Ensemble and probabilistic hydro-meteorological forecasts: predictive uncertainty, verification and decision making 

This session brings together scientists, forecasters, practitioners and stakeholders interested in exploring the use of ensemble hydro-meteorological forecast techniques in hydrological applications: e.g., flood control and warning, reservoir operation for hydropower and water supply, transportation, and agricultural management. It will address the understanding of sources of predictability and quantification and reduction of predictive uncertainty of hydrological extremes in deterministic and ensemble hydrological forecasting. Uncertainty estimation in operational forecasting systems is becoming a more common practice. However, a significant research challenge and central interest of this session is to understand the sources of predictability and development of approaches, methods and techniques to enhance predictability (e.g. accuracy, reliability etc.) and quantify and reduce predictive uncertainty in general. Ensemble data assimilation, NWP preprocessing, multi-model approaches or hydrological postprocessing can provide important ways of improving the quality (e.g. accuracy, reliability) and increasing the value (e.g. impact, usability) of deterministic and ensemble hydrological forecasts. The models involved with the methods for predictive uncertainty, data assimilation, post-processing and decision-making may include machine learning models, ANNs, catchment models, runoff routing models, groundwater models, coupled meteorological-hydrological models as well as combinations (multimodel) of these. Demonstrations of the sources of predictability and subsequent quantification and reduction in predictive uncertainty at different scales through improved representation of model process (physics, parameterization, numerical solution, data support and calibration) and error, forcing and initial state are of special interest to the session.

Co-organized by NH1
Convener: Jan Verkade | Co-conveners: Trine Jahr Hegdahl, Albrecht Weerts, Shaun HarriganECSECS, Kolbjorn Engeland
| Thu, 26 May, 15:55–18:30 (CEST)
Room 2.44

Presentations: Thu, 26 May | Room 2.44

Chairpersons: Jan Verkade, Albrecht Weerts
Presentation form not yet defined
Jens Grundmann and Andy Philipp

Reliable warnings and forecasts of extreme precipitation and resulting floods are an important prerequisite for disaster mangers to initiate flood defence measures. Thus, disaster managers are interested in extended lead times, which can be obtained by employing forecast of numerical weather models as driving data for hydrological models. To portray the inherent uncertainty of weather model output, ensemble hydro-meteorological forecasts can be used, which offers the opportunity of probability based decision making for disaster managers. However, especially for changing weather systems under unstable atmospheric conditions and for small, fast-responding catchments, the signals of extreme precipitation in the forecasting models may change quickly in magnitude and ensemble spread for successive forecast in expectation of an approaching event.

With this contribution, we analyse the behaviour and reliability of ensemble hydro-meteorological forecasts depending on their lead time in order to derive appropriate indicators for decision making. We use results of our operational web-based demonstration platform for ensemble hydrological forecasting in small catchments, which is established for three pilot regions with different hydrological settings in Saxony, Germany. The demonstration platform processes ensemble forecasts of the ICON/COSMO-D2-EPS product of the German Weather Service, which provides an ensemble of 20 members each three hours, for lead times up to 27 hours.  Each member is evaluated regarding specific extreme precipitation thresholds. If these thresholds are exceeded in a specific region, rainfall-runoff models for the associated catchments are used to propagate the meteorological uncertainty into the resulting runoff, followed by statistical post processing and visualization. In addition, different options for the visualization of the uncertainty information were developed to monitor the behaviour and reliability of the forecast ensemble over successive forecast lead times. These options contain exceedance probabilities for thresholds in rainfall and resulting runoff and were discussed with decision makers regarding their applicability for decision making. First results are presented for observed extreme events in the small pilot regions.

How to cite: Grundmann, J. and Philipp, A.: Analysis of ensemble forecasts over successive forecast lead times for decision support in flood management, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6373, https://doi.org/10.5194/egusphere-egu22-6373, 2022.

On-site presentation
Adele Young, Biswa Bhattacharya, Faisal Mahood, Emma Daniels, and Chris Zevenbergen

High-resolution Quantitative Precipitation Forecasts (QPF)  are essential to accurately forecast the magnitude, timing and location of precipitation and as input for pluvial flood forecasting using urban drainage models. However, there are challenges of producing high-resolution forecast capable of capturing the spatial and temporal variability of rainfall needed for urban flood modelling and the uncertainty associated with meteorological forecast and urban flood models. Therefore there is a challenge to balance data availability, model uncertainty, resolution, forecast lead-time and computational demands, especially in data-scarce regions.

