Flash Flood Forecasting Research Articles

Forecasting extreme hourly rainfall in South Africa for disaster risk reduction: thresholds and return periods

Highest and percentile values determined for daily, hourly and 5-min rainfall data (July 1994 to June 2021) from 64 automatic weather stations across South Africa were used to define extreme hourly and 5-min rainfall intensity. Internationally, 99.9th and 99.99th percentiles are typically considered as thresholds for hourly and sub-hourly extreme rainfall when forecasting for disaster risk reduction assessments. In South Africa (SA), the average of the 99.9th percentile for hourly rainfall values is 29.9 mm/h. This represents a good indicator of extreme hourly rainfall in SA and is a useful threshold for forecasting flash floods. The average highest of the hourly rainfalls for SA, 53.9 mm/h, should be a good indicator of more extreme hourly rainfalls for the country. The average of the 99.99th percentile for 5-min rainfall values is 12.8 mm/5 min, which equates to 2.6 mm/min. Significantly, the 5-min rainfall data is used to establish South African categories based on rainfall intensity and total rainfall, whereby an event can be classified as a cloudburst, downpour or shower. Using the newly established local categories, the severe thunderstorm of 4 April 2000 at Hoedspruit that produced 132.2 mm in 25 min from an intensifying upper air trough system was classified as a cloudburst. Interestingly, the 66.2 mm recorded in 5 min during this event makes it the world record holder for all-time highest 5-min rainfall, passing the previous world 5-min rainfall record of 63.0 mm in 5 min recorded at Porto Bello, Panama, on 29 November 1911. Return periods of expected maximum daily, hourly and 5-min rainfall, based on yearly highest values, were also calculated for South Africa. This study presents expected maximums for 5-min rainfall in return periods of 10, 25, 50 and 100 years, a first for South Africa, which can inform strategies for disaster risk reduction.

A time-space varying distributed unit hydrograph (TS-DUH) for operational flash flood forecasting using publicly-available datasets

Although flash floods originate in upstream (headwater) catchments causing hazards locally, they often occur at short time scales and affect widely while relatively small size catchments respond to intense rainfall systems. Due to the heavy cost of computing and parameterization, high resolution spatially distributed hydrological models are not effectively deployed for global or large-scale flash flood forecasting. This research proposes a Time-Space varying Distributed Unit Hydrograph (TS-DUH) based on publicly-available data for efficient global scale flash flood forecasting. TS-DUH estimates the runoff travel time (and flow velocity) from each location to the catchment outlet based on topographic and hydroclimatic characteristics, while it delineates the runoff-drainage process by emphasizing the heterogeneous and dynamic runoff contribution corresponding to rainfall and soil moisture variations. The excess rainfall is estimated by two methods: the Soil Conservation Service’s (SCS) curve (CN-Runoff), and the widely used Global Flood Monitoring System (GFMS) which provides both long-term (2000-present) well-archived and real-time online global runoff (GFMS-Runoff) prediction based on a distributed hydrological model. TS-DUH is first validated based on 6,324 flash flood events from 281 small-to-medium-sized catchments broadly distributed across the CONUS, using “perfect” precipitation input (i.e., NLDAS-2 for this study). TS-DUH calibrated with GFMS-Runoff shows better performance than using CN-Runoff, with a probability of detection (POD) value of 0.9 for all the events and 71% of them having a KGE metric above 0.5. More importantly, with the real-time satellite rainfall-driven GFMS-Runoff, the long-term (2003–2020) TS-DUH simulation shows better performance than GFMS, indicating promising value in global flash flood prediction of the TS-DUH-integrated GFMS.

Toward Flash Flood Modeling Using Gradient Resolving Representative Hillslopes

AbstractIt is increasingly acknowledged that the acceleration of the global water cycle, largely driven by anthropogenic climate change, has a disproportionate impact on sub‐daily and small‐scale hydrological extreme events such as flash floods. These events occur thereby at local scales within minutes to hours, typically in response to high‐intensity rainfall events associated with convective storms. In the present work, we show that by employing physically based representative hillslope models that resolve the main gradients controlling overland flow hydrology and hydraulics, we can get reliable simulations of flash flood response in small data‐scarce catchments. To this end, we use climate reanalysis products and transfer soil parameters previously obtained for hydrological predictions in an experimental catchment in the same landscape. The inverted mass balance of flood reservoirs downstream is employed for model evaluation in these nearly ungauged basins. We show that our approach using representative hillslopes and climate data sets can provide reasonable uncalibrated estimates of the overland runoff response (flood magnitude, storm volume, and event runoff coefficients) in three of the four catchments considered. Given that flash floods typically occur at scales of a few km2 and in ungauged places, our results have implications for operational flash flood forecasting and open new avenues for using gradient resolving physically based models for the design of small and medium flood retention basins around the world.

