Luzi Bernhard

Stream water temperatures influence water quality, with effects on aquatic biodiversity, drinking-water provision, electricity production, agriculture, and recreation. Therefore, stakeholders would benefit from an operational forecasting service that would support timely action. Deep-learning models are well-suited to providing probabilistic forecasts at individual stations of a monitoring network. Here, we train and evaluate several state-of-the-art models using 10 years of data from 54 stations across Switzerland. Static catchment features, time of the year, meteorological observations from the past 64 d, and their ensemble forecasts for the following 32 d are included as predictors in the models to estimate daily maximum water temperature over the next 32 d. Results show that the temporal fusion transformer (TFT) model performs best, with a continuous rank probability score (CRPS) of 0.70 °C averaged over all lead times, stations, and 90 forecasts distributed over 1 year. The TFT is followed by the recurrent neural network encoder–decoder, with a CRPS of 0.74 °C, and the neural hierarchical interpolation for time series, with a CRPS of 0.75 °C. These deep-learning models outperform other simpler models trained at each station: random forest (CRPS = 0.80 °C), multi-layer perceptron neural network (CRPS = 0.81 °C), and autoregressive linear model (CRPS = 0.96 °C). The average CRPS of the TFT degrades from 0.38 °C at lead a time of 1 d to 0.90 °C at a lead time of 32 d, largely driven by the uncertainty of the meteorological ensemble forecasts. In addition, TFT water temperature predictions at new and ungauged stations outperform those from the other models. When analyzing the importance of model inputs, we find a dominant role of observed water temperature and future air temperature, while including precipitation and time of the year further improves predictive skill. Operational probabilistic forecasts of daily maximum water temperature are generated twice per week with our TFT model and are publicly available at https://www.drought.ch/de/impakt-vorhersagen-malefix/wassertemperatur-prognosen/ (last access: 20 March 2025). Overall, this study provides insights into the extended-range predictability of stream water temperature and into the applicability of deep-learning models in hydrology.

Medium to subseasonal hydrological forecasts contain more information relevant to water and environmental management tasks than climatological forecasts. However, extracting this information at the most appropriate level of accuracy and spatiotemporal resolution remains a difficulty. Many studies show that the skill of the extended range forecasts with daily resolution tends toward zero after 7-14 days for small mountainous catchments. Beyond that forecast horizon the application of highly sophisticated pre- and postprocessing methods generally produce limited gains. Consequently, current forecasting techniques cannot effectively represent forecast extremes at extended ranges such as anomalously high and low runoff or soil moisture. To tackle these deficiencies, this study analyzes the value of tercile forecasts for weekly aggregates of runoff and soil moisture forecasts available at a daily resolution for Switzerland. The forecasts are classified into three categories: below, above, and normal conditions, which are derived from long-term simulations and correspond approximately to climatological conditions. To achieve improved reliability and skill of the predicted tercile probabilities, a nonparametric probabilistic classification method has been tested. It is based on Gaussian process (GP), which is attractive in machine learning (ML) applications because of its ability to estimate the predictive uncertainty. The outcome of these postprocessed forecasts was compared to preprocessing methods where the meteorological predictions are statistically corrected before passing to the hydrological model. Our results indicate that tercile forecasts of weekly aggregates produce a suitable skill up to 3 weeks lead time using the preprocessed input and up to 4 weeks lead time using the GP method.

There is a growing need for reliable medium to extended range hydrological forecasts in water and environmental management (e.g. hydro-power and agricultural production). The objective of this paper is a first assessment of the skill of hydrological forecasts based on Numerical Weather Predictions (NWPs) in comparison to the skill of forecasts based on climatology for monthly forecasts with daily resolutions and to identify possibilities of improvement by post-processing the hydrological forecasts. Various hydrological relevant model variables, such as the surface and subsurface runoff and the soil water content, will be analysed for entire Switzerland. The spatially aggregated predictions of these variables are compared to daily simulations and to long-term daily averages of simulations driven by meteorological observations (i.e. climatology). Besides this comparison of forecasts with simulations for model variables without direct measurements available, the skill of the monthly stream-flow forecasts is estimated at four catchments with discharge measurements. Additionally post-processing methods have been applied to remove bias and dispersion errors and to estimate the predictive uncertainty of the stream-flow. Some results of various verification measures like variants of the Geometric Mean for ratios of spatial aggregates and the Continuous Rank Probability Skill Score (CRPSS) will be shown. Apart from the indication of a strong diversity of upper limits of the forecast skill depending on catchment characteristics, the results of NWPs are generally superior to climatological predictions and could be applied gainfully for various kinds of long-term water management planning.

Eine Studie zur Wasserverfügbarkeit unter verschiedenen Klimaszenarien. Das Engadin zählt trotz der vielerorts sichtbaren Wasservorkommen zu den trockensten Gebieten der Schweiz. In niederschlagsarmen Sommern kann es in einigen Teileinzugsgebieten bereits heute zu Engpässen im Wasserangebot kommen. Zukünftige Klimaänderungen könnten bestehende Herausforderungen in der Bewirtschaftung von Wasserressourcen weiter verschärfen.

