Multi-method global sensitivity analysis of flood inundation models

Global sensitivity analysis is a valuable tool in understanding flood inundation models and deriving decisions on strategies to reduce model uncertainty. In this paper, a sensitivity analysis of a one-dimensional flood inundation model (HEC-RAS) on the River Alzette, Luxembourg, is presented. It is impossible to define sensitivity in a unique way and different methods can lead to a difference in ranking of importance of model factors. In this paper five different methods (Sobol, Kullback–Leibler entropy, Morris, regionalised sensitivity analysis and regression) are applied and the outcomes on selected examples compared. It is demonstrated that the different methods lead to completely different ranking of importance of the parameter factors and that it is impossible to draw firm conclusions about the relative sensitivity of different factors. Moreover, the uncertainty inherent in the sensitivity methods is highlighted.     

A fast simplified model for predicting river flood inundation probabilities in poorly gauged areas

The capability of a simple kinematic‐storage model (KSM) is analysed to be used as a tool for a Decision Support System for the evaluation of probability inundation maps in near real time in poorly gauged areas. KSM simulates the floodplain as a storage and assumes no exchange of momentum with the channel. For the in‐bank flow, the storage is modified through a coefficient for taking the variations of channel cross sections into account. The generalized likelihood uncertainty estimation approach is used for addressing the probability flood maps along with their associated uncertainties. The model is tested on two river reaches along the Tiber River in Central Italy where observed inundation maps are available for two recent flood events. Despite the inherent uncertainties present in the input data and in the model structure, the results show that the model reproduces reasonably well, in terms of both precision and accuracy, the observed inundated areas. Tests were performed at different digital elevation model resolutions, showing a small effect of the geometry on the maximum performance obtained. The very low computational times, the parsimony of the model and its low sensitivity to the quality of the geometry representation of the channel and the floodplain makes KSM very appealing for flood forecasting and early warning system applications in poorly gauged and inaccessible areas. Copyright © 2014 John Wiley & Sons, Ltd.     

ROC‐based calibration of flood inundation models

The use of spatial patterns of flood inundation (often obtained from remotely sensed imagery) to calibrate flood inundation models has been widespread over the last 15 years. Model calibration is most often achieved by employing one or even several performance measures derived from the well‐known confusion matrix based on a binary classification of flooding. However, relatively early on, it has been recognized that the use of commonly reported performance measures for calibrating flood inundation models (such as the F measure) is hampered because the calibration procedure commonly utilizes only one possible solution of a wet/dry classification of a remote sensing image [most often acquired by a synthetic aperture radar (SAR)] to calibrate or validate models and are biased towards either over‐prediction or under‐prediction of flooding. Despite the call in several studies for an alternative statistic, to this date, very few, if any, unbiased performance measure based on the confusion matrix has been proposed for flood model calibration/validation studies. In this paper, we employ a robust statistical measure that operates in the receiver operating characteristics (ROC) space and allows automated model calibration with high identifiability of the best model parameter set but without the need of a classification of the SAR image. The ROC‐based method for flood model calibration is demonstrated using two different flood event test cases with flood models of varying degree of complexity and boundary conditions with varying degree of accuracy. Verification of the calibration results and optional SAR classification is successfully performed with independent observations of the events. We believe that this proposed alternative approach to flood model calibration using spatial patterns of flood inundation should be employed instead of performance measures commonly used in conjunction with a binary flood map. © 2013 California Institute of Technology. Hydrological Processes © 2013 John Wiley & Sons, Ltd.     

Estimation of uncertainty propagation in flood inundation mapping using a 1‐D hydraulic model

A need for more accurate flood inundation maps has recently arisen because of the increasing frequency and extremity of flood events. The accuracy of flood inundation maps is determined by the uncertainty propagated from all of the variables involved in the overall process of flood inundation modelling. Despite our advanced understanding of flood progression, it is impossible to eliminate the uncertainty because of the constraints involving cost, time, knowledge, and technology. Nevertheless, uncertainty analysis in flood inundation mapping can provide useful information for flood risk management. The twin objectives of this study were firstly to estimate the propagated uncertainty rates of key variables in flood inundation mapping by using the first‐order approximation method and secondly to evaluate the relative sensitivities of the model variables by using the Hornberger–Spear–Young (HSY) method. Monte Carlo simulations using the Hydrologic Engineering Center's River Analysis System and triangle‐based interpolation were performed to investigate the uncertainty arising from discharge, topography, and Manning's n in the East Fork of the White River near Seymour, Indiana, and in Strouds Creek in Orange County, North Carolina. We found that the uncertainty of a single variable is propagated differently to the flood inundation area depending on the effects of other variables in the overall process. The uncertainty was linearly/nonlinearly propagated corresponding to valley shapes of the reaches. In addition, the HSY sensitivity analysis revealed the topography of Seymour reach and the discharge of Strouds Creek to be major contributors to the change of flood inundation area. Copyright © 2014 John Wiley & Sons, Ltd.     

