WegenerNet high-resolution weather and climate data 2007 to 2019

This paper describes the latest reprocessed data record (version 7.1) over 2007 to 2019 from the WegenerNet climate station networks, which since 2007 provide measurements with very high spatial and temporal resolution of hydrometeorological variables for two regions in the state of Styria, southeastern Austria: 1) the WegenerNet Feldbach Region, in the Alpine forelands of southeastern Styria, which extends over an area of about 22 km × 16 km and comprises 155 meteorological stations placed on a tightly spaced grid, with an average spatial density of one station per ∼2 km and a temporal sampling 5 of 5 min; and 2) the WegenerNet Johnsbachtal, which is a smaller "sister network" of the WegenerNet Feldbach Region in the mountainous Alpine region of upper Styria that extends over an area of about 16 km × 17 km and comprises 13 meteorological stations and one hydrographic station, at altitudes ranging from below 600 m to over 2100 m and with a temporal sampling of 10 min. These networks operate on a long-term basis and continuously provide quality-controlled station time series for a multitude of hydrometeorological near-surface and surface variables, including air temperature, relative humidity, 10 precipitation, wind speed and direction, wind gust speed and direction, soil moisture, soil temperature, and others like pressure and radiation variables at a few reference stations. In addition, gridded data are available at a resolution of 200 m × 200 m for air temperature, relative humidity, precipitation and heat index, for the Feldbach Region, and at a resolution of 100 m× 100 m for the wind parameters for both regions. Here we describe this dataset (the most recent reprocessing version 7.1), in terms of the measurement site and station characteristics as well as the data processing from raw data (level 0) via quality-controlled 15 basic station data (level 1) to weather and climate data products (level 2). In order to showcase the practical utility of the data we also include two illustrative example applications and briefly summarize and refer to scientific uses in a range of previous studies. The dataset is published as part of the University of Graz Wegener Center’s WegenerNet data repository under the DOI https://doi.org/10.25364/WEGC/WPS7.1:2020.1 (Fuchsberger et al., 2020) and is continuously extended.

1 Introduction 20 While climate model simulations can achieve kilometer-scale resolution for both general circulation models (e.g., Miyamoto et al., 2013;Klocke et al., 2017) and regional climate models (eg., Prein et al., 2015;Kendon et al., 2017;Leutwyler et al., 2017;Fuhrer et al., 2018), there is a lack of ground observation data for verifying model outputs and studying weather and climate at this resolution. Common meteorological networks cover scales of 10 km or larger [e.g. interstation distance in Switzerland (of MeteoSwiss-operated rain gauges) is 10-15 km (Wüest et al., 2009), and in Germany the Deutsche Wetterdienst (DWD) were incorporated into the WPS data processing. Over several years, stations from the other providers in the region were gradually added to the network as well, once both the data exchange contracts and the long-term access to the actual data had been finalised. A complete list of the station characteristics can be found in Table 2.
The instrumentation differs from station to station and includes sensors for air temperature, relative humidity, precipitation, wind parameters, snow parameters, radiation parameters, air pressure, and hydrographic parameters. A list of all measured 130 parameters at each station, together with the sensor heights, can be found in Table 3. Some of the data are available from 2007, and others only from subsequent years (depending on the individual station's construction dates). The sampling interval is 10 min at all stations, except 5 min at all WEGC stations (as of Oct. 2019, when the interval was changed in connection with an upgrade of the data logger hardware) and AHYD stations. At the NPG stations, the sampling interval was 30 min between 2007 and 2014. See Table A2 for a complete list of all installed sensors including their record period, height, and sampling 135 interval. Just as for the WegenerNet FBR, we do not give the exact day of installation in this table, but only the year. For a detailed overview, which is timed down to the hour, please refer to the Station Data → Stations → Sensor Details and Station Data → Download → Sensor list CSV file sections at the WegenerNet data portal (www.wegenernet.org).

Data processing and monitoring
The acquisition, processing, and visualization of the stations' data are conducted automatically by the WegenerNet Processing 140 System (WPS), that was originally introduced by Kabas (2012) and Kirchengast et al. (2014). Each release of the WPS is assigned a new version number in order that the data produced with different releases can be archived and reproduced if needed. The data described in this paper were generated using version 7.1 of the WPS. Table 4 shows an overview of the steps involved in the processing and defines their output products. The WPS consists of four main parts, which are described in the following sections. Except for the proprietary software provided by the data logger manufacturers, the WPS was developed 145 entirely by the WEGC, using Python as the primary programming language and PHP and JavaScript for some parts where using Python was not feasible.

