Dr. Konrad Bogner

Across the globe, there has been an increasing interest in improving the predictability of subseasonal hydrometeorological forecasts, as they play a valuable role in medium- to long-term planning in many sectors, such as agriculture, navigation, hydropower, and emergency management. However, these forecasts still have very limited skill at the monthly time scale; hence, this study explores the possibilities for improving forecasts through different pre- and postprocessing techniques at the interface with a Precipitationn–Runoff–Evapotranspiration Hydrological Response Unit Model (PREVAH). Specifically, this research aims to assess the benefit of European weather regime (WR) data within a hybrid forecasting setup, a combination of a traditional hydrological model and a machine learning (ML) algorithm, to improve the performance of subseasonal hydrometeorological forecasts in Switzerland. The WR data contain information about the large-scale atmospheric circulation in the North Atlantic–European region, and thus allow the hydrological model to exploit potential flow-dependent predictability. Four hydrological variables are investigated: total runoff, baseflow, soil moisture, and snowmelt. The improvements in the forecasts achieved with the pre- and postprocessing techniques vary with catchments, lead times, and variables. Adding WR data has clear benefits, but these benefits are not consistent across the study area or among the variables. The usefulness of WR data is generally observed for longer lead times, e.g., beyond the third week. Furthermore, a multimodel approach is applied to determine the “best practice” for each catchment and improve forecast skill over the entire study area. This study highlights the potential and limitations of using WR information to improve subseasonal hydrometeorological forecasts in a hybrid forecasting system in an operational mode.

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.

Ende August und Anfang Oktober 2020 waren die Südschweiz und angrenzende Bergkantone von zwei Unwettern betroffen, welche teils zu beträchtlichen Abflüssen, jedoch verhältnismässig geringen Schäden führten. Im Sinne einer Standortbestimmung wurden die beiden Unwetterereignisse zum Anlass genommen, das Funktionieren der Warnkette an der Schnittstelle zwischen den Bundesfachstellen und den Kantonen zu überprüfen. Zu diesem Zweck wurden die hydrometeorologischen Vorhersagen vor und während den beiden Unwettern sowie die daraus abgeleiteten Warnungen auf ihre Qualität und gegenseitige Abstimmung analysiert. Des Weiteren wurden mit Verantwortlichen der betroffenen Kantone Interviews geführt, um ihren Umgang mit den Warnungen und ihre Einschätzung zum Funktionieren der Warnkette zu erfahren. Schliesslich wurden in einem gemeinsamen Workshop potentielle Optimierungsmassnahmen erarbeitet.
Die Analysen und Rückmeldungen haben aufgezeigt, dass die Warnkette während den beiden Unwetterereignissen grundsätzlich gut funktioniert hat. In beiden Fällen war das Vorgehen der einzelnen Kantone bezüglich Informationsempfang, interner Beurteilung und Massnahmenentscheidungen sowie der Kontakt mit den Bundesstellen sehr unterschiedlich. Insgesamt wurde der Informationsfluss des Bundes während der beiden Ereignisse von den Kantonen als zielführend eingeschätzt. Die Analyse der Unwetter im August und Oktober 2020 hat aber auch punktuelle Optimierungsmöglichkeiten aufgezeigt und beispielsweise beim kommunikativen Zusammenspiel der Starkniederschlags- und Hochwasserwarnungen Fragen aufgeworfen.

A la fin août et au début octobre 2020, le sud de la Suisse et les cantons voisins ont été touchés par deux intempéries qui ont entraîné des débits parfois considérables, mais relativement peu de dégâts. Dans l'optique d'un état des lieux, ces deux événements ont été l'occasion de vérifier le fonctionnement de la chaîne d'alerte à l'interface entre les services spécialisés de la Confédération et les cantons. A cet effet, les prévisions hydrométéorologiques avant et pendant les deux intempéries, ainsi que les alertes qui en ont découlé, ont été analysées quant à leur qualité et leur coordination mutuelle. Par ailleurs, des entretiens ont été menés avec des responsables des cantons concernés afin de connaître leur gestion des alertes et leur évaluation du fonctionnement de la chaîne d'alerte. Enfin, des mesures d’optimisation potentielles ont été élaborées dans le cadre d'un séminaire commun.
Les analyses et les retours d'information ont montré que la chaîne d'alerte avait en principe bien fonctionné pendant les deux événements. Dans les deux cas, la procédure des différents cantons concernant la réception des informations, l'évaluation interne et les décisions relatives aux mesures, ainsi que le contact avec les services fédéraux, étaient très différents. Dans l'ensemble, les cantons ont estimé que le flux d’informations de la Confédération pendant les deux événements était conforme aux objectifs. Cependant, l'analyse des intempéries d'août et d'octobre 2020 a également montré des possibilités d'optimisation et soulevé des questions, par exemple concernant l'interaction en matière de communication entre les alertes aux fortes précipitations et aux crues.

