Flash Flood Forecasting Research Articles

Evaluation of two hydrometeorological ensemble strategies for flash-flood forecasting over a catchment of the eastern Pyrenees

Abstract. This study aims at evaluating the performances of flash-flood forecasts issued from deterministic and ensemble meteorological prognostic systems. The hydrometeorological modeling chain includes the Weather Research and Forecasting Model (WRF) forcing the rainfall-runoff model MARINE dedicated to flash floods. Two distinct ensemble prediction systems accounting for (i) perturbed initial and lateral boundary conditions of the meteorological state and (ii) mesoscale model physical parameterizations have been implemented on the Agly catchment of the eastern Pyrenees with three subcatchments exhibiting different rainfall regimes. Different evaluations of the performance of the hydrometeorological strategies have been performed: (i) verification of short-range ensemble prediction systems and corresponding streamflow forecasts, for a better understanding of how forecasts behave; (ii) usual measures derived from a contingency table approach, to test an alert threshold exceedance; and (iii) overall evaluation of the hydrometeorological chain using the continuous rank probability score, for a general quantification of the ensemble performances. Results show that the overall discharge forecast is improved by both ensemble strategies with respect to the deterministic forecast. Threshold exceedance detections for flood warning also benefit from large hydrometeorological ensemble spread. There are no substantial differences between both ensemble strategies on these test cases in terms of both the issuance of flood warnings and the overall performances, suggesting that both sources of external-scale uncertainty are important to take into account.

A Flood Forecasting Model based on Wireless Sensor and Actor Networks

Flood forecasting is a challenging area of research that can help to save precious lives by timely intimating about the flood possibilities. The role of advancements in computing and allied technologies has moved the research towards a new horizon. Researchers from all over the world are using Artificial Neural Networks (ANN), Global Information Systems (GIS), and Wireless Sensor Networks (WSN) based schemes for flash flood forecasting and hydrological risk analysis. ANN and GIS-based solutions are much costly whereas the analysis and prediction using WSN require much less cost for infrastructure deployment. It will enable the use of flood prediction mechanisms in the third world and poor countries. New variation in storage capacity can be a vital source to eliminate or reduce the chance of flood. By considering this observation, it is proposed to develop a generic flood prediction scheme that can manage the system as per resources and environmental conditions. A heterogeneous WSN has considered where powerful Collector Nodes (CN) continuously takes values from member sensor nodes in the region. CN transmits the alerts to Administration Unit (AU) when threshold values are crossed. It is proposed to save the threshold values from all water sources like storage capacity, water inflow, and outflow in the repository for decision making. Moreover, environmental parameters including water level, humidity, temperature, air pressure, rainfall, soil moisture, etc. are considered for the analysis of flood prediction. We have also considered the evaluation of these parameters in specified boundary values that were not considered in existing schemes. In this research study, we have used Arc GIS and NS2 simulation tools to analyze the parameters and predict the probability of the occurrence of a flood.

The Coupling of NSSL Warn-on-Forecast and FLASH Systems for Probabilistic Flash Flood Prediction

AbstractThe goal of the National Oceanic and Atmospheric Administration’s (NOAA) Warn-on-Forecast (WoF) program is to provide frequently updating, probabilistic model guidance that will enable National Weather Service (NWS) forecasters to produce more continuous communication of hazardous weather threats (e.g., heavy rainfall, flash floods, damaging wind, large hail, and tornadoes) between the watch and warning temporal and spatial scales. To evaluate the application of this WoF concept for probabilistic short-term flash flood prediction, the 0–3-h rainfall forecasts from NOAA National Severe Storms Laboratory’s (NSSL) experimental WoF System (WoFS) were integrated as the forcing to the NWS operational hydrologic modeling core within the Flooded Locations and Simulated Hydrographs (FLASH) system. Initial assessment of the potential impacts of probabilistic short-term flash flood forecasts from this coupled atmosphere–hydrology (WoFS-FLASH) modeling system were evaluated in the 2018 Hydrometeorology Testbed Multi-Radar Multi-Sensor Hydrology experiment held in Norman, Oklahoma. During the 3-week experiment period, a total of nine NWS forecasters analyzed three retrospective flash flood events in archive mode. This study will describe specifically what information participants extracted from the WoFS-FLASH products during these three archived events, and how this type of information is expected to impact operational decision-making processes. Overall feedback from the testbed participants’ evaluations show promise for the coupled NSSL WoFS-FLASH system probabilistic flash flood model guidance to enable earlier assessment and detection of flash flood threats and to advance the current warning lead time for these events.

