Operational Flood Forecasting Research Articles

Enhancing flood forecasting and warning precision through multi-task deep learning approaches

ABSTRACT This study proposes a multi-task deep learning model for simultaneous prediction of time-series water levels and flood risk thresholds, aiming to enhance flood forecasting precision. Using AutoKeras, single-task and multi-task models were optimised to predict water levels 10–360 min ahead based on 720 min of prior data. The multi-task model consistently outperformed the single-task model across multiple evaluation metrics, including correlation coefficients, RMSE, Nash–Sutcliffe efficiency, and Kling–Gupta efficiency scores. Real-time prediction tests on actual rainfall events further validated the multi-task model's improved accuracy and applicability in operational flood forecasting. The study demonstrates significant progress in flood prediction methodologies, offering a more comprehensive approach to forecasting and categorising flood incidents.

Fast urban inundation simulation with RIM2D for flood risk assessment and forecasting

IntroductionUrban pluvial flooding is a growing concern worldwide as consequence of rising urban population and climate change induced increases in heavy rainfall. Easy-to-implement and fast simulation tools are needed to cope with this challenge.MethodsThis study describes the development of the parsimonious, GPU-accelerated hydraulic model RIM2D for urban pluvial flood simulations. This is achieved by considering the built-up urban area as flow obstacles, and by introducing capacity-based approaches to consider urban drainage by infiltration on pervious surfaces and sewer drainage from roofs and sealed surfaces. The model performance was analyzed by simulating 8 heavy rainfall events in a test area in the city of Dresden, Germany. For these events detailed discharge measurements of sewer discharge are available, providing a unique dataset for evaluating the sewer drainage simulation, which is of high importance for realistic pluvial inundation simulations in urban areas.Results and discussionWe show that the model simulates the temporal dynamics of the sewer discharge and the sewer volume within acceptable ranges. Moreover, the erratic variation of the simulated to measured sewer discharge suggests that the deviations from the measurements are caused by the precipitation input rather than the model simplifications. We conclude that RIM2D is a valid tool for urban inundation simulation. Its short simulation runtimes allow probabilistic flood risk assessments and operational flood forecasts.

Application and Research of Liuxihe Model in the Simulation of Inflow Flood at Zaoshi Reservoir

Floods occur frequently in China, and watershed floods are caused mainly by intensive rainfall, but the spatial distribution of this rainfall is often very uneven. Thus, a watershed hydrological model that enables a consideration of a heterogeneous spatial distribution of rainfall is needed. In this study, a flood forecasting scheme based on the Liuxihe model is established for the Zaoshi Reservoir. The particle swarm optimization (PSO) algorithm is used to optimize the model parameters for flood simulation, and the model’s performance is assessed by a comparison with measured flood data. The spatial distributions of rainfall selected for this study are non-uniform, with much greater rainfall in some areas than in others in some cases. Rainfall may be concentrated in the middle of the basin, in the reservoir area, or in the upstream portion of the basin. The Liuxihe-model-based flood inflow forecasting scheme for the Zaoshi Reservoir demonstrates an excellent simulation effect, with an average peak simulation accuracy of 96.3%, an average peak time of 1.042 h early, and an average Nash–Sutcliffe coefficient of 0.799. Under the condition of an uneven spatial distribution of rainfall, the Liuxihe model simulates floods well. The PSO algorithm significantly improves the model’s simulation accuracy, and its practical application requires only the selection of a typical flood for parameter optimization. Thus, the flood simulation effect of the Liuxihe model is ideal for the watershed above the Zaoshi Reservoir, and the scheme developed in this study can be applied for operational flood forecasting.