Different postprocessing techniques are frequently employed to improve the outcome of ensemble forecasting models. The main reason is to compensate for biases caused by errors in model structure or initial conditions, and as a correction for under- or overdispersed ensembles. Here we use the ensemble model output statistics method to postprocess the ensemble output from a continental-scale hydrological model, LISFLOOD, as used in the European Flood Awareness System (EFAS). We develop a method for local calibration and interpolation of the postprocessing parameters and compare it with a more traditional global calibration approach for 678 stations in Europe based on long-term observations of runoff and meteorological variables. For the global calibration we also test a reduced model with only a variance inflation factor. Whereas the postprocessing improved the results for the first 1–2 days lead time, the improvement was less for increasing lead times of the verification period. This was the case both for the local and global calibration methods. As the postprocessing is based on assumptions about the distribution of forecast errors, we also present an analysis of the ensemble output that provides some indications of what to expect from the postprocessing.

The quality of the forecasts, that is, the accuracy in predicting the observed streamflow, affects the decisions that can be taken thus determining the success or failure of hydropower operations, that is, the so‐called forecast value. Although preprocessing techniques can be employed to improve forecast quality, the corresponding improvement in forecast value to hydropower is not straightforward to anticipate because of the complex relationship between quality and value, which depends on the hydrometeorological regime and water system features. The objective of this paper is to demonstrate the value to hydropower reservoir operations of preprocessed (i.e., bias corrected and downscaled) subseasonal forecasts and to compare the forecast quality and value across and within seasons. We use the forecast ensembles provided by the European Center for Medium‐Range Weather Forecasts (ECMWF). We assess forecast quality in terms of bias, Continuous Ranked Probability Score (CRPS), and Continuous Ranked Probability Skill Score (CRPSS), and forecast value in terms of mean revenue and avoided unproductive spill. We consider daily subseasonal hydrometeorological forecasts covering lead times up to 1 month and a quantile mapping technique to preprocess the hydrometeorological forecasts. Forecasts are then used in a rolling horizon set up to optimize hydropower operations of the Verzasca hydropower system in the Alps (CH). Results show that preprocessing is essential to improve both forecast quality and value. Although hydropower reservoir operations benefit from considering forecasts all the yearlong, the relationship between forecast quality and value is complex and strongly depends on the metrics used to assess forecast quality and value and on the season.

Reliable predictions of the energy consumption and production is important information for the management and integration of renewable energy sources. Several different Machine Learning (ML) methodologies have been tested for predicting the energy consumption/production based on the information of hydro-meteorological data. The methods analysed include Multivariate Adaptive Regression Splines (MARS) and various Quantile Regression (QR) models like Quantile Random Forest (QRF) and Gradient Boosting Machines (GBM). Additionally, a Nonhomogeneous Gaussian Regression (NGR) approach has been tested for combining and calibrating monthly ML based forecasts driven by ensemble weather forecasts. The novelty and main focus of this study is the comparison of the capability of ML methods for producing reliable predictive uncertainties and the application of monthly weather forecasts. Different skill scores have been used to verify the predictions and their uncertainties and first results for combining the ML methods applying the NGR approach and coupling the predictions with monthly ensemble weather forecasts are shown for the southern Switzerland (Canton of Ticino). These results highlight the possibilities of improvements using ML methods and the importance of optimally combining different ML methods for achieving more accurate estimates of future energy consumptions and productions with sharper prediction uncertainty estimates (i.e., narrower prediction intervals).