Flash Flood Forecasting Based on Long Short-Term Memory Networks

Flash floods occur frequently and distribute widely in mountainous areas because of complex geographic and geomorphic conditions and various climate types. Effective flash flood forecasting with useful lead times remains a challenge due to its high burstiness and short response time. Recently, machine learning has led to substantial changes across many areas of study. In hydrology, the advent of novel machine learning methods has started to encourage novel applications or substantially improve old ones. This study aims to establish a discharge forecasting model based on Long Short-Term Memory (LSTM) networks for flash flood forecasting in mountainous catchments. The proposed LSTM flood forecasting (LSTM-FF) model is composed of T multivariate single-step LSTM networks and takes spatial and temporal dynamics information of observed and forecast rainfall and early discharge as inputs. The case study in Anhe revealed that the proposed models can effectively predict flash floods, especially the qualified rates (the ratio of the number of qualified events to the total number of flood events) of large flood events are above 94.7% at 1–5 h lead time and range from 84.2% to 89.5% at 6–10 h lead-time. For the large flood simulation, the small flood events can help the LSTM-FF model to explore a better rainfall-runoff relationship. The impact analysis of weights in the LSTM network structures shows that the discharge input plays a more obvious role in the 1-h LSTM network and the effect decreases with the lead-time. Meanwhile, in the adjacent lead-time, the LSTM networks explored a similar relationship between input and output. The study provides a new approach for flash flood forecasting and the highly accurate forecast contributes to prepare for and mitigate disasters.

Performance Assessment of Sub-Daily and Daily Precipitation Estimates Derived from GPM and GSMaP Products over an Arid Environment

Precipitation is a critical variable for comprehending various climate-related research, such as water resources management, flash flood monitoring and forecasting, climatic analyses, and hydrogeological studies, etc. Here, our objective was to evaluate the rainfall estimates obtained from Global Precipitation Mission (GPM), and Global Satellite Mapping of Precipitation (GSMaP) constellation over an arid environment like the Sultanate of Oman that is characterized by a complex topography and extremely variable rainfall patterns. Global Satellite-based Precipitation Estimates (GSPEs) can provide wide coverage and high spatial and temporal resolutions, but evaluating their accuracy is a mandatory step before involving them in different hydrological applications. In this paper, the reliability of the Integrated Multi-satellitE Retrievals for the GPM (IMERG) V04 and GSMaP V06 products were evaluated using the reference in-situ rain gauges at sub-daily (e.g., 6, 12, and 18 h) and daily time scales during the period of March 2014–December 2016. A set of continuous difference statistical indices (e.g., mean absolute difference, root mean square error, mean difference, and unconditional bias), and categorical metrics (e.g., probability of detection, critical success index, false alarm ratio, and frequency bias index) were used to evaluate recorded precipitation occurrences. The results showed that the five GSPEs could generally delineate the spatial and temporal patterns of rainfall while they might have over- and under-estimations of in-situ gauge measurements. The overall quality of the GSMaP runs was superior to the IMERG products; however, it also encountered an exaggeration in case of light rain and an underestimation for heavy rain. The effects of the gauge calibration algorithm (GCA) used in the final IMERG (IMERG-F) were investigated by comparison with early and late runs. The IMERG-F V04 product did not show a significant improvement over the early (i.e., after 4 h of rainfall observations) and late (i.e., after 12 h of rainfall observations) products. The results indicated that GCA could not reduce the missed precipitation records considerably.