Ensemble precipitation forecasts are used to capture uncertainties of meteorological forecasting in flood models. This research aims to evaluate the skill of a downscaled ensemble precipitation forecast over the coastal city of Alexandria, Egypt which experiences extreme rainfall and flooding from winter storms. A Weather Research Forecast (WRF) convection-permitting model was initialised using the Global Ensemble Forecast System (GEFS) which provides 21 ensemble members (1 degree archived). The model was run using three domains with horizontal grid resolutions of 30km, 10 km and 3.3 km at a 24h leadtime). For the 3.3 km horizontal grid, ensemble members were coupled with a 1D Mike urban model to evaluate the meteorological uncertainty representation and propagation.

In the absence of sufficient rainfall and flow gauge data, results were verified against Multi-Source Weighted-Ensemble Precipitation (MSWEP) satellite-derived product and further compared with the ECMWF ensemble prediction system precipitation forecast. 1D flood simulations were evaluated against 1D- 2D hydrodynamic simulations run with MSWEP data.

Ensembles showed varying probability of detection for different severity events. In general, the majority of ensemble rainfall values resulted in flooding greater than the flooding simulated from the satellite observed rainfall. Although deterministic forecast also indicated flooding and threshold exceedance, the number of ensemble members exceeding critical thresholds has the benefit of providing decision-makers with the probability of threshold exceedance and likelihood of flooding to trigger protective actions. A study such as this provides knowledge for understanding, future applications and limitations of using high-resolution ensemble Quantitative Precipitation Forecasts (QPFs) and the importance of capturing the spatial and temporal variability of rainfall in urban drainage models. Additionally, the potential use of MSWEP for the verification of ensemble forecasts in ungauged and data-scarce regions is investigated.

How to cite: Young, A., Bhattacharya, B., Mahood, F., Daniels, E., and Zevenbergen, C.: High-Resolution Ensemble Precipitation for Pluvial Flood Forecasting in the Urban Data Scarce city of Alexandria, Egypt , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2596, https://doi.org/10.5194/egusphere-egu22-2596, 2022.

Virtual presentation
Sakila Saminathan and Subhasis Mitra

Reliable and accurate precipitation forecast information is needed for various disaster management and mitigation purposes. Spatio-temporal variability of forecast and uncertainty in the NWP models reduces the skill and reliability of the forecasts, hampering greater uptake for various purposes. This study aims to quantify the performance of short to medium range (1 to 7 days) precipitation forecast information from four different NWP models over the Indian sub-continent. The precipitation forecasts from these four models, namely Climate Forecast System version 2 (CFSv2), European Centre for Medium Range Weather Forecasts (ECMWF), Global Ensemble Forecast System (GEFS), and Indian Institute of Tropical Meteorology (IITM), has been assessed using different precipitation indices namely number of rainy days, accumulated precipitation, consecutive wet days, and consecutive dry days. The indices are evaluated for all the models using the evaluation metrics Heidke Skill Scores (HSS) for different seasons and basins. HSS for different indices shows that monthly HSS value was around 0.2 for the consecutive wet days while being 0.4 for the consecutive dry days showing that model's performance was good for the consecutive dry days than consecutive wet days. Results also show that the models are able to capture the number of rainy days and accumulated precipitation satisfactorily. The assessment of models and indices for monsoon and non-monsoon season showed better performance in the non-monsoon season. The evaluation of models and indices spatially over different basins in India showed that the performance was good in the central region (i.e., Narmada and Tapti basin). Overall, the forecasts from the ECMWF performed better compared to GEFS, CFSv2, and IITM. 

How to cite: Saminathan, S. and Mitra, S.: Intermodel comparison of Short to Medium Range Precipitation Forecasts over the Indian Sub-Continent, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-220, https://doi.org/10.5194/egusphere-egu22-220, 2022.

