NH1.2

The disastrous flood of July 2021 in Germany has shown that forecasts of river discharge or water levels at selected gauges do not provide sufficient information for timely and location specific warning of the population and targeted disaster management actions. The gauge forecasts as well as the available flood hazard maps were insufficient to assess the flood severity in downstream areas. In order to provide more actionable flood forecasts, the hydraulic model RIM2D was developed and setup for the Ahr river. It solves the inertial formulation of the shallow water equations on a regular grid, and is highly parallelized on Graphical Processor Units (GPUs). Moreover, the modelling concept is parsimonious and allows for fast model setup. We show that hydraulic simulations driven by the available hydrological gauge forecasts would have been feasible with short simulation duration. It would be possible to provide spatially explicit forecasts of inundation depths and flow velocities with sufficient lead time. Moreover, we also show that impact forecasts indicating human instability in water and building failure hazard can be additionally provided in operational mode. We argue that using these hydraulic and impact forecasts would have had a substantial impact on the flood alertness of the population and responsible authorities, enabling a better early warning and disaster management. This could eventually save lives during future extreme flash floods.

How to cite: Apel, H., Vorogushyn, S., and Merz, B.: Operational hydraulic flood impact forecasting with RIM2D for improved disaster management, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-1424, https://doi.org/10.5194/egusphere-egu22-1424, 2022.

With the acceleration of urbanization, urban pluvial flooding seriously threatens urban sustainable development and human life. It is widely accepted that various landscape elements contribute to the magnitude of urban pluvial flooding. Considerable efforts investigated the universal mechanism of urban pluvial flooding by regarding the whole study area as spatial homogeneous while ignoring its local specific mechanism. The spatially heterogeneous effects of landscape elements on urban pluvial flooding remain poorly understood. Additionally, it is still unclear how the interactive effects of landscape elements affect urban pluvial flooding. In most practical situations, urban pluvial flooding is affected by multiple factors, rather than by a single factor alone. These shortcomings make it impossible to formulate urban pluvial flooding mitigation measures based on the relative contribution of various landscape elements on urban pluvial flooding. To shed some light on this topic, an innovative method that integrated the all subsect regression model, cubist regression tree, and geographical detector model is presented to spatially explicit the heterogeneous forces driving urban pluvial flooding variation and identify the pluvial flooding dominant driving forces with different local conditions. By comparing with two other commonly used regression methods (global regression model, spatial lag model), the proposed method can fully quantify this spatial non-stationarity mechanism and spatially explicit the local driving forces. Urban pluvial flooding dominant driving factors and their contribution vary with the local site conditions. Even for the same dominant factor, its contribution to pluvial flooding varies considerably in different watersheds. Based on this, local authorities can develop site-specific urban pluvial flooding mitigation strategies according to the dominant factors in different areas. The results of this study extend our scientific understanding of the site-specific mechanism of urban pluvial flooding, providing useful information for formulating more targeted and effective urban pluvial flooding mitigation strategies with different local conditions, rather than a “one-size-fits-all” policy.

How to cite: Zhang, Q., Wu, Z., Guo, G., and Tarolli, P.: How to develop site-specific urban pluvial flooding mitigation strategies? A new approach to investigating the spatial heterogeneous driving forces of urban pluvial flooding, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-1573, https://doi.org/10.5194/egusphere-egu22-1573, 2022.

The monitoring of river water-levels is essential to study floods and mitigate their risks. However, it is difficult to obtain accurate measurements of river water-levels: indeed, the river gauges commonly used to measure these levels can be overwhelmed during flood events, and their number is declining globally [1,2]. This means that the monitoring and study of floods relying on gauge station measurements can only be based on sparse and possibly inaccurate river water-level data distributed unevenly along the rivers, sometimes several kilometres away from the location of interest.

We investigate if deep learning can be used to monitor river water-levels in a more flexible and efficient way. More specifically, we apply two deep learning approaches on river cameras, which are CCTV cameras commonly used to monitor the surroundings of rivers and could be easily installed at new locations. The first approach [3,4] relies on transfer learning to train water segmentation networks able to find the water pixels within the camera images and use the number of water pixels within (regions of) the images to monitor the relative evolution of the river water-level. The second approach is based on the creation of a large dataset of 32,715 images annotated with distant gauge water-level data in order to accurately train networks able to produce river water-level indexes independent from the field of view of the cameras. 