Traditional ensemble streamflow prediction (ESP) systems are known to provide a valuable baseline to predict streamflows at the subseasonal to seasonal timescale. They exploit a combination of initial conditions and past meteorological observations, and can often provide useful forecasts of the expected streamflow in the upcoming month. In recent years, numerical weather prediction (NWP) models for subseasonal to seasonal timescales have made large progress and can provide added value to such a traditional ESP approach. Before using such meteorological predictions two major problems need to be solved: the correction of biases, and downscaling to increase the spatial resolution. Various methods exist to overcome these problems, but the potential of using NWP information and the relative merit of the different statistical and modelling steps remain open. To address this question, we compare a traditional ESP system with a subseasonal hydrometeorological ensemble prediction system in three alpine catchments with varying hydroclimatic conditions and areas between 80 and 1700 km2. Uncorrected and corrected (pre-processed) temperature and precipitation reforecasts from the ECMWF subseasonal NWP model are used to run the hydrological simulations and the performance of the resulting streamflow predictions is assessed with commonly used verification scores characterizing different aspects of the forecasts (ensemble mean and spread). Our results indicate that the NWP-based approach can provide superior prediction to the ESP approach, especially at shorter lead times. In snow-dominated catchments the pre-processing of the meteorological input further improves the performance of the predictions. This is most pronounced in late winter and spring when snow melting occurs. Moreover, our results highlight the importance of snow-related processes for subseasonal streamflow predictions in mountainous regions.

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.

Subseasonal predictions bridge the gap between medium-range weather forecasts and seasonal climate predictions. This time horizon is of crucial importance for many planning purposes, including energy production and agriculture. The verification of such predictions is normally done for areal averages of upper-air parameters. Only few studies exist that verify the forecasts for surface parameters with observational stations, although this is crucial for real-world applications, which often require such predictions at specific surface locations. With this study we provide an extensive station-based verification of subseasonal forecasts against 1,637 ground based observational time series across Europe. Twenty years of temperature and precipitation reforecasts of the European Centre for Medium-Range Weather Forecasts Integrated Forecasting System are used to analyze the period of April 1995 to March 2014. A lead time and seasonally dependent bias correction is performed to correct the daily temperature and precipitation forecasts at all stations individually. Two bias correction techniques are compared, a mean debiasing method and a quantile mapping approach. Commonly used skill scores characterizing different aspects of forecast quality are computed for weekly aggregated forecasts with lead times of 5–32 days. Overall, promising skill is found for temperature in all seasons except spring. Temperature forecasts tend to show higher skill in Northern Europe and in particular around the Baltic Sea, and in winter. Bias correction is shown to be essential in enhancing the forecast skill in all four weeks for most of the stations and for both variables with QM generally performing better.

The enhanced availability of many different hydro-meteorological modelling and forecasting systems raises the issue of how to optimally combine this great deal of information. Especially the usage of deterministic and probabilistic forecasts with sometimes widely divergent predicted future streamflow values makes it even more complicated for decision makers to sift out the relevant information. In this study multiple streamflow forecast information will be aggregated based on several different predictive distributions, and quantile forecasts. For this combination the Bayesian model averaging (BMA) approach, the non-homogeneous Gaussian regression (NGR), also known as the ensemble model output statistic (EMOS) techniques, and a novel method called Beta-transformed linear pooling (BLP) will be applied. By the help of the quantile score (QS) and the continuous ranked probability score (CRPS), the combination results for the Sihl River in Switzerland with about 5 years of forecast data will be compared and the differences between the raw and optimally combined forecasts will be highlighted. The results demonstrate the importance of applying proper forecast combination methods for decision makers in the field of flood and water resource management.

