Real-time Flood Prediction Research Articles

Optimizing flood predictions by integrating LSTM and physical-based models with mixed historical and simulated data

The current flood forecasting models heavily rely on historical measured data, which is often insufficient for robust predictions due to practical challenges such as high measurement costs and data scarcity. This study introduces a novel hybrid approach that synergistically combines the outputs of traditional physical-based models with historical data to train Long Short-Term Memory (LSTM) networks. Specifically, the NAM hydrological model and the HD hydraulic model are employed to simulate flood processes. Focusing on the Jinhua basin, a typical plains river area in China, this research evaluates the efficacy of LSTM models trained on measured, mixed, and simulated datasets. The LSTM architecture includes multiple layers, with optimized hyperparameters tailored for flood forecasting. Key performance indicators such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Peak-relative Error (PRE) are employed to assess the predictive accuracy of the models. The findings demonstrate that LSTM models trained on mixed datasets with a simulated-to-measured data ratio of less than 2:1 consistently achieve superior performance, exhibiting significantly lower RMSE and MAE values compared to models trained on mixed data with larger data ratios. This highlights the advantage of integrating measured and simulated data, leveraging the strengths of both data types to enhance model accuracy. Despite its advantages, the approach has limitations, including dependence on the quality of simulated data and potential computational complexity. However, the development of this hybrid model marks a significant advancement in flood forecasting, offering a promising solution to the challenges of computational efficiency and data scarcity. Potential applications of this approach include real-time flood prediction and risk management in other flood-prone regions, providing a robust framework for integrating diverse data sources to improve forecasting accuracy.

Flood Water Depth Prediction with Convolutional Temporal Attention Networks

Robust and accurate flood hazard maps are essential for early warning systems and flood risk management. Although physically based models are effective in estimating pluvial flooding, the computational burden makes them difficult to use for real-time flood prediction. In contrast, data-driven models can provide faster flood predictions if trained offline. While most studies have focused on predicting maximum water depth, in this study, we predict pixel-wise water depth maps for entire catchments at a lead time of 2 h. To that end, we propose a deep learning approach that uses a sequence encoding network with temporal self-attention. We also adapt the popular hydrological performance metric Nash–Sutcliffe efficiency (NSE) as our loss function. We test the effectiveness and generalizability of our method using a new dataset called SwissFlood, which consists of 100 catchments and 1500 rainfall events extracted from real observations in Switzerland. Our method produces 2 m spatial resolution flood maps with absolute error as low as 27 cm for water depth exceeding 1 m.

Examining the Feasibility of GeoAI and IoT for Smart Flood Early Warning Systems in Local Communities for Caribbean Urban Spaces

Over the last few decades, flooding has resulted in many problems that significantly impact countries in the Caribbean. This has been especially challenging in urban areas where widespread damage has occurred. In addition, given that over 50% of the world's population lives in urban areas, these locations are deemed to be vulnerable to climate-related disaster events that would further exacerbate the challenges in the region. These urban spaces in the Caribbean have limited access to real-time flood monitoring data for formulating and supporting policies for disaster practitioners to coordinate timely preparedness and mitigation efforts. While flooding is complex, with a series of negative impacts on social and economic sectors, it is essential to provide a basis to support decision-making information on vulnerability and resilience through early warning systems (EWS). However, the main obstacle in creating early warnings in the Caribbean is the suitability and availability of data for real-time flood prediction. While many methods of flood models have been applied in the traditional ways, the methods are complex and not readily adaptable to the context of the local community spaces, especially with limited available data. The Caribbean urban landscape is dynamically changing, which requires careful monitoring for predicting flooding. In a disaster, informing the community is critical for managing flooding risk. The adoption of GeoAI and IoT prepares the next frontier of models for usage in the Caribbean to close the gaps of modeling in data scarcity in urban areas. Consequently, there is a research gap from the perspective of short-term forecasting for sudden rainfall events in urban spaces in the Caribbean. Given the early warning system culture in the Caribbean has an environment of real-time data of scarce resources, it is necessary to forge an approach for real-time forecasting flooding impact in urban spaces. The paper provides a preliminary analysis of the feasibility of machine learning and IoT use in supporting EWS in Caribbean urban spaces.