Hydrological classification schemes are important tools for assessing the impacts of a changing climate on the hydrology of a region. In this paper, we present bivariate mapping as a simple means of classifying hydrological data for a quantitative and qualitative assessment of temporal change. Bivariate mapping consists of classifying map objects into discrete classes based on the values of two variables. We demonstrate the application of bivariate mapping to distributed hydro-climatic model outputs for the whole of Switzerland with a cell size of 200 m and compared the resulting bivariate maps with an existing classification of Swiss river regimes. The effects of scale were investigated by comparing these raster maps with a map showing the same variables aggregated to sub-basins with a mean area of 40 km². Finally, maps of the current state were compared with predictions for future periods based on various model chains and greenhouse gas emission scenarios. For the map comparisons, four measures of association and two measures of agreement were used.
Of all the variable pairs tested, a bivariate map combining runoff and snowmelt contribution to runoff obtained the highest similarity scores with the map of river regimes, which suggests a strong link between the combination of these variables and intra-annual streamflow variations. Also, this classification offers new insights, as it includes absolute values of runoff, which are often ignored in classification schemes. Comparing current-state maps with future predictions indicated that the magnitude of change is reflected in the patterns of bivariate maps, with lower agreement scores for predictions further away in time or when higher greenhouse gas emissions are assumed. Furthermore, a visualization of the spatial distribution of agreement scores allows a qualitative assessment of the magnitude of change for different regions, and an analysis of the differences in spatial patterns of predictions. based on different model chains or emission scenarios.

To evaluate how summer low flows and droughts are affected by the winter snowpack, a Tri-National effort will analyse data from three catchments: Alpbach (Prealps, central Switzerland), Gudjaretis-Tskali (Little Caucasus, central Georgia), and Kamenice (Jizera Mountains, northern Czech Republic). Two GIS-based rainfall-runoff models will simulate over 10 years of runoff in streams based on rain and snowfall measurements, and further meteorological variables. The models use information on the geographical settings of the catchments together with knowledge of the hydrological processes of runoff generation from rainfall, looking particularly at the relationship between spring snowmelt and summer droughts. These processes include snow accumulation and melt, evapotranspiration, groundwater recharge in spring that contributes to (the) summer runoff, and will be studied by means of the environmental isotopes ¹⁸O and ²H. Knowledge about the isotopic composition of the different water sources will allow to identify the flow paths and estimate the residence time of snow meltwater in the subsurface and its contribution to the stream. The application of the models in different nested or neighbouring catchments will explore their potential for further development and allow a better early prediction of low-flow periods in various mountainous zones across Europe. The paper presents the planned activities including a first analysis of already available dataset of environmental isotopes, discharge, snow water equivalent and modelling experiments of the (already) available datasets.

For most forest types in the European Alps, little is known about mixture effects on stand productivity. The comparability of studies on mixture effects often suffers from the open methodological question of whether the results depend on the definition of compositional proportion. In this study, data from the Swiss National Forest Inventory were used to investigate how the growth of Norway spruce (Picea abies (L.) Karst.) and silver fir (Abies alba Mill.) is modified by the admixture of the other species and if the mixture effect depends on site, climate, age, or stand density. Stocking proportion (proportion by area) as well as the proportion of relative density index, stem number, basal area, stem volume, and aboveground biomass were used to define compositional proportion, and the results were compared. At low-quality sites, Norway spruce grew faster in basal area as its relative share of composition increased, but this pattern diminished as the site quality increased. At cooler sites, silver fir grew faster as its share of composition decreased, but the pattern reversed at warmer sites. Overyielding was predicted only for 16% of the 679 sites used for this study. Beneficial effects of species mixture were overestimated when species-specific stocking potentials were not considered in the definition of compositional proportion.

– For all greenhouse gas scenarios, a shift in runoff regime type is projected to occur in most catchments in Switzerland, with lower summer runoff and higher winter runoff, but little change in the total annual volume.

– Limiting emissions to RCP3PD levels would reduce the impacts on mean winter and summer runoff by approximately half in comparison with scenario A2 for the period 2085 – but even under scenario RCP3PD runoff seasonality is likely to be altered.

– For groundwater temperature change, the medium estimates for the period 2085 under the three greenhouse gas scenarios range from +1 to +3°C. The warming, which could pose problems for groundwater quality and drinking water production, is accentuated in certain aquifers, especially in the mid- to long term (periods 2060 and 2085). Situations of no change in groundwater temperature, however, cannot be excluded given the assessed levels of uncertainty in the projections.

In recent years Switzerland has experienced some unprecedented drought situations. At a political level solutions have been requested for early recognition of hydrological droughts. A prototype information platform has been developed to guide water resources management during situations where water resources drop below critical levels. The development was steered by stakeholders from national administrations and different economic sectors. Since June 2013 the platform has presented daily updated real-time information on several drought indicators including precipitation, streamflow, lake levels, groundwater levels, soil moisture deficit, snow resources, dryness in forests and stream temperatures. For three basins, ensemble forecasts of runoff, soil moisture, snowpack and groundwater storage have been provided. Furthermore, a nationwide operational hydrological simulation at 600 × 600 m resolution gives indications on local water resources deficits. Information for each variable has been used to create automatic "awareness maps" for nine large regions. Three levels of information with increasing detail and complexity can be accessed by over 180 registered users. The operators of the platform give interpreted comments on the content of the platform each week-day. The test phase of the platform will last until the end of 2014. 