A probabilistic model for real‐time flood warning based on deterministic flood inundation mapping

Deterministic flood inundation mapping is valuable for the investigation of detailed flood depth and extent. However, when these data are used for real‐time flood warning, uncertainty arises while encountering the difficulties of timely response, message interpretation and performance evaluation that makes statistical analysis necessary. By incorporating deterministic flood inundation map outputs statistically by means of logistic regression, this paper presents a probabilistic real‐time flood warning model determining region‐based flood probability directly from rainfall, being efficient in computation, clear in message, and valid in physical meaning. The calibration and validation of the probabilistic model show a satisfactory overall correctness rate, with the hit rate far surpassing the false alarm rate in issuing flood warning for historical events. Further analyses show that the probabilistic model is effective in evaluating the level of uncertainty lying within flood warning which can be reduced by several techniques proposed in order to improve warning performance. Copyright © 2011 John Wiley & Sons, Ltd.     

Bayesian updating of flood inundation likelihoods conditioned on flood extent data

Previously we have detailed an application of the generalized likelihood uncertainty estimation (GLUE) procedure to estimate spatially distributed uncertainty in models conditioned against binary pattern data contained in flood inundation maps. This method was applied to two sites where a single consistent synoptic image of inundation extent was available to test the simulation performance of the method. In this paper, we extend this to examine the predictive performance of the method for a reach of the River Severn, west‐central England. Uniquely for this reach, consistent inundation images of two major floods have been acquired from spaceborne synthetic aperture radars, as well as a high‐resolution digital elevation model derived using laser altimetry. These data thus allow rigorous split sample testing of the previous GLUE application. To achieve this, Monte Carlo analyses of parameter uncertainty within the GLUE framework are conducted for a typical hydraulic model applied to each flood event. The best 10% of parameter sets identified in each analysis are then used to map uncertainty in flood extent predictions using the method previously proposed for both an independent validation data set and a design flood. Finally, methods for combining the likelihood information derived from each Monte Carlo ensemble are examined to determine whether this has the potential to reduce uncertainty in spatially distributed measures of flood risk for a design flood. The results show that for this reach and these events, the method previously established is able to produce sharply defined flood risk maps that compare well with observed inundation extent. More generally, we show that even single, poor‐quality inundation extent images are useful in constraining hydraulic model calibrations and that values of effective friction parameters are broadly stationary between the two events simulated, most probably reflecting their similar hydraulics. Copyright © 2004 John Wiley & Sons, Ltd.     

Microwave remote sensing of flood inundation

Flooding is one of the most costly natural disasters and thus mapping, modeling and forecasting flood events at various temporal and spatial scales is important for any flood risk mitigation plan, disaster relief services and the global (re-)insurance markets. Both computer models and observations (ground-based, airborne and Earth-orbiting) of flood processes and variables are of great value but the amount and quality of information available varies greatly with location, spatial scales and time. It is very well known that remote sensing of flooding, especially in the microwave region of the electromagnetic spectrum, can complement ground-based observations and be integrated with flood models to augment the amount of information available to end-users, decision-makers and scientists. This paper aims to provide a concise review of both the science and applications of microwave remote sensing of flood inundation, focusing mainly on synthetic aperture radar (SAR), in a variety of natural and man-made environments. Strengths and limitations are discussed and the paper will conclude with a brief account on perspectives and emerging technologies.     

Unstructured grid generation using LiDAR data for urban flood inundation modelling

Inundation disasters, caused by sudden water level rise or rapid flow, occur frequently in various parts of the world. Such catastrophes strike not only in thinly populated flood plains or farmland but also in highly populated villages or urban areas. Inundation of the populated areas causes severe damage to the economy, injury, and loss of life; therefore, a proper management scheme for the disaster has to be developed. To predict and manage such adversity, an understanding of the dynamic processes of inundation flow is necessary because risk estimation is performed based on inundation flow information. In this study, we developed a comprehensive method to conduct detailed inundation flow simulations for a populated area with quite complex topographical features using LiDAR (Light Detection and Ranging) data. Detailed geospatial information including the location and shape of each building was extracted from the LiDAR data and used for the grid generation. The developed approach can distinguish buildings from vegetation and treat them differently in the flow model. With this method, a fine unstructured grid can be generated representing the complicated urban land features precisely without exhausting labour for data preparation. The accuracy of the generated grid with different grid spacing and grid type is discussed and the optimal range of grid spacing for direct representation of urban topography is investigated. The developed method is applied to the estimation of inundation flows, which occurred in the basin of the Shin‐minato River. A detailed inundation flow structure is represented by the flow model, and the flow characteristics with respect to topographic features are discussed. Copyright © 2010 John Wiley & Sons, Ltd.     

