Renaud Hostache

A Change Detection Approach to Flood Mapping in Urban Areas Using TerraSAR-X

Very high resolution synthetic aperture radar (SAR) sensors represent an alternative to aerial photography for delineating floods in built-up environments where flood risk is highest. However, even with currently available SAR image resolutions of 3 m and higher, signal returns from man-made structures hamper the accurate mapping of flooded areas. Enhanced image processing algorithms and a better exploitation of image archives are required to facilitate the use of microwave remote-sensing data for monitoring flood dynamics in urban areas. In this paper, a hybrid methodology combining backscatter thresholding, region growing, and change detection (CD) is introduced as an approach enabling the automated, objective, and reliable flood extent extraction from very high resolution urban SAR images. The method is based on the calibration of a statistical distribution of “open water” backscatter values from images of floods. Images acquired during dry conditions enable the identification of areas that are not “visible” to the sensor (i.e., regions affected by “shadow”) and that systematically behave as specular reflectors (e.g., smooth tarmac, permanent water bodies). CD with respect to a reference image thereby reduces overdetection of inundated areas. A case study of the July 2007 Severn River flood (UK) observed by airborne photography and the very high resolution SAR sensor on board TerraSAR-X highlights advantages and limitations of the method. Even though the proposed fully automated SAR-based flood-mapping technique overcomes some limitations of previous methods, further technological and methodological improvements are necessary for SAR-based flood detection in urban areas to match the mapping capability of high-quality aerial photography.

High-Resolution 3-D Flood Information From Radar Imagery for Flood Hazard Management

This paper presents a remote-sensing-based steady-state flood inundation model to improve preventive flood-management strategies and flood disaster management. The Regression and Elevation-based Flood Information eXtraction (REFIX) model is based on regression analysis and uses a remotely sensed flood extent and a high-resolution floodplain digital elevation model to compute flood depths for a given flood event. The root mean squared error of the REFIX, compared to ground-surveyed high water marks, is 18 cm for the January 2003 flood event on the River Alzette floodplain (G.D. of Luxembourg), on which the model is developed. Applying the same methodology on a reach of the River Mosel, France, shows that for some more complex river configurations (in this case, a meandering river reach that contains a number of hydraulic structures), piecewise regression is required to yield more accurate flood water-line estimations. A comparison with a simulation from the Hydrologic Engineering Centers River Analysis System hydraulic flood model, calibrated on the same events, shows that, for both events, the REFIX model approximates the water line reliably

Probabilistic Flood Mapping Using Synthetic Aperture Radar Data

Probabilistic flood mapping offers flood managers, decision makers, insurance agencies, and humanitarian relief organizations a useful characterization of uncertainty in flood mapping delineation. Probabilistic flood maps are also of high interest for data assimilation into numerical models. The direct assimilation of probabilistic flood maps into hydrodynamic models would be beneficial because it would eliminate the intermediate step of having to extract water levels first. This paper introduces a probabilistic flood mapping procedure based on synthetic aperture radar (SAR) data. Given a SAR image of backscatter values, we construct a total histogram of backscatter values and decompose this histogram into probability distribution functions of backscatter values associated with flooded (open water) and non-flooded pixels, respectively. These distributions are then used to estimate, for each pixel, its probability of being flooded. The new approach improves on binary SAR-based flood mapping procedures, which do not inform on the uncertainty in the pixel state. The proposed approach is tested using four SAR images from two floodplains, i.e., the Severn River (U.K.) and the Red River (U.S.). In all four test cases, reliability diagrams, with error values ranging from 0.04 to 0.23, indicate a good agreement between the SAR-derived probabilistic flood map and an independently available validation map, which is obtained from aerial photography.

Change detection approaches for flood extent mapping: How to select the most adequate reference image from online archives?

Abstract Synthetic Aperture Radar images are routinely used for delineating flooded areas. Processing algorithms are often based on change detection techniques that enable a comparison of backscattering signals between the flood image and a reference image. However, as of today, there is little guidance on how to rapidly and reliably extract the most adequate reference image from an online archive. Our study proposes a method that allows the automatic and objective identification of the best reference image with respect to a given flood image. The proposed method consists of two processing steps. First, a subset of archived candidate images acquired on the same track, with the same polarization and in the same period of year as the flood image is created. Next, site-specific time series of regional backscattering values are established and the effects of flooding on the backscattering behaviour are statistically evaluated. We propose two complementary anomaly indexes and their combination in a single index as a means to identify the most adequate reference image for flooding-related change detection applications. The reliability of the proposed method is demonstrated in three representative case studies targeting the flood prone areas of the Severn River (United Kingdom), the Red River (United States) and the Meghna River (Bangladesh).

Flood Hazard Mapping Combining Hydrodynamic Modeling and Multi Annual Remote Sensing data

This paper explores a method to combine the time and space continuity of a large-scale inundation model with discontinuous satellite microwave observations, for high-resolution flood hazard mapping. The assumption behind this approach is that hydraulic variables computed from continuous spatially-distributed hydrodynamic modeling and observed as discrete satellite-derived flood extents are correlated in time, so that probabilities can be transferred from the model series to the observations. A prerequisite is, therefore, the existence of a significant correlation between a modeled variable (i.e., flood extent or volume) and the synchronously-observed flood extent. If this is the case, the availability of model simulations over a long time period allows for a robust estimate of non-exceedance probabilities that can be attributed to corresponding synchronously-available satellite observations. The generated flood hazard map has a spatial resolution equal to that of the satellite images, which is higher than that of currently available large scale inundation models. The method was applied on the Severn River (UK), using the outputs of a global inundation model provided by the European Centre for Medium-range Weather Forecasts

A Hierarchical Split-Based Approach for Parametric Thresholding of SAR Images: Flood Inundation as a Test Case