Level 0 processor: Command Receive Archiving System (CRAS)
In the first processing step, raw data generated by the stations' data loggers are received and stored into a database by the CRAS. 150 In order to achieve this, the following tasks are executed: 1. The data loggers in the field collect measurements at the sampling interval specified in Tables A1 and A2 Another part of this task is the automatic surveillance of all heated precipitation gauges. These gauges have additional hardware such as temperature and heating-fan rotation sensors to monitor the proper functioning of the heating system. In case of a detected malfunction, the corresponding sensor is automatically marked as not-heated in the database, and the WegenerNet team is informed per email and can thus initiate immediate actions to remedy the problem. 165 The last part, humidity-sensor problem detection, focuses on the issue that humidity sensors are known to be the most errorprone in the network. If the sensor surface is contaminated by dust and dirt, a specific faulty behavior shows a negative offset at high humidity levels, which gets worse as contamination increases, and finally results in an inverse measurement of the values.
This behavior and its detection are described in detail in Scheidl et al. (2017). The algorithms presented therein have been implemented in the WPS, which enables the early detection of problematic humidity sensors and the reliable posterior flagging 170 of bad sensors.

Level 1 processor: Quality Control System (QCS)
The QCS checks the data for their technical and physical plausibility, and flags all values that did not pass a certain check. It is run hourly, and processes all data that are new since the last run. The process analyzes single measurement values and up to three-hourly periods thereof. 175 Each release of the QCS is assigned a new version number in order that the data produced with different releases can be archived and reproduced if needed. The data described in this paper were generated using version 5 of the QCS, generating level 1 (L1) data version 5. The QCS consists of seven main processing steps (called QCS layers), which are described in detail in the following paragraphs. with n being the number of the respective QCS layer. Thus a QF in layer 0 gets an integer value of 2 0 = 1 = binary 0000 0001, layer 1 gets 2 1 = 2 = binary 0000 0010, layer 2 gets 2 2 = 4 = binary 0000 0100, etc. An example for a combination of 185 QFs would be 0111 0000 (integer 2 4 + 2 5 + 2 6 = 112), which would result from failed QCS checks in layers 4, 5, and 6.
If there are insufficient data for executing a certain rule, the respective data values are marked by a so-called no-ref flag instead of a QF. Finally, after a layer has been processed, a checked flag is set. All flags follow the same encoding as described above. The output of the QCS, the L1 data, consists of the L0 data values plus quality flags, no-ref flags, and checked flags.
All checks are executed for each station of the network and for each sensor that is mounted on the station.

190
-Layer 0 (operations check) checks if a certain station or sensor is in operation (status set manually by the maintenance crew) and sets QF (0) if not.
-Layer 1 (availability check) checks if data from a certain sensor are available and sets QF (1) if not.
- precipitation measurements at an hourly mean temperature below 2 • C. Flagging all potential snowfall at unheated stations, it has major implications for the precipitation data in winter, since it reduces the effective number of precipitation gauges from 155 to 14 at temperatures below 2 • C.
The second rule (reference-precip check) checks for consistency between the measurements of the three precipitation gauges installed at the reference station. In the third to sixth rules (windsensor-, wind-mean-boe-, windvalues-, and 210 windboe-dir check), wind data are screened for inconsistencies. For example, the data must satisfy (gust value > speed value) and (gust value and direction = speed value and direction). If any of the checks fail, QF (5) is set.
-Layer 6 (interstational check) deals with the consistency of measurements between a station and its neighbors.  Scheidl et al. (2017). Further details on the other rules can be found in Scheidl (2014).

225
If any of the checks fail, QF (6) is set.
This layer is applied to WegenerNet FBR data only, because the WegenerNet JBT station density is too low for the interstational checks to run reliably.
-Layer 7 (external reference check) compares net radiation and humidity data between ZAMG and WegenerNet FBR stations. If the data deviate too much, QF (7) is set.

230
This layer is applied to WegenerNet FBR data only, because ZAMG stations near the WegenerNet JBT region are too distant from the other stations.

