HS4.3 | Probabilistic hydro-meteorological forecasts: ensembles, assimilation, predictive uncertainty, verification and decision making
EDI
Probabilistic hydro-meteorological forecasts: ensembles, assimilation, predictive uncertainty, verification and decision making
Co-organized by NH1
Convener: Ruben Imhoff | Co-conveners: Trine Jahr Hegdahl, Albrecht Weerts, Annie Yuan-Yuan Chang, Fredrik Wetterhall
Orals
| Mon, 24 Apr, 14:00–17:55 (CEST)
 
Room 2.31
Posters on site
| Attendance Mon, 24 Apr, 10:45–12:30 (CEST)
 
Hall A
Orals |
Mon, 14:00
Mon, 10:45
This session brings together scientists, forecasters, practitioners and stakeholders interested in exploring the use of ensemble hydro-meteorological forecast and data assimilation 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.

Orals: Mon, 24 Apr | Room 2.31

Chairpersons: Ruben Imhoff, Annie Yuan-Yuan Chang, Fredrik Wetterhall
14:00–14:05
Development of hydro-meteorological forecasting systems
14:05–14:15
|
EGU23-282
|
ECS
|
On-site presentation
Gwyneth Matthews, Hannah Cloke, Sarah Dance, Cinzia Mazzetti, and Christel Prudhomme

Floods are the most common and disastrous natural hazards, but early warning systems can help mitigate the damage by increasing preparedness. However, the products from these early warning systems must be skilful and actionable to be useful in the event of a flood. The European Flood Awareness System (EFAS), part of the European Commission's Copernicus Emergency Management Service, provides complementary flood forecasts to EFAS partners across the whole of Europe. One forecast product provided by EFAS is the ‘post-processed forecast product’ which is generated for the location of approximately 1600 river gauge stations where sufficient historic and near-real time river discharge observations are available. The aim of this product is to provide an error-adjusted forecast up to a maximum lead-time of 15 days. However, the post-processing methodology and the product design of the post-processed forecast product has not evolved over the past few years and therefore may not satisfy user’s changing requirements nor benefit from recent scientific advances. Following a skill assessment of the EFAS post-processed forecasts and a consultation with the EFAS partners a roadmap for future developments of the EFAS post-processed forecast product was designed. Here, we present this roadmap, and the results of the first stages, which include increasing the temporal resolution to 6-hourly timesteps, improvements to the post-processing methodology to better account for the different hydroclimatic regimes across Europe, and changing the post-processed forecast product to make it more locally relevant and useful to the EFAS Partners.

How to cite: Matthews, G., Cloke, H., Dance, S., Mazzetti, C., and Prudhomme, C.: Developing a user-focused flood forecast product for a continental-scale system, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-282, https://doi.org/10.5194/egusphere-egu23-282, 2023.

14:15–14:25
|
EGU23-8578
|
Highlight
|
On-site presentation
Andy Wood, Josh Sturtevant, Naoki Mizukami, and Guoqiang Tang

Water-related applications and decisionmaking for flood forecasting and seasonal water management commonly rely on hydrologic modeling and forecasting that must provide accurate information over large domains as well as at local watershed scales.  We present progress on an experimentally operational hydrologic forecasting system being developed in a project between NCAR and the US Army Corps of Engineers to increase situational awareness in the US Columbia River basin of the Pacific Northwest, where myriad management concerns include flood risk mitigation, hydropower generation, navigation, water supply, recreation, fisheries and environmental management. The components of the system arise from process-oriented hydrologic modeling, analysis and prediction research that has been developed over the last decade in a collaboration between NCAR, federal US water agencies, and several academic institutions. In particular, a calibrated, watershed-based SUMMA hydrologic model and MizuRoute channel routing model are run in both retrospective and real time modes to provide 3-hourly timestep ensemble flood forecasts for short to medium range lead times, as well as ensemble seasonal streamflow and water supply forecasts up to a 1-year lead time. A 36-member meteorological forcing analysis is used to initialize the model states, while ensemble meteorological forecasts from GEFS, sub-seasonal-to-seasonal (S2S) climate forecasts and ESP are used to drive future flow predictions. We present the current status of the system, which runs in real time at NCAR, and discuss different elements of the forecast approach, including model calibration, ensemble initialization, data assimilation, downscaling of NWP, S2S climate forecast use, post-processing, and hindcasting. We also discuss project links to a related streamflow forecasting testbed initiative through the new NOAA Cooperative Institute for Research to Operations in Hydrology (CIROH)

How to cite: Wood, A., Sturtevant, J., Mizukami, N., and Tang, G.: A new process-oriented ensemble hydrological prediction system for flood prediction and water management in the US Pacific Northwest, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-8578, https://doi.org/10.5194/egusphere-egu23-8578, 2023.

14:25–14:35
|
EGU23-10284
|
ECS
|
On-site presentation
|
Louise Arnal, Martyn P. Clark, Vincent Fortin, Alain Pietroniro, Vincent Vionnet, Paul H. Whitfield, and Andy W. Wood

Seasonal streamflow forecasts are critical for many different sectors - e.g., water supply management, hydropower generation, and irrigation scheduling. Initial hydrological conditions (e.g., snow storage and soil moisture) are an important source of hydrological predictability on seasonal timescales. Snowmelt is the main source of runoff generation in high-latitude and/or high-altitude headwaters basins across North America, and the basins downstream. As a result, data-driven forecasting from snow observations is a well-established approach for operational seasonal streamflow forecasting in the USA and Canada.

The aim of this work is to benchmark the skill of probabilistic seasonal streamflow forecasts across North America. To this end, we developed a reproducible data-driven workflow and implemented it for basins with a nival regime across North America. The workflow uses snow water equivalent measurements from the Canadian historical Snow Water Equivalent dataset (CanSWE), 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 neighboring snow and precipitation stations. Principal Component Analysis is then used to define a small set of orthogonal predictor variables. These principal components are used as predictors in a regression model to generate ensemble hindcasts of streamflow volumes for basins across North America. 

Preliminary results for 93 nival basins and 17 glacial basins across Canada suggest that this forecasting method has the ability to provide skilful hindcasts (i.e., better than streamflow climatology) during the snowmelt season with up to 2-3 months lead. The results of this study provide a reference against which alternative forecasting methods (e.g., process-based forecasting models or machine learning approaches) can be assessed in the future.

This work is a contribution of the recently launched Cooperative Institute for Research to Operations in Hydrology (CIROH) initiative that aims to develop next-generation water prediction capabilities. The CIROH program and the Global Water Futures (GWF) program are advancing capabilities for probabilistic streamflow forecasting over North America.

How to cite: Arnal, L., Clark, M. P., Fortin, V., Pietroniro, A., Vionnet, V., Whitfield, P. H., and Wood, A. W.: A reproducible data-driven workflow for probabilistic seasonal streamflow forecasting over North America, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-10284, https://doi.org/10.5194/egusphere-egu23-10284, 2023.

14:35–14:40
Verification methods and hydro-meteorological forecasting system evaluations
14:40–14:50
|
EGU23-4349
|
On-site presentation
Jan Verkade

The reliability diagram is often used to assess the reliability of a set of probabilistic forecasts. It plots the observed relative frequency of an event against its predicted probability. Reliability is evaluated by assessing the distance of the plotting positions from the diagonal which designates 'perfect reliability'. The reliability diagram is easy to construct and to understand. However, there is a caveat: it is not immediately clear which distance from the diagonal would still be considered 'reliable'. Due to finite sample size, even a perfectly reliable forecasting system could result in a reliability diagram where not all points are on the diagonal. Various authors have proposed visual guidance that allows a forecaster to assess whether the observed relative frequencies fall within the variations that can be expected even when a forecasting system is perfectly reliable. These include 'consistency bars', a modified version of the reliability diagram which is plotted on probability paper and  a 'standardized reliability diagram' based on the Normal transform of the Poisson binomial distribution. The present contribution provides another method to visualize the expected deviation from the diagonal: confidence intervals based on the Poisson binomial distribution. The application is demonstrated in various case studies.