Presentation form not yet defined
Alireza Askarinejad, Mélanie Trudel, and Marie-Amélie Boucher

Recent studies have shown that probabilistic forecasts are superior to
deterministic forecasts in terms of quality, reliability, and representing the
uncertainty of future states. One of the most well-known and widely used
tools for assessing the performance of (probabilistic) forecast systems is the
continuous ranked probability score (CRPS). This metric is employed to
evaluate the forecasting system when only forecast uncertainty is
considered. In addition to multiple sources of uncertainty in a forecasting
system (such as initial conditions, model structure and parameters, and
boundary conditions), the uncertainty can also originate from observations
(e.g., streamflow). However, this uncertainty, which has rarely been
explored in previous research, should also be regarded in evaluating the
forecasting system. A version of the CPRS is redefined and analyzed to
overcome this important flaw, considering the observation's uncertainty. To
estimate the uncertainty associated with streamflow observations, the
Bayesian Rating curve method (BaRatin) is utilized. This study focuses on
comparing the different versions of the CRPS in considering the
uncertainties of forecasts and observations. Three types of streamflow
forecasting systems are used in this study: deterministic forecasts, raw
ensemble forecasts (applying meteorological ensemble forecasts as inputs to
the hydrological model), and post-processed ensemble forecasts (postprocessing
of hydrological model outputs using weighted ensemble dressing
method). The assessment is performed for short-term forecasts (lead times of
1 to 5 days) for the Au Saumon watershed in southern central Quebec,
Canada. It is found that considering observation uncertainty has a significant
effect on the values of CRPS compared to when only forecast uncertainty is
considered. In addition, CRPS changes in probabilistic forecasts are more
than deterministic ones. Our results also point out that using the modified
version of the CRPS can help end-users better understand and evaluate their
forecasting system.

How to cite: Askarinejad, A., Trudel, M., and Boucher, M.-A.: Comparing different versions of the continuous ranked probability score to account for forecast or observation uncertainty, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10728, https://doi.org/10.5194/egusphere-egu22-10728, 2022.

Virtual presentation
Ankit Singh and Sanjeev jha

This study evaluates the deterministic and ensemble quantitative precipitation forecasts (QPFs) obtained from four Numerical Weather Prediction (NWPs) models over the Indian region during the monsoon period (June to September) for the years 2011 to 2020. We considered 18 river basins and 14 Agro climatic zones to compare the skill of the forecasts with the observation data. From The Observing System Research and Predictability Experiment Interactive Grand Global Ensemble (TIGGE) archives, we obtained QPFs from Environment and Climate Change Canada (ECCC), European Centre for Medium-Range Weather Forecasts (ECMWF), Korea Meteorological Agency (KMA), and National Centres for Environmental Prediction (NCEP) with 1 to 5 day lead time at a spatial resolution of 0.50. The Integrated Multi-satellite Retrievals for Global Precipitation Measurement (IMERG) data for the same time period is used as observation data. Deterministic (RMSE, NSE, and CC) and dichotomous (POD and FAR) assessment have been performed to evaluate the skill of the QPF(s). Our result shows that overall the performance of ECMWF ensembles mean is better than the other NWPs model, as the NSE and CC value is more close to 1. The river basins in the southern part of the country (Godavari, Krishna and Cauveri River Basins) have the higher error (RMSE more than 100 and NSE close to 0) compared to Brahmputra, Ganga, and Barak River basins. The errors are less in those agro-climatic zones which has high elevation where the rainfall is less. The detailed result of the ongoing research will be presented at the conference. 

How to cite: Singh, A. and jha, S.: Evaluation of ensemble precipitation forecasts from NWP models in Indian River basins and agro-climatic zones, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12242, https://doi.org/10.5194/egusphere-egu22-12242, 2022.

Coffee break
Chairpersons: Jan Verkade, Albrecht Weerts
On-site presentation
Seán Donegan, Conor Murphy, Ciaran Broderick, Dáire Foran Quinn, Saeed Golian, and Shaun Harrigan

Ensemble streamflow prediction (ESP) is a well-established and widely used approach to hydrological forecasting, the application of which requires a hydrological model that can contribute to forecast skill by providing: (i) accurate initial hydrological conditions; and (ii) accurate transformation of climate to river flow signals. It is widely known that there exists a relationship between ESP skill and the hydrological regime of a catchment, and several studies have correlated forecast quality with sets of catchment descriptors. The choice of hydrological model is therefore significant. Whilst a parsimonious structure may be preferable for efficiency, potential skill could be lost if the model’s simplicity means it cannot adequately reproduce key hydrological processes in the catchment. This work seeks to examine the contribution of hydrological model complexity to forecast skill. Using a parsimonious model as a reference, we investigate if additional model complexity adds forecast skill at different lead times and initialisation months through the use of models with different structures and parametric complexity. Forecast skill is evaluated within a hindcast experiment for a selection of Irish river catchments using the continuous ranked probability skill score. Results are presented for our reference model, GR4J (Génie Rural à 4 paramètres Journalier), and our complex model, SMART (Soil Moisture Accounting and Routing for Transport). The performance of each model is viewed in the context of its ability to reproduce key hydrological signatures known to control forecast quality in Ireland (e.g., baseflow index).