Quantitative Precipitation Estimation Using Weather Radar Data and Machine Learning Algorithms for the Southern Region of Brazil

In addressing the challenges of quantitative precipitation estimation (QPE) using weather radar, the importance of enhancing the rainfall estimates for applications such as flash flood forecasting and hydropower generation management is recognized. This study employed dual-polarization weather radar data to refine the traditional Z–R relationship, which often needs higher accuracy in areas with complex meteorological phenomena. Utilizing tree-based machine learning algorithms, such as random forest and gradient boosting, this research analyzed polarimetric variables to capture the intricate patterns within the Z–R relationship. The results highlight machine learning’s potential to improve the precision of precipitation estimation, especially under challenging weather conditions. Integrating meteorological insights with advanced machine learning techniques is a remarkable achievement toward a more precise and adaptable precipitation estimation method.

Generalized Structure of Group Method of Data Handling: Novel Technique for Flash Flood Forecasting

Generalized Structure of Group Method of Data Handling: Novel Technique for Flash Flood Forecasting

Deep Convolutional LSTM for improved flash flood prediction

Flooding remains one of the most devastating and costly natural disasters. As flooding events grow in frequency and intensity, it has become increasingly important to improve flood monitoring, prediction, and early warning systems. Recent efforts to improve flash flood forecasts using deep learning have shown promise, yet commonly-used techniques such as long short term memory (LSTM) models are unable to extract potentially significant spatial relationships among input datasets. Here we propose a hybrid approach using a Convolutional LSTM (ConvLSTM) network to predict stream stage heights using multi-modal hydrometeorological remote sensing and in-situ inputs. Results suggest the hybrid network can more effectively capture the specific spatiotemporal landscape dynamics of a flash flood-prone catchment relative to the current state-of-the-art, leading to a roughly 26% improvement in model error when predicting elevated stream conditions. Furthermore, the methodology shows promise for improving prediction accuracy and warning times for supporting local decision making.

Urban hydrological model (UHM) developed for an urban flash flood simulation and analysis of the flood intensity sensitivity to urbanization

An urban hydrological model (UHM) forced by radar-observed rainfalls was developed for producing rainstorm-related urban inundation maps for obtaining flash flood forecasts. The parameter sets of the land use and land cover (LULC) and urban drainage capacity derive from satellite multispectral images and high spatial resolution GIS datasets, relating to the urban hydrology and hydraulic properties. The hydrodynamics model was based on the simplified shallow water equation (SWE) neglecting the convective acceleration and pressure terms in the St. Venant equation. The intense convective rainstorm that induced heavy flash floods on 21 Jul. 2012 and the large-scale stratiform precipitation that occurred from 19 Jul. to 21 Jul. 2016 in Beijing were deliberately selected to perform flood-simulation and model-validation case studies. The simulation of severe flooding scenarios in the 2012 event was fairly reproduced and verified using media reports. In the case studies, the sensitivity tests verified that the flood intensity (1) would be enhanced slightly on pure impervious surfaces, (2) would be reduced significantly on pure pervious surfaces, and (3) would increase by 30%–60% without a drainage system. No false alarm could be perceived in the simulation of the 2016 event in a prolonging stratiform precipitation. The case studies confirmed the perspectives of using the hydrological model fed by QPEs/QPFs for flash flood forecasting and concluded that the hourly precipitation surpassing 30 mm/h would be indicative in impending flash floods in Beijing metropolis.

Assessing the ability of a new seamless short-range ensemble rainfall product to anticipate flash floods in the French Mediterranean area

Abstract. Flash floods have dramatic economic and social consequences, and efficient adaptation policies are required to reduce their impacts, especially in the context of global change. Developing more efficient flash flood forecasting systems can largely contribute to these adaptation requirements. The aim of this study was to assess the ability of a new seamless short-range ensemble quantitative precipitation forecast (QPF) product, called PIAF-EPS (Prévision Immédiate Agrégée Fusionnée ensemble prediction system) and recently developed by Météo-France, to predict flash floods when used as input to an operational hydrological forecasting chain. For this purpose, eight flash flood events that occurred in the French Mediterranean region between 2019 and 2021 were reanalysed, using a hydrological-modelling chain similar to the one implemented in the French Vigicrues Flash operational flash flood monitoring system. The hydrological forecasts obtained from PIAF-EPS were compared to the forecasts obtained with different deterministic QPFs from which PIAF-EPS is directly derived. The verification method applied in this work uses scores calculated on contingency tables and combines the forecasts issued on each 1 km2 pixel of the territory. This offers a detailed view of the forecast performances, covering the whole river network and including the small ungauged rivers. The results confirm the added value of the ensemble PIAF-EPS approach for flash flood forecasting, in comparison to the different deterministic scenarios considered.

An IoT-based system for monitoring and forecasting flash floods in real-time

An IoT-based system for monitoring and forecasting flash floods in real-time

Towards Improved Flash Flood Forecasting over Dire Dawa, Ethiopia Using WRF-Hydro

Flash floods are increasingly frequent worldwide. Recent flooding in eastern Ethiopia resulted in casualties, destruction of property and interruptions of service. National flash flood forecasts made today primarily consider precipitation, putting less emphasis on surface processes. Enhancing accurate flash flood forecasts by accounting for surface processes and hydrological models together with a deeper understanding of heavy precipitation mechanisms is of paramount importance. To this end, an uncoupled WRF-Hydro model was calibrated for eastern Ethiopia to simulate extreme floods. Sensitivity analysis for August 2006 showed that infiltration runoff, hydraulic soil conductivity and saturated volumetric soil moisture with parameter values of 0.1, 1.5 and 1.0 produced realistic streamflow distribution. Extreme floods in March 2005 and April 2007 were further studied. The results showed that WRF-Hydro replicates temporal and spatial patterns well. Analysis using observational/reanalysis data revealed associated physical processes. Precipitation during these events exceeded long-term climatology and spanned wider areas in eastern Ethiopia. These heavy precipitation events are associated with strong upper-level westerly jet streams and rainfall-conducive circulation anomalies at lower levels. Positive outcomes from WRF-Hydro suggest operational implementation for flood monitoring and early warning systems in forecasting centers.

Correction of Fused Rainfall Data Based on Identification and Exclusion of Anomalous Rainfall Station Data

High-quality rainfall data are crucial for accurately forecasting flash floods and runoff simulations. However, traditional correction methods often overlook errors in rainfall-monitoring data. We established a screening system to identify anomalous stations using the Hampel method, Grubbs criterion, analysis of surrounding measurement stations, and radar-assisted verification. Three rainfall data-fusion methods were used to fuse rainfall station data with radar quantitative precipitation estimation data; the accuracies of the fused data products with and without anomalous data identification were compared. Validation was performed using four 2012 rainfall events in Hebei Province. The 08:00–19:00 July 3 rainfall event had the highest number of anomalous stations (11.5% of the total), while the 01:00–17:00 August 9 event had the lowest number (7.8%). By comparing stations deemed to be anomalous with stations that were actually anomalous, we determined that the accuracy of reference station determination using Hampel’s method and Grubbs’ test was 94.2%. Radar-assisted validation improved the average accuracy of anomalous station identification during the four typical rainfall events from 89.7 to 93.7%. Excluding anomalous data also significantly impacted the efficacy of rainfall-data fusion, as it improved the quality of the rainfall station data. Among the performance indicators, 95% improved after the exclusion of anomalous data for all four rainfall events.

Damages Associated with Excessive Rainfall Outlooks (ERO) and Missed Flash Floods

Abstract Flash flooding is the most damaging and deadly type of flooding event in the continental United States (CONUS), and one of the deadliest hazards worldwide. The Weather Prediction Center’s (WPC) Excessive Rainfall Outlook (ERO) is used to highlight regions at risk of receiving excessive rainfall that can lead to flash flooding. While EROs have been validated by the WPC across several metrics, an analysis of flash floods that were not forecast by EROs, which we define as missed flash floods, has not been performed, nor have damages associated with missed flash floods been examined. Using EROs, flash flood data from the Unified Flooding Verification System (UFVS), and flash flood damage data from the National Centers for Environmental Information (NCEI) Storm Events Database, this research investigates the characteristics of missed flash floods. We find that missed damages occur most frequently in summer and least frequently in winter, but that only 9.2% of flash flood damages and 23.2% of damaging events are missed by the ERO. Missed damaging flash flood events occur frequently in the U.S. Southwest. In this region, missed damages and missed events are primarily attributed to the North American monsoon (NAM). We also investigate forecast flash floods by ERO risk category. High risks incur the most damages despite having the fewest number of damaging events, indicating that the ERO is able to distinguish higher-impact events. The ERO predicts damaging flash floods well, although the Southwest and urban areas broadly could be investigated further to ensure that the potential of damaging flash floods are accurately forecast.

An Efficient Ensemble Technique for Hydrologic Forecasting Driven by Quantitative Precipitation Forecasts

Abstract Uncertainty in quantitative precipitation forecasts (QPFs) from numerical weather prediction (NWP) models manifests in errors in the amounts of rainfall, storm structure, storm location, and timing, among other precipitation characteristics. In flash flood forecasting applications, errors in the QPFs can translate into significant uncertainty in forecasts of surface water flows and their impacts. In particular, the QPF errors in location and structure result in errors on flow paths, which can be highly detrimental in identifying locations susceptible to flash flood impacts. To account for this type of uncertainty, the neighboring pixel ensemble technique (NPET) was devised and implemented as a postprocessing algorithm of deterministic or ensemble outputs from a distributed hydrologic model. The aim of the technique is to address displaced hydrologic responses resulting from location biases in QPFs using a probabilistic approach. NPET identifies a sampling region surrounding each forecast pixel and builds an ensemble of surface water flow values considering the pixel’s physiographic similarities. The probabilistic information produced with NPET can be calibrated through a set of tunable parameters that are adjusted to account for NWP-specific QPF error characteristics. The utility of NPET is demonstrated for the Ellicott City flash flood event on 27 May 2018, using products and tools routinely used in the U.S. National Weather Service for warning operations. Results from this case demonstrate that NPET effectively conveys uncertainty information about QPF precipitation location in a hydrologic context. Significance Statement This study introduces a new method suitable for operational use called the neighboring pixel ensemble technique (NPET). NPET is an algorithm that generates ensemble-based streamflow forecasts accounting for the location uncertainties in quantitative precipitation forecasts (QPFs) without the requirement of multiple hydrologic model runs. NPET is capable of this feat through probabilistic assimilation of a priori QPF displacement information and its uncertainty. The application of NPET with the Flooded Locations and Simulated Hydrographs (FLASH) project shows the technique could be beneficial for flash flood warning operations in the U.S. National Weather Service (NWS). It is envisioned that the application of NPET with Warn-on-Forecast System (WoFS)-forced FLASH outputs will further enhance the quality of flash flood forecasts that support NWS warning operations.

Research on Parameter Regionalization of Distributed Hydrological Model Based on Machine Learning

In the past decade, more than 300 people have died per year on average due to mountain torrents in China. Mountain torrents mostly occur in ungauged small and medium-sized catchments, so it is difficult to maintain high accuracy of flood prediction. In order to solve the problem of the low accuracy of flood simulation in the ungauged areas, this paper studies the influence of different methods on the parameter regionalization of distributed hydrological model parameters in hilly areas of Hunan Province. According to the terrain, landform, soil and land use characteristics of each catchment, we use Shortest Distance, Attribute Similarity, Support Vector Regression, Generative Adversarial Networks, Classification and Regression Tree and Random Forest methods to create parameter regionalization schemes. In total, 426 floods of 25 catchments are selected to calibrate the model parameters, and 136 floods of 8 catchments are used for verification. The results showed that the average values of the Nash–Sutcliffe coefficients of each scheme were 0.58, 0.64, 0.60, 0.66, 0.61 and 0.68, and the worst values were 0.27, 0.31, 0.25, 0.43, 0.35 and 0.59. The random forest model is the most stable solution and significantly outperforms other methods. Using the random forest model to regionalize parameters can improve the accuracy of flood simulation in ungauged areas, which is of great significance for flash flood forecasting and early warning.

An efficient flash flood forecasting system for the un-gaged Meghna basin using open source platform Delft-FEWS

The northeast region of Bangladesh is highly vulnerable to flash floods which often damage the only crop of that region, the boro rice. Developing an efficient and real-time flash flood forecasting system for that region is found essential. The traditional model-centric approach requires plenty of time and manpower to effectively model flash floods. This paper discusses an open-source data-centric integrated flash flood modelling approach using a platform named Delft-FEWS. It is based on three inter-dependent models, e.g., meteorological, hydrologic, and hydrodynamic. The model was calibrated and validated with the flash flood data of 2010-12 and 2013-15, respectively. The performance of the model in real-time was assessed in the context of the 2021 flash flood season. In both cases, it performed satisfactorily. The entire system helps the practitioners to work in a flexible environment to generate flash flood forecasting and early warnings.

Computational Machine Learning Approach for Flood Susceptibility Assessment Integrated with Remote Sensing and GIS Techniques from Jeddah, Saudi Arabia

Floods, one of the most common natural hazards globally, are challenging to anticipate and estimate accurately. This study aims to demonstrate the predictive ability of four ensemble algorithms for assessing flood risk. Bagging ensemble (BE), logistic model tree (LT), kernel support vector machine (k-SVM), and k-nearest neighbour (KNN) are the four algorithms used in this study for flood zoning in Jeddah City, Saudi Arabia. The 141 flood locations have been identified in the research area based on the interpretation of aerial photos, historical data, Google Earth, and field surveys. For this purpose, 14 continuous factors and different categorical are identified to examine their effect on flooding in the study area. The dependency analysis (DA) was used to analyse the strength of the predictors. The study comprises two different input variables combination (C1 and C2) based on the features sensitivity selection. The under-the-receiver operating characteristic curve (AUC) and root mean square error (RMSE) were utilised to determine the accuracy of a good forecast. The validation findings showed that BE-C1 performed best in terms of precision, accuracy, AUC, and specificity, as well as the lowest error (RMSE). The performance skills of the overall models proved reliable with a range of AUC (89–97%). The study can also be beneficial in flash flood forecasts and warning activity developed by the Jeddah flood disaster in Saudi Arabia.

Integration of Satellite Precipitation Data and Deep Learning for Improving Flash Flood Simulation in a Poor-Gauged Mountainous Catchment

Satellite remote sensing precipitation is useful for many hydrological and meteorological applications such as rainfall-runoff forecasting. However, most studies have focused on the use of satellite precipitation on daily, monthly, or larger time scales. This study focused on flash flood simulation using satellite precipitation products (IMERG) on an hourly scale in a poorly gauged mountainous catchment in southwestern China. Deep learning (long short-term memory, LSTM) was used, merging satellite precipitation and gauge observations, and the merged precipitation data were used as inputs for flood simulation based on the HEC-HMS model, compared with the gauged precipitation data and original IMERG data. The results showed that the application of original IMERG data used directly in the HEC-HMS hydrological model had much lower accuracy than that of gauged data and merged data. The simulation using the merged precipitation in HEC-HMS exhibited much better performances than gauged data. The mean NSE improved from 0.84 to 0.87 for calibration and 0.80 to 0.84 for verification, while the lower NSE improved from 0.81 to 0.84 for calibration and 0.73 to 0.86 for verification, which showed that accuracy and robustness were both significantly improved. Results of this study indicate the advances of remote sensing precipitation with deep learning for flash flood forecasting in mountainous regions. It is likely that more significant improvements can be made in flash flood forecasting by employing multi-source remote sensing products and deep learning merging methods considering the impact of complex terrain.

Improving Hourly Precipitation Estimates for Flash Flood Modeling in Data-Scarce Andean-Amazon Basins: An Integrative Framework Based on Machine Learning and Multiple Remotely Sensed Data

Accurate estimation of spatiotemporal precipitation dynamics is crucial for flash flood forecasting; however, it is still a challenge in Andean-Amazon sub-basins due to the lack of suitable rain gauge networks. This study proposes a framework to improve hourly precipitation estimates by integrating multiple satellite-based precipitation and soil-moisture products using random forest modeling and bias correction techniques. The proposed framework is also used to force the GR4H model in three Andean-Amazon sub-basins that suffer frequent flash flood events: upper Napo River Basin (NRB), Jatunyacu River Basin (JRB), and Tena River Basin (TRB). Overall, precipitation estimates derived from the framework (BC-RFP) showed a high ability to reproduce the intensity, distribution, and occurrence of hourly events. In fact, the BC-RFP model improved the detection ability between 43% and 88%, reducing the estimation error between 72% and 93%, compared to the original satellite-based precipitation products (i.e., IMERG-E/L, GSMAP, and PERSIANN). Likewise, simulations of flash flood events by coupling the GR4H model with BC-RFP presented satisfactory performances (KGE* between 0.56 and 0.94). The BC-RFP model not only contributes to the implementation of future flood forecast systems but also provides relevant insights to several water-related research fields and hence to integrated water resources management of the Andean-Amazon region.

Improvement of Heavy Rainfall Simulated with SST Adjustment Associated with Mesoscale Convective Complexes Related to Severe Flash Flood in Luwu, Sulawesi, Indonesia

Flash flooding is an important issue as it has a devastating impact over a short time and in a limited area. However, predicting flash floods is challenging because they are connected to convection systems that rapidly evolve and require a high-resolution forecasting system. In addition, modeling a case study of a mesoscale convective complex (MCC) is the key to improving our understanding of the heavy rainfall systems that trigger flash floods. In this study, we aim at improving modeling skills to simulate a heavy rainfall system related to flash-flood-producing MCCs. We simulated a heavy rainfall event related to a severe flash flood in Luwu, Sulawesi, Indonesia, on 13 July 2020. This flood was preceded by persistent heavy rainfall from 11 to 13 July 2020. In this case, we investigated the role of sea surface temperature (SST) in producing the persistent heavy rainfall over the region. Therefore, we explore the physical and dynamic processes that caused the heavy rainfall using a convection-permitting model with 1 km resolution and an experiment comparing the situation with and without updated SST. The results show that the heavy rainfall was modulated by the development of a pair of MCCs during the night. The pair of MCCs was triggered by a meso-low-pressure system with an anti-cyclonic circulation anomaly over the Makassar Strait and was maintained by the warm front passing between the sea and land over central Sulawesi. This front was characterized by moist–warm and cold–dry low-level air, which may have helped extend the lifetime of the MCCs. The north-westward propagation of the MCCs was due to the interaction between predominantly a south-easterly monsoon and SST anomalies. This study suggested that the long-lived (>10 h) MCCs (>80,000 km2 cloud shield) and persistent precipitation are reproduced well in the updated SST scenario in the WRF model. This relatively simple technique in the running model provides a new strategy for improving flash flood forecasting by better predicting rainfall as an input in the hydrological model. Our findings also indicated a long-lived MCC maintained by back-building mechanisms from night to morning inland as an exceptional MCC, which does not correspond to a previous study.

Towards Predicting Flood Event Peak Discharge in Ungauged Basins by Learning Universal Hydrological Behaviors with Machine Learning

AbstractIn the hydrological sciences, the outstanding challenge of regional modeling requires to capture common and event-specific hydrologic behaviors driven by rainfall spatial variability and catchment physiography during floods. The overall objective of this study is to develop robust understanding and predictive capability of how rainfall spatial variability influences flood peak discharge relative to basin physiography. A machine learning approach is used on a high-resolution dataset of rainfall and flooding events spanning 10 years, with rainfall events and basins of widely varying characteristics selected across the continental United States. It overcomes major limitations in prior studies that were based on limited observations or hydrological model simulations. This study explores first-order dependencies in the relationships between peak discharge, rainfall variability, and basin physiography, and it sheds light on these complex interactions using a multi-dimensional statistical modeling approach. Amongst different machine learning techniques, XGBoost is used to determine the significant physiographical and rainfall characteristics that influence peak discharge through variable importance analysis. A parsimonious model with low bias and variance is created which can be deployed in the future for flash flood forecasting. The results confirm that although the spatial organization of rainfall within a basin has a major influence on basin response, basin physiography is the primary driver of peak discharge. These findings have unprecedented spatial and temporal representativeness in terms of flood characterization across basins. An improved understanding of sub-basin scale rainfall spatial variability will aid in robust flash flood characterization as well as with identifying basins which could most benefit from distributed hydrologic modeling.