How to cite: Verkade, J.: Confidence intervals for the reliability diagram, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-4349, https://doi.org/10.5194/egusphere-egu23-4349, 2023.

14:50–15:00
|
EGU23-5735
|
ECS
|
On-site presentation
Thibault Hallouin, François Bourgin, Charles Perrin, Maria-Helena Ramos, and Vazken Andréassian

Whether they refer to it as validation, verification, or evaluation, hydrological practitioners regularly need to compute performance metrics to measure the differences between observed and simulated/predicted streamflow time series. While the metrics used are often the same (MAE, NSE, KGE, Brier, CRPS, etc.), the tools used to compute them are seldom the same. In some cases, specific tools are not used and the computation of the metrics are directly hand written in the scripts used to analyse model outputs. In addition, the computation of performance metrics is often accompanied with a variety of pre- and post-processing steps that are rarely documented (e.g. handling of missing data, data transformation, selection of events, uncertainty estimation). This can be error prone and hinder the reproducibility of published results. The sharing of tools computing these performance metrics is likely limited by the variety of programming environments in the hydrological community, and by well-established practices in operational environments that are difficult to modify. In order to enable the sharing between researchers and practitioners and move towards more reproducible hydrological science, we argue that an evaluation tool for streamflow predictions must be polyglot (i.e. that it must be usable in several programming languages) and that it must not only compute the performance metrics themselves, but also the pre- and post-processing steps required to compute them. To this end, we present a new open source, polyglot, and compiled tool for the evaluation of deterministic and probabilistic streamflow predictions. The tool, named “evalhyd”, can be used in Python, in R, and as a command line tool. We will present the concept behind its development and illustrate how it works in practice through examples from operational streamflow predictions in France. We will also discuss further steps and remaining challenges in the evaluation of hydrological model predictions.

How to cite: Hallouin, T., Bourgin, F., Perrin, C., Ramos, M.-H., and Andréassian, V.: A polyglot tool for the evaluation of deterministic and probabilistic streamflow predictions, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5735, https://doi.org/10.5194/egusphere-egu23-5735, 2023.

15:00–15:10
|
EGU23-14792
|
On-site presentation
Long-term based evaluation of a real-time regional ensemble flow forecasting system in Catalonia
(withdrawn)
Xinyu li, Carles Corral, Marc Berenguer, and Daniel Sempere-Torres
15:10–15:20
|
EGU23-2950
|
On-site presentation
|
Vincent Fortin, Silvia Innocenti, Étienne Gaborit, Dorothy Durnford, Setigui Keita, Jacob Bruxer, Marie-Amélie Boucher, Shaun Harrigan, Ervin Zsoter, Milena Dimitrijevic, Caroline Sévigny, Nicole O'Brien, and Natalie Gervasi

Environment and Climate Change Canada (ECCC) is in the process of deploying a continental-scale hydrological prediction system known as the National Surface and River Prediction System (NSRPS). Operating at a resolution of 30 arc seconds, NSRPS currently generates Ensemble Streamflow Predictions (ESPs) for over four million grid points. Issued once per day, the ensemble is composed of twenty members and provides 16-day forecasts. NSRPS differs from other hydrological forecasting systems in operational use in Canada by being continental in scope and by relying on an Earth System Modelling (ESM) approach for prediction. In order to assess the value of forecasts issued by NSRPS, a comparison is performed with a similar ESP system available over all of Canada: the Global Flood Awareness System (GloFAS) from the European Centre for Medium-Range Weather Forecasts (ECMWF). The evaluation focusses on the Great Lakes and St. Lawrence watershed as well as the Nelson and Churchill watersheds, each over one million km² in size. 393 streamflow stations are identified where NSRPS and GloFAS agree on the watershed delineation. The comparison is limited to the Spring, Summer and Fall of 2022 due to NSRPS forecast data availability. For most stations, NSRPS performs better than GloFAS in terms of Continuous Ranked Probability Score (CRPS), but the median of the potential CRPS across the 393 stations is very similar for days 3-16. Both systems suffer from a lack of spread, particularly for short lead times, but the problem is slightly more acute for GloFAS. Bayesian Model Averaging (BMA) is explored in order to obtain calibrated probabilistic forecasts that perform better than both NSRPS and GloFAS. 

How to cite: Fortin, V., Innocenti, S., Gaborit, É., Durnford, D., Keita, S., Bruxer, J., Boucher, M.-A., Harrigan, S., Zsoter, E., Dimitrijevic, M., Sévigny, C., O'Brien, N., and Gervasi, N.: Evaluation of continental-scale ensemble hydrological forecasts from Environment and Climate Change Canada: a comparison with forecasts from the Global Flood Awareness System (GloFAS), EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-2950, https://doi.org/10.5194/egusphere-egu23-2950, 2023.

15:20–15:30
|
EGU23-5511
|
Highlight
|
On-site presentation
Jesús Casado Rodríguez, Corentin Carton De Wiart, Stefania Grimaldi, and Peter Salamon

The European Flood Awareness System (EFAS) of the Copernicus Emergency Management Service is an operational forecasting system whose aim is to raise awareness about floods in European transnational rivers. It produces probabilistic, medium-range discharge forecasts twice a day by running the open-source hydrological model LISFLOOD with four different meteorological forcings, two deterministic forecasts from the DWD (German Weather Service) and the ECMWF (European Centre for Medium Range Weather Forecasts), respectively, and two probabilistic forecasts from ECMWF and the Cosmo Consortium (COSMO-LEPS). Based on these forecasts, flood notifications are issued to the EFAS partners if a set of criteria is met: contributing area larger than 2000 km², lead time from 48 to 240 h, at least one deterministic model exceeds the discharge threshold (5-year return period), and at least one probabilistic model predicts 30% exceedance probability of that discharge threshold for three or more consecutive forecasts.

The operational EFAS is being regularly updated, so the configuration of EFAS has changed since the time these notification criteria were defined. For instance, the temporal resolution has increased from daily to 6-hourly, and the spatial resolution is planned to improve from 5km to approximately 1.5 km (1 arcminute).

This study aims at assessing the skill of the notification criteria above presented with the current system setup, and to derive a new set of criteria that optimizes the notification skill. We will focus on three research questions: (i) how can we combine the different models (deterministic and probabilistic) into a grand ensemble and what probability threshold optimizes skill? (ii) Is the persistence criterion (i.e. 3 consecutive forecasts need to provide persistent predictions of high flood risk) adding to the skill both at shorter and larger lead times? (iii) Can we reduce the contributing area threshold without compromising skill?

The study will make use of reanalysis, driven by meteorological observations, and forecast data at over 2300 stations across Europe for a time span from October 2020, which was the release time of the last major change in the EFAS setup, until present. By comparing the reanalysis data with the simulated discharge threshold, a total of 1327 “observed” flood events have been identified in the 2 years from October 2020 to October 2022. The “notified” events will be computed by comparing the forecast data against the notification criteria; we will compute skill metrics (f1, Hanssen-Kuipers) at each daily lead time for different combinations of meteorological forcing and notification criteria in order to find the procedure that maximizes the skill of EFAS notifications and to assess the above research questions.

The outcome of this study will be applied to the EFAS operational system, directly impacting the preparedness of the relevant authorities in future flood events.