Advancing operational flood forecasting, early warning and risk management with new emerging science: Gaps, opportunities and barriers in Kenya

AbstractKenya and the wider East African region suffer from significant flood risk, as illustrated by major losses of lives, livelihoods and assets in the most recent years. This is likely to increase in future as exposure rises and rainfall intensifies under climate change. Accordingly, flood risk management is a priority action area in Kenya's national climate change adaptation planning. Here, we outline the opportunities and challenges to improve end‐to‐end flood early warning systems, considering the scientific, technical and institutional/governance dimensions. We demonstrate improvements in rainfall forecasts, river flow, inundation and baseline flood risk information. Notably, East Africa is a ‘sweetspot’ for rainfall predictability at sub‐seasonal to seasonal timescales for extending forecast lead times beyond a few days and for ensemble flood forecasting. Further, we demonstrate coupled ensemble flow forecasting, new flood inundation simulation, vulnerability and exposure data to support Impact based Forecasting (IbF). We illustrate these advances in the case of fluvial and urban flooding and reflect on the potential for improved flood preparedness action. However, we note that, unlike for drought, there remains no national flood risk management framework in Kenya and there is need to enhance institutional capacities and arrangements to take full advantage of these scientific advances.

A novel smoothing-based long short-term memory framework for short-to medium-range flood forecasting

ABSTRACT A novel smoothing-based long short-term memory (Smooth-LSTM) framework for flood forecasting up to five days ahead is proposed, and compared with the benchmark LSTM (LSTM) model, an Artificial Neural Network (ANN) model, and the conceptual Nedbør Afstrømnings Model (MIKE11 NAM)-Hydrodynamic (HD) (MIKE) hydrological model. This framework was tested in the typical middle Mahanadi River basin (India), which has a tropical monsoon-type climate. Variation of training loss indicated the LSTM network has a higher learning ability at smaller network and batch sizes. The Smooth-LSTM model could predict streamflow with higher Nash-Sutcliffe efficiency of 0.82–0.87 at up to five days lead time with a better reproduction of the observed crucial high peak floods, whereas the corresponding MIKE, ANN and LSTM model-based forecasts were acceptable only up to four-, three- and one-day lead times, respectively. Overall, the Smooth-LSTM model is found to be robust in operational flood forecasting, with lower uncertainty and the least sensitivity to redundant input information.

Evaluation of parametric postprocessing of ensemble precipitation forecasts of the NCMRWF for the Vishwamitri River Basin

AbstractPostprocessing of the ensemble precipitation data improves the bias and uncertainty induced in the numerical weather prediction (NWP) due to perturbations of the initial condition of atmospheric models. The evaluation of the NWP of short-range quantitative weather forecasts provided by the NCMRWF archived in TIGGE, for the Vishwamitri River Basin. The aim of the study is to perform univariate statistical postprocessing using six parametric methods and compare to find the best suitable approach among censored non-homogeneous logistic regression (cNLR), Bayesian model averaging (BMA), logistic regression (logreg), heteroscedastic logistic regression (hlogreg), heteroscedastic extended logistic regression (HXLR), and ordered logistic regression (OLR) methods for the Vishwamitri River Basin. The Brier score (BS), the area under curve (AUC) of receiver operator characteristics (ROC) plots, Brier decomposition, and reliability plots were used for the verification of the probabilistic forecasts. In agreement with the BS and AUC, the cNLR approach for postprocessing performed very well for calibration at all five grids and is preferred, whereas BMA and hlogreg approaches showed relatively poor performance for the Vishwamitri basin. The best post-processed ensemble precipitation will further be employed as an input for the generation of the hydrological forecasts in operational flood forecasting in the Vishwamitri River Basin.