Parametric thresholding algorithms applied to synthetic aperture radar (SAR) imagery typically require the estimation of two distribution functions, i.e., one representing the target class and one its background. They are eventually used for selecting the threshold that allows binarizing the image in an optimal way. In this context, one of the main difficulties in parameterizing these functions originates from the fact that the target class often represents only a small fraction of the image. Under such circumstances, the histogram of the image values is often not obviously bimodal and it becomes difficult, if not impossible, to accurately parameterize distribution functions. Here we introduce a hierarchical split-based approach that searches for tiles of variable size allowing the parameterization of the distributions of two classes. The method is integrated into a flood-mapping algorithm in order to evaluate its capacity for parameterizing distribution functions attributed to floodwater and changes caused by floods. We analyzed a data set acquired during a flood event along the Severn River (U.K.) in 2007. It is composed of moderate (ENVISAT-WS) and high (TerraSAR-X)-resolution SAR images. The obtained classification accuracies as well as the similarity of performance levels to a benchmark obtained with an established method based on the manual selection of tiles indicate the validity of the new method.

A user-driven case-based reasoning tool for infilling missing values in daily mean river flow records

Missing data in river flow records represent a loss of information and a serious drawback in water management. In this work, we introduce gapIt, a user-driven case-based reasoning tool for infilling gaps in daily mean river flow records. Given a set of flow time series, gapIt builds a database of artificial gaps for which it computes several flow estimates, to find the best combinations of infilling algorithm and automatically selected donor station(s), according to state-of-the-art performance indicators. We obtained satisfactory results with Nash-Sutcliffe >0.7 for more than half of the ~5000 synthetic gaps of various lengths and positions, randomly created along the available records. gapIt was evaluated on 24 daily river discharge time series recorded in Luxembourg over seven years from 01/01/2007 to 31/12/2013. We also discuss the benefits of coupling this approach with user-expertise for an improved infilling of real data gaps. We propose a user-driven case-based reasoning tool to infill gaps in flow records.With data analytics the tool finds the best combination: algorithm - donor stations.Satisfactory results were obtained on synthetic gaps of different characteristics.The tool bridges a gap between data-driven and user-expertise approaches.

From SAR-derived flood mapping to water level data assimilation into hydraulic models

This paper describes a fully automatic processing chain that makes use of SAR images for retrieving water stage information to be assimilated into a hydraulic forecasting model. This chain is composed of three steps: flood extent delineation, water stage retrieval and data assimilation of stage information into a hydraulic model. The flood-mapping step is addressed with a fully automatic algorithm, based on image statistics and applicable to all existing SAR datasets. Uncertainty on the flood extent map is represented with an ensemble of flood extent maps, obtained following a bootstrap methodology. Water stage observations are then retrieved by intersecting the flood shoreline with the floodplain topography. The ensemble of flood extent maps allows extracting multiple water levels at any river cross section of the hydraulic model, thereby taking into account the uncertainty associated with the floodmapping step. Finally, data assimilation consists in integrating uncertain observations, i.e. SAR-derived water stages, with uncertain hydraulic model simulations. The proposed processing chain was applied to two case studies. For the test case of June 2008 on the Po River (Italy), only low resolution but freely available satellite data were used. For the January 2011 flood on the Sure River (Luxembourg), higher resolution data were used and obtained at a cost. The results show that with the assimilation of SAR-derived water stages significant improvements can be achieved in the forecasting performance of the hydraulic model.

Flood hazard mapping combining high resolution multi-temporal SAR data and coarse resolution global hydrodynamic modelling

A new method for flood hazard mapping that integrates global flood inundation modeling and microwave remote sensing is presented. It combines the time and space continuity of a global inundation model with the limited revisit time but high spatial resolution of satellite observations. The availability of model simulations over a long time period allows a robust estimate of non-exceedance probabilities that can be attributed to the corresponding satellite observations. The resulting flood hazard map will have a spatial resolution equal to that of the used satellite images, generally higher than that of the global inundation model. This can theoretically be done for any point in the world, allowing the estimation of flood hazard at a global scale, provided that a sufficient number of remote sensing images are available. The method is tested on the Severn River (UK), with a high number of flood events observed by ENVISAT ASAR. The global ECMWF flood inundation model is considered for this study.

Near‐Real‐Time Assimilation of SAR‐Derived Flood Maps for Improving Flood Forecasts

Shortto medium-range flood forecasts are central to predicting and mitigating the impact of flooding across the world. However, producing reliable forecasts and reducing forecast uncertainties remains challenging, especially in poorly gauged river basins. The growing availability of synthetic aperture radar (SAR)-derived flood image databases (e.g., generated from SAR sensors such as Envisat advanced synthetic aperture radar) provides opportunities to improve flood forecast quality. This study contributes to the development of more accurate global and near real-time remote sensing-based flood forecasting services to support flood management. We take advantage of recent algorithms for efficient and automatic delineation of flood extent using SAR images and demonstrate that near real-time sequential assimilation of SAR-derived flood extents can substantially improve flood forecasts. A case study based on four flood events of the River Severn (United Kingdom) is presented. The forecasting system comprises the SUPERFLEX hydrological model and the Lisflood-FP hydraulic model. SAR images are assimilated using a particle filter. To quantify observation uncertainty as part of data assimilation, we use an image processing approach that assigns each pixel a probability of being flooded based on its backscatter values. Empirical results show that the sequential assimilation of SAR-derived flood extent maps leads to a substantial improvement in water level forecasts. Forecast errors are reduced by as much as 50% at the assimilation time step, and improvements persist over subsequent time steps for 24 to 48 hr. The proposed approach holds promise for improving flood forecasts at locations where observed data availability is limited but satellite coverage exists.