Post-processing has received much attention during the last couple of years within the hydrological community, and many different methods have been developed and tested, especially in the field of flood forecasting. Apart from the different meanings of the phrase "post-processing" in meteorology and hydrology, in this paper, it is regarded as a method to correct model outputs (predictions) based on meteorological (1) observed input data, (2) deterministic forecasts (single time series) and (3) ensemble forecasts (multiple time series) and to derive predictive uncertainties. So far, the majority of the research has been related to floods, how to remove bias and improve the forecast accuracy and how to minimize dispersion errors. Given that global changes are driving climatic forces, there is an urgent need to improve the quality of low-flow predictions, as well, even in regions that are normally less prone to drought. For several catchments in Switzerland, different post-processing methods were tested with respect to low stream flow and flooding conditions. The complexity of the applied procedures ranged from simple AR processes to more complex methodologies combining wavelet transformations and Quantile Regression Neural Networks (QRNN) and included the derivation of predictive uncertainties. Furthermore, various verification methods were tested in order to quantify the possible improvements that could be gained by applying these post-processing procedures based on different stream flow conditions. Preliminary results indicate that there is no single best method, but with an increase of complexity, a significant improvement of the quality of the predictions can be achieved.

For many years, meteorological models have been run with perturbated initial conditions or parameters to produce ensemble forecasts that are used as a proxy of the uncertainty of the forecasts. However, the ensembles are usually both biased (the mean is systematically too high or too low, compared with the observed weather), and has dispersion errors (the ensemble variance indicates a too low or too high confidence in the forecast, compared with the observed weather). The ensembles are therefore commonly post-processed to correct for these shortcomings. Here we look at one of these techniques, referred to as Ensemble Model Output Statistics (EMOS) (Gneiting et al., 2005). Originally, the post-processing parameters were identified as a fixed set of parameters for a region. The application of our work is the European Flood Awareness System (http://www.efas.eu), where a distributed model is run with meteorological ensembles as input. We are therefore dealing with a considerably larger data set than previous analyses. We also want to regionalize the parameters themselves for other locations than the calibration gauges. The post-processing parameters are therefore estimated for each calibration station, but with a spatial penalty for deviations from neighbouring stations, depending on the expected semi-variance between the calibration catchment and these stations. The estimated post-processed parameters can then be used for regionalization of the postprocessing parameters also for uncalibrated locations using top-kriging in the rtop-package (Skoien et al., 2006, 2014). We will show results from cross-validation of the methodology and although our interest is mainly in identifying exceedance probabilities for certain return levels, we will also show how the rtop package can be used for creating a set of post-processed ensembles through simulations.

In the last decade operational probabilistic ensemble flood forecasts have become common in supporting decision-making processes leading to risk reduction. Ensemble forecasts can assess uncertainty, but they are limited to the uncertainty in a specific modeling system. Many of the current operational flood prediction systems use a multimodel approach to better represent the uncertainty arising from insufficient model structure. This study presents a multimodel approach to building a global flood prediction system using multiple atmospheric reanalysis datasets for river initial conditions and multiple TIGGE forcing inputs to the ECMWF land surface model. A sensitivity study is carried out to clarify the effect of using archive ensemble meteorological predictions and uncoupled land surface models. The probabilistic discharge forecasts derived from the different atmospheric models are compared with those from the multimodel combination. The potential for further improving forecast skill by bias correction and Bayesian model averaging is examined. The results show that the impact of the different TIGGE input variables in the HTESSEL/Catchment-Based Macroscale Floodplain model (CaMa-Flood) setup is rather limited other than for precipitation. This provides a sufficient basis for evaluation of the multi model discharge predictions. The results also highlight that the three applied reanalysis datasets have different error characteristics that allow for large potential gains with a multimodel combination. It is shown that large improvements to the forecast performance for all models can be achieved through appropriate statistical postprocessing (bias and spread correction). A simple multimodel combination generally improves the forecasts, while a more advanced combination using Bayesian model averaging provides further benefits.

This study applies statistical postprocessing to ensemble forecasts of near-surface temperature, 24 h precipitation totals, and near-surface wind speed from the global model of the European Centre for Medium-Range Weather Forecasts (ECMWF). The main objective is to evaluate the evolution of the difference in skill between the raw ensemble and the postprocessed forecasts. Reliability and sharpness, and hence skill, of the former is expected to improve over time. Thus, the gain by postprocessing is expected to decrease. Based on ECMWF forecasts from January 2002 to March 2014 and corresponding observations from globally distributed stations, we generate postprocessed forecasts by ensemble model output statistics (EMOS) for each station and variable. Given the higher average skill of the postprocessed forecasts, we analyze the evolution of the difference in skill between raw ensemble and EMOS. This skill gap remains almost constant over time indicating that postprocessing will keep adding skill in the foreseeable future.