Projecting changes in flood event runoff coefficients under climate change

Climate change is projected to impact on the magnitude, spatial location, and timing of large precipitation events, with evidence for this already present in the historical record. However, determining the impacts of these precipitation changes on floods is non-trivial as floods are a result of many compounding factors. For example, changes in the timing and spatial distribution of rainfalls influence soil moisture conditions prior to the flood-causing precipitation event, which in turn change the peak and volume of the resultant flood. Despite this, the impact of climate-changed precipitation on runoff and flood volumes has yet to be quantified historically or examined under different scenarios of climate change.It has previously been suggested that in Australia, the impact of increased precipitation on floods have been partially offset by decreases in soil moisture prior to storms. Here, we seek to quantify the change in flood-event runoff coefficients resulting from climate change across Australia, where it is hoped that such information can be used to help inform the projections of floods resulting from climate-change-induced precipitation changes. We use an event runoff coefficient to assess the flood runoff response to precipitation events containing the annual maximum three-day rainfall event. The assessment was applied to 467 unimpaired catchments in Australia ranging across arid, temperate, and tropical climates. Catchments in temperate climates largely show decreasing trends in event runoff coefficients while increasing trends were observed in tropical regions. Statistically significant trend magnitudes of up to 0.13 per decade were found.Runoff outputs from a semi-distributed landscape model were then used to assess projected runoff responses under different scenarios of future climates. The outputs of modelled runoff were first validated and shown to adequately capture the observed spatial and temporal trends in flood event runoff coefficients despite underestimating the trend magnitude. The projections showed that the median proportion of rainfall converted to runoff from large rainfall events will decrease into the future across over 80% of unimpaired catchments in Australia under climate change scenarios of RCP 4.5 and RCP 8.5. Future work will investigate how these reductions in flood volume are best represented by rainfall loss models used widely in operational flood forecasts and design applications.

Hydrological model preselection with a filter sequence for the national flood forecasting system in Kenya

AbstractThe choice of model for operational flood forecasting is not simple because of different process representations, data scarcity issues, and propagation of errors and uncertainty down the modeling chain. An objective decision needs to be made for the choice of the modeling tools. However, this decision is complex because all parts of the process have inherent uncertainty. This paper provides a model selection with a filter sequence for flood forecasting applications in data scarce regions, using Kenya as an example building on the existing literature, concentrating on six aspects: (i) process representation, (ii) model applicability to different climatic and physiographic settings, (iii) data requirements and model resolution, (iv) ability to be downscaled to smaller scales, (v) availability of model code, and (vi) possibility of adoption of the model into an operation flood forecasting system. In addition, we review potential models based on the proposed criteria and apply a decision tree as a filter sequence to provide insights on the possibility of model applicability. We summarize and tabulate an evaluation of the reviewed models based on the proposed criteria and propose the potential model candidates for flood applications in Kenya. This evaluation serves as an objective model preselection criterion to propose a modeling tool that can be adopted in development and operational flood forecasting to the end‐users of an early warning system that can help mitigate the effects of floods in data scarce regions such as Kenya.

Hydrological performance of the ERA5 reanalysis for flood modeling in Tunisia with the LISFLOOD and GR4J models

Study regionThis study is focused on 6 catchments in Tunisia, in North Africa, under a semi-arid climate where daily river discharge and precipitation observations are available between 1992 and 2006. Study focusThe distributed hydrological model LISFLOOD has been compared with a lumped model, GR4J, to reproduce river runoff and floods, using observed rainfall interpolated from rain gauges and from the ERA5 reanalysis. New hydrological insights for the region under studyLISFLOOD and GR4J with their default parameters have little skill in Tunisia, but when calibrated, both hydrological models provide similar performance to reproduce river discharge,with a mean KGE of 0.65 for GR4J driven by observed rainfall, and 0.59 with ERA5, and similarly a mean KGE of 0.59 for LISFLOOD driven by observed rainfall and 0.48 with ERA5. All model simulations perform very similarly using either observed rainfall data from rain gauges interpolated over the basins or ERA5 rainfall, indicating its value as an alternative to observations in data-scarce areas such as Tunisia. The development of regional databases of river discharge observations in North Africa could allow the calibration of such hydrological models on a larger scale for operational flood modeling and forecasting.