How to cite: Casado Rodríguez, J., Carton De Wiart, C., Grimaldi, S., and Salamon, P.: A skill assessment of the European Flood Awareness System notifications, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5511, https://doi.org/10.5194/egusphere-egu23-5511, 2023.

15:30–15:40
Coffee break
Chairpersons: Trine Jahr Hegdahl, Albrecht Weerts
Statistical and machine-learning based processing tools for observations and probabilistic systems
16:15–16:20
16:20–16:30
|
EGU23-1536
|
ECS
|
On-site presentation
Nibedita Samal, Akshay Singhal, Ankit Singh, and Sanjeev Kumar Jha

The Ganga and Godavari are major rivers of India and are known for satisfying the agricultural needs of most part of the country. In the past few decades, these basins have seen increased geohazard scenarios such as floods, flash floods, landslides, etc. The availability of fine-scale precipitation data is a necessity for accurate monitoring and routine issuance of flood warnings. Downscaling of precipitation is a challenging task due to the complex topography of the basin, seasonality of the Indian rainfall, and large-scale influence of meteorological variables. In this study, we set up a Multiple-Point Statistics (MPS) based statistical downscaling approach using available precipitation data of the previous years to generate precipitation data for future at a finer resolution. The MPS approach uses the Training Image (TI) as input, hence we investigate into the adequate length of the past data record required for setting up the statistical model. We also investigate whether the length of data used as a TI in one River basin has any similarity in the other River basin. Further, what is the minimum year of data required to set up the statistical model. This is done by diving the datasets into five sets of TIs with each succeeding set larger than the previous one. This study uses datasets from High Asia Refined Analysis (HAR) (30×30 km) and the Integrated Multi-satellitE Retrievals for GPM (IMERG) (10×10 km) as the reanalysis and observation data respectively for a time period of 2001 to 2014. The idea is to explore if MPS is able to reproduce proper spatial features even with smaller TI data. The work is significant as it will benefit the hydrologists and water resource managers. The work is in progress and the results of the study will be presented at the conference.

How to cite: Samal, N., Singhal, A., Singh, A., and Jha, S. K.: Statistical generation of fine-resolution precipitation data in Ganga and Godavari River basins of India using limited training datasets, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-1536, https://doi.org/10.5194/egusphere-egu23-1536, 2023.

16:30–16:40
|
EGU23-5632
|
ECS
|
On-site presentation
|
Ben Maybee, Cathryn Birch, Steven Boeing, Thomas Willis, Linda Speight, Aurore Porson, Kay Shelton, Charlie Pilling, and Mark Trigg

Surface water flooding (SWF) presents a significant risk to livelihoods, which is projected to increase under climate change. However, forecasting the intense convective rainfall that causes most SWF on the temporal and spatial scales required for effective flood forecasting remains extremely challenging. National scale flood forecasts are currently issued for England and Wales by the Flood Forecasting Centre (FFC). The forecasts are well regarded amongst flood responders, although they feel they would benefit from more location-specific information.

We have developed an enhanced, regional-scale surface water flood forecast system driven by post-processed ensemble rainfall forecasts. We apply a neighbourhood post-processing method to generate percentile-based reasonable worst case rainfall scenarios from the UK operational Met Office Global and Regional Ensemble Prediction System (MOGREPS-UK), a 2.2km horizontal resolution, convection-permitting operational ensemble system that provides forecasts at up to 5 days lead time. Enhanced surface water flood forecasts are then generated by conducting look-ups of meteorological inputs against catchment-level hydrological reference data from the national Environment Agency Risk of SWF mapping database. In this manner the likely severity of flooding associated with forecast rainfall events is assessed by reference to the driving hyetographs for local-scale hydrological modelling, which is available nationally.

Evaluation of the forecasts is informed by both quantitative assessment and qualitative user feedback. We tested the new forecast system over Northern England over summer 2022 and held a co-development workshop with professional and volunteer flood responders, in which we presented participants with existing and new forecasts for recent case-study flood events. We found that responders would routinely use the enhanced forecasts if they were offered as a complement to existing operational provision, with the enhanced information having the strongest impact on decision making for severe, high impact flood events. Responders valued having access to more localised forecast information, which was viewed as useful for decision making, despite the necessity of accepting a higher degree of forecast uncertainty.

We evaluated the SWF forecasts over a historical 10-year period for days with observed SWF events across Northern England and, to assess false alarms, we verified them against SWF forecasts produced using radar observations for several summers’ continuous daily forecasts. The method is effective at forecasting impacts from higher impact flood events, although still generally over-estimates the extent of affected areas. The results of quantitative skill assessment will form a key basis for determining future operational deployment across England and Wales, which we will discuss the feasibility of and requisite next steps.

How to cite: Maybee, B., Birch, C., Boeing, S., Willis, T., Speight, L., Porson, A., Shelton, K., Pilling, C., and Trigg, M.: Surface Water Flood forecasting using reasonable worst case scenarios from ensemble rainfall forecasts, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5632, https://doi.org/10.5194/egusphere-egu23-5632, 2023.

16:40–16:50
|
EGU23-16465
|
ECS
|
Highlight
|
Virtual presentation
Raquel Gómez-Beas, María José Polo, María Fátima Moreno, Manuel del Jesus, and Cristina Aguilar

 

The operation of hydropower systems is carried out based on operation rules and environmental flows requirements. The effects of the highly temporal variability of hydrological regime in Mediterranean areas are more pronounced in Run-of-River hydropower systems, located in mountainous areas as they must often cease operation due to flow rates either below the turbine minimum discharge and the environmental flow requirements, or over the turbine maximum discharge. Conversely, regulated basins with storage systems are more resilient to changes in the short-medium term. In any case, having a forecast operational tool with delimited uncertainty and sufficient reliability would mean an improvement in the hydropower production planning, as well as a decrease in opportunity costs.

A stochastic flow forecast tool is applied in two selected Mediterranean hydropower systems (Southern Spain). In particular, the mini hydropower plants in Poqueira river, as representative of Run-of-River systems; and Los Hurones plant, with a reservoir, are presented. The forcing agents of the increase in humidity were first identified, being the snow and rainfall regimes in Poqueira, and the atmospheric pressure and NAO index in Los Hurones respectively. Secondly, the statistical modelling of dependent variables was carried out with parametric and non-parametric approaches to, finally, generate the probability distribution functions of occurrence of the flow regime. This structure of Bayesian dynamics forecast of water inputs to the plants on a week-month and month-season scale allows the forecast based on observable and verifiable antecedent conditions in quasi-real time.

The operationality of the hydropower plants refers to the probability of producing energy, so that its complementary value, probability of failure, is defined as the number of days in which the plant is not operating. Failure frequency and the associated operationality were calculated from the 250 stochastic replications of the 20-year period of the forcing agent in the selected case study. Including the 250 replications of the inputs allows considering the effect of different combinations of wet and dry years on the variables analysed, and provides the uncertainty associated with both, a certain value of operationality, and a fixed value of the hydrological variable at the desired time scale.

Results reveal that the higher operationality in Poqueira is given between April and May when the snowmelt produces greater flows, with a 25% probability of having less than 4 days of failure, lower than in the winter months (December to February), with a 25% probability of having 8-18 days of failure. In Los Hurones, with a 25% probability, the lowest failure will be 8-12 days between April and June, being significantly higher the rest of the year. Operational graphs obtained from the uncertainty analysis allow estimating how to plan the operation of hydroelectric plants to maximize its production based on the data observed in previous weeks and months.