How to cite: Donegan, S., Murphy, C., Broderick, C., Foran Quinn, D., Golian, S., and Harrigan, S.: Do more complex hydrological models produce more skilful streamflow forecasts?, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12775, https://doi.org/10.5194/egusphere-egu22-12775, 2022.

On-site presentation
Marie-Amélie Boucher, Jean Odry, Vincent Fortin, Simon Lachance-Cloutier, Richard Turcotte, and Dominic Roussel

Global or large-scale hydrological forecasting systems covering entire countries, continents and even the entire planet are growing in popularity. As more large-scale hydrological forecasting systems emerge, it is likely that they will co-exist with pre-existing local forecasting systems. It is the case for instance in Canada, where most provinces have their own streamflow forecasting system, while the new NSRPS will eventually cover the whole country using a 1km by 1km grid. Those province, for instance Quebec, built their own forecasting systems on hydrological models configured for river catchments rather than a regular grid. Using this situation as a starting point and a case study, we propose a Bayesian framework for merging the forecasts from two systems. Within this Bayesian framework, the large-scale prior information comes from the NSRPS. This prior information is then updated using forecasts from the government of Quebec and the associated likelihood. In order to account for forecast uncertainty, this work is carried out using a probabilistic approach for both the NSRPS and Quebec’s Système de Prévision Hydrologique (SPH). While SPH produces probabilistic forecasts by default, the preliminary version of the NSRPS that we had access to is deterministic. Consequently, forecasts from the NSRPS had to be dressed into an ensemble in order to use them as prior distribution within the Bayesian merging framework. Alternative prior distributions (climatology, Markov chain) are also considered instead of those obtained from the NSRPS. Since both forecasting systems include ungauged sites, a version of this Bayesian merging framework based on regional statistics was also developed and tested using cross-validation. Our results show that the merged forecasts perform at least as well as the best individual system, for both gauged and ungauged basins. For longer lead times, merged forecasts can even outperform individual systems. Considering that the NSRPS relies on a non-calibrated model with no data assimilation, those results show that there could be important practical gains in merging large scale hydrological forecasts with local scale forecasts.

How to cite: Boucher, M.-A., Odry, J., Fortin, V., Lachance-Cloutier, S., Turcotte, R., and Roussel, D.: Bayesian merging of large scale and local scale hydrological forecasts, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-3228, https://doi.org/10.5194/egusphere-egu22-3228, 2022.

On-site presentation
Silvia Barbetta, Bhabagrahi Sahoo, Bianca Bonaccorsi, Tommaso Moramarco, Trushnamayee Nanda, Chandranath Chatterjee, and Ezio Todini

The impact of flood events is usually approached through structural measures, such as riverbanks and dams able to mitigating, although not fully eliminating flooding risk. Therefore, complementary non-structural measures, mainly real-time Flood Forecasting and Warning Systems (FFWSs), usually combined with operational decision support systems, must be developed to improve the population safety and resilience. Flood forecasting models, essential components of FFWSs, provide deterministic forecasts of discharge or water levels at critical sections on forecast horizons to support the decision-makers activities. Unfortunately, under the uncertainty of future events, predictions must be probabilistic, to be effective and to guarantee the required robustness to the decision makers (Todini, 2017).

Many studies are available in the literature on generating probabilistic forecasting starting from a deterministic forecast and considering the error distribution. Alternatively, the introduction of the Hydrological Uncertainty Processor (Krzysztofowicz, 1999) has posed the basis for the estimation of the predictive uncertainty, PU, that is the probability of occurrence of a future value conditional on all the available information, usually provided by forecasting models.

In this context, for estimating the PU, Todini (2008) proposed the Model Conditional Processor (MCP) which allows for the analytical treatment of the multivariate probability densities after converting both observations and model predictions into the Normal space. Afterwards, MCP was extended to the multi-model approach (Barbetta et al., 2017) enabling a decision based on “multiple forecasts” of different deterministic models at the same time.