Snow Multidata Mapping and Modeling (S3M) 5.1: a distributed cryospheric model with dry and wet snow, data assimilation, glacier mass balance, and debris-driven melt

Abstract. By shifting winter precipitation into summer freshet, the cryosphere supports life across the world. The sensitivity of this mechanism to climate and the role played by the cryosphere in the Earth's energy budget have motivated the development of a broad spectrum of predictive models. Such models represent seasonal snow and glaciers with various complexities and generally are not integrated with hydrologic models describing the fate of meltwater through the hydrologic budget. We present Snow Multidata Mapping and Modeling (S3M) v5.1, a spatially explicit and hydrology-oriented cryospheric model that simulates seasonal snow and glacier evolution through time and that can be natively coupled with distributed hydrologic models. Model physics include precipitation-phase partitioning, snow and glacier mass balances, snow rheology and hydraulics, a hybrid temperature-index and radiation-driven melt parametrization, and a data-assimilation protocol. Comparatively novel aspects of S3M are an explicit representation of the spatial patterns of snow liquid-water content, the implementation of the Δh parametrization for distributed ice-thickness change, and the inclusion of a distributed debris-driven melt factor. Focusing on its operational implementation in the northwestern Italian Alps, we show that S3M provides robust predictions of the snow and glacier mass balances at multiple scales, thus delivering the necessary information to support real-world hydrologic operations. S3M is well suited for both operational flood forecasting and basic research, including future scenarios of the fate of the cryosphere and water supply in a warming climate. The model is open source, and the paper comprises a user manual as well as resources to prepare input data and set up computational environments and libraries.

Ecological aspects of forecasting the hydrological situation

The article considers peculiarities of prediction of dangerous hydrological phenomena on water bodies of the Voronezh region. The types of activities performed by the Federal Service for Hydrometeorology and Environmental Monitoring (Federal State Budgetary Institution (FSBI)"Voronezh Regional Center for Hydrometeorology and Environmental Monitoring") are presented. The schematic functional diagram of the system of automated monitoring of hydrological characteristics and operational forecasting of high waters and floods, proposed.

Continental Hydrologic Intercomparison Project, Phase 1: A Large‐Scale Hydrologic Model Comparison Over the Continental United States

AbstractHigh‐resolution, coupled, process‐based hydrology models, in which subsurface, land‐surface, and energy budget processes are represented, have been applied at the basin‐scale to ask a wide range of water science questions. Recently, these models have been developed at continental scales with applications in operational flood forecasting, hydrologic prediction, and process representation. As use of large‐scale model configurations increases, it is exceedingly important to have a common method for performance evaluation and validation, particularly given challenges associated with accurately representing large domains. Here, we present phase 1 of a comparison project for continental‐scale, high‐resolution, processed‐based hydrologic models entitled the Continental Hydrologic Intercomparison Project (CHIP). The first phase of CHIP is based on past Earth System Model intercomparisons and comprised of a two‐model proof of concept comparing the ParFlow‐CONUS hydrologic model, version 1.0 and a NOAA US National Water Model configuration of WRF‐Hydro, version 1.2. The objectives of CHIP phase 1 are: (a) describe model physics and components, (b) design an experiment to ensure a fair comparison, and (b) assess simulated streamflow with observations to better understand model bias. To our knowledge, this is the first comparison of continental‐scale, high‐resolution, physics‐based models which incorporate lateral subsurface flow. This model intercomparison is an initial step toward a continued effort to unravel process, parameter, and formulation differences in current large‐scale hydrologic models and to engage the hydrology community in improving hydrology model configuration and process representation.

Управление информационными процессами в системах моделирования природных объектов

The paper proposes a methodological approach and software for managing information processes in systems for modeling natural objects on the example of a system for operational forecasting of fl oods. The proposed approach is based on application of cyber-physical systems, the industrial Internet and fog computing concepts. The results of testing the technologies for managing information processes demonstrate increasing both reliability of modeling and effi ciency of using the fl ood forecasting system computing resources.

Operational flood forecasting and warning under the changing environment in China

Abstract. Flooding has been the most severe hazard in China from time immemorial, due to its special geographical, climatic and socio-economic conditions. In recent years, against the background of global climate change and rapid urbanization, extreme hydro-meteorological events have obviously increased in China, in turn affecting the sustainable social and economic development. This paper analyses how modern flood risk management has evolved from the early “build and protect” flood control approach to a broader flood risk management approach by analyzing what has happened in China and the role of major floods in this evolution. The development of flood forecasting models and systems in terms of how they have informed decision-making as flood risk management has evolved over the years. The challenging of recognizing and dealing with forecast uncertainty and flood risk in decision-making is also analyzed in the paper.