Temperature spike check
Since some of the EE08-05 temperature sensors that were installed in the WegenerNet FBR in 2018 have been found to produce unnatural spikes in the measurement data, a new check was implemented to detect these spikes. This temperature spike check 235 compares the spike magnitude T s to the derivative of the (non-flagged) temperature values of the last 120 minutes before the spike occurred. Fig. 2 shows an example of such spikes and those detected by the spike check.
The check flags any temperature value T s,i if where T s is the spike magnitude, and T j the (non-flagged) temperature values of the last 120 minutes (nominally 24 data 240 values).
Note that this check is only executed for sensors that are manually marked as problem sensors by the WegenerNet team. The DPG processes all unflagged L1 data (i.e. the best-quality data) and interpolates the gaps resulting from flagged time steps using three different interpolation schemes. It also generates gridded fields of temperature, humidity, and precipitation data for 245 the WegenerNet FBR. The DPG is run hourly, after the QCS has finished writing the L1 data.
Each release of the DPG is assigned a new version number in order that the data produced with different releases can be archived and reproduced if needed. The data described in this paper were generated using DPG version 7, generating L2 basis data version 7. The term basis data refers to the temporal resolution (base period) of the L2 data, which is 5 min for the WegenerNet FBR and 10 min for the WegenerNet JBT. Level 2 basis data are generated in near-real time with a data latency 250 of less than 2 hours between measurement and storage into the database.
The DPG consists of six main processing steps described in detail below.
1. On reading the L1 data, the DPG applies homogenization factors for temperature and humidity data according to Ebner  2. In the second processing step, data with a higher temporal resolution than the region's base period is resampled, generally by averaging (except summation for precipitation, vector mean for wind speed and maximum for peak gust).
3. The next step is the interpolation of the data gaps. The type of interpolation depends on the length of the interpolation pe-260 riod and the measured parameter; if a gap is shorter than the value defined in Table 5, the missing values are interpolated linearly between the two adjacent valid data values and a Data Product Flag (DP-Flag) of 1 is set for the interpolated time steps (see Table 6 for definition of DP-Flags).
Due to the high station density of the WegenerNet FBR, interpolation can also be carried out spatially for the main parameters air temperature, humidity, and precipitation, which are measured at all stations. If a gap is longer than defined 265 in Table 5, the values for temperature and humidity are interpolated from the values of the surrounding stations, using a linear inverse distance weighting (IDW) algorithm. The temperature values are projected to a reference altitude of 300 m before interpolation, using a lapse rate calculated from all stations' temperature data and their altitudes for every 1 h time window. Precipitation data are interpolated using an inverse distance squared weighting (IDSW) algorithm. Further, a DP-Flag of 2 is set for all neighbor-station-interpolated data. Additionally, the spatial maximum, minimum, average, and standard deviation are calculated for all gridded fields. Preprint. Discussion started: 3 November 2020 c Author(s) 2020. CC BY 4.0 License.
Temperature grids are stored in three separate fields, one calculated at a constant reference altitude of 300 m, and two terrain-following fields. For these two fields, one is calculated at the average terrain altitude of each grid point, and one at the altitude of the center of each grid point. In this context a digital elevation model (DEM) with 10 m resolution is used for calculating the altitudes. The altitude dependence of the temperature is accounted for by using the calculated lapse rate as described in step 3.
The DP-Flags of the fields are also stored as 2D-grids, serving as spatially resolved quality indicators. The DP-Flag 280 values of each station are interpolated onto the grid using the same interpolation algorithm as for the corresponding measurement values. Additionally, the spatial average of each DP-Flag grid is calculated. Two examples of gridded DP-Flag data can be found in (Fuchsberger et al., 2018, Fig. 2.1 therein).
5. WegenerNet FBR air temperature, humidity, and precipitation data that could not be interpolated from the surrounding stations in step 3 are then interpolated from the gridded data calculated in step 4. Temperature data are interpolated from 285 the reference altitude (300 m) grid and then projected back to the stations' altitudes using the lapse rate calculated in step 3. Furthermore, a DP-Flag of 3 is set for the interpolated time steps.
6. If a gap could not be interpolated, the missing time steps are filled with an empty value (not a number, NaN) and are marked by a DP-Flag of 4.

290
The L2+ processor generates data derived from several input variables, including one or several L2 data variables and other sources like soil texture data, land use data, etc. It is run hourly, after the DPG processing has finished.