Short‐term flood inundation prediction using hydrologic‐hydraulic models forced with downscaled rainfall from global NWP

A short‐term flood inundation prediction model has been formulated based on the combination of the super‐tank model, forced with downscaled rainfall from a global numerical weather prediction model, and a one‐dimensional (1D) hydraulic model. Different statistical methods for downscaled rainfall have been explored, taking into account the availability of historical data. It has been found that the full implementation of a statistical downscaling model considering physically‐based corrections to the numerical weather prediction model output for rainfall prediction performs better compared with an altitudinal correction method. The integration of the super‐tank model into the 1D hydraulic model demonstrates a minimal requirement for the calibration of rainfall–runoff and flood propagation models. Updating the model with antecedent rainfall and regular forecast renewal has enhanced the model's capabilities as a result of the data assimilation processes of the runoff and numerical weather prediction models. The results show that the predicted water levels demonstrate acceptable agreement with those measured by stream gauges and comparable to those reproduced using the actual rainfall. Moreover, the predicted flood inundation depth and extent exhibit reasonably similar tendencies to those observed in the field. However, large uncertainties are observed in the prediction results in lower, flat portions of the river basin where the hydraulic conditions are not properly analysed by the 1D flood propagation model. Copyright © 2013 John Wiley & Sons, Ltd.     

Urban flood hazard mapping using change detection on wetness transformed images

ABSTRACT

The present study demonstrates the use of a new approach for delineating the accurate flood hazard footprint in the urban regions. The methodology involves transformation of Landsat Thematic Mapper (TM) imagery to a three-dimensional feature space, i.e. brightness, wetness and greenness, then a change detection technique is used to identify the areas affected by the flood. Efficient thresholding of the normalized difference image generated during change detection has shown promising results in identifying the flood extents which include standing water due to flood, sediment-laden water and wetness caused by the flood. Prior to wetness transformations, dark object subtraction has been used in lower wavelengths to avoid errors due to scattering in urban areas. The study shows promising results in eliminating most of the problems associated with urban flooding, such as misclassification due to presence of asphalt, scattering in lower wavelengths and delineating mud surges. The present methodology was tested on the 2010 Memphis flood event and validated on Queensland floods in 2011. The comparative analysis was carried out with the widely-used technique of delineating flood extents using thresholding of near infrared imagery. The comparison demonstrated that the present approach is more robust towards the error of omission in flood mapping. Moreover, the present approach involves less manual effort and is simpler to use.

Editor Z.W. Kundzewicz; Associate editor A. Viglione     

Evaluation of on-line DEMs for flood inundation modeling

Recent and highly accurate topographic data should be used for flood inundation modeling, but this is not always feasible given time and budget constraints so the utility of several on-line digital elevation models (DEMs) is examined with a set of steady and unsteady test problems. DEMs are used to parameterize a 2D hydrodynamic flood simulation algorithm and predictions are compared with published flood maps and observed flood conditions. DEMs based on airborne light detection and ranging (LiDAR) are preferred because of horizontal resolution, vertical accuracy (∼0.1 m) and the ability to separate bare-earth from built structures and vegetation. DEMs based on airborne interferometric synthetic aperture radar (IfSAR) have good horizontal resolution but gridded elevations reflect built structures and vegetation and therefore further processing may be required to permit flood modeling. IfSAR and shuttle radar topography mission (SRTM) DEMs suffer from radar speckle, or noise, so flood plains may appear with non-physical relief and predicted flood zones may include non-physical pools. DEMs based on national elevation data (NED) are remarkably smooth in comparison to IfSAR and SRTM but using NED, flood predictions overestimate flood extent in comparison to all other DEMs including LiDAR, the most accurate. This study highlights utility in SRTM as a global source of terrain data for flood modeling.     

