the attribute table's header. The time series of observations are provided in additional text tables: each sensor is represented by one file that is named according to the unique sensor ID. The first column in any such file holds the date and time (UTC, ISO 8601). Any other columns represents the measured variables which are documented in the file header, including the physical units.

Time series data with a fixed spatiotemporal resolution (e.g., data from SoilNet) are provided in netCDF files, which include an explicit documentation of all dimensions and observational variables but which are also accompanied by an overview text table that contains key attributes of the measurement locations.

Any exceptions from these data models (e.g., for the CRNS roving, the land use and soil data, the digital elevation model, or the remote sensing data) are explicitly elaborated in the corresponding subsections and by using metadata files. For polygon data, we use the ESRI shapefile and the GeoJSON format, and for gridded data we use the GeoTIFF format.

Further details of the data repository are given in the code and data availability section (Sect. 6).

The permanent meteorological instrumentation at the Rott headwater catchment (also known as the TERENO site Fendt) consists of an eddy-covariance flux tower and several precipitation sensors. Besides the high-frequency measurements of the eddy-flux system (the data of which are available through the Integrated Carbon Observation System, ICOS, and TERENO), standard climate variables are recorded every minute. The meteorological observations and their respective devices selected for the presented data set are listed in Table .

During the period from mid-May to mid-July 2019, a total of 24 stationary CRNS sensors were operated, though not all sensors were measuring at all times. CRNS sensor no. 8 in the northeast had already been installed as part of TERENO infrastructure, and it continued operation after the JFC. The temporal data availability is illustrated in Fig. b. Data gaps and differing periods of data availability are due to various reasons, including different dates for installation/deinstallation, sensor maintenance, power failures, or logger configuration, to name a few.

Table  gives an overview of all stationary CRNS sensors that were used during the campaign. Figure  illustrates the proportions of the different sensor models. Out of the 24 sensors, 18 have been manufactured by Hydroinnova LLC (Albuquerque, NM, USA). Those instruments are based on neutron-sensitive detector gases, which are either 3He gas (CRS-1000, CRS-2000) or 10BF3-enriched gas (CRS-1000-B, CRS-2000-B). Five sensors (StX-140-5-15) were manufactured by StyX Neutronica GmbH (Dossenheim, Germany) and feature a new experimental design, based on solid 10B-lined converters. One of the 24 sensors was manufactured by Lab-C LLC (Tucson, AZ, USA) and also constitutes a new approach that uses a multiwire proportional chamber with solid 6Li as a neutron converter. Figure  exemplifies time series of raw neutron count rates from six CRNS sensors of different types (manufacturers) and thus different sensitivities.

For all CRNS sensors, the detection chamber is surrounded by a so-called moderator material in order to thermalize (i.e., slow down) epithermal neutrons for detection seefor details. Six sensor units were equipped with an additional detection unit without a moderator (i.e., a bare counter) in order to directly count thermal neutrons (see Table ). This approach was motivated by recent studies which show that the ratio of thermal to epithermal neutron count rates is useful to distinguish specific hydrogen pools, namely vegetation biomass or snow .

The CRNS sensor units were equipped with varying meteorological sensors for air temperature, relative humidity, and air pressure, all of which are required to correct for atmospheric effects on epithermal neutron count rates. For some units, such sensors were only available internally in the logger box (making the observations less representative of the sensor footprint), while some units also featured external meteorological sensors. A more detailed specification of the placement of meteorological sensors is included in the metadata of the individual CRNS sensor units. All CRNS sensors were set up to record neutron counts and additional variables at a temporal interval of 20 min.

At 19 of the 24 CRNS sensor locations a profile probe (40 or 100 cm maximum measurement depth; see Sect.  for further details) was installed to monitor the vertical distribution of soil moisture.

The locations of the CRNS sensors are shown in Fig. . The sampling design underlying the placement of sensor units was subject to various scientific and practical – partly antagonizing – constraints

to achieve a maximum coverage of the catchment area in order to capture soil moisture storage dynamics in the root zone as an important dynamic part of the catchment's water balance and a key control of groundwater recharge;

The trade-off between the aim of maximum coverage versus maximum overlap was resolved by increasing the density of CRNS sensors in the northwest, where the permanent SoilNet allows for optimal validation.