Wind Product Generator (WPG)
The WPG generates high-resolution 100 m × 100 m wind fields and peak gust fields for the WegenerNet FBR and WegenerNet JBT, using the California Meteorological Model (CALMET) as a core tool. It is run hourly, after the DPG processing has 295 finished, and was derived in recent years by Schlager et al. (2017Schlager et al. ( , 2018. Input variables of the model are meteorological observations, terrain elevations taken from a 100 m resolution DEM, and land use information. Meteorological variables include temperature, air pressure, wind speed and wind direction, all taken from WegenerNet station observations. The wind fields are generated for two height levels (10 m and 50 m above ground) with a temporal resolution of 30 minutes.
Gust fields are generated for a height of 10 m above ground in a separate process described in Schlager et al. (2018, Sect. 3 Similarly to the QCS and DPG, the releases of the WPG are versioned for each new release. The version used for generating the described data is WPG v7.1.

Value-Added Product Generator (VAPG)
The VAPG generates data derived from L2 data and complementary data sources. It currently contains two main processing 310 steps: Soil moisture data generation Soil moisture time series data are derived from L2 pF value data and soil-related metadata like soil texture, humus content, and dry bulk density, using the Mualem-Van Genuchten equation (Mualem, 1976;Van Genuchten, 1980). A detailed description of this process can be found in Fuchsberger and Kirchengast (2013).

315
The pF values were the only source of information about soil moisture in the WegenerNet FBR until 2013, when replacement of pF-meters by Stevens HydraProbe II sensors (capable of measuring soil moisture directly) began. Since then, pF-meters at all soil stations have been replaced or supplemented by HydraProbe II sensors (see Table A1 for details on the sensor mount dates). For keeping a cross-comparison capacity, stations 27 and 77 have both types of sensors installed.
Heat index data generation 320 Gridded heat index (apparent temperature) fields are generated from L2 temperature and humidity fields using an equation developed by Schoen (2005), where HI is the heat index ( • C), T the air temperature ( • C), and D the dewpoint ( • C).
The dewpoint therein is calculated using an equation based on the Magnus-Tetens formula (Barenbrug, 1974;Schoen, 2005), a = 17.27, b = 237.3, and RH is the relative humidity expressed as a (dimensionless) fraction between 0 and 1.

330
As an auxiliary classification, for user convenience, the heat index is also categorized into a scheme of five danger classes (based on NOAA, 2020), consisting of comfortable (20 • C < HI < 27 • C), caution (27 • C < HI < 32 • C), extreme caution (32 • C < HI < 41 • C), danger (41 • C < HI < 54 • C), and extreme danger (HI > 54 • C). All L2 data, including value-added data products, are aggregated to half-hourly, hourly, and daily (weather data), and monthly, seasonal, and annual (climate data) products. The aggregation is generally done by averaging the data, with the exception of wind speed and direction (the vector mean is used for consistent averaging), peak gust data (the maximum value is used), and precipitation data (the sum is taken). For temperature data, the maximum and minimum values are also calculated. Weather data are generated by aggregating the basis data, and climate data are generated by aggregating the daily data.

Weather and climate data quality indicators
A special algorithm is used for calculating quality indicators for weather and climate data, the so-called flagged percentage of data (FP ). The FP indicates how much of the data has been interpolated or flagged missing in the DPG, but it does not indicate how the data have been interpolated (i.e. which DP-Flag they have received). It is calculated for station time series data using the following equations (taken from Fuchsberger et al., 2018). 345 a) for continuous data, like temperature and humidity: where N [DP-Flag > 0] is the number of flagged basis data values, and N total is the total number of basis data values within the given timespan of a weather or climate data product.
b) for precipitation data: where P f lagged is the precipitation sum over all flagged basis data values, i.e.
and P total is the total precipitation sum for a given timespan (T ) of a weather or climate data product, i.e.

355
where index i runs over the timespan T up to the final value N T .
In order to ensure that only "wet data" (data with a reasonable minimum precipitation) are flagged, FP P is set to zero if the flagged precipitation sum is 0, or the maximum precipitation is below a certain low limit, i.e., if P f lagged ≡ 0 or max(P i ) < 0.21 mm, then Weather and climate data quality grids likewise use the flagged percentage of data FP , as described above. However, instead of a calculation per station, FP is now calculated per grid point. The same equations (Eq. 5 to 8) are used for this purpose, only now with the threshold DP-Flag > 0.5 instead of DP-Flag > 0 in Eqs. 5 and 7, in order to allow moderate deviations from zero for these interpolated grid point DP-Flag values.
Since flagging on the grid also applies to "wet data" only, FP P is set to zero if the areal mean precipitation is below a certain 365 low limit, i.e., if P total (x, y) < 0.21 mm then FP P = 0 , where P total (x, y) denotes the average precipitation over all grid points (x, y).
Examples showing gridded FP -fields and the associated data for precipitation can be found in Fuchsberger et al. (2018, Figs. 2.2 and 2.3 therein).

Data products and auxiliary data
This section gives an overview of the data products generated by the WPS and shows details of their format. As explained in Sect. 4, all data are available in four processing levels: Raw data (L0 data), quality controlled raw data (L1 data), quality controlled and interpolated data (L2 data), and data derived from several input variables (L2+ data). The latter two (L2 and L2+ data) include time-aggregated (from half-hourly to annual) weather and climate data products.

375
The parameters available in L0, L1, and L2 are shown in Table 1 (for the WegenerNet FBR) and Data are available both as time series in CSV format (see Table 7 for a description of the CSV format) and, for L2 and L2+ 380 only, as gridded data in NetCDF format (see Tables A3 and A4 for a list of variables available in the NetCDF files). from basic level 2 data and auxiliary data or models, by using functional dependencies or modeling (see also Sect. 4.4).

Level
The L2 and L2+ time series data can be plotted in the data portal and downloaded as CSV files. The data are available in basis data resolution (5 minutes for WegenerNet FBR and 10 minutes for WegenerNet JBT), and time-aggregated as halfhourly, hourly, daily, monthly, seasonal, and annual data.
Level 2 and 2+ gridded data products are available in NetCDF format. See Table A3 for a list of all parameters stored in 395 the NetCDF files. The data are available in a base temporal resolution of 5 minutes for WegenerNet FBR and 10 minutes for WegenerNet JBT, and time-aggregated as half-hourly, hourly, daily, monthly, seasonal, and annual data. The spatial resolution of the grids is 200 m × 200 m.
Gridded wind field data products are available in NetCDF format and can be downloaded via the data portal. See Table A4 for a list of all parameters stored in the wind NetCDF files. The data are available in a base temporal resolution of 30 minutes, 400 and time-aggregated as hourly, daily, monthly, seasonal, and annual data. The spatial resolution of the grids is 100 m × 100 m.

Auxiliary data
In addition to the meteorological data described above, a growing number of auxiliary data are available for the WegenerNet and can be downloaded via the data portal. These data currently include a Digital Elevation Model (DEM), landuse/landcover data, and hydro-pedological soil characteristics, all provided by state offices of the regional government of Styria, Austria. Landuse/landcover data are available at a horizontal resolution of 100 m and cover the WegenerNet FBR and its surroundings. They originate from a project to classify the hydro-pedological characteristics of the Raab valley and southeastern Styria 410 (Klebinder et al., 2017). Additional data from this project are available, but due to copyright reasons they must be requested from the WEGC. They include: soil type (content of silt, clay, and sand), saturated hydraulic conductivity k sat , total pore volume, air capacity, permanent wilting point, available water capacity, the Mualem-Van Genuchten parameters (θ r , θ s , α, and n), runoff coefficients, and soil moisture distribution.
6 Example applications 415 We decided to show just two "arbitrary" examples for illustration, out of the many possible uses of the WegenerNet dataset.
Section 6.1 presents a multi-variable view of meteorological data for a storm event caused by a midlatitude cyclone in 2017, and Sect. 6.2 presents high-resolution gridded precipitation and temperature data for a strong convective event. We note that the WegenerNet data, including predecessors of the version 7.1 dataset, have been applied in a wide variety of scientific uses so far, such as the studies summarized in Sect. 2 above.

Outlook
In 2020, the WegenerNet FBR network will be upgraded with three major new observing components, expanding it from a 2D ground station hydrometeorological network into a 3D open-air laboratory for climate change research at very high resolution.