Real-time peak flow prediction based on signal matching

Real-time peak flow prediction under heavy precipitation is critically important for flood emergency evacuation planning and management. In the case of emergency evacuation, every second matters as a slightly longer lead time could save more lives and reduce the associated social, economic, and health impacts. Here, we present a model (named SIGMA) based on the principle of signal matching to facilitate real-time peak flow prediction at sub-hourly scales (e.g., minutes to seconds). The SIGMA model divides the target watershed into small zones and the heavy precipitation falling into each zone is collected into a small water tank. As the water tank moves downstream and arrives in the watershed outlet, it will discharge the collected precipitation and generate a small single-pulse streamflow signal. By combining all small signals coming from all zones within the watershed, we will be able to generate a synthesized peak flow signal. The proposed model is applied to simulate the peak flow events observed in a real-world watershed to verify its effectiveness in real-time flood prediction. The results suggest that the presented model can reasonably predict three key aspects of a peak flow event, including the peak flow rate, the arrival time of peak flow, and the duration of the peak flow event. The proposed model is demonstrated to be effective in real-time flood prediction and can be used to support flood emergency evacuation planning and management.

Hydrologic Assessment of IMERG Products Across Spatial Scales Over Iowa

Abstract IMERG provides state-of-the-art satellite-based precipitation estimates that combine observations from multiple satellite platforms. This study evaluates IMERG products by examining hydrologic simulations of streamflow at a range of spatial scales. The main objective of this study is to assess the predictive utility of the near-real-time product (IMERG-Early). The assessment also includes the IMERG-Final product that is not available in real time. The authors used MRMS precipitation estimates and USGS streamflow observation data as references for the precipitation and streamflow evaluations during a 5-yr period (2016–20). The precipitation evaluation results show that IMERG-Early yields significant overestimations, particularly during warm months, with higher variability in its conditional distributions, whereas the performance of IMERG-Final seems unbiased. The authors performed hydrologic simulations using the Iowa Flood Center’s Hillslope Link Model with three precipitation forcing products, i.e., MRMS, IMERG-Early, and IMERG-Final. The simulation results reveal that IMERG-Early leads to high hit and false alarm rates due to its overestimation in precipitation and has almost no skill, as measured by the overall performance metric Kling–Gupta efficiency (KGE), in streamflow prediction regarding basin scales ranging from 10 to 30 000 km2. This indicates that the product requires a bias correction before it is useful for real-time flood prediction. The streamflow prediction performance of IMERG-Final seems comparable to that of MRMS at spatial scales greater than 100 km2. This scale limitation is attributable to the IMERG’s product spatial resolution that is inadequate to capture the small-scale variability of precipitation.

Machine Learning Framework Supervised by Hydraulic Mechanical Models for Real-time Pluvial Flood Prediction

Real-time flood prediction in urban areas is an important tool for city emergency planning. Earlier studies suggest two approaches to predict flooding: a supervised machine learning approach based on observed data and a modeling approach for urban environments based on hydraulics. However, the first approach can only be applied in areas where there is sufficient data on flooding and is not accurate enough for prediction. The second approach can provide accurate predictions even for cities that have never experienced flooding, but models of complex urban environments are not suitable for real-time prediction owing to significant computational complexity. Therefore, we propose a third approach, machine learning supervised by a program. This approach consists of training a lightweight neural network using an integrated flood analysis program composed of multiple hydrologically based models. We demonstrate that this trained neural network is 19.8 times faster and and has accuracy comparable to that of the previous modeling approaches.