Monitoring soil moisture at the catchment scale – A novel approach combining antecedent precipitation index and radar-derived rainfall data

Knowledge about soil moisture is important for event-based rainfall-runoff models but monitoring conditions at the catchment scale is not a trivial task. Soil moisture is highly variable in space and time, particularly in dry climates with seasonal and spatially heterogeneous rainfall. Point measurements are difficult to upscale, and remotely sensed (RS) data often lack in spatial or temporal resolution for local or regional studies. Longer latency periods – the time required before data becomes available – of some RS data make them less applicable to time-sensitive analyses such as flash flood forecasting. This study evaluated a novel approach for estimating catchment-scale volumetric soil moisture using an antecedent precipitation index (API) -based model. The model was trained and tested using in-situ soil moisture measurements collected during a 3-month field sampling campaign in a 142 km2 study area in central New Mexico. The calibrated model was applied at the catchment scale to produce soil moisture grids from radar-derived rainfall estimates. Model performance, resolution and latency were compared to satellite-based soil moisture estimates. Benefits of the proposed new method include high spatial resolution (1 × 1 km or less depending on the precipitation data source) and high prediction accuracy (root mean square errors 0.014–0.018 m3/m3). Given the short latency period for radar-derived rainfall data, the method has potential for use in operational flood risk assessment and forecasting.

Evaluation of Radar-Gauge Merging Techniques to Be Used in Operational Flood Forecasting in Urban Watersheds

Demand for radar Quantitative Precipitation Estimates (QPEs) as precipitation forcing to hydrological models in operational flood forecasting has increased in the recent past. It is practically impossible to get error-free QPEs due to the intrinsic limitations of weather radar as a precipitation measurement tool. Adjusting radar QPEs with gauge observations by combining their advantages while minimizing their weaknesses increases the accuracy and reliability of radar QPEs. This study deploys several techniques to merge two dual-polarized King City radar (WKR) C-band and two KBUF Next-Generation Radar (NEXRAD) S-band operational radar QPEs with rain gauge data for the Humber River (semi-urban) and Don River (urban) watersheds in Ontario, Canada. The relative performances are assessed against an independent gauge network by comparing hourly rainfall events. The Cumulative Distribution Function Matching (CDFM) method performed best, followed by Kriging with Radar-based Error correction (KRE). Although both WKR and NEXRAD radar QPEs improved significantly, NEXRAD Level III Digital Precipitation Array (DPA) provided the best results. All methods performed better for low- to medium-intensity precipitation but deteriorated with the increasing rainfall intensities. All methods outperformed radar only QPEs for all events, but the agreement is best in the summer.

Current status and advancement suggestions of ice-jam flood hazard and risk assessment

In many northern rivers, ice-jam flooding can be more severe than open-water flooding, leading to human casualties, damages to property and infrastructure, and adverse impacts on the ecology. Consequently, ice jam related flooding is a major concern for many riverside communities, water authorities, insurance companies, and government agencies. Ice-jam flood hazard delineation and risk analysis are important measures for flood preparation, mitigation, and management strategies. Although methodologies and techniques for open-water flood hazard and risk assessment are well established, methodologies and techniques for ice-jam flood hazard and risk assessment are often unavailable or less developed. In addition to this, a considerable number of studies have been conducted in the context of flood management, but a very limited number of studies have been carried out in real-time flood risk analysis during operational flood forecasting. In this paper, the current status of ice-jam flood hazard delineation and risk analysis is discussed. A framework for real-time risk analysis for operational flood forecasting is also discussed. Finally, current limitations and future requirements for developing effective ice-jam flood hazard delineation and risk analysis methodologies are provided.

Effective flood forecasting at higher lead times through hybrid modelling framework

Artificial neural network has been acknowledged as a promising tool for accurately forecasting the streamflow. However, several constraints limit its application in operational hydrology; the primary one being, the non-availability (practically not available) of observed streamflow to be used as input to the model. In addition, progressive reduction in forecast accuracy along with an increase in forecast lead time makes the ANN model less amenable for operational flood forecasting. This study proposes a hybrid model (M4) that combines the strength of a physically-based distributed hydrological model and a data-driven model (ANN), which purges the above concerns. The proposed model was tested using two different case studies (ST-I and ST-II), with respective time steps as fifteen minutes and one day. While the first case study demonstrated the application of the model on a real-time streamflow forecasting scenario, the second one represented a continuous streamflow simulation. The performance of the hybrid model was evaluated, and the results showed that the proposed hybrid model (M4) performed reasonably well at higher lead times (NSE = 0.91 for ST-I and 0.77 for ST-II at time step 't+4'). The proposed model (M4), was further tested for its ability to work with forecasted rainfall (synthetically generated), using the data of ST-I and the model performed well with NSE of 0.95. Though the performance of the models was drawn only up to 't+4', it was illustrated that the proposed hybrid model could be used to generate forecasted streamflow hydrograph corresponding to a full flood event well in advance. This characteristic of the proposed model enhances its utility in operational flood forecasting.

Time Variant Sensitivity Analysis of Hydrological Model Parameters in a Cold Region Using Flow Signatures

The complex terrain, seasonality, and cold region hydrology of the Nelson Churchill River Basin (NCRB) presents a formidable challenge for hydrological modeling, which complicates the calibration of model parameters. Seasonality leads to different hydrological processes dominating at different times of the year, which translates to time variant sensitivity in model parameters. In this study, Hydrological Predictions for the Environment model (HYPE) is set up in the NCRB to analyze the time variant sensitivity analysis (TVSA) of model parameters using a Global Sensitivity Analysis technique known as Variogram Analysis of Response Surfaces (VARS). TVSA can identify parameters that are highly influential in a short period but relatively uninfluential over the whole simulation period. TVSA is generally effective in identifying model’s sensitivity to event-based parameters related to cold region processes such as snowmelt and frozen soil. This can guide event-based calibration, useful for operational flood forecasting. In contrast to residual based metrics, flow signatures, specifically the slope of the mid-segment of the flow duration curve, allows VARS to detect the influential parameters throughout the timescale of analysis. The results are beneficial for the calibration process in complex and multi-dimensional models by targeting the informative parameters, which are associated with the cold region hydrological processes.

Operationalizing a flood forecasting decision support system for Ayeyarwady river, Myanmar

ABSTRACT Myanmar is one of the developing countries of South East Asia (SEA), wherein millions of people are exposed to risk of flooding every year. The Department of Meteorology and Hydrology (DMH), Myanmar is thus, bestowed with the responsibility of flood forecasting within the country for early warning. This study presents a new web-based Decision Support System (DSS), for DMH’s use in operational flood forecasting. The new DSS, that has a forecast lead time of 24–72 h, is a significant improvement over the existing forecasting systems used by DMH, which had a lead time of 4–24 h. Moreover, the new DSS couples Hydrological Engineering Centre Hydrological Modelling System (HEC-HMS) with Weather and Research (WRF) rainfall forecast for Ayeyarwady river basin (ARB), the principle river basin of the country. In addition to the hydro-meteorological coupling, core modules of the DSS include rainfall forecasting, bias correction and flow error correction. The hydrological model setup of the DSS was calibrated and validated for the periods 2008–2012 and 2013–2014 with observed rainfall. The DSS was then actively deployed and tested operationally by DMH using bias-corrected WRF rainfall forecasts for the period 2015–2017. Performance of the system is very good in forecasting flows and floods within ARB, and also provides additional lead time to DMH for issuing the flood advisories. Hence, the novel web-based flood forecasting DSS, is an important applied research contribution, that is being successfully used by the operational staff of DMH for short-term flood forecasting in Myanmar and has the potential to improve flood resiliency of similar agencies in the SEA region.