 

Acknowledgments: This work has been funded by project TED2021-130937A-I00, ENFLOW-MED "Incorporating climate variability and water quality aspects in the implementation of environmental flows in Mediterranean catchments" with the economic collaboration of MCIN/AEI/10.13039/501100011033 and European Union "NextGenerationEU"/Plan de Recuperación, Transformación y Resiliencia.

How to cite: Gómez-Beas, R., Polo, M. J., Moreno, M. F., del Jesus, M., and Aguilar, C.: Stochastic flow forecast tool in Mediterranean watersheds for hydropower plants management at operational time scales, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-16465, https://doi.org/10.5194/egusphere-egu23-16465, 2023.

16:50–17:00
|
EGU23-17141
|
On-site presentation
Mariana Damova and Stanko Stankov

The exploitation of rivers and hydropower reservoirs involves daily monitoring of  the water resources, the meteorological conditions, the status of  the river banks, the flood areas, etc. As maximum river discharge often results in flooding, it is of importance to provide with timely and reliable forecasts of discharge and water levels. Predicting river discharge and water levels has been a subject of hydrological modelling and a topic of serious research. However, only in recent years scholars and practitioners have turned to consider earth observation data for their studies, mainly to compare evidence of flood mapping. We present an approach of using earth observation data to feed AI architectures – EO4AI – and produce forecasts for discharge and water level with significant degrees of accuracy.  Our starting point is that river discharge and water levels depend on a variety of meteorological and environmental factors like precipitations, snow cover, soil moisture, vegetation index and satellite data offer rich variety of datasets, supplying this information. We adopt a pipeline of deep learning architectures consisting of GAN, CNN, LSTM and EA to actually generate forecasts for river discharge and water level by using historic satellite data of the meteorological features listed above, and in-situ measurements for water level and discharge. The satellite data are provided by ADAM  via the NoR service of ESA. ADAM provides data access to satellite datasets from different satellites with semantic relevance for the construction of sediment transport and deposition forecast model as discussed above.  We explain the purpose of the pipeline components. Our forecast models are calibrated for 3, 5, 7, 30 days ahead, and our experiments provide predictions for one year ahead with each of the calibrated models. We discuss experiment results carried out with data from the Danube and Arda rivers, including three dams from cascade Arda and compare them with predictions derived with other methods. We demonstrate the viability of the approach and the reliability of the forecasting results. We further show how the forecasts can be used in hydrodynamic modelling context, for early warning applications and for routine water resources management and monitoring tasks.

How to cite: Damova, M. and Stankov, S.: Forecasting Discharge and Water Levels of Rivers and Dams using Earth Observation and AI, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-17141, https://doi.org/10.5194/egusphere-egu23-17141, 2023.

17:00–17:10
|
EGU23-6796
|
ECS
|
On-site presentation
Yiheng Du, Ilaria Clemenzi, and Ilias Pechlivanidis

A key challenge in continental and global hydro-climate services deals with the incomplete (or even lack of) incorporation of local knowledge and data from the users. Here, we demonstrate the regional skill of seasonal forecasts from large-scale hydro-climate services, while we present a framework that accounts for local data and with the use of machine-learning enhances the seasonal forecasts by better capturing the local information. Five European case studies subject to different hydro-climate conditions and user needs are selected. We test our framework using the E-HYPE hydrological model forced by bias-adjusted ECMWF SEAS5 seasonal meteorological forecasts. We firstly assess the skill of seasonal hydrological forecasts using pseudo-reality and “real” local observations as reference. The skill assessment is driven by the local needs and hence it is conducted for different target hydro-climatic variables and conditions (i.e. floods and droughts). This first evaluation sets the benchmark for quantifying the added value from a machine-learning enhanced hydro-climate service. We next introduce a post-processing workflow to take advantage of the available local observations and potentially improve the forecasting skill. Here, quantile mapping and machine-learning post-processors are tested in the case study areas to further tune the output from the European hydro-climate service towards the local observations. Results from these hybrid seasonal forecasts show potentials to meet the local conditions and consequently address the user expectations from the service. The current work is highlighting the way forward for machine-learning enhanced services that allow tailoring large-scale hydro-climate services using local knowledge and data.

How to cite: Du, Y., Clemenzi, I., and Pechlivanidis, I.: Enhancing the seasonal forecasts from large-scale hydro-climate services to better meet the local conditions, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-6796, https://doi.org/10.5194/egusphere-egu23-6796, 2023.

17:10–17:20
|
EGU23-10361
|
ECS
|
On-site presentation
Freya Saima Aguilar Andrade, Richard Arsenault, and Annie Poulin

Hydrological forecasts often contain biases or uncertainty that make them less useful to water resources system managers. They can, however, be further improved using post-processing methods. Post-processing has the capability to reduce overall bias and improve the uncertainty quantification (spread), in order to enhance the usefulness of the forecasts in decision-making. In this study, a Quantile Mapping (QM) post-processing method was implemented on meteorological forecasts assessed in three different configurations: A monthly, a seasonal, and an annual quantile mapping schemes. The evaluation was carried out over 22 watersheds with different basin areas in the south of Canada. Post-processing methods were trained on ECMWF operational forecasts from 2015-2019 inclusively, then applied on forecasts from 2020 and fed to 8 assimilated hydrological models on each catchment. The hydrological forecasts for the year 2020 were generated at a lead time of 8 days and a timestep of 6 hours. The methodology and results were evaluated using the Continuous Ranked Probability Score (CRPS) metric. Results show that all three QM combinations improve the performance of the forecasts at the most distant lead times, showing significant improvements from day 4. The annual QM implementation was shown to perform the best, followed by seasonal or monthly, depending on the watershed.

How to cite: Aguilar Andrade, F. S., Arsenault, R., and Poulin, A.: Application of weather post-processing methods for operational ensemble hydrological forecasting on multiple catchments in Canada, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-10361, https://doi.org/10.5194/egusphere-egu23-10361, 2023.

17:20–17:30
Predictive uncertainty estimation
17:30–17:40
|
EGU23-7873
|
ECS
|
On-site presentation
Bin Xiong, Shuchen Zheng, Lihua Xiong, and Chong-Yu Xu

Flood risk has been increasing in many basins of the world, due to the global water cycle change driving by the global climate warming. To deal with the nonstationary properties of hydrological extremes, some new concepts, methods and models on flood frequency analysis and risk assessment are developed and applied. However, the robustness of nonstationary frequency analysis models, e.g. those based on the Generalized Additive Models for Location, Scale and Shape, is yet a big concern because the uncertainty of the parameters introduced by the methods and its impact on design flood values are difficult to quantify. This study aims to develop sensitivity degree indexes to assess the robustness of the nonstationary estimation of flood risk rates and their attributions, based on classical and Bayesian statistics, respectively. The results of the case study showed that the proposed method was efficient in identifying significant driving factors of nonstationary flood frequency; the results of the sensitivity index based on the Bayesian statistics showed that the uncertain degree of the nonstationary flood risk estimation increases with uncertain degree of the nonstationary model parameters as expected, but the sensitivity degree is decreased. It is indicated that the degree of influence of model parameters uncertainty on the risk estimation results is model dependent. This study will benefit the application of nonstationary frequency analysis methods in the flood risk assessment and flood design inference fields.

Keywords: Flood frequency analysis; Flood risk; Non-stationarity; Attribution

*This work was supported by the Research Council of Norway (FRINATEK Project 274310).

How to cite: Xiong, B., Zheng, S., Xiong, L., and Xu, C.-Y.: Sensitivity Analysis of Flood Risk Estimation under Nonstationary Conditions: A Case Study of the Weihe River, China, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-7873, https://doi.org/10.5194/egusphere-egu23-7873, 2023.