Uncertainty analysis of flood inundation modelling using GLUE with surrogate models in stochastic sampling

A generalized likelihood uncertainty estimation (GLUE) method incorporating moving least squares (MLS) with entropy for stochastic sampling (denoted as GLUE‐MLS‐E) was proposed for uncertainty analysis of flood inundation modelling. The MLS with entropy (MLS‐E) was established according to the pairs of parameters/likelihoods generated from a limited number of direct model executions. It was then applied to approximate the model evaluation to facilitate the target sample acceptance of GLUE during the Monte‐Carlo‐based stochastic simulation process. The results from a case study showed that the proposed GLUE‐MLS‐E method had a comparable performance as GLUE in terms of posterior parameter estimation and predicted confidence intervals; however, it could significantly reduce the computational cost. A comparison to other surrogate models, including MLS, quadratic response surface and artificial neural networks (ANN), revealed that the MLS‐E outperformed others in light of both the predicted confidence interval and the most likely value of water depths. ANN was shown to be a viable alternative, which performed slightly poorer than MLS‐E. The proposed surrogate method in stochastic sampling is of practical significance in computationally expensive problems like flood risk analysis, real‐time forecasting, and simulation‐based engineering design, and has a general applicability in many other numerical simulation fields that requires extensive efforts in uncertainty assessment. Copyright © 2014 John Wiley & Sons, Ltd.     

Investigating sea-state effects on flash flood hydrograph and inundation forecasting

A common source of uncertainty in flood inundation forecasting is the hydrograph used. Given the role of sea-air-hydro-land chain processes on the water cycle, flood hydrographs in coastal areas can be indirectly affected by sea state. This study investigates sea-state effects on precipitation, discharge, and flood inundation forecasting implementing atmospheric, ocean wave, hydrological, and hydraulic-hydrodynamic coupled models. The Chemical Hydrological Atmospheric Ocean wave System (CHAOS) was used for coupled hydro-meteorological-wave simulations ‘accounting’ or ‘not accounting’ the impact of sea state on precipitation and, subsequently, on flood hydrograph. CHAOS includes the WRF-Hydro hydrological model and the WRF-ARW meteorological model two-way coupled with the WAM wave model through the OASIS3-MCT coupler. Subsequently, the 2D HEC-RAS hydraulic-hydrodynamic model was forced by the flood hydrographs and map the inundated areas. A flash flood event occurred on 15 November 2017 in Mandra, Attica, Greece, causing 24 fatalities, and damages was selected as case study. The calibration of models was performed exploiting historical flood records and previous studies. Human interventions such as hydraulic works and the urban areas were included in the hydraulic modelling geometry domain. The representation of the resistance caused by buildings was based on Unmanned Aerial System (UAS) data while the local elevation rise method was used in the urban-flood simulation. The flood extent results were assessed using the Critical Success Index (CSI), and CSI penalize. Integrating sea-state affected the forecast of precipitation and discharge peaks, causing up to +24% and from −8% to +36% differences, respectively, improving inundation forecast by 4.5% and flooding additional approximately 70 building blocks. The precipitation forcing time step was also highlighted as significant factor in such a small-scale flash flood. The integrated multidisciplinary methodological approach could be adopted in operational forecasting for civil protection applications facilitating the protection of socio-economic activities and human lives during similar future events.     

Assessing the reliability of probabilistic flood inundation model predictions

An ability to quantify the reliability of probabilistic flood inundation predictions is a requirement not only for guiding model development but also for their successful application. Probabilistic flood inundation predictions are usually produced by choosing a method of weighting the model parameter space, but previous study suggests that this choice leads to clear differences in inundation probabilities. This study aims to address the evaluation of the reliability of these probabilistic predictions. However, a lack of an adequate number of observations of flood inundation for a catchment limits the application of conventional methods of evaluating predictive reliability. Consequently, attempts have been made to assess the reliability of probabilistic predictions using multiple observations from a single flood event. Here, a LISFLOOD‐FP hydraulic model of an extreme (>1 in 1000 years) flood event in Cockermouth, UK, is constructed and calibrated using multiple performance measures from both peak flood wrack mark data and aerial photography captured post‐peak. These measures are used in weighting the parameter space to produce multiple probabilistic predictions for the event. Two methods of assessing the reliability of these probabilistic predictions using limited observations are utilized; an existing method assessing the binary pattern of flooding, and a method developed in this paper to assess predictions of water surface elevation. This study finds that the water surface elevation method has both a better diagnostic and discriminatory ability, but this result is likely to be sensitive to the unknown uncertainties in the upstream boundary condition. Copyright © 2015 John Wiley & Sons, Ltd.     

On flood probabilities of East Alpine rivers

The flood data of five stations in the East Alps with relatively long observation series are represented by means of Fisher-Tippett II or III distributions. The observations of the longest observation periods are fitted well by the type II curves. A comparison of calculations with only a part of the observations shows that the magnitude of the curvature parameter k of the Fisher-Tippett curves increases with increasing observation period. The series at hand are not long enough to see if there is a convergence to a limiting value of k.A trend test showed an increasing trend in all of the Danube stations after about 1860 and nearly no trend at all in the observations from the southern rivers.The variance spectra yielded the result that the series may be considered as realizations of autoregressive processes of the first order. Superposed to the red-noise spectra there are waves with periods of 2.35 and 8 years in the case of the Danube stations and with 5.7 and 6.7 years in the case of the southern rivers.     