An approach for exploring the impacts of climate change on tree migration and on the natural water balance of two large Swiss catchments (Rhône and Ticino) has been developed. We run the spatio-temporal forest model TreeMig for the period 1400–2100 with climate input from observations and from the A1B Scenario ECHAM5-REMO (from the "ENSEMBLES"-Suite). We post-processed the outcomes of TreeMig concerning forest biomass and leaf-area-index (LAI) and modified the forest coverage and LAI used in the hydrological model PREVAH. Every fifth computational year, forest coverage and LAI have been updated. PREVAH has been recently used to assess climate change impacts on water resources for the whole of Switzerland for the control period 1980–2009 and for the scenario periods 2021–2050 and 2070–2099. Analyses show a degradation of forest for the 2070–2099 period in the lowest elevation ranges (<1200 m) as a result of increased drought stress. This results in lower LAI, reduced interception and increased discharge as compared to a baseline run without implementation of forest scenarios. For elevation ranges above 1500 m, TreeMig predicts higher biomasses and LAI. The simulated tree-line elevation might increase by up to 200 m. PREVAH computes in these areas higher evapotranspiration and less runoff as compared to the baseline run. Consideration of forest change yields in both areas changes in simulated discharge of –5 to –10 mm per year and changes in evapotranspiration of +5 to +10 mm per year. However, changes at specific elevation bands are up to four times higher.

Zukünftige Situationen mit lokaler, temporärer Wasserknappheit erfordern eine integrale Information der Wassernutzer bezüglich aktueller und bevorstehender Trockenheit. Weltweit gibt es hierfür schon einige Beispiele und Finanzierungsmodelle, aber in der Schweiz ist erst vor Kurzem eine entsprechende Informationsplattform - drought.ch - entwickelt worden. Der ganze Planungs- und Entwicklungsprozess, welcher eine umfassende Diskussion zwischen Forschenden und Vertretern der Bundesämter und Wassernutzungssektoren beinhaltet hat, wird in diesem Artikel dargelegt. Ein Testbetrieb von drought.ch ist ab diesem Sommer geplant.

During the last few years hydrological models have been applied to estimate current and future spatial and temporal availability of water resources in different areas of the European Alps. In parallel, large efforts have been allocated for the development and operational application of hydrological ensemble prediction systems. WSL has been active in all these fields and is now testing opportunities of combining the outcomes of climate impacts studies with operational forecasting.
In the framework of the CCHYDRO Project a water resources climatology for the control period 1980-2009 and scenarios for two 30-years periods in the future have been created. Results point out, that the seasonal accumulation of snow water will drastically reduce by the end of the 21st century and have large impact on the seasonal distribution of water resources in Switzerland.
From the NRP61 project DROUGHT-CH tools for improving the prediction of water resources availability with lead times of up to 1 month and climatologybased extrapolations for up to 120 days have been realized. First applications show that, depending on the quality of the initial conditions, the forecasts of water storage and low-flow can yield added value for stakeholders concerning lead times of over 20 days, while the forecasts of high discharge and floods show no added value for lead times larger than 10 days.
Our next goal is the establishment of a system for the early detection of anomalies in the temporal and spatial availability of water resources in Switzerland.

We present the derivation and validation of a 250 m x 250 m gridded daily minimum and maximum temperature and daily total precipitation data set for Switzerland that covers the 1961 - 2008 period and accounts for the complexity of the Alpine terrain. We investigate the spatial and temporal development of daily temperature and precipitation for the last decades as influenced by climate change and natural variability. To verify and assess the generation of climatic surfaces, we use the Daymet Model (Thornton et al., 1997). Predictions of minimum and maximum temperatures receive very high correlations with observed values, with a smaller bias for maximum temperature. Based on a regionalization of precipitation by means of a combined principal component and cluster analysis, we explore to what degree the model uncertainty is affected by precipitation characteristics. Additionally we compare the simulated continuous daily precipitation surfaces with the modified interpolation scheme PRISM (Parameter-elevation Regression on Independent Slope Model) (Daly et al., 1994).

Wie wirken sich Luftschadstoffe, die Klimaänderung und andere Belastungen langfristig auf den Schweizer Wald aus? Auf diese Frage möchten WSL und BUWAL in dem Programm «Langfristige Waldkökosystem-Forschung» (LWF) eine Antwort finden (Kräuchi, 1997). Auf 17 über die Schweiz verteilten 2 ha grossen Beobachtungsflächen werden deshalb seit 1995 wichtige physikalische und biologische Standorteigenschaften und ihre Veränderungen erfasst. Dazu gehören der Boden, die Schadstoffeinträge, die Witterung im Wald und im Freiland, die Vegetation, die Nährstoffflüsse, das Wachstum, die Sterberaten und die Nutzung der Bäume, die Baumverjüngung und Kronenmarkmale der Bäume.