17:40–17:50
|
EGU23-14944
|
ECS
|
Virtual presentation
Mehnaza Akhter, Munir Ahmad Nayak, and Manzoor Ahmad Ahanger

Streamflow simulated from hydrological models is associated with uncertainty from a variety of sources. The chief sources of uncertainty are: i) errors in the measurement of the observed inputs to the hydrologic models, say precipitation, discharge, and temperature observations, ii) calibration uncertainty associated with algorithms used to estimate the parameters of the model iii) structural uncertainty, associated with incomplete or approximate representation of the catchment with hydrologic models. Operational forecasts generally ignore these uncertainties for important management decisions in water resources, for example, in issuing flood warnings. However, several works have shown that these uncertainties can substantially impact large streamflow forecasts made through hydrologic models. In this work, we explore different error models for estimating the relative contribution of individual error sources to overall uncertainty in the streamflow simulations. Four hydrologic models are used to estimate error distributions at various flow quantiles due to individual sources. The strategy can be adopted to improve the sources contributing to these uncertainties for future predictions from these systems. The approach may be used to reduce the major sources of uncertainty, which will help in reducing the computational efforts in estimating the uncertainties in streamflow simulations.

How to cite: Akhter, M., Nayak, M. A., and Ahanger, M. A.: Estimation of different uncertainties in simulated streamflow from hydrological models, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-14944, https://doi.org/10.5194/egusphere-egu23-14944, 2023.

17:50–17:55

Posters on site: Mon, 24 Apr, 10:45–12:30 | Hall A

Chairpersons: Ruben Imhoff, Annie Yuan-Yuan Chang, Albrecht Weerts
Development and verification of hydro-meteorological forecasting systems
A.85
|
EGU23-15775
Husain Najafi, Rakovec Oldrich, Pallav Kumar Shrestha, Stephan Thober, and Luis Samaniego

An experimental hydrological forecasting system has been developed for Germany (https://www.ufz.de/HS2SForcasts4Germany) at high resolution (1 km). Since early 2021, the hydrological forecasting system provides operational ensemble forecasts for soil moisture droughts at sub-seasonal time scale (HS2S). In the next year, it will be upgraded to streamflow and inundation areas. The mesoscale Hydrologic Model (mHM- www.ufz.de/mhm) with the Multiscale Parameter Regionalization scheme [1,2,3] is used to simulate hydrological forecasts across German catchments. This model is forced with the extended large atmospheric ensemble forecasts from the European Centre for Medium-Range Weather Forecast (ECMWF). The soil moisture index is updated twice per week with associated uncertainties of the initial atmospheric conditions. The initial conditions are obtained with the DWD precipitation and temperature data, similar to the German Drought Monitor (www.ufz.de/droughtmonitor). The hydrological forecasting system was also evaluated for 2021 summer flood in west Germany [4]. The system has shown promising results in flood forecasting as well. This system is based on the EDgE system [5] and can easily be developed across other regions around the world.

Refrences

[1] Samaniego L., Kumar R., & Attinger, S. (2010). Multiscale parameter regionalization of a grid-based hydrologic model at the mesoscale. Water Resour. Res., 46,W05523, doi:10.1029/2008WR007327. WRR Editors' Choice Award 2010

[2] Samaniego L., et al. (2021). mesoscale Hydrologic Model. Zenodo. doi:10.5281/zenodo.1069202, https://doi.org/10.5281/zenodo.1069202

[3] Kumar, R., Samaniego, L., & Attinger, S. (2013). Implications of distributed hydrologic model parameterization on water fluxes at multiple scales and locations. Water Resour. Res., 49(1), 360-379. https://doi.org/10.1029/2012WR012195.

[4] Najafi, H., Rakovec, O., Kumar Shrestha, P., Thober, S., & Samaniego, L. (2022). Post-Assessment of ECMWF-mHM ensemble flood forecasting for 2021 summer flood in west Germany. 2022 AGU Fall meeting. Chicago, IL & online everywhere.

[5] Samaniego et al. (2019). Hydrological Forecasts and Projections for Improved Decision-Making in the Water Sector in Europe. BAMS, 100(12), 2451–2472. https://doi.org/10.1175/BAMS-D-17-0274.1

How to cite: Najafi, H., Oldrich, R., Shrestha, P. K., Thober, S., and Samaniego, L.: A seamless hydrologic forecasting system for Germany, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-15775, https://doi.org/10.5194/egusphere-egu23-15775, 2023.

A.86
|
EGU23-2956
|
Highlight
Ingrid Petry, Fernando Fan, Vinicius Siqueira, Walter Collischonn, Rodrigo Paiva, Erik Quedi, Cléber Gama, Reinaldo Silveira, Camila Freitas, and Cássia Aver

Society’s increasing demand for water and the need for its long-term management have motivated efforts toward improving seasonal streamflow forecasts. Currently, seasonal climate forecasts are routinely issued in meteorological centers around the world, generating information for decision-making and seasonal streamflow forecasting (SSF) studies that are becoming more frequent. Seasonal streamflow forecast skill derives from land surface initial conditions and atmospheric boundary conditions that mostly depend on large-scale climate phenomena (such as ENSO). Thus, seasonal rainfall predictions produced by dynamic climate models that represent ocean-atmosphere interactions may have a positive impact on streamflow forecasts. In South America, seasonal streamflow forecasts are essential for the hydropower sector, which is responsible for ~65% of the electric energy produced in countries such as Brazil. In this work, we assessed seasonal streamflow forecasts over South America based on a continental-scale application of a hydrologic-hydrodynamic model and precipitation forecasts from the ECMWF's fifth generation seasonal forecast system (SEAS5). Seasonal streamflow forecasts (SEAS5-SF) were evaluated against a reference model run and forecast skill was estimated relative to the Ensemble Streamflow Prediction (ESP) method. The bias correction of SEAS5 predicted precipitation improved the performance of the seasonal streamflow forecasts, frequently turning negative skill results into near null to positive skill. Results indicate that the ESP remains a hard-to-beat method for seasonal streamflow forecasting in South America. SEAS5-SF skill was found to be dependent on initialization month, season, basin and forecast lead time, with greater skill on the initialization month lead time. Rivers where the forecast skill is higher were Amazon, Araguaia, Tocantins and Paraná.

 

Acknowledgments: This work presents part of the results obtained during the project granted by the Brazilian Agency of Electrical Energy (ANEEL) under its Research and Development program Project PD 6491-0503/2018 – “Previsão Hidroclimática com Abrangência no Sistema Interligado Nacional de Energia Elétrica” developed by the Paraná State electric company (COPEL GeT), the Meteorological System of Paraná (SIMEPAR) and the RHAMA Consulting company. The Hydraulic Research Institute (IPH) from the Federal University of Rio Grande do Sul (UFRGS) contribute to part of the project through an agreement with the RHAMA company (IAP-001313).

How to cite: Petry, I., Fan, F., Siqueira, V., Collischonn, W., Paiva, R., Quedi, E., Gama, C., Silveira, R., Freitas, C., and Aver, C.: Evaluation of Seasonal Streamflow Forecasts over South American Large Rivers, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-2956, https://doi.org/10.5194/egusphere-egu23-2956, 2023.

A.87
|
EGU23-4846
|
ECS
James Bennett, David Robertson, Durga Lal Shrestha, Kim Robinson, and Andrew Schepen

For streamflow forecasting, calibration of ensemble numerical weather prediction (NWP) models has long been considered a necessary evil. Necessary, because NWP forecasts are usually too biased to force calibrated hydrological models, they often produce unreliable ensembles and may produce forecasts that are less accurate than simple climatology at longer lead times. Evil, because calibration adds complexity to any forecasting system and the calibration process destroys spatial, temporal and inter-variable correlations in the ensemble, which then must be reconstructed in various and usually unsatisfying ways. As ensemble NWPs improve, the degree to which calibration is ‘necessary’ declines.