With the aim to shed light on the benefits of using PU, the multi-model MCP is applied to discharge forecasts at sites along Indian rivers. Specifically, a data-driven model, i.e. a novel Wavelet-based Non-linear AutoRegressive with eXogenous inputs (WNARX) model and the grid-based semi-distributed VIC hydrological model are used to this end. The future estimates of the river discharge coming into artificial reservoirs, provided by VIC and WNARX models (Nanda et al., 2019) at the same time, are used to feed simultaneously the MCP; thus, showing the benefits in terms of improved effectiveness of the future prediction. The analysis is performed for the Hirahud dam along the Manhanadi River: the results indicate that the methodology could be able to provide effective probabilistic real-time inflow forecasting to be used during significant floods as an appropriate support for the artificial reservoir management.


Barbetta S., Coccia G., Moramarco T., Brocca L., and Todini E. (2017). Improving the effectiveness of real-time flood forecasting through Predictive Uncertainty estimation: the multi-temporal approach, J. of Hydrol., 51, 555-576. 

Krzysztofowicz, R. 1999. Bayesian theory of probabilistic forecasting via deterministic hydrologic model, Water Resour. Res., 35, 2739–2750.

Nanda, T., Sahoo, B., Chatterjee, C. (2019). Enhancing real-time streamflow forecasts with wavelet-neural network-based error-updating schemes and ECMWF meteorological predictions in Variable Infiltration Capacity model. J. Hydrol., 575, pp. 890–910.

Todini, E. A model conditional processor to assess predictive uncertainty in flood forecasting. Int. J. River Basin Manag. 2008, 6, 123–137.

Todini E. Flood Forecasting and Decision Making in the new Millennium. Where are We?, Water Resour Manage. 2017, doi:10.1007/s11269-017-1693-7, pp.1-19.


How to cite: Barbetta, S., Sahoo, B., Bonaccorsi, B., Moramarco, T., Nanda, T., Chatterjee, C., and Todini, E.: Addressing effective real-time flood forecasting for upstream artificial reservoirs through predictive uncertainty  , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4458, https://doi.org/10.5194/egusphere-egu22-4458, 2022.

Virtual presentation
Martin Widmann, Michael Angus, Andrew Orr, and Gregor Leckebusch

Accurate predictions of heavy precipitation in India are vital for impact-orientated forecasting, and an essential requirement for mitigating the impact of damaging flood events. Operational forecasts from non-convection-permitting models can have large biases in the intensities of heavy precipitation, and while convection-permitting models can perform better, their operational use over large areas is not yet feasible. Statistical postprocessing can reduce these biases for relatively little computational cost, but few studies have focused on postprocessing forecasts of monsoonal rainfall.

As part of the UK Weather and Climate Science for Service Partnership India (WCSSP India), the HEavy Precipitation Forecast Postprocessing over India (HEPPI) project has evaluated and compared two popular postprocessing methods: Univariate Quantile Mapping (UQM) and Ensemble Model Output Statistics (EMOS). The project focuses on the suitability of the methods for postprocessing heavy rainfall in India. Both methods are applied to daily precipitation in the National Centre for Medium Range Weather Forecasting (NCMWF) 12km forecast for the 2018 and 2019 monsoon seasons. The evaluation is based on day 1 forecasts and fitting the methods individually for each location.

UQM leads by construction to precipitation distributions close to the observed ones, while EMOS optimises the spread of the postprocessed ensemble without guaranteeing realistic rainfall distributions, and it is not a priori clear which method is better suited for practical applications. The methods are therefore compared with respect to several aspects: local distributions, representation of temporal variability using the Continuous Ranked Probability Score, ensemble spread using Rank Histograms, and exceedance of heavy precipitation thresholds using Brier Scores, Reliability Diagrams, and Receiver Operating Characteristics curves.

EMOS performs not only best, as expected, with respect to correcting under- or overdispersive ensembles, but also with respect to scores for temporal variability, both for the whole range of rainfall values and specifically for heavy rainfall. UQM performs best, as expected, with respect to the local precipitation distributions. The ROC results are inconclusive and location dependent, although both postprocessing methods consistently outperform the raw forecast. These findings are independent of the choice of gridded precipitation data sets used for model fitting and validation.

We recommend EMOS for operational application, as from a user perspective a good performance in forecasting values at a given time, in particular heavy precipitation events, can be expected to be more important than achieving a close match between the forecasted and observed local precipitation distributions.


How to cite: Widmann, M., Angus, M., Orr, A., and Leckebusch, G.: Postprocessing of precipitation forecasts over India with Quantile Mapping  and Ensemble Model Output Statistics, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7191, https://doi.org/10.5194/egusphere-egu22-7191, 2022.