Neutron count rates at different observation locations were obtained with different CRNS sensor types. Even within the same sensor type, the effective sensitivity varies. Between all sensors used in this study, the sensitivity between the least and the most sensitive sensor (CRS-1000 and Lab-C, respectively) is expected to vary by an order of magnitude. In order to compare neutron count rates observed at different locations, these count rates need to be normalized to a standard level. Given that the effective sensitivity of the instrument is unknown, we need to introduce a reference standard. For that purpose, we placed a mobile CRNS sensor (calibrator, Hydroinnova, no. 20 in Table , basically consisting of two combined CRS-1000 systems) just beside each stationary CRNS sensor for a period of at least 24 h. The ratio between the average count rates of the stationary CRNS sensor and the calibrator – the sensitivity factor – was used to standardize the count rates for the entire time series to the calibrator level. For those CRNS sensors which we could not collocate with the calibrator, we used the average sensitivity factors obtained for the same instrument type. The resulting sensitivity factors are included in Table . Furthermore, a CRNS rover unit was collocated with most of the stationary CRNS sensors on June 27, which provides a further sensitivity reference (see Sect. ).

During the campaign, portable cosmic-ray neutron sensors were used

to study the spatial variability of neutron intensity and soil moisture in the whole catchment and particularly in between the stationary sensors,

We used three types of mobile devices:

UFZ rover (installed in a Land Rover Defender) equipped with a moderated CRNS-RV unit (Hydroinnova LLC, Albuquerque, USA) based on 3He gas see for details. In contrast to previous studies, we applied an additional polyethylene shield of 5 cm thickness below the detector to reduce local effects. Air temperature and humidity were recorded with sensors attached outside the car. Three consecutive measurements underwent a moving-average filter to account for the moving footprint and to reduce the relative statistical uncertainty . The average epithermal neutron count rate across all campaigns was 9788±2141 cph prior to corrections.

All detector systems were set to integrate neutron counts over 10 s, and the driving speed ranged from 10 to 100 m min-1. Within the 2 months of the field campaign, roving measurements took place on 13 d (see Fig. ).

The neutron intensity at ground level strongly depends on the incoming cosmic-ray neutron flux. In order to account for variations of that incoming flux, researchers typically use neutron monitor (NM) recordings from the Neutron Monitor Database (http://www.nmdb.eu, last access: 11 May 2020) as a baseline. For that purpose, and recommended selecting a neutron monitor with a cutoff rigidity similar to the study location. Consequently, we suggest neutron monitors Irkutsk and Jungfraujoch as potential candidates for the reference flux:

Fendt site, Rcut=4.14±0.02 GV calculation based on, altitude 595 m;

Nevertheless, open questions about the suitability of neutron monitor data for local CRNS applications still exist, as pointed out by and . To build the basis for a more thorough analysis in future studies, we take advantage of additional neutron detector instruments during the study period in Fendt, the so-called mini-NM , and Bonner spheres . Both instruments were installed close to CRNS probe no. 15 and measure different neutron energies due to different thickness of the polyethylene shielding (PE).

Combinations of multiple Bonner spheres can be used to estimate the spectral flux distribution (i.e., the energy spectrum) of secondary cosmic-ray neutrons . The Bonner spheres are spherical 3He proportional counters (3.3 cm diameter, type SP9, Centronic Ltd.) surrounded by polyethylene shields of 3, 5, 6, 8, 9, 10, and 12 in. (7.62, 12.7, 15.24, 20.23, 22.86, 25.4, and 30.48 cm), respectively. An additional bare detector without any surrounding material was used to get a high response to thermal neutrons. To increase the response to high-energy neutrons (>20 MeV), the 9 in. detector has been modified by adding a 0.5 in. lead shell .

Figure  shows a comparison of the count rates from the remote neutron monitors at Jungfraujoch and Irkutsk, the local mini-NM, and the local Bonner spheres. Future studies will investigate the environmental factors that have an effect on the various neutron energies and which of the detectors is most suitable for the incoming neutron corrections.

As pointed out in Sect. , the estimation of volumetric soil moisture from epithermal neutron count rates typically requires a sufficient number of soil moisture and bulk density measurements in the sensor footprint. These measurements commonly originate from thermogravimetry (the retrieval of soil moisture as the mass difference of a soil core before and after oven-drying) but may as well be acquired with other measurement techniques. The more reference measurements within the footprint are available, the more reliably the effects of spatial variability of soil moisture can be accounted for in the calibration.

In this study, we combined several measurement techniques. The Rott headwater catchment study site at Fendt is partly equipped with permanent soil moisture measurement devices (SoilNet, see Sect. , and Fig. ), which provided long-term spatially and temporally dense records. For the observation period, the existing network was temporarily extended by additional wireless sensors (see Sect. ) and profile probes (see Sect. ).

In addition to these continuous observations, manual sampling was carried out in an intensive campaign from 25 to 26 June 2019 (see Sect. ). In that campaign, soil samples were extracted directly at the sensor locations from various depths for the thermogravimetric measurement of water content and other soil properties. Furthermore, vertical profiles of FDR measurements were carried out at these and additional locations.