Here we investigate recent versions of two ensemble NWP models – the European Centre for Medium-range Weather Forecasts ensemble NWP (ECMWF-ens) and the Bureau of Meteorology’s Australian Community Climate and Earth-System Simulator Global Ensemble (ACCESS-GE) NWP. The models are tested over Tasmania, where CSIRO is working with Hydro Tasmania, Australia’s largest generator of hydropower, to establish new ensemble streamflow forecasting systems. Tasmania is mountainous and temperate and features strong rainfall gradients. We apply an existing calibration method – the Catchment-scale Hydrological Precipitation Processor (CHyPP) – which uses a Bayesian Joint Probability model to calibrate ensemble precipitation forecasts.

We show that CHyPP improves reliability in both the ECMWF-ens and ACCESS-GE ensembles, but these improvements come at the cost of a slight reduction in skill at short lead times. Uncalibrated ACCESS-GE forecasts generally produce more biased and less reliable forecasts than ECMWF-ens, and we conclude that calibration is necessary for the ACCESS-GE model, both to reduce biases and improve reliability. However, the improvements in bias from calibrating the ECMWF-ens are negligible in some catchments, with the main benefit being improved reliability at longer lead times. This brings into question the need for calibration of the ECMWF-ens model with CHyPP. We note that these findings may not hold outside the Tasmanian catchments tested, where high resolution ensemble NWP forecasts generally perform well. We discuss the implications of these findings with respect to streamflow forecasts.

How to cite: Bennett, J., Robertson, D., Shrestha, D. L., Robinson, K., and Schepen, A.: Are ensemble NWP forecasts now so good that calibration is unnecessary?, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-4846, https://doi.org/10.5194/egusphere-egu23-4846, 2023.

A.88
|
EGU23-7722
|
Yongshin Lee, Andres Peñuela-Fernandez, Francesca Pianosi, and Miguel Rico-Ramirez

Due to the intensified impact of climate change, the intensity and severity of catastrophic droughts is increasing all over the world. South Korea had also suffered from extreme droughts, including a recent a drought that prolonged from 2013 to 2015 and caused nation-wide crop failures. As one of the measures to anticipate droughts and mitigate damages, past studies have evaluated the use of seasonal forecasts in other regions. However, few studies have focussed the assessment at catchment-scale, which is more suitable for practical water management, and no studies were found on the assessment of seasonal forecasts over South Korea.

Firstly, we assessed the skill of Seasonal Precipitation Forecasts (SPFs) over the 20 catchments in South Korea where the largest reservoirs are located, over the period 2011 to 2020. Ensemble SPFs from 4 weather forecasting centres (ECMWF, UK Met Office, Météo France and DWD) were evaluated, and the skill quantified using the Continuous Ranked Probability Skill Score (CRPSS). We analysed how the skill of the seasonal meteorological forecasts varies across the seasons and years, before and after bias correction, and if the skill can be linked to catchments characteristics. In doing so, we developed a methodology and a Python package to implement it, which is freely available for future applications to other regions (https://github.com/uobwatergroup/seaform.git). The results showed that amongst the four forecasting centres, SPFs by ECMWF were the most skilful in South Korea. In particular, they generally outperformed climatology for up to 2 months of lead time and during the Wet season of drier years for all the lead times. We also found that linear bias correction is useful to correct systematic seasonal biases and there is no significant correlation between the catchment characteristics and forecast skill. Additionally, we investigated the possibility of anticipating dry years from ENSO indices and the forecasts themselves, but we found no significant link.

Secondly, we looked at how skill in seasonal meteorological forecasts propagates into the skill of hydrological forecasts (SHFs). We used the lumped hydrological Tank model to generate ensembles of reservoir inflow from ECMWF’s seasonal forecasts data (precipitation, Evapotranspiration and temperature). Again, we quantified the skill (CRPSS) of SHFs at different lead times, seasons and in wet and dry year. The results showed that the skill of SHFs is highly dependent on the skill of SPFs, and it mimics the seasonal and annual (dry and wet years) features of precipitation forecasts. We also tested 4 different types of processing methods (raw, pre-processing, post-processing, pre/post-processing) found that pre-processing method which corrects bias of weather forcings is the most useful to improve forecast skill.

How to cite: Lee, Y., Peñuela-Fernandez, A., Pianosi, F., and Rico-Ramirez, M.: Assessing seasonal meteorological and hydrological forecasts across South Korea, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-7722, https://doi.org/10.5194/egusphere-egu23-7722, 2023.

A.89
|
EGU23-9579
Alessandro Ceppi, Enrico Gambini, Giovanni Ravazzani, Gabriele Lombardi, Stefania Meucci, and Marco Mancini

Among the natural disasters floods are identified with the greatest impact on highly urbanized areas in economic terms and loss of human lives. Flooding phenomena are more often observed due not only to significant weather events, but also to less intense, but more frequent episodes that undermine the urban drainage system and its interconnections with the river network.

In this context, an alert system has been developed to predict possible river floods for the Seveso, Olona and Lambro rivers (SOL) 24 - 36 hours in advance with a technology based on the sequential functioning of hydrological meteorological, hydraulic engineering calculation models, and visualization on web-GIS.

The proposed flood warning system is, in fact, composed of a physically based and spatially distributed hydrological model for the rainfall-runoff transformation, fed by both observed and forecasted atmospheric forcings of various deterministic and probabilistic meteorological models such as the GFS, Bolam, Moloch, Cosmo I2 and I5, Cosmo-Leps, and WRF.

The data measured on the ground are daily provided by a citizen scientist observation network of the Meteonetwork association and by the official Arpa Lombardia network.

In recent years, this decision support system has also been integrated with a real-time hydrological monitoring and alert network (MoCAP, an Italian acronym which stands for municipal monitoring for flood alerts), developed for the civil protection of the Bovisio-Masciago town, which is located along the Seveso River.

This study describes the benchmark analysis of the coupled forecasting and monitoring systems for local civil protection purposes and its relative performance in the last five years (2018-2022) of functioning.

How to cite: Ceppi, A., Gambini, E., Ravazzani, G., Lombardi, G., Meucci, S., and Mancini, M.: Five years of real time hydro-meteorological forecasts and monitoring for local civil protection: the SOL and MOCAP warning systems, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9579, https://doi.org/10.5194/egusphere-egu23-9579, 2023.

A.90
|
EGU23-9645
|
Highlight
Jens Grundmann, Michael Wagner, and Andy Philipp

Flood forecasting and warning for small catchments are challenging due to the short response time of the catchments on heavy rainfall events. Thus, disaster managers are interested in extended lead times to initiate flood defence measures, which can be obtained by employing forecasts 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.

By this contribution, we introduce the next steps to improve our operational web-based demonstration platform for ensemble hydrological forecasting in small catchments of Saxony, Germany (http://howa-innovativ.hydro.tu-dresden.de/WebDemoLive/). In its current configuration, it uses the Icon-D2-EPS numerical weather prediction product of the German Weather Service (DWD) and provides a hydrologic forecast ensemble of 20 members each three hours, for lead times up to 27 hours. The system is established for three pilot regions with different hydrological settings in Saxony, Germany. Within the second funding period of the HoWa-project improvements are planned for three main parts of the hydro-meteorological ensemble prediction platform considering a) the for observed and forecasted precipitation input, b) the hydrological forecast modelling, and c) the post-processing, visualisation and communication of results including their uncertainty.