On-site presentation
Daniele Dalla Torre, Andrea Menapace, Ariele Zanfei, and Maurizio Righetti

Data-driven methods are widely adopted to forecast short-term streamflow with lead time up to a few days. Flood risk mitigation, multi-use water management and hydropower plants schedule are the most common fields to use forecasting results. Increasing the accuracy and limiting the uncertainty of the predictions are common needs and also this work would evaluates these aspects combining regional climate models and machine learning techniques. Thus, the research question addressed regards the suitability of the machine learning algorithm fed by the ICON forecasting regional climate model for short lead time streamflow prediction in a small and complex Alpine environment.

A data-driven forecasting procedure is used for streamflow forecasting on a lead time of two days in small Alpine catchments of the Alto Adige Province (Italy). Bias correction of the ICON prediction data inputs against the historical data and the machine learning module compose the two steps data-driven methodology that we propose. Historical time series of precipitation and temperature provided by weather stations have been used for training the machine learning algorithms, while the ICON prediction data of precipitation and temperature have been adopted for testing them. The use of historical data has been essential for collecting a reasonable amount of data required for algorithm learning. The methodology performance evaluation is on the meteorological correction and on the hydrological forecasting.

This first assessment shows promising results for two-day head streamflow prediction even in the context of small catchments with complex orography. This finding suggests that the merging of robust data-driven methodologies with high-resolution detailed weather prevision inputs can be a consistent breakthrough for reliable hydrological short-term forecasting. In conclusion, the flexibility of machine learning and ensemble climate prediction allows for adequate management of uncertainty along the prediction procedure, which is crucial in hydrological applications.

How to cite: Dalla Torre, D., Menapace, A., Zanfei, A., and Righetti, M.: Assessment of a short-term machine learning streamflow forecasting in small Alpine catchments leveraging Deutscher Wetterdienst ICON climate forecasting model, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12149, https://doi.org/10.5194/egusphere-egu22-12149, 2022.

On-site presentation
Katherine Kowal, Louise J. Slater, Alan García-López, and Anne F. Van Loon

Seasonal forecasts present an opportunity to enhance preparedness for hydrometeorological extremes in Central America. Many seasonal forecasts are publicly available, but their comparative value is not well understood, especially over the Central American region. Knowing how best to combine the different seasonal forecast models on offer, or when and where to trust them, requires further study. This evaluation compares seasonal rainfall forecasts over Central America with a focus on hydrometeorological extremes using two of the globally leading ensembles: the Copernicus Climate Change Service seasonal forecasting system (C3S), and the North American Multimodel Ensemble (NMME). We compare the two multimodel ensemble means, eleven individual model means, and model member predictions of monthly and seasonal rainfall over different months, locations, and lead times to better understand their relative forecast quality and identify potential regional predictability limits at the seasonal scale. Direct rainfall forecasts from the models are compared with indirect dynamical-statistical forecasts using large-scale climate precursors within a statistical rainfall prediction system. Results show that C3S and NMME exhibit similar regional variability in their direct rainfall forecasts, revealing the influence of important climate mechanisms on rainfall predictability in the region, which originate in both the Pacific and Atlantic Oceans. The models with the best skill also vary depending on the season, subregion, and lead time assessed. The relative accuracy of indirect versus direct forecasts is still under consideration but we expect their accuracy to vary geographically and seasonally, depending on the associations between the regional climate precursors (e.g. El Niño Southern Oscillation and Tropical North Atlantic variability) and local rainfall. Overall, the models compared can provide useful information on upcoming rainfall, but their regional and seasonal variability affect their usefulness for different types of forecasting applications.

How to cite: Kowal, K., Slater, L. J., García-López, A., and Van Loon, A. F.: Dynamic vs. Hybrid Seasonal Rainfall Forecasts over Central America: A Comparative Evaluation of C3S and NMME, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-367, https://doi.org/10.5194/egusphere-egu22-367, 2022.

On-site presentation
Sandra Margrit Hauswirth, Marc F.P. Bierkens, Vincent Beijk, and Niko Wanders

Ensemble hydrological forecasts are important for operational water management and near future planning, even more so in times of increased extreme events such as floods and droughts. Especially the latter requires a planning horizon of several weeks to months to optimize water availability. Having a flexible forecasting framework that can deliver this information in a fast and computational efficient manner is critical. In this study we are exploring a new hybrid framework, combining machine learning models with seasonal (re)forecasting information, in a hindcasting experiment to evaluate the potential of data driven approaches for seasonal forecasting purposes.