River Discharge Simulation in the High Andes of Southern Ecuador Using High-Resolution Radar Observations and Meteorological Station Data

The prediction of river discharge using hydrological models (HMs) is of utmost importance, especially in basins that provide drinking water or serve as recreation areas, to mitigate damage to civil structures and to prevent the loss of human lives. Therefore, different HMs must be tested to determine their accuracy and usefulness as early warning tools, especially for extreme precipitation events. This study simulated the river discharge in an Andean watershed, for which the distributed HM Runoff Prediction Model (RPM) and the semi-distributed HM Hydrologic Modelling System (HEC-HMS) were applied. As precipitation input data for the RPM model, high-resolution radar observations were used, whereas the HEC-HMS model used the available meteorological station data. The obtained simulations were compared to measured discharges at the outlet of the watershed. The results highlighted the advantages of distributed HM (RPM) in combination with high-resolution radar images, which estimated accurately the discharges in magnitude and time. The statistical analysis showed good to very good accordance between observed and simulated discharge for the RPM model (R2: 0.85–0.92; NSE: 0.77–0.82), whereas for the HEC-HMS model accuracies were lower (R2: 0.68–0.86; NSE: 0.26–0.78). This was not only due to the application of means values for the watershed (HEC-HMS), but also to limited rain gauge information. Generally, station network density in tropical mountain regions is poor, for which reason the high spatiotemporal precipitation variability cannot be detected. For hydrological simulation and forecasting flash floods, as well as for environmental investigations and water resource management, meteorological radars are the better choice. The greater availability of cost-effective systems at the present time also reduces implementation and maintenance costs of dense meteorological station networks.

Simulation of Hydraulic Structures in 2D High-Resolution Urban Flood Modeling

Urban flooding as a result of inadequate drainage capacity, failure of flood defenses, etc. is usually featured with highly transient hydrodynamics. Reliable and efficient prediction and forecasting of these urban flash floods is still a great technical challenge. Meanwhile, in urban environments, the flooding hydrodynamics and process may be influenced by flow regulation and flood protection hydraulic infrastructure systems, such as sluice gates, which should be effectively taken into account in an urban flood model. However, direct simulation of hydraulic structures is not a current practice in 2D urban flood modeling. This work aims to develop a robust numerical approach to directly simulate the effects of gate structures in a 2D high-resolution urban flood model. A new modeling component is developed and fully coupled to a finite volume Godunov-type shock-capturing shallow water model, to directly simulate the highly transient flood waves through hydraulic structures. Different coupling approaches, i.e., flux term coupling and source term coupling, are implemented and compared. A numerical experiment conducted for an analytical dam-break test indicates that the flux term coupling approach may lead to more accurate results, with the calculated RMSE against water level 28%–38% less than that produced by the source term coupling approach. The flux term coupling approach is therefore adopted to improve the current urban flood model, and it is further tested by reproducing the laboratory experiments of flood routing in a flume with partially open sluice gates, conducted in the hydraulic laboratory at the Zhejiang Institute of Hydraulics and Estuary, China. The numerical results are compared favorably with experimental measurements, with a maximum RMSE of 0.0851 for all the individual tests. The satisfactory results demonstrate that the flood model implemented with the flux coupling approach is able to accurately simulate the flow through hydraulic structures, with enhanced predictive capability for urban flood modeling.

An Operational Forecasting System for Flash Floods in Mountainous Areas in Taiwan

Flash floods are different from common floods because they occur rapidly over short time scales, and they are considered to be one of the most devastating natural hazards worldwide. Mountainous areas with high population densities are particularly threatened by flash floods because steep slopes generate high flow velocities. Therefore, there is a great need to develop an operational forecasting system (OFS) for better flash flood prediction and warning in mountainous regions. This study developed an OFS through the integration of meteorological, hydrological, and hydrodynamic models. Airborne light detection and ranging (LiDAR) data were used to generate a digital elevation model (DEM). The OFS employs high-density and high-accuracy airborne LiDAR DEM data to simulate rapid water level rises and flooding as the result of intense rainfall within relatively small watersheds. The water levels and flood extent derived from the OFS are in agreement with the measured and surveyed data. The OFS has been adopted by the National Science and Technology Center for Disaster Reduction (NCDR) for forecasting flash floods every six hours in a mountainous floodplain in Taiwan. The 1D and 2D visualization of the OFS is performed via the National Center for Atmospheric Research Command Language (NCL).