We show that both approaches can be used as sources of river water-level data. The first approach is able to produce river water-level indexes highly correlated with ground truth river water-levels (Pearson correlation coefficient >0.94). While the second approach is less accurate (Pearson correlation coefficients between 0.8 and 0.94), it is able to produce calibrated indexes independent from the field of view of the camera. 

[1] Mishra, A. K., and Coulibaly, P. (2009), Developments in hydrometric network design: A review, Rev. Geophys., 47, RG2001, doi:10.1029/2007RG000243.

[2] Global Runoff Data Center (2016).  Global runoff data base, temporal distribution of available discharge data.  https://www.bafg.de/SharedDocs/Bilder/Bilder_GRDC/grdcStations_tornadoChart.jpg. Last visited:2021-04-26.

[3] Vandaele, R., Dance, S. L., & Ojha, V. (2020, September). Automated water segmentation and river level detection on camera images using transfer learning. In DAGM German Conference on Pattern Recognition (pp. 232-245). Springer, Cham.

[4] Vandaele, R., Dance, S. L., & Ojha, V. (2021). Deep learning for automated river-level monitoring through river camera images: an approach based on water segmentation and transfer learning. Hydrology and Earth System Sciences, 25(8), 4435-4453.

How to cite: Vandaele, R., Dance, S. L., and Ojha, V.: Deep learning approaches to study floods through river cameras, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2344, https://doi.org/10.5194/egusphere-egu22-2344, 2022.

Flood is one of the most devastating natural disasters that cause enormous socioeconomic and environmental destruction. The severity of flood losses has evoked emphasis on more comprehensive and vigorous flood modeling techniques for alleviating flood damages. Flood vulnerability in Navsari is intensifying due to urbanization, industrialization, and population growth. Although there has been a significant increase in research on flood assessment at a local scale in Navsari, there remains a lack of tools developed which utilize the risk map of the city. In response to this prerequisite, in this study we have employed a GIS-based Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) multi-criteria analysis model to develop a flood risk map for Navsari city in Gujarat, India, to determine the vulnerable areas that are more susceptible to flooding. To estimate the extent of flood hazard, vulnerability, and risk intensities in terms of area covered, the city was divided into ten zones (i.e. NC1 to NC10) and classified into five classes: very high, high, moderate, low, and very low. A total of seven hazard forming spatial layers (i.e. slope, elevation, soil, rainfall, flow accumulation, distance to a river, and drainage density) and seven vulnerability forming spatial layers (i.e. female population, population density, land use, household, distance to hospital, road network density, and literacy rate) were appraised for evaluating the risk of flooding. The generated flood risk map has been compared with the extent of flood calculated based on field data collected from thirty-six random places. The outcome of the model unveiled the capability of the TOPSIS model since it capitulate low RMSE value varied between 0.95 to 0.43 and high R square value ranged from 0.78 to 0.95. The zones indicated under ‘high’ and ‘very high’ categories (i.e. NC8, NC6, NC4, NC1, NC7, and NC10) demand abrupt flood control action to alleviate the severity of flood risk and subsequent damages. The approach implemented in the study can be applied to any flood-sensitive region around the globe to accurately evaluate the risk of flood. Lastly, flood risk mapping using TOPSIS based geospatial techniques divulge the novel and efficacious approach, especially for data-sparse regions.

How to cite: Pathan, A. I., Agnihotri, Dr. P. G., Said, Dr. S., Patel, Dr. D., Prieto, Dr. C., Mohsini, U., Patidar, N., Gandhi, Dr. P., Jariwala, K., Đurin, B., Azimi, M. Y., Rasuli, J., Dummu, K., Raaj, S., Shaikh, A. A., and Salihi, M.: Flood risk mapping using multi-criteria analysis (TOPSIS) model through geospatial techniques- A case study of the Navsari city, Gujarat, India, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2418, https://doi.org/10.5194/egusphere-egu22-2418, 2022.

Floods are natural hazards with severe socio-economic and environmental impacts on affected areas and societies every year. A chain of different processes being involved in a flooding - characterized by precipitation, topography, land use etc. - complicates the understanding of the dynamics of a flood. However, the prediction of probabilities, flood hazards, flooding extