Groundwater data

Figure 1 shows a map with the locations of the monitoring wells included in the dataset after quality control. The color of the data points denotes the altitude of the land surface at the location of the monitoring well. The dataset covers a substantial portion of Switzerland, with most of the monitoring wells located in the northern regions. This area, also known as the Swiss Plateau, is the most densely populated region of the country and is characterized by intensive agricultural activity. In contrast to the Alpine regions, it features extensive alluvial aquifers that are more suitable for the installation of piezometers and represent groundwater resources of greater practical relevance for regional water use. The combination of favorable hydrogeological conditions and the increased demand for groundwater monitoring—driven by higher population density and land use pressure—likely explains the denser distribution of wells in these areas. The areas where the effect of urbanization on the density of groundwater monitoring networks is particularly clear are, for example, the canton of Geneva and Basel-Stadt, as shown in the inset plots in Fig. 1.

Fig. 1Fig. 1

Locations of the monitoring wells in the database, colored by the altitude of the land surface at the monitoring well. The labels denote the different cantons of Switzerland.

The spatial distribution of the monitoring wells can be linked to the elevation of the land surface, with fewer wells at increasing altitudes in the Alpine region. This is in part the result of the hydrogeological settings of the area, which is dominated by hard and fractured rock aquifers, where it is substantially more difficult to install monitoring wells and groundwater resources are typically more localized and discontinuous. Moreover, water supply in these areas is more dependent on spring discharge, which is not (yet) part of this database. It is worth noting, however, that many monitoring wells are located in alluvial aquifers of mountain valleys (i.e., the Rhone, Inn, and Rhine valleys) in the south and east. Another important reason for the spatial data gaps is that no data was provided or that the time series were too short for some regions. For some of these regions without data, the FOEN data provides some coverage. Already, the database covers a substantial portion of Switzerland, and future updates are expected to further expand its spatial coverage, helping to close remaining gaps and achieve even broader national representation. However, a completely uniform spatial distribution of monitoring wells is unlikely, reflecting the uneven distribution of groundwater resources and Switzerland’s geological and topographic complexity.

Temporal availability

Figure 2 shows the availability of data for all 985 well locations. Each row is a time series, and the color scale indicates the number of measurements per month, where yellow denotes months with a high number of observations (e.g., daily measurements), while purple indicates months with only a small number of measurements. The first measurement in the database was made on the 15th of November in 1924, and the number of wells gradually increased over time. In particular, from the 1970s onward, the number of available measurements increased sharply, possibly related to the Federal Water Protection Act [Gewässerschutzgesetz, GSchG38], which may have led to increased monitoring efforts. The majority of the dataset thus focuses on more recent decades, and only a few time series provide longer-term information. The measurement period ranges between 5 (the minimum length requirement) and 100 years, with an average length of 28 years of data. Figure 2 shows that the measurement length is highly variable, and there are relatively few measurements before 1970.

Fig. 2Fig. 2

Data availability of the groundwater level time series. Each row is a time series, and the color denotes the temporal resolution of the data.

Throughout the measurement period, the method to measure groundwater level may have changed for individual monitoring wells (i.e., purple goes to yellow). For example, the measurements were first conducted manually and later with a pressure transducer3. As a result, the temporal resolutions of the groundwater level measurements change over time for many monitoring wells. The most common measurement frequencies are once per day, once per week, and once per month. As is common in groundwater data, many of the groundwater level time series have irregular time steps between the groundwater level measurements. These data collection results also highlight the need to deal with irregular time steps between measurements when analyzing and modeling (long-term) groundwater data. Based on this characteristic, different subsets of the data may have to be gathered from this data set for formal analyses, that is, if analysis methods require constant time steps between measurements.

A gap-filled version of the data from 1980 to 2023 is provided as well for analyses requiring equidistant time steps. Because the original data have varying frequencies between the cantons (i.e., daily, weekly, or monthly data), the gap-filling algorithm is run for each canton separately. Consequently, the imputed data is provided for each canton separately with the highest possible frequency. It is stressed here that many methods of data imputation exist, and the results of subsequent analyses using the gap-filled data will depend on the method chosen. The imputation performance metrics for the gap-filling models generally show good agreement with the observed data across cantons (i.e., the median NSE over all wells was above 0.75). Nonetheless, we strongly suggest that, where possible, for example, for tasks such as model evaluation and process-based analyses, the original measurements rather than the gap-filled data are used.

Environmental and climatic descriptors

Figure 3 shows overview maps for six of the environmental and climatic descriptors that were collected. For the climate descriptors, Fig. 3 shows the average temperature (Fig. 3a), the annual precipitation sum (Fig. 3b), the annual potential evaporation (Fig. 3c), and the Aridity Index (Fig. 3d), all calculated for the 30-year period 1993-2023. The data shown in these figures indicates that there is some variability in the data, particularly from north to south. The largest amount of precipitation is measured in the Alps, which also coincides with lower temperatures and potential evaporation. Based on the Aridity Index (AI), all of Switzerland is classified as humid. Nonetheless, spatial variability is evident, with higher AI values in inner-Alpine valleys and southern regions and lower values across the Swiss Plateau and northern Alpine foreland, reflecting strong topographic and climatic gradients.

Fig. 3Fig. 3

Maps with examples of various environmental and climatic descriptors for the study area.

For environmental descriptors, Fig. 3 shows the depth to the water table (DTW, Fig. 3e) and the distance to the river (Fig. 3f). For the largest portion of the wells, the depth to the water table is relatively small, on the order of meters up to twenty meters. Only a small percentage (~17%) has a depth to the water table exceeding 20 meters. Similarly, most monitoring wells are located within a few kilometers of a river—an expected pattern, as many are situated in unconsolidated alluvial aquifers that typically develop along river corridors and are hydraulically connected to surface water systems.

Groundwater dynamics

For each of the 985 groundwater level time series, 27 groundwater signatures were computed to characterize groundwater behavior at each monitoring well. Figure 4 shows examples of the values of two groundwater signatures. The plot on the left shows the autocorrelation time, the number of days it takes the autocorrelation function to drop below 0.9. This is interpreted as a measure of the memory of the system [e.g.39]. The larger the value (i.e., number of days), the larger the memory. Higher values (blue) indicate slower, more persistent groundwater responses, while lower values (yellow) reflect more dynamic systems. A spatial pattern emerges, with longer autocorrelation times predominantly found in the northern lowland regions, particularly the Swiss Plateau, where larger unconsolidated aquifers are common. In contrast, shorter autocorrelation times are more frequent in alpine and southern areas, likely reflecting more dynamic groundwater systems with faster responses to recharge events.

Fig. 4Fig. 4

Examples of two of the groundwater signatures for the monitoring wells. The left plot shows the autocorrelation time, and the right plot shows the average day of the year with the maximum groundwater level.

The plot on the right shows the average day of the year with the maximum groundwater level having a clear spatial pattern. Particularly in the northeastern part of Switzerland, the groundwater level peaks early in the year (i.e., January to March), while towards the Alps, the groundwater level peaks later (April-June). These patterns can be partially linked to the dominant recharge processes, where winter precipitation in the Alps falls more as snow and melts in spring, while in the north, the winter precipitation infiltrates and causes an earlier rise and subsequent peaking40. Another probable reason is the influence of the rivers that also peak in spring. Future research can further investigate the relationships between groundwater behavior and descriptors, such as those shown above.