Cosmic-ray neutron sensors are affected by all hydrogen pools within the footprint. Therefore, water stored in plants and hydrogen as a component of the plant tissue had to be quantified. The applied methods differ for grassland (more dynamic due to mowing operations) and forest (higher total biomass).

Thermal imagery was acquired covering a 60 m radius of 14 different CRNS sensors (see Fig. ) as an indicator for soil moisture distribution within the footprint area of highest contribution to the CRNS signal . The times and dates of every flight are summarized in Fig. .

For the data acquisition the UAS MK Okto XL 6S12 (HiSystems GmbH, Moormerland, Germany) equipped with a radiometric calibrated FLIR Tau 2 336 (FLIR Systems, Inc., Wilsonville, OR, USA) was used. This thermal camera uses a VOx microbolometer focal plane array which is sensitive to wavelengths from 7.5 to 13.5 µm. Manufacturers specify an accuracy of ±5 ∘C. The camera was upgraded with an external shutter system (TeAx Technology GmbH, Wilnsdorf, Germany) to achieve a better radiometric accuracy and avoid the vignetting effect in the individual scenes (Rainer Schlepphorst, personal communication, 5 May 2019). Further features are its 9 mm focal length and a sensor resolution of 336 pixels × 256 pixels, leading to a 35∘×27∘ field of view.

Flights were performed at an altitude of 100 m and a speed of approximately 5 m s-1, with a horizontal and vertical scene overlap of 82 % × 84 %. Within each area of interest, 5 to 10 ground control plates were distributed to enable accurate georectification of the geomatic products. Their positions were determined with a Leica Zeno GG04 (Leica Geosystems AG, Heerbrugg, Switzerland) DGPS antenna with subpixel accuracy.

Three steps of data preprocessing were performed before the creation of orthomosaics: for each individual scene capture, the camera creates multiple frames. The frame with the highest image quality according to the Agisoft Image Quality tool was selected for further processing. The extracted frames were temperature corrected as proposed by

7TUAScorr=TUAS-Tair+Tair_mean, with TUAScorr and TUAS being the corrected and uncorrected measured temperature by the camera, respectively, Tair the air temperature at the moment of image capture, and Tair_mean the mean of the air temperature during the flight. Air temperature of the meteorological data (see Sect. ) served as input for the correction. The orthomosaics for each flight were created in Metashape Professional (version 1.5.3, Agisoft LLC, St. Petersburg, Russia). The application of a Wallis filter increased the contrast within the individual frames and thus enabled automatic feature recognition by the image alignment algorithms. The filtered images are again substituted by the temperature-corrected frames when building the orthomosaic in mosaic mode. The root mean square error (RMSE) of the positions of the ground control plates is included in the metadata of each flight.

Discharge observations, which are, for example, needed for setting up hydrological or land surface models of the study area, were derived from water level measurements (Datalogger Type 575-II, HT-Hydrotechnik, Obergünzburg, Germany), taken at the outlet of the Rott headwater catchment in the north (see Fig. ) with 10 min resolution. The rating curve was derived from salt-tracer–electric conductivity measurements. Due to the lack of discharge reference values for channel water levels above 40 cm, discharge rates exceeding 0.1 m3 s-1 are increasingly uncertain.

For hydrological modeling and water balance assessments, information about the groundwater dynamics are also of interest. The saturated zone at the Fendt valley bottom can be differentiated into a shallow aquifer situated on top of and between the Quaternary sediment layers and a thicker, deeper confined aquifer. For each groundwater layer, a hydraulic head measurement is available during the campaign period. The observation well for the shallow aquifer is located in the vicinity of the climate station, and the one for the deeper aquifer is situated below the country road, south of the SoilNet at the center between CRNS sensor nos. 12, 18, and 24 (see Fig. ). The time series were recorded with 3-hourly (Datalogger Type 575-II, HT-Hydrotechnik, Obergünzburg, Germany) and 15 min (Hobo U20L-04, Onset, Bourne, MA, USA) resolution for the shallow and deep aquifer, respectively. As the well opening for the shallow aquifer is flush with the ground surface and located in a small depression, strong precipitation events with ponding can cause a direct water flow into the well tube. Therefore, sharp peaks should be interpreted carefully and the groundwater temperature data should be considered as indicators for such events, too. Negative peaks in the time series are due to local pumping tests.

The Cosmic Sense joint field campaign 2019 was carried out at the same time as the ScaleX 2019 campaign of KIT Campus Alpin, which involved additionally the MOSES (Modular Observation Solutions for Earth Systems) test campaign for the heat wave event chain. From these activities, complementary measurements were performed, including sensible and latent heat fluxes, net ecosystem exchange, turbulence statistics, planetary boundary layer depth, surface temperature, surface emissivity, surface thermal infrared images, vertical profiles of wind speed and direction, and water vapor and air temperature lidar profiles at Hohenpeißenberg. As of now, access to these data needs to be requested from the individual project leaders of ScaleX 2019. Further description and contact information is available at https://scalex.imk-ifu.kit.edu (last access: 11 June 2020).