Scenario-Based Real-Time Flood Prediction with Logistic Regression

This study proposed a real-time flood extent prediction method to shorten the time it takes from the flood occurrence to an alert issuance. This method uses logistic regression to generate a flood probability discriminant for each grid constituting the study area, and then predicts the flood extent with the amount of runoff caused by rainfall. In order to generate the flood probability discriminant for each grid, a two-dimensional (2D) flood inundation model was verified by applying the Typhoon Chaba, which caused great damage to the study area in 2016. Then, 100 probability rainfall scenarios were created by combining the return period, duration, and time distribution using past observation rainfall data, and rainfall-runoff–inundation relation databases were built for each scenario by applying hydrodynamic and hydrological models. A flood probability discriminant based on logistic regression was generated for each grid by using whether the grid was flooded (1 or 0) for the runoff amount in the database. When the runoff amount is input to the generated discriminant, the flood probability on the target grid is calculated by the coefficients, so that the flood extent is quickly predicted. The proposed method predicted the flood extent in a few seconds in both cases and showed high accuracy with 83.6~98.4% and 74.4~99.1%, respectively, in the application of scenario rainfall and actual rainfall.

Development and Evaluation of ICT Operation Support System for Urban Flood Control Facilities

This study was conducted to develop and evaluate the prediction accuracy and effectiveness of “ICT operation support system for urban flood control facilities,” which is installed in Eba catchment area in Hiroshima city, Japan. This system consists of real-time facilities monitoring technology, rainfall data from eXtended RAdar Information Network (XRAIN), and Real-time flood prediction technology. High prediction accuracy is crucial for effective control facility management using ICT operation support system. In this study, ICT operation support system was installed for effective operation of stormwater pump. The prediction accuracy and effectiveness of this system were evaluated based on the past rainfall record from XRAIN. The system takes only three minutes to distribute the urban flood prediction information, and it has been operated stably during the operation. Flood risk reduction effect can be better expected in case of central concentrated rainfall pattern.

Development of a Real-time Flood Prediction Model with Deep Learning and Physical Simulation

Development of a Real-time Flood Prediction Model with Deep Learning and Physical Simulation

Real-Time Flood Disaster Prediction System by Applying Machine Learning Technique

As rainfall intensity varies irregularly, urban floods can cause extreme damage. Furthermore, they are extremely nonlinear phenomena that are complex to analyze. Therefore, a classification-based real-time flood prediction model for urban areas is constructed in this study, by combining a numerical analysis model based on hydraulic theory with a machine learning model. Flood databases are constructed in advance for different rainfall scenarios using the Environmental Protection Agency-Storm Water Management Model (EPA-SWMM) and a two-dimensional inundation model. The flood depth data for each map grid are divided into five categories based on the average flood depth using the Latin hypercube sampling (LHS) and probabilistic neural network (PNN) classification techniques for higher-precision flood range prediction. A model is constructed to predict the representative cumulative volume if the observed rainfall is entered. For spatial expansion of the flood depth with the predicted representative cumulative volume, a system capable of generating a real-time flood map is constructed by linking the cumulative volume of each grid with the representative cumulative volume using linear and nonlinear regression. When compared with the results of a verified two-dimensional (2D) flood model, the developed-model goodness-of-fit is 85%, with a required run time of 1 min 12 s. Using the developed system, rainfall-induced flooding can potentially be predicted, facilitating disaster risk management and minimizing damage to property and health.

Advancing real-time flood prediction in large estuaries: iFLOOD a fully coupled surge-wave automated web-based guidance system

Real-time flood forecasting computational frameworks that can dynamically integrate oceanic, coastal and estuarine processes are becoming essential to provide accurate and timely information for emergency response and planning in largely populated estuaries during extreme events. This study presents a newly developed real-time total water flood guidance system that is fully automated based on the coupled surge-wave (ADCIRC + SWAN) model and provides water level forecasts in the Chesapeake Bay for a lead-time of 84 h twice a day displayed on a web-based public interface. This system improved the current total water level predictions in the Bay (RMSE < 0.12 m) when compared to the existing operational forecasting systems over the period of 6 months (Jan’19-Jun’19). Furthermore, we demonstrated that a bias correction scheme and a multi-member ensemble forecast improve the overall flood prediction. Results suggests that this framework can improve our current capacity to predict total water levels in large estuaries.