Le programme HYMEX – Connaissances et prévision des pluies intenses et crues rapides en région méditerranéenne

Le programme international de recherche HyMeX (Hydrological cycle in the Mediterranean eXperiment, www.hymex.org), composante du chantier multi-organisme MISTRALS lancé pour 10 ans en 2010, a pour objectif de progresser dans la compréhension du cycle de l'eau en Méditerranée, et notamment d'améliorer les connaissances et la prévision des risques hydrométéorologiques. L'étude et la prévision des épisodes de pluies intenses et crues rapides qui affectent régulièrement le pourtour méditerranéen sont au cœur du programme HyMeX. Après une présentation de la stratégie générale d'observations qui a combiné campagnes de mesures, renforcement d'observations pendant plusieurs automnes et collectes de données socio-hydrologiques après les évènements majeurs qu'a connu la région sur la période, une revue d'ensemble des avancées scientifiques réalisées depuis le lancement d'HyMeX est proposée tant en termes de connaissances et modélisation des épisodes méditerranéens de pluie intense et de leurs impacts, qu'en termes d'évaluation et amélioration des méthodes et systèmes de prévision, et plus particulièrement les systèmes de prévision d'ensemble hydrométéorologiques à courte échéance basés sur le modèle de prévision météorologique AROME de Météo-France.

Using nowcasting technique and data assimilation in a meteorological model to improve very short range hydrological forecasts

Abstract. Forecasting flash floods some hours in advance is still a challenge, especially in environments made up of many small catchments. Hydrometeorological forecasting systems generally allow for predicting the possibility of having very intense rainfall events on quite large areas with good performances, even with 12–24 h of anticipation. However, they are not able to predict the exact rainfall location if we consider portions of a territory of 10 to 1000 km2 as the order of magnitude. The scope of this work is to exploit both observations and modelling sources to improve the discharge prediction in small catchments with a lead time of 2–8 h. The models used to achieve the goal are essentially (i) a probabilistic rainfall nowcasting model able to extrapolate the rainfall evolution from observations, (ii) a non-hydrostatic high-resolution numerical weather prediction (NWP) model and (iii) a distributed hydrological model able to provide a streamflow prediction in each pixel of the studied domain. These tools are used, together with radar observations, in a synergistic way, exploiting the information of each element in order to complement each other. For this purpose observations are used in a frequently updated data assimilation framework to drive the NWP system, whose output is in turn used to improve the information as input to the nowcasting technique in terms of a predicted rainfall volume trend; finally nowcasting and NWP outputs are blended, generating an ensemble of rainfall scenarios used to feed the hydrological model and produce a prediction in terms of streamflow. The flood prediction system is applied to three major events that occurred in the Liguria region (Italy) first to produce a standard analysis on predefined basin control sections and then using a distributed approach that exploits the capabilities of the employed hydrological model. The results obtained for these three analysed events show that the use of the present approach is promising. Even if not in all the cases, the blending technique clearly enhances the prediction capacity of the hydrological nowcasting chain with respect to the use of input coming only from the nowcasting technique; moreover, a worsening of the performance is observed less, and it is nevertheless ascribable to the critical transition between the nowcasting and the NWP model rainfall field.

Real‐time monitoring and estimation of the discharge of flash floods in a steep mountain catchment

AbstractSynchronously and accurately estimating the flood discharges and dynamic changes in the fluid density is essential for hydraulic analysis and forecasting of flash floods, as well as for risk assessment. However, such information is rare for steep mountain catchments, especially in regions that are hotspots for earthquakes. Therefore, six hydrological monitoring sites were established in the main stream and tributaries of the 78.3‐km2 Longxi River catchment, an affected region of the Wenchuan earthquake region in China. Direct real‐time monitoring equipment was installed to measure the flow depths, velocities, and fluid total pressures of the flood hydrographs. On the basis of field measurements, real‐time mean cross‐sectional velocities during the flood hydrographs could be derived from easily obtainable parameters: cross‐sectional maximum velocities and the calibrated dimensionless parameter Kh. Real‐time discharges were determined on the basis of a noncontact method to establish the effective rating curves of this mountainous stream, ranging from 1.46 to 386.34 m3/s with the root mean square errors of ≤10.22 m3/s. Compared with the traditional point‐velocity method and empirical Manning's formula, the proposed noncontact method was reliable and safe for monitoring whole flood hydrographs. Additionally, the real‐time fluid density during the flood hydrographs was calculated on the basis of the direct monitoring parameters for fluid total pressures and water depths. During the flood hydrograph, transient flow behaviour with higher fluid density generally occurred downstream during the flood peak periods when the flow was in the supercritical flow regime. The observed behaviour greatly increased the threat of damage to infrastructure and human life near the river. Thus, it is important to accurately estimate flood discharge and identify for fluid densities so that people at risk from an impending flash flood are given reliable, advanced warning.