Long-term hydrometeorological and ecological data for the TERENO Pre-Alpine Observatory and the Rott headwater catchment (Fendt) are available via the TERENO Data Discovery Portal (https://www.tereno.net/ddp/, last access: 27 May 2020). Further details about the TERENO Pre-Alpine Observatory are available in and via the KIT Campus Alpin's website (https://www.imk-ifu.kit.edu/tereno.php, last access: 21 May 2020).

A high-resolution digital elevation (DEM) or terrain model (DTM) can be helpful, particularly for hydrological applications, e.g., for identifying flow paths or estimating the depth of the groundwater table, or for forest biomass estimation (see Sect. ). DEM and DTM products at various horizontal resolutions and vertical accuracy levels can be obtained from the corresponding state agency, the Bayerisches Landesamt für Digitalisierung, Breitband und Vermessung (https://www.ldbv.bayern.de), e.g., the DEM1 product with a resolution of 1 m × 1 m, and a vertical accuracy better than 20 cm.

Soil maps (Bodenübersichtskarte 1 : 25 000) for Bavaria are available in vector (shapefile) format published under CC BY-3.0 license by the Bavarian Environmental Agency (LfU, https://www.lfu.bayern.de/umweltdaten/geodatendienste/pretty_downloaddienst.htm?dld=uebk25, last access: 18 July 2019).

During fieldwork and for visualization, we used OpenStreetMap data layers available via http://download.geofabrik.de (last access: 4 June 2020), namely land use, waterways, and traffic. The data are distributed under an ODbL license (https://www.openstreetmap.org/copyright, last access: 4 June 2020).

The German Weather Service (DWD) provides open climate data via https://opendata.dwd.de (last access: 3 March 2020). These include comprehensive open-access observations of climate variables at the climate station Hohenpeißenberg (identifier 2290, 977 m a.s.l.), which is located about 5 km southwest of the study area.

The heart of the data set is the dense cluster of CRNS sensors. Although an in-depth data analysis is, by definition, beyond the scope of this data publication, we would still like to give an impression of the potential of this methodology as well as of the temporal and spatial variability of soil moisture in the context of this study. In order to convey such an impression, we applied a standard processing workflow to estimate soil moisture from neutron count rates. As we only consider this an illustration, not a study result, we only briefly outline the corresponding steps in the following:

Standardize neutron count rates to a common sensitivity level, which is the sensitivity of the calibrator sensor (see Table ).

It should be noted that the above processing workflow uses only parts of the entire data set in order to roughly characterize soil moisture patterns in space and time. The scientific potential of the data set, however, is in combining the various observations at various scales. Figure  illustrates the pronounced dynamics of soil moisture as well as the spatial variability, i.e., within the catchment area and over the campaign duration. From 20 to 22 May, an unusually intense and persistent rainfall event resulted in more than 125 mm of rainfall in less than 48 h, followed by another 40 mm towards the end of May. That sequence of events led to saturated conditions in the second half of May. Over June and July, the catchment was subject to substantial drying, interrupted by occasional but intense rainfall events. The median volumetric soil moisture dropped from a maximum of 65 % to a minimum of 37 %. Over the same period, volumetric soil moisture always strongly varied in space: at the driest period around early July, the wettest 5 % of the catchment still exceeded a soil moisture of 45 %. The wettest parts of the catchment are located around CRNS sensor no. 23 (see Figs.  and ) and are characterized by peat soils and very shallow groundwater tables.

The mobile roving campaign on 27 June (Figs.  and ) brings the spatial patterns to a higher level of detail. The spatial distribution of soil moisture in the Rott headwater catchment is visible from the CRNS network, but only with the mobile measurements is it possible to cross the fields between the stationary sensors and to significantly extend spatial coverage and resolution. The campaign was conducted with the UFZ rover on a hand wagon, and the data have been aggregated on a regular 30 m grid. Neutron data have undergone basic atmospheric corrections (see Sect. ), corrections for soil properties using the SoilGrids database , and a reduction by 10 % at locations governed by forest-type land use in order to account for the biomass effect . The overall wetness gradient in the area has been confirmed, but the mobile measurements revealed a much larger wet spot in the north, which could not have been resolved by the stationary sensors due to a limited number of sensors located in that area. Besides the wet soil in the peat area and despite the downward shielding of the roving sensor, the observations are still clearly influenced by the lower water content of the roads, which are typically slightly elevated over the surrounding areas seefor a description of this effect. In prospective analyses of the data for the determination of field soil moisture, it will be necessary to correct for this road effect, e.g., by applying the procedure developed by .