455
The following new atmospheric 3D observation components, as shown in a most recent WegenerNet FBR instrumentation map in Fig. 6, have started or will soon start operations during 2020: a) A polarimetric X-band Doppler weather radar for studying precipitation parameters in the troposphere above the ground network, such as rain rate, hydrometeor classification, Doppler velocity, and approximate drop size distribution and number. It can provide 3D volume data (at about 1 km × 1 km horizontal and 500 m vertical resolution, and 2.5-min The WegenerNet JBT network will be expanded with a new station in the Enns valley, measuring air temperature, humidity, 475 precipitation, radiation, and wind. This will allow more accurate wind modeling in the WegenerNet JBT area in particular, following a recommendation by Schlager et al. (2018Schlager et al. ( , 2019 based on their JBT wind field quality evaluations.
Data from these new components will be made available via the WegenerNet data portal, and a description of the components is planned for a future paper.
Regarding improvements to the processing system, a version of the QCS which is capable of processing daily time periods This will lead to improvements of the quality of precipitation data for cases like blocked gauges or melting snow, where the QCS algorithms gain additional information from the larger sample size.
Another future improvement will be the implementation of a new algorithm for the calculation of temperature lapse-rates in the DPG (see also step 3 of Sect. 4.3) which was developed in the course of a master's thesis (Hocking, 2020). per year), expanding the time range documented in this paper. The DOIs will be made available at the data portal's entry page upon release, and can be used together with this paper to properly cite the WegenerNet dataset.

Conclusions
This paper provided an overview of the current state of the WegenerNet climate station networks Feldbach Region (FBR) and Johnsbachtal (JBT), their processing system, and their data products. It is the first major update to a WegenerNet introduction Author contributions. J.F. and G.K. designed the paper. J.F. wrote the first draft, is maintaining and continuously improving the WegenerNet processing system, which is used to generate the data, and prepared the dataset publication. G.K. advised the publication in all aspects, 500 provides advancement ideas and guides the overall concept and design of the WegenerNet, and has substantially contributed to the manuscript text. T.K. supported the data and information collection for the manuscript, advised on application aspects, and provided feedback and comments to the paper.
Competing interests. The authors declare that they have no conflict of interest.
processing system and advising on the data portal. Sincere thanks also go to Robert Galovic, Albert and Alois Neuwirth, Andreas Pilz, Daniel Scheidl, Heimo Truhetz, and all previous team members for their various valuable contributions to the project since its inception by 2005. Special thanks to Thomas Hocking, who also provided some proofreading and valuable comments on the paper.

510
WegenerNet funding is provided by the Austrian Ministry for Science and Research, the University of Graz, the state of Styria (which also included European regional development funds), and the city of Graz; detailed information can be found online (www.wegcenter.at/wegenernet, last access: 4 September 2020).
18 https://doi.org/10.5194/essd-2020-302              Figure 6. The WegenerNet Feldbach Region (FBR) 3D Open-Air Laboratory for Climate Change Research, consisting of the ground station network as described in this paper, and additional 3D observing system components including an X-band precipitation radar, a tropospheric profiling and cloud structure radiometer pair, and a GNSS station network called GNSS-StarNet, comprising 6 GNSS receivers for water vapor sensing. The legend on the right explains all map characteristics. b Soil parameters include soil temperature, soil moisture, pF value, and electric conductivity; see Table A1 for details on installed sensors and Sect. 4 for details on the conversion from pF value to soil moisture. c Depth of soil sensors has changed from -0.3 m to -0.2 m over the years; see Table A1 for details. d Station 44 is a silo rooftop station in the Raab valley measuring temperature and relative humidity at a height of 53 m, precipitation at 51 m and wind parameters at 55 m. e "Solid precipitation" indicates that the corresponding stations are equipped with heated rain gauges, and can therefore measure snow and other forms of frozen precipitation in addition to liquid precipitation. Note that for reasons of simplicity liquid precipitation is referred to as just "precipitation" in this table. f Wind parameters comprise speed, direction, gust speed, and gust direction.  Preprint. Discussion started: 3 November 2020 c Author(s) 2020. CC BY 4.0 License.  3 3  3  3  3  2  2  2  3 -------503  4  4  2  10  10  10  10  2  2  2  -4  ------504  3  3  -6  6  6  6  3  -3  --------505  3  3  -6  6  6  6  3  -3  ----- .5 -5.5 --5.5 --a T air temperature, rh relative humidity, P precipitation, v wind speed, φ wind direction, vg peak gust, φg peak gust direction, Qg global radiation, Qr reflected radiation, Qn net radiation, p air pressure, sd snow depth, swe snow water equivalent, Tsn snow temperature, Ts surface temperature, wl water level, Qw water discharge, and  Preprint. Discussion started: 3 November 2020 c Author(s) 2020. CC BY 4.0 License. Preprint. Discussion started: 3 November 2020 c Author(s) 2020. CC BY 4.0 License.     Precipitation parameters: Preprint. Discussion started: 3 November 2020 c Author(s) 2020. CC BY 4.0 License. Preprint. Discussion started: 3 November 2020 c Author(s) 2020. CC BY 4.0 License.