We focussed on 5 different ML methods, which are used to predict discharge and surface water levels of various stations at a national scale (the Netherlands). Input from the European Flood Awareness System and SEAS5 serve as boundary conditions. The ensemble hydrological hindcasts were then evaluated against climatological baseline hindcast with commonly used scores such as anomaly correlation coefficient (ACC), brier skill score (BSS) and continuously ranked probability score (CRPS).

We observed consistently skilful predictions for the first lead months throughout the year for all station and model combinations. Early spring and summer months show increased skill up to several months as a result of snow dynamics that were captured. Furthermore, we show that the choice of ML model only has a limited impact on the overall forecast performance.

With our study we show that a hybrid framework is able to bring location specific skilful seasonal forecast information with global seasonal forecast inputs. At the same time our hybrid approach is flexible and fast, and as such a hybrid framework could easily be adapted to make it even more interesting to water managers and their needs.

How to cite: Hauswirth, S. M., Bierkens, M. F. P., Beijk, V., and Wanders, N.: The potential of a hybrid framework including data driven approaches for hydrological forecasting, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-302, https://doi.org/10.5194/egusphere-egu22-302, 2022.

Virtual presentation
Louise Arnal, Martyn Clark, Vincent Fortin, Alain Pietroniro, Vincent Vionnet, Paul Whitfield, and Andy Wood

Seasonal streamflow forecasts represent critical operational inputs for water sectors and society, for instance for spring flood early warning, water supply, hydropower generation, and irrigation scheduling. Initial hydrological conditions (e.g., snow cover and soil moisture) are an important driver of hydrological predictions on these timescales. In high-latitude and/or high-altitude basins across North America, and the basins downstream of these headwaters, snow is one of the main sources of runoff generation. As a result, data-driven forecasting from snow observations is a well-established approach for operational seasonal streamflow forecasting in the USA (Fleming et al., 2021) and Canada (Zahmatkesh et al., 2019).

As part of the Global Water Futures programme (GWF), we are advancing capabilities for probabilistic streamflow forecasting over North America. The first aim of this work is to benchmark probabilistic seasonal streamflow predictability across the continent. To this end, a data-driven probabilistic seasonal streamflow hindcasting system is being developed and implemented for basins with a nival regime across North America. It uses snow water equivalent measurements from the recent update of the Canadian historical Snow Water Equivalent dataset (CanSWE, 1928–2020; Vionnet et al., 2021) and the Natural Resources Conservation Service (NRCS) manual snow surveys and the SNOTEL automatic snow pillow in the USA. These datasets are gap filled using quantile mapping based on neighbouring snow and precipitation stations (SCDNA dataset; Tang et al., 2020), and subsequently transformed into principal components. These principal components are then used as predictors into a regression model, to generate ensemble hindcasts of streamflow volumes for basins across North America. Preliminary results indicate that this approach is skilful (i.e., better than streamflow climatology) for basins across the Canadian Rockies during the snowmelt season.


Fleming, S. W., Garen, D. C., Goodbody, A. G., McCarthy, C. S., and Landers, L. C.: Assessing the new Natural Resources Conservation Service water supply forecast model for the American West: A challenging test of explainable, automated, ensemble artificial intelligence. Journal of Hydrology, 602, https://doi.org/10.1016/j.jhydrol.2021.126782, 2021.

Tang, G., Clark, M. P., Newman, A. J., Wood, A. W., Papalexiou, S. M., Vionnet, V., and Whitfield, P. H.: SCDNA: a serially complete precipitation and temperature dataset for North America from 1979 to 2018, Earth Syst. Sci. Data, 12, 2381–2409, https://doi.org/10.5194/essd-12-2381-2020, 2020.

Vionnet, V., Mortimer, C., Brady, M., Arnal, L., and Brown, R.: Canadian historical Snow Water Equivalent dataset (CanSWE, 1928–2020), Earth Syst. Sci. Data, 13, 4603–4619, https://doi.org/10.5194/essd-13-4603-2021, 2021.

Zahmatkesh, Z., Sanjeev Kumar, J., Coulibaly, P., and Stadnyk, T.: An overview of river flood forecasting procedures in Canadian watersheds, Canadian Water Resources Journal / Revue canadienne des ressources hydriques, 44, 3, https://doi.org/10.1080/07011784.2019.1601598, 2019.

How to cite: Arnal, L., Clark, M., Fortin, V., Pietroniro, A., Vionnet, V., Whitfield, P., and Wood, A.: A benchmark for probabilistic seasonal streamflow forecasting over North America, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6280, https://doi.org/10.5194/egusphere-egu22-6280, 2022.