In terms of precipitation input we are going to incorporate radar based nowcasting for short term forecasts of the next two hours. Furthermore, we will enlarge the maximum forecast period to 48 hours by exploring the full range of the Icon-D2-EPS NWP forecasts. In addition, forecasts will be updated more frequent. Regarding the hydrological forecasting features will be implemented for flood reservoir operation, and the numbers of catchments/regions will be increased. Finally, a new web-based visualisation dash board will be developed to allow for user oriented analysis and configuration. First steps towards these improvements will be presented.

How to cite: Grundmann, J., Wagner, M., and Philipp, A.: Towards improved hydro-meteorological ensemble forecasting for flood warning in small catchments in Saxony, Germany, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-9645, https://doi.org/10.5194/egusphere-egu23-9645, 2023.

A.91
|
EGU23-14249
|
ECS
Nathalli Rogiski da Silva, Reinaldo Bomfim da Silveira, Camila Freitas, Cassia Silmara Aver Paranhos, André Luiz de Campos, and Leandro Ávila Rangel

The Electric Energy Company of Parana (COPEL GeT), the Meteorological System of Parana (SIMEPAR) and RHAMA Consulting company are undertaking the research project PD-6491-0503/2018 for the development of a hydrometeorological seasonal forecasting for Brazilian reservoirs. The project, sponsored by the Brazilian Electricity Regulatory Agency (ANEEL) under its research and development programme, aims the forecasting of streamflow, at temporal scales ranging from 1 to 270 days, at hydro power enterprises, which are integrated by the National Power System Operator (ONS) through its Interconnected System (SIN). In the present work, we verify the precipitation seasonal product from SEAS5 from ECMWF against three references, namely model climatology, ERA5 reanalysis and in-situ observations. In order to achieve the results, we extract the values from the model, respectively to the closest location of observations within Brazilian rain gauge network, corresponding to hydro power plants, and compare them to the observed values and ERA5 results, for the period from 2000 to 2020. The accuracy measurement was performed by settling a contingency matrix to estimate the probability of detection (POD), probability of false detection (POFD), the ROC curve, the area under the ROC (AUC) and other related metrics. The statistics are gathered by monthly and by season and by considering three quantile thresholds of rainfall distribution for forecasting, computed for 153 reservoirs of the SIN. The results describe a good performance of SEAS5 for either monthly or seasonal forecast if compared to climatology or ERA5, but less accuracy if compared to the rain gauges, mainly for low quantiles. Despite this, by considering the large extension of the country and its climate diversity, we noticed the SEAS5 is quite promising for using on hydrological forecasting at seasonal scale.

How to cite: Rogiski da Silva, N., Bomfim da Silveira, R., Freitas, C., Silmara Aver Paranhos, C., Luiz de Campos, A., and Ávila Rangel, L.: Verification of ECMWF SEAS5 precipitation seasonal forecasting using ERA5 and observations for Brazilian Hidro Power Plants, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-14249, https://doi.org/10.5194/egusphere-egu23-14249, 2023.

A.92
|
EGU23-16025
|
ECS
Cléber Gama, Vinicius Siqueira, Arthur Kolling Neto, Rodrigo Paiva, Fernando Fan, Walter Collischonn, Erik Quedi, Ingrid Petry, Reinaldo Silveira, Camila Freitas, and Cassia Paranhos

Short-to-medium range streamflow forecasting is essential for planning and operating hydropower plants (HPPs). The Brazilian National Interconnected System (SIN) is composed of more than 150 HPPs that are located over a wide range of climate and hydrological conditions. Forecasts of natural inflow into the SIN reservoirs are important to establish optimal operating rules to reduce costs with other energy sources, therefore influencing the prices in the energy market. The objective of this work is twofold: (i) evaluate the skill of ensemble streamflow forecasts for the SIN hydropower plants based on continental-scale hydrological modeling (MGB-SA) and medium-range ECWMF rainfall forecasts (MGB-ECMWF), and (ii) compare the MGB-ECMWF forecasts to those produced operationally by the Electric System National Operator (ONS). The MGB-ECMWF predictions were additionally bias-corrected and updated using quantile mapping and auto-regressive model approaches, and were assessed in the period from 2015 to 2020 in terms of weekly averages. The forecast skill was estimated relative to both streamflow climatology and persistency using the CRPS metric, while the comparison between MGB-ECMWF and operational forecasts was performed using deterministic metrics typically adopted by ONS. The skill of MGB-ECMWF forecasts was substantially improved (especially in the first week) by the use of output correction methods, which were demonstrated to be essential for quantitative streamflow forecasting using a continental-scale hydrological model. The relative performance between ONS and MGB-ECMWF forecasts was quite variable (exhibiting positive and negative values) over the geographical extent of the SIN, although in several locations the MGB-ECMWF forecasts have performed equal to or even better than those issued by ONS. Finally, the results presented here provide insights for investigations and applications of streamflow forecasts using continental-scale modeling and simple output correction techniques, which can bring benefits, for example, in the optimization of the reservoir operation and electricity generation.

Acknowledgments: This work presents part of the results obtained during the project granted by the Brazilian Agency of Electrical Energy (ANEEL) under its Research and Development program Project PD 6491-0503/2018 – “Previsão Hidroclimática com Abrangência no Sistema Interligado Nacional de Energia Elétrica” developed by the Paraná State electric company (COPEL GeT), the Meteorological System of Paraná (SIMEPAR) and the RHAMA Consulting company. The Hydraulic Research Institute (IPH) from the Federal University of Rio Grande do Sul (UFRGS) contribute to part of the project through an agreement with the RHAMA company (IAP-001313).

How to cite: Gama, C., Siqueira, V., Kolling Neto, A., Paiva, R., Fan, F., Collischonn, W., Quedi, E., Petry, I., Silveira, R., Freitas, C., and Paranhos, C.: Medium range ensemble streamflow forecasts for hydropower dams of the Brazilian National Interconnected System, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-16025, https://doi.org/10.5194/egusphere-egu23-16025, 2023.

Statistical and machine-learning based processing tools for observations and probabilistic systems
A.93
|
EGU23-613
|
ECS
Prasad Deshpande, Shivam Tripathi, and Arnab Bhattacharya

Fog, an essential component of the hydrological cycle, is frequently experienced in North India during winter. The reduced visibility due to fog causes many accidents and delays in trains and flights, leading to loss of health and economy. Hence real-time detection and forecast of fog are crucial for mitigating these losses. The study proposes an algorithm to detect fog using satellite observations. The algorithm consists of Bayesian Neural Networks containing weights as probability distributions, unlike ordinary neural networks that treat weights as deterministic parameters. This algorithm provides prediction uncertainty. Both epistemic (data-dependent) and aleatoric (model-dependent) uncertainties are modelled. The final output is the percentage chances of fog which can be suitably thresholded into fog/no-fog. In this study, in situ airport weather records (METAR) are used as reference observations, whereas satellite observations are obtained from the 6 bands of the INSAT-3D geostationary satellite (with a spatial resolution of 4 km). Sub-hourly data of wintertime observations from 2016 to 2020 for seven cities spread across North India are used to train and validate the proposed methodology. The model performs better than the INSAT-3D fog product developed by ISRO. The critical success index of INSAT-3D fog product and the proposed method are 0.17 and 0.44, respectively, whereas Cohen’s Kappa values are 0.22 and 0.50, respectively. The uncertainty analysis shows that aleatoric uncertainty is generally higher than epistemic uncertainty. Moreover, for observations having higher aleatoric uncertainty, the epistemic uncertainty is also high, showing a positive correlation. The real-time predictions are disseminated on the website (www.fog.iitk.ac.in) for the public and scientists. This work is a part of the Fog Prediction using Data Science project.