Data-Driven Approach for the Rapid Simulation of Urban Flood Prediction

Flooding due to the increase of heavy rainfall caused even larger damage in metropolitan areas. Therefore, numerical simulation and probabilistic models have been used for flood prediction, but the methodologies for real-time flood prediction by drainage district in metropolitan areas are still not sufficient. In this study, a flood scenario database was established by using one- and two-dimensional hydraulic analysis models to propose a realtime urban flood prediction method by drainage districts in metropolitan areas. Flood prediction models were constructed for each drainage district through the Nonlinear Auto-Regressive with eXogenous inputs and Self-Organizing Map (NARX-SOM). Suggested prediction model is a data-driven model because it is based on flood database which composed with diverse flood simulation results. To evaluate the predictive capacity of the models, flood prediction was performed for the actual heavy rainfall in 2010 and 2011 that caused severe flooding in Seoul, Republic of Korea. Flood prediction models for a total of 24 drainage districts were constructed, and it was found that the goodness of fit on the flood area ranged from 68.7 to 89.7%. In terms of the expected inundation map, the predictive power was found to be high when the SOM result with 5 × 5 dimension was mainly used. Through this study, it was possible to identify the predictive capability of the NARX-SOM flood prediction algorithm. The time for inundation map prediction for each area was within two minutes, but the one- and two-dimensional flood simulation usually takes 60–80 minutes. Moreover, when the calculated goodness of fit was examined, the proposed method was found to be a practical methodology that can be helpful in improving flood response capabilities.

Application of 2D numerical simulation for rating curve development and inundation area mapping: a case study of monsoon dominated Dwarkeswar river

ABSTRACT While the implementation of real-time flood prediction and warning systems based on real-time continuous rainfall and streamflow data collection has been implemented in a number of river basins in a number of countries such systems require a reliable network of telemetered rainfall and streamflow gauges within the river basin. Typical of a developing country, the southern part of West Bengal has very few gauging stations with, most stations recording only gauge height data. There is a significant lack of local-level information on rating curves and inundation mapping which can be used for flood management purposes and to identify under what conditions areas of the floodplain need to be evacuated to protect human life. Under these circumstances, stand-alone flood modelling can be undertaken to generate rating curves at key locations and inundation maps which can be used during floods to assist emergency managers implement management actions and to decide when to evacuate communities at risk from rising floodwaters. In this study, the monsoon dominated Dwarkeswar River near Arambag Town, Hooghly District, West Bengal was selected for the prediction of a rating curve and inundation area. The August 2016 flood was simulated using a HEC-RAS (Version 5.0.7) rain-on-grid model. The simulated rating curve and inundation area were validated against the river stage data and flooded area observed in the same year. It was concluded that the model was able to simulate the rating curve and inundation area with an acceptable level of accuracy and that this demonstrated the feasibility of applying this approach to other river basins with a limited number of gauges in order to improve flood management and evacuation practices.

Real-Time Urban Inundation Prediction Combining Hydraulic and Probabilistic Methods

Damage caused by flash floods is increasing due to urbanization and climate change, thus it is important to recognize floods in advance. The current physical hydraulic runoff model has been used to predict inundation in urban areas. Even though the physical calculation process is astute and elaborate, it has several shortcomings in regard to real-time flood prediction. The physical model requires various data, such as rainfall, hydrological parameters, and one-/two-dimensional (1D/2D) urban flood simulations. In addition, it is difficult to secure lead time because of the considerable simulation time required. This study presents an immediate solution to these problems by combining hydraulic and probabilistic methods. The accumulative overflows from manholes and an inundation map were predicted within the study area. That is, the method for predicting manhole overflows and an inundation map from rainfall in an urban area is proposed based on results from hydraulic simulations and uncertainty analysis. The Second Verification Algorithm of Nonlinear Auto-Regressive with eXogenous inputs (SVNARX) model is used to learn the relationship between rainfall and overflow, which is calculated from the U.S. Environmental Protection Agency’s Storm Water Management Model (SWMM). In addition, a Self-Organizing Feature Map (SOFM) is used to suggest the proper inundation area by clustering inundation maps from a 2D flood simulation model based on manhole overflow from SWMM. The results from two artificial neural networks (SVNARX and SOFM) were estimated in parallel and interpolated to provide prediction in a short period of time. Real-time flood prediction with the hydraulic and probabilistic models suggested in this study improves the accuracy of the predicted flood inundation map and secures lead time. Through the presented method, the goodness of fit of the inundation area reached 80.4% compared with the verified 2D inundation model.