Flash Flood Forecasting Using Support Vector Regression Model in a Small Mountainous Catchment

Flash floods in mountainous catchments are often caused by the rainstorm, which may result in more severe consequences than plain area floods due to less timescale and a fast-flowing front of water and debris. Flash flood forecasting is a huge challenge for hydrologists and managers due to its instantaneity, nonlinearity, and dependency. Among different methods of flood forecasting, data-driven models have become increasingly popular in recent years due to their strong ability to simulate nonlinear hydrological processes. This study proposed a Support Vector Regression (SVR) model, which is a powerful artificial intelligence-based model originated from statistical learning theory, to forecast flash floods at different lead times in a small mountainous catchment. The lagged average rainfall and runoff are identified as model input variables, and the time lags associated with the model input variables are determined by the hydrological concept of the time of response. There are 69 flash flood events collected from 1984 to 2012 in a mountainous catchment in China and then used for the model training and testing. The contribution of the runoff variables to the predictions and the phase lag of model outputs are analyzed. The results show that: (i) the SVR model has satisfactory predictive performances for one to three-hours ahead forecasting; (ii) the lagged runoff variables have a more significant effect on the predictions than the rainfall variables; and (iii) the phase lag (time difference) of prediction results significantly exists in both two- and three-hours-ahead forecasting models, however, the input rainfall information can assist in mitigating the phase lag of peak flow.

Windshield wipers on connected vehicles produce high-accuracy rainfall maps

Connected vehicles are poised to transform the field of environmental sensing by enabling acquisition of scientific data at unprecedented scales. Drawing on a real-world dataset collected from almost 70 connected vehicles, this study generates improved rainfall estimates by combining weather radar with windshield wiper observations. Existing methods for measuring precipitation are subject to spatial and temporal uncertainties that compromise high-precision applications like flash flood forecasting. Windshield wiper measurements from connected vehicles correct these uncertainties by providing precise information about the timing and location of rainfall. Using co-located vehicle dashboard camera footage, we find that wiper measurements are a stronger predictor of binary rainfall state than traditional stationary gages or radar-based measurements. We introduce a Bayesian filtering framework that generates improved rainfall estimates by updating radar rainfall fields with windshield wiper observations. We find that the resulting rainfall field estimate captures rainfall events that would otherwise be missed by conventional measurements. We discuss how these enhanced rainfall maps can be used to improve flood warnings and facilitate real-time operation of stormwater infrastructure.

Flash-Flood Potential assessment in the upper and middle sector of Prahova river catchment (Romania). A comparative approach between four hybrid models

An accurate assessment of Flash-Flood Potential for certain areas is mandatory for the improvement of flash-flood forecast and warnings. The main aim of the present study is represented by the calculation of Flash-Flood Potential Index within the upper and the middle sector of Prahova river catchment (Romania) by using 4 hybrid models: Logistic Regression-Frequency Ratio (LR-FR) model, Logistic Regression-Weights of Evidence (LR-WoE) model, Support Vector Machine-Frequency Ratio (SVM-FR) model and Support Vector Machine-Weights of Evidence (SVM-WoE). The identification of areas affected by torrential phenomena represents the first step performed in the present research. These areas with a total surface of 260 km2 were divided into training areas (70%) and validating areas (30%). By the mean of Linear Support Vector Machine (LSVM) model, 10 flash-flood conditioning factors were selected and further used for the Flash-Flood Potential assessment. Based on the spatial relationship between areas affected by torrential phenomena and flash-floods conditioning factors characteristics, the FR and WoE coefficients were calculated. In order to be integrated into Logistic Regression and Support Vector Machine (RBF) analysis, these values were standardized. According to the results of the 4 hybrid models used for FFPI calculation, the high and very high Flash-Flood Potential are spread over 33% of the study area. The model performance assessment and results validation were carried out by the mean of the three different methods: i) relative frequency distribution of torrential phenomena pixels within FFPI classes; ii) ROC Curve (Success Rate and Prediction Rate) and AUC value; iii) statistical measures represented by Sensitivity, Specificity and Accuracy.