How to cite: Deshpande, P., Tripathi, S., and Bhattacharya, A.: Bayesian neural network-based satellite fog detection, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-613, https://doi.org/10.5194/egusphere-egu23-613, 2023.

A.94
|
EGU23-11632
Trine Jahr Hegdahl, Thordis Thorarinsdottir, and Kolbjørn Engeland

Small catchments have a quick flood response subject to intense precipitation. Previous studies show a lack of predictability for rain induced floods in these catchments. The aims of this study are to (i) apply processing techniques that focus on improving high to extreme precipitation forecasts, and (ii) evaluate how and if the spatial distribution of precipitation within a catchment affects the flood forecast, and the ultimate effect on flood predictability. Even though preprocessing precipitation has shown to improve flood prognosis, high (intense) precipitation values are still difficult to forecast correctly. In this study, we use precipitation forecasts from the regional weather forecasting model AROME-MetCoOp (MEPS, a 30-member lagged ensemble with a grid resolution of 2.5 km) and the global European Center of Medium-Range Weather Forecasts Integrated Forecasting System (ECMWF HRes and ENS, grid resolution of ~8km and for the ensemble ~16 km). MEPS serves as the reference forecast and is used in the operational flood forecasting system in Norway. For the ECMWF HRes and ENS we will apply techniques focusing on the high precipitation values. We will use Bayesian Model Averaging and apply a sampling approach that ensures that the tail of the posterior distribution is represented. We will also use a quantile regression method that employs an extreme value distribution in the tail. To assess the streamflow forecasts from all ensemble forecasts, a gridded HBV model run at a 3 hourly temporal resolution is used. 

The performance of flood forecasts for the different preprocessing approaches for the precipitation ensemble forecasts will be evaluated. For intense precipitation events the spatial distribution of precipitation within a catchment will be evaluated with an emphasis on the ultimate effect on estimating flood peaks for small and quick responding catchments. 

How to cite: Hegdahl, T. J., Thorarinsdottir, T., and Engeland, K.: Preprocessing intense precipitation forecasts to improve flood predictability for small and quick responding catchments, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11632, https://doi.org/10.5194/egusphere-egu23-11632, 2023.

A.95
|
EGU23-11744
Jungsoo Yoon, Seokhwan Hwang, and Narae Kang

Numerical weather prediction (NWP) provided by the Korea Meteorological Administration (KMA) has rainfall predictions such as typhoons, so it simulates the time point relatively well, but the rainfall intensity of heavy rain, such as the peak of precipitation, is inaccurate to use for flood forecasting. Various methods have been tried for peak smoothing or under-estimation limits due to limitations due to the temporal and spatial scale of the prediction field, but a solution that can be used in practice has not been found. In order to solve this problem, this study developed a technique for correcting the temporal distribution of meteorological forecast data using the representative temporal distribution extracted based on a large amount of past observation data. In order to solve the peak smoothing problem of numerical forecasting, after merging radar quantative precipitation forecasting (QPF) and NWP, the abnormal distribution of precipitation around the peak was corrected using the standard time distribution based on observation data. As a result of correction for typhoon Hinnamno attack in 2022, the accuracy was improved from 68% of the actual rainfall before correction to 85% due to improvement in the peak.

 

Acknowledgement : This research was supported by a grant(2022-MOIS61-002) of ‘Development Risk Prediction Technology of Storm and Flood for Climate Change based on Artificial Intelligence’ funded by Ministry of Interior and Safety(MOIS, Korea).

 

How to cite: Yoon, J., Hwang, S., and Kang, N.: Development of forecasting rainfall accuracy correction method based on observation scenario, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11744, https://doi.org/10.5194/egusphere-egu23-11744, 2023.

A.96
|
EGU23-17334
|
ECS
|
Sung Wook An, Jung Ryel Choi, and Byung Sik Kim

Water resource management requires long-term historical discharge data, and physical hydrology models were widely used. Recently, in the field of water resources, various studies using artificial neural networks have been conducted. In this paper, long-term discharge was estimated using meteorological data and LSTM (Long Short-Term Memory). Study area is selected as Osipcheon watershed in Korea. Observed meteorological data and discharge data were collected for 10 years to training period (2011-2018) and testing period (2019–2020). The potential evaporation data was calculated by Hargreaves formula equation. And NSE (Nash– Sutcliffe Efficiency), RMSE (Root Mean Square Error), and MSE (Mean Square Error) were used to compare LSTM results and observed discharge during the training, test and total period. As a result, NSE, RMSE, and MSE were satisfactory during the total period which showed a high possibility of using the LSTM deep learning technique in the water resource area.

Acknowledgment: This research was supported by a grant(2022-MOIS61-001) of Development Risk Prediction Technology of Storm and Flood For Climate Change based on Artificial Intelligence funded by Ministry of Interior and Safety(MOIS, Korea).

How to cite: An, S. W., Choi, J. R., and Kim, B. S.: Simulation of long-term rainfall runoff using a Long Short-Term Memory (LSTM) networks: Case of Osipcheon watershed in Korea, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-17334, https://doi.org/10.5194/egusphere-egu23-17334, 2023.

A.97
|
EGU23-17521
|
ECS
I-Hsiu Chuang, Gwo-Fong Lin, Ming-Jui Chang, and Yuan-Fu Zeng

Taiwan is located in the subtropical monsoon region, both typhoons and vigorous convection caused by strong southwesterly flow develop seasonal compound disasters.
In addition, the response time for early warning systems of the reservoirs and downstream riverbanks has been shortened due to higher frequency and greater intensity of short-duration rainfall events in recent years. Past studies pointed out that the current water level forecast does not consider the outflow discharge of the reservoir. Therefore, this study proposes a downstream water level forecasting model that considers the outflow discharge of the reservoir, and the model is provided to relevant hazard mitigation centers.
This research has selected the water level of the Taipei bridge as target status and collected data of typhoon and storm events from 2014-2021. These data included the precipitation in the watershed of upstream of Taipei bridge, outflow discharge of Shimen reservoir, outflow discharge of Feitsui reservoir, and tidal of Tamsui river estuary as the alternative factors. Subsequently, building several models based on multiple machine learning, such as RNN, SVM, and LSTM to interface with the constant-quantity rainfall forecast of the Central Weather Bureau, then produce the forecast in the future 12 hours with Multi-Step Forecasting (MSF) about the water level of Taipei bridge.
The result shows that SVM, RNN, LSTM forecast in the future 1 hour precisely, which statistical values of CC are more than 0.97, and root mean square errors of water level are around 0.2 m. As the forecast time is longer, the statistical values of CC decrease around 0.93 and root mean square errors of water level increase around 0.3 m.
However, LSTM is able to learn dependencies between the time series and get more precise outcomes than the SVM and RNN, which is not outstanding initially then performs the best at the last. The proposed water level forecast is proved to improve the accuracy of the forecast in the future 12 hours about the water level of Taipei bridge. Moreover, by coordinating the Quantitative Precipitation Forecast (QPF) and warning water level, the model provides early warning of the future twelve-hour water level, which is not only beneficial to evacuation and operating traversing dock-gate and evacuation gates efficiently, but also conducive to reducing the risk of losses in life and property.
Keywords: Water level, Quantitative Precipitation Forecast, Machine learning, Multi-Step Forecasting

How to cite: Chuang, I.-H., Lin, G.-F., Chang, M.-J., and Zeng, Y.-F.: Water level forecasting of Reservoir downstream by machine learning, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-17521, https://doi.org/10.5194/egusphere-egu23-17521, 2023.