Virtual presentation
Dmitri Kavetski, David McInerney, Mark Thyer, Richard Laugesen, Fitsum Woldemeskel, Narendra Tuteja, and George Kuczera

Sub-seasonal streamflow forecasts are used in a wide range of water resource management and planning applications. Practical interest includes forecasts of high flows (e.g., for managing flood events) and low flows (e.g., for managing environmental flows). However, this work reveals that while probabilistic forecasts evaluated over the full flow range can appear statistically reliable, performance specifically for high/low flows can suffer from notable under/over-estimation of forecast uncertainty, respectively. To address this challenge we consider a flow-dependent (FD) nonparametric representation of hydrological forecasting errors, and employ this representation to enhance the existing Multi-Temporal Hydrological Residual Error (MuTHRE) forecasting model. In a case study with 11 Australian catchments, the new MuTHRE-FD model achieves practically significant improvements over the original MuTHRE model in the reliability of forecasted cumulative volumes for high flows out to 7 days, low flows out to 2 days, and mid flows for majority of lead times in the range of 1-30 days. The improved performance of the MuTHRE-FD model provides forecast users with increased confidence in using sub-seasonal streamflow forecasts for applications across a range of flow magnitudes and lead times.

How to cite: Kavetski, D., McInerney, D., Thyer, M., Laugesen, R., Woldemeskel, F., Tuteja, N., and Kuczera, G.: Improving sub-seasonal forecasts of high and low flows using a flow-dependent nonparametric model, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-13078, https://doi.org/10.5194/egusphere-egu22-13078, 2022.

Climate change and floods
Presentation form not yet defined
Miru Seo, Sunghun Kim, jihye Kwon, and Junhaeng Heo

Probable maximum precipitation (PMP) means the maximum precipitation that can occur under the most severe weather conditions at specific area and rainfall duration in watershed. Greenhouse gas emissions in the atmosphere have increased due to industrialization caused by economic development and population growth. As a result, natural disaster damage from climate change is rapidly increasing because of many abnormal climates and phenomena. Futhemore, PMP has been increased due to such climate change. There are several methods for estimating PMP; statistical method, hydrometeorlogical method, and encelope method. In this study, statistical PMP was calculated using observed data up to 2020, and future PMP was estimated using the RCP 4.5 and RCP 8.5 scenarios up to 2100. The Hershfield’s method was used to calculate the statistical PMP, World meteorological organization (WMO) introduced the statistical method suggested by Hershfield (1961) in which frequency factor was 15. However, the frequency factor of 15 was reported to be too large in the area with heavy rainfall and too small in a dry area. Therefore, Hershfield (1965) suggested the range of 5 ~ 20 as a frequency factor.  In this study, PMPs for observed(historical) data and simulated data from RCP 4.5 and RCP 8.5 scenarios were calculated. Then the frequency factors were compared with those suggested by Hershfield. Finally, the derived statistical PMPs were compared with those from hydrometeorlogical method.

How to cite: Seo, M., Kim, S., Kwon, J., and Heo, J.:  Statisrical Probable Maximum Precipitation using RCP 4.5 and RCP 8.5 scenarios, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10937, https://doi.org/10.5194/egusphere-egu22-10937, 2022.

Presentation form not yet defined
Heechul Kim, Miru Seo, Taewon Lee, and Junhaeng Heo

Recently, extreme hydrological phenomena are increasing rapidly due to abnormal climate caused by global warming, and many damages are occurring as the change of precipitation characteristics. The intensity-duration-frequency(IDF) curve is widely applied in practice for designing the hydro-infrastructures. In addition, it is important to predict future changes in rainfall intensity due to climate change.

For this purpose, this study intends to derive the IDF curve, for future periods. In this study, the RCP scenario, a climate change scenario, was used based on historical data (1975-2020) and future rainfall data (2021-2100). Using these data, the stationary and nonstationary regions in the Korean are classified using regional frequency analysis, and the rainfall quantiles for non-stationary regions was calculated using the GEV(1,0,0) model with time varying location parameter. Finally, IDF curves for the historical and future data were derived and analyzed.


How to cite: Kim, H., Seo, M., Lee, T., and Heo, J.: IDF curves in nonstationary regions using regional frequency analysis and RCP scenarios in south korea, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10805, https://doi.org/10.5194/egusphere-egu22-10805, 2022.