Flood Routing Model with Particle Filter-Based Data Assimilation for Flash Flood Forecasting in the Micro-Model of Lower Yellow River, China

Reliable real-time flood forecasting is a challenging prerequisite for successful flood protection. This study developed a flood routing model combined with a particle filter-based assimilation model and a one-dimensional hydrodynamic model. This model was applied to an indoor micro-model, using the Lower Yellow River (LYR) as prototype. Real-time observations of the water level from the micro-model were used for data assimilation. The results show that, compared to the traditional hydrodynamic model, the assimilation model could effectively update water level, flow discharge, and roughness coefficient in real time, thus yielding improved results. The mean water levels of the particle posterior distribution are closer to the observed values than before assimilation, even when water levels change greatly. In addition, the calculation results for different lead times indicate that the root mean square error of the forecasting water level gradually increases with increasing lead time. This is because the roughness value changes greatly in response to unsteady water flow, and the incurring error accumulates with the predicted period. The results show that the assimilation model can simulate water level changes in the micro-model and provide both research method and technical support for real flood forecasting in the LYR.

Flash Flood Verification: Pondering Precipitation Proxies

Abstract Quantitative precipitation estimate (QPE) exceedances of numerous different heavy precipitation thresholds—including spatially varying average recurrence interval (ARI) and flash flood guidance (FFG) thresholds—are compared among each other and against reported and warned flash floods to quantify existing deficiencies with QPEs and to identify best practices for using QPE for flash flood forecasting and analysis. QPEs from three different sources—NCEP Stage IV Precipitation Analysis (ST4), Climatology Calibrated Precipitation Analysis (CCPA), and Multi-Radar Multi-Sensor (MRMS) QPE—are evaluated across the United States from January 2015 to June 2017. In addition to evaluating different QPE sources, threshold types, and magnitudes, QPE accumulation interval lengths from hourly to daily are considered. Systematic errors with QPE sources are identified, including a radar distance dependence on extreme rainfall frequency in MRMS, spurious occurrences of locally extreme precipitation in the complex terrain of the West in ST4, and insufficient QPEs for many legitimate heavy precipitation events in CCPA. Overall, flash flood warnings and reports corresponded to each other far more than any QPE exceedances. Correspondence between all sources was at a maximum in the East and worst in the West, with ST4, CCPA, and MRMS QPE exceedances locally yielding maximal correspondence in the East, Plains, and West, respectively. Surprisingly, using a fixed 2.5 in. (24 h)−1 proxy outperformed shorter accumulation exceedances and the use of ARIs and FFGs. On regional scales, different ARI exceedances achieved superior performance to the selection of any fixed threshold; FFG exceedances were consistently too rare to achieve optimal correspondence with observed flash flooding.

Flash-Flood Forecasting in an Andean Mountain Catchment—Development of a Step-Wise Methodology Based on the Random Forest Algorithm

Flash-flood forecasting has emerged worldwide due to the catastrophic socio-economic impacts this hazard might cause and the expected increase of its frequency in the future. In mountain catchments, precipitation-runoff forecasts are limited by the intrinsic complexity of the processes involved, particularly its high rainfall variability. While process-based models are hard to implement, there is a potential to use the random forest algorithm due to its simplicity, robustness and capacity to deal with complex data structures. Here a step-wise methodology is proposed to derive parsimonious models accounting for both hydrological functioning of the catchment (e.g., input data, representation of antecedent moisture conditions) and random forest procedures (e.g., sensitivity analyses, dimension reduction, optimal input composition). The methodology was applied to develop short-term prediction models of varying time duration (4, 8, 12, 18 and 24 h) for a catchment representative of the Ecuadorian Andes. Results show that the derived parsimonious models can reach validation efficiencies (Nash-Sutcliffe coefficient) from 0.761 (4-h) to 0.384 (24-h) for optimal inputs composed only by features accounting for 80% of the model’s outcome variance. Improvement in the prediction of extreme peak flows was demonstrated (extreme value analysis) by including precipitation information in contrast to the use of pure autoregressive models.