Development of flood hazard maps of Bangladesh using NOAA-AVHRR images with GIS

Abstract

Flood hazard maps were developed using remote sensing (RS) data for the historical event of the 1988 flood with data of elevation height, and geological and physiographic divisions. Flood damage depends on the hydraulic factors which include characteristics of the flood such as the depth of flooding, rate of the rise in water level, propagation of a flood wave, duration and frequency of flooding, sediment load, and timing. In this study flood depth and “flood-affected frequency” within one flood event were considered for the evaluation of flood hazard assessment, where the depth and frequency of the flooding were assumed to be the major determinant in estimating the total damage function. Different combinations of thematic maps among physiography, geology, land cover and elevation were evaluated for flood hazard maps and a best combination for the event of the 1988 flood was proposed. Finally, the flood hazard map for Bangladesh and a flood risk map for the administrative districts of Bangladesh were proposed.     

Influence of uncertain boundary conditions and model structure on flood inundation predictions

In this study, the GLUE methodology is applied to establish the sensitivity of flood inundation predictions to uncertainty of the upstream boundary condition and bridges within the modelled region. An understanding of such uncertainties is essential to improve flood forecasting and floodplain mapping. The model has been evaluated on a large data set. This paper shows uncertainty of the upstream boundary can have significant impact on the model results, exceeding the importance of model parameter uncertainty in some areas. However, this depends on the hydraulic conditions in the reach e.g. internal boundary conditions and, for example, the amount of backwater within the modelled region. The type of bridge implementation can have local effects, which is strongly influenced by the bridge geometry (in this case the area of the culvert). However, the type of bridge will not merely influence the model performance within the region of the structure, but also other evaluation criteria such as the travel time. This also highlights the difficulties in establishing which parameters have to be more closely examined in order to achieve better fits. In this study no parameter set or model implementation that fulfils all evaluation criteria could be established. We propose four different approaches to this problem: closer investigation of anomalies; introduction of local parameters; increasing the size of acceptable error bounds; and resorting to local model evaluation. Moreover, we show that it can be advantageous to decouple the classification into behavioural and non-behavioural model data/parameter sets from the calculation of uncertainty bounds.     

Probabilistic evaluation of flood hazard in urban areas using Monte Carlo simulation

The goal of the presented research was the derivation of flood hazard maps, using Monte Carlo simulation of flood propagation at an urban site in the UK, specifically an urban area of the city of Glasgow. A hydrodynamic model describing the propagation of flood waves, based on the De Saint Venant equations in two‐dimensional form capable of accounting for the topographic complexity of the area (preferential outflow paths, buildings, manholes, etc.) and for the characteristics of prevailing imperviousness typical of the urban areas, has been used to derive the hydrodynamic characteristics of flood events (i.e. water depths and flow velocities). The knowledge of the water depth distribution and of the current velocities derived from the propagation model along with the knowledge of the topographic characteristics of the urban area from digital map data allowed for the production of hazard maps based on properly defined hazard indexes. These indexes are evaluated in a probabilistic framework to overcome the classical problem of single deterministic prediction of flood extent for the design event and to introduce the concept of the likelihood of flooding at a given point as the sum of data uncertainty, model structural error and parameterization uncertainty. Copyright © 2011 John Wiley & Sons, Ltd.     

Development of a global flood risk index based on natural and socio-economic factors

Abstract

A global flood risk index (FRI) is established, based on both natural and social factors. The advanced flood risk index (AFRI) is the expectation of damage in the case of a single flood occurrence, estimated by a linear regression-based approach as a function of hazard and vulnerability metrics. The resulting equations are used to predict potential flood damage given gridded global data for independent variables. It is new in the aspect that it targets floods by units of events, instead of a long-term trend. Moreover, the value of the AFRI is that it can express relative potential flood risk with the process of flood damage occurrence considered. The significance of this study is that not only the hazard parameters which contribute directly to flood occurrence, but vulnerability parameters which reflect the conditions of the region where flood occurred, including its residential and social characteristics, were shown quantitatively to affect flood damage.

Citation Okazawa, Y., Yeh, P., Kanae, S. & Oki, T. (2011) Development of a global flood risk index based on natural and socioeconomic factors. Hydrol. Sci. J. 56(5), 789–804.