River flood prediction using fuzzy neural networks: an investigation on automated network architecture.

Urban floods are one of the most devastating natural disasters globally and improved flood prediction is essential for better flood management. Today, high-resolution real-time datasets for flood-related variables are widely available. These data can be used to create data-driven models for improved real-time flood prediction. However, data-driven models have uncertainty stemming from a number of issues: the selection of input data, the optimisation of model architecture, estimation of model parameters, and model output. Addressing these sources of uncertainty will improve flood prediction. In this research, a fuzzy neural network is proposed to predict peak flow in an urban river. The network uses fuzzy numbers to account for the uncertainty in the output and model parameters. An algorithm that uses possibility theory is used to train the network. An adaptation of the automated neural pathway strength feature selection (ANPSFS) method is used to select the input features. A search and optimisation algorithm is used to select the network architecture. Data for the Bow River in Calgary, Canada are used to train and test the network.

Assessment of Ensemble Flood Forecasting with Numerical Weather Prediction by considering Spatial Shift of Rainfall Fields

This study evaluated the ensemble flood forecasting by using ensemble forecast outputs of the Numerical Weather Prediction (NWP) model with a distributed hydrologic model. Ensemble NWP rainfall with a 2 km horizontal resolution and 30 hour forecast time was assessed whether it can produce suitable rainfall compared with the deterministic rainfall forecast during the typhoon TALAS, 2011. The ensemble flood forecast based on ensemble NWP rainfall was also assessed for hydrological applications. This study presents a methodology using the spatial movement of ensemble rainfall field in order to improve the accuracy of ensemble flood prediction considering the location correction of Quantitative Precipitation Forecast (QPF). This study was conducted using the largest flood events occurred by typhoon TALAS in 2011 with regard to the Futatsuno (356.1 km2) and Nanairo (182.1 km2) dam catchments in the Shingu river basin (2,360 km2) located in Kii Peninsula of Japan. As a result, the flood prediction using the spatial movement of ensemble rainfall field could improve the under-predicted interval of the flood prediction using the original ensemble rainfall. In addition, it was able to reproduce the peak discharge in the first and second prediction intervals for both catchments and to cover the observed value. The method we proposed in this study can be used in hydrological applications such as real-time flood prediction including flood prediction and warning systems, optimized discharge flow for dam operation.

Discharge estimation and forecasting by MODIS and altimetry data in Niger-Benue River

Flooding is one of the most devastating natural hazards in the world and its forecast is essential in flood risk reduction and disaster response decision. The lack of adequate monitoring networks, especially in developing countries prevents near real-time flood prediction that could help to reduce the loss of lives and economic damages. In the last few years, increasing availability of multi-satellite sensors induced to develop new techniques for retrieving river discharge and especially in supporting discharge nowcasting and forecasting activities. Recently, the potential of radar altimetry to estimate water levels and discharge in ungauged river sites with good accuracy has been demonstrated. However, the considerable benefit derived from this technique is attenuated by the low revisit time of the satellite (10 or 35days, depending on the satellite mission) causing delays on the predicting operations. For this reason, sensors with a higher temporal resolution such as the MODerate resolution Imaging Spectroradiometer (MODIS), working in visible/Infra-Red bands, can support flood forecasting.In this study, we performed the forecast of river discharge by using MODIS and we compared it with the radar altimetry and in-situ data along the Niger-Benue River in Nigeria to develop an operational flood forecasting scheme that could help in rapid emergency response and decision making processes. In the first step, four MODIS products (daily and, 8-day from the TERRA and AQUA satellites) at two gauged sites were used for discharge estimation. Secondly, the capability of remote sensing sensors to forecast discharge a few days (~4days) in advance at a downstream section using MODIS is analyzed and also compared with the one obtained by the use of radar altimetry by ENVISAT and Jason-2.The results confirmed the capability of the MODIS data to estimate river discharge with performance indices >0.97 and 0.95 in terms of coefficient of correlation and Nash Sutcliffe efficiency. In particular, RMSE does not exceed 1300m3/s and the fractional RMSE ranges between 0.15 and 0.23. For the forecasting exercise, both altimetry and MODIS provide satisfactory results with positive coefficient of persistence considering 4days of lead time (>0.34). Although altimetry was found to be more accurate in the forecasting of river discharge (RMSE ~350m3/s), the much higher temporal resolution of MODIS guarantees a continuity that is more suitable to address operational activities.