Simple and modular integrated modeling of storm drain network with gridded distributed hydrologic model via grid-rendering of storm drains for large urban areas

For accurate flash flood forecasting and effective stormwater planning and management in urban areas, it is necessary to model not only the natural channel systems but also the large and complex networks of storm drains. In this work, we describe a modular storm drain model that can easily be coupled with existing gridded distributed hydrologic models for real-time flash flood forecasting, and stormwater planning and management for large urban areas (>100 km2). A salient feature of our approach is the use of the equivalent storm drain network (ESDN) which approximates the actual network on the same grid as that of the distributed hydrologic model, thereby rendering coupling simple and modular. In the integrated model, storm-drain flow occurs through the simplified network coarsened to the resolution of the distributed hydrologic model without full 1D-2D dynamics or very detailed man-made structures and features modeled. The integrated model is applied to a 144.6 km2 area of five urban catchments in the Cities of Arlington and Grand Prairie in Texas, US. Comparisons at selected locations using kinematic-wave flow simulations show that the ESDN approximates the flow through the original network very well with attendant savings in computational amount and reduction in modeling complexities. The impact of the storm drain network is assessed via a combination of simulation experiments, sensitivity analysis and limited comparison with observed flow. It is shown that the storm drain network in highly urbanized catchments in the study area is very effective in reducing surface flow at most locations for about 30 min following onset of significant rainfall, that the existing stormwater infrastructure would lose effectiveness for approximately 30% of the study area with a 15% increase in imperviousness relative to the current conditions, and that significant uncertainties exist in partitioning of surface flow into storm drain and natural channel flows due to sensitivity to inlet flow modeling. With greatly reduced computing cost and modeling complexity, the proposed approach offers a practical solution for integrated stormwater modeling for large urban areas for a wide range of applications. The proposed approach cannot, however, resolve flow at sub-grid scales and hence is not appropriate for very detailed modeling for small areas.

Ensemble model to enhance robustness of flash flood forecasting using an Artificial Neural Network: case-study on the Gardon Basin (south-eastern France)

During the last few decades neural networks have been increasingly used in hydrological modelling for their fundamental property of parsimony and of universal approximation of non-linear functions. For the purpose of flash flood forecasting, feed-forward and recurrent multi-layer perceptrons appear to be efficient tools. Nevertheless, their forecasting performances are sensitive to the initialization of the network parameters. We have studied the cross-validation efficiency to select initialization providing the best forecasts in real time situation. Sensitivity to initialization of feed-forward and recurrent models is compared for one-hour lead-time forecasts. This study shows that cross-validation is unable to select the best initialization. A more robust model has been designed using the median of several models outputs; in this context, this paper analyses the design of the ensemble model for both recurrent and feed-forward models.

How to Improve Fault Tolerance in Disaster Predictions: A Case Study about Flash Floods Using IoT, ML and Real Data.

The rise in the number and intensity of natural disasters is a serious problem that affects the whole world. The consequences of these disasters are significantly worse when they occur in urban districts because of the casualties and extent of the damage to goods and property that is caused. Until now feasible methods of dealing with this have included the use of wireless sensor networks (WSNs) for data collection and machine-learning (ML) techniques for forecasting natural disasters. However, there have recently been some promising new innovations in technology which have supplemented the task of monitoring the environment and carrying out the forecasting. One of these schemes involves adopting IP-based (Internet Protocol) sensor networks, by using emerging patterns for IoT. In light of this, in this study, an attempt has been made to set out and describe the results achieved by SENDI (System for dEtecting and forecasting Natural Disasters based on IoT). SENDI is a fault-tolerant system based on IoT, ML and WSN for the detection and forecasting of natural disasters and the issuing of alerts. The system was modeled by means of ns-3 and data collected by a real-world WSN installed in the town of São Carlos - Brazil, which carries out the data collection from rivers in the region. The fault-tolerance is embedded in the system by anticipating the risk of communication breakdowns and the destruction of the nodes during disasters. It operates by adding intelligence to the nodes to carry out the data distribution and forecasting, even in extreme situations. A case study is also included for flash flood forecasting and this makes use of the ns-3 SENDI model and data collected by WSN.