Implementation of Real-Time Flood Prediction and its Application to Dam Operations by Data Integration Analysis System

Despite recent advances in hydrological models and observation technology, the prediction of floods using advanced models and data has not yet been fully implemented for practical use. The major issues in prediction originate from the underlying uncertainty of the initial conditions of the basin and the accuracy of the precipitation forecast. Effective transmission of flood information to corresponding authorities is also necessary when considering countermeasures against an oncoming flood. We present in this article a data archive and model integrated system to overcome these issues. The system realizes flood forecasting by employing a land surface model coupled with hydrological model and an ensemble precipitation forecast model to address the accuracy of initial conditions and precipitation. While the Water and Energy Budget Based Distributed Hydrological Model (WEB-DHM) rigorously estimates the physical state of the basin, the ensemble precipitation forecast model analyzes historical errors in forecasts and returns precipitation ensembles reflecting the uncertainty in the forecast specifically regarding the target basin. A combination of these models yields an ensemble of streamflow forecasts. We further develop a virtual reservoir simulator to enhance the proactive use of forecast information to support decision-making by reservoir managers. These models are integrated into the Data Integration Analysis System (DIAS). The feasibility of the system for practical use is tested against data from recent typhoon events.

Collaborative modelling for active involvement of stakeholders in urban flood risk management

Abstract. This paper presents an approach to enhance the role of local stakeholders in dealing with urban floods. The concept is based on the DIANE-CM project (Decentralised Integrated Analysis and Enhancement of Awareness through Collaborative Modelling and Management of Flood Risk) of the 2nd ERANET CRUE funding initiative. The main objective of the project was to develop and test an advanced methodology for enhancing the resilience of local communities to flooding. Through collaborative modelling, a social learning process was initiated that enhances the social capacity of the stakeholders due to the interaction process. The other aim of the project was to better understand how data from hazard and vulnerability analyses and improved maps, as well as from the near real-time flood prediction, can be used to initiate a public dialogue (i.e. collaborative mapping and planning activities) in order to carry out more informed and shared decision-making processes and to enhance flood risk awareness. The concept of collaborative modelling was applied in two case studies: (1) the Cranbrook catchment in the UK, with focus on pluvial flooding; and (2) the Alster catchment in Germany, with focus on fluvial flooding. As a result of the interactive and social learning process, supported by sociotechnical instruments, an understanding of flood risk was developed amongst the stakeholders and alternatives for flood risk management for the respective case study area were jointly developed and ranked as a basis for further planning and management.

Real-time forecasting urban drainage models: full or simplified networks?

Lead time between rainfall prediction results and flood prediction results obtained by hydraulic simulations is one of the crucial factors in the implementation of real-time flood forecasting systems. Therefore, hydraulic simulation times must be as short as possible, with sufficient spatial and temporal flood distribution modelling accuracy. One of the ways to reduce the time required to run hydraulic model simulations is increasing computational speed by simplifying the model networks. This simplification can be conducted by removing and changing some secondary elements using network simplification techniques. The emphasis of this paper is to assess how the level of urban drainage network simplification influences the computational time and overall simulation results' accuracy. The models used in this paper comprise a sewer network and an overland flow drainage system in both 1D/1D and 1D/2D approaches. The 1D/1D model is used as the reference model to generate several models with different levels of simplifications. The results presented in this paper suggest that the 1D/2D models are not yet suitable to be used in real-time flood prediction applications due to long simulation time, while on the other hand, the simplified 1D/1D models show that considerable reductions in simulation time can be achieved without compromising simulation results (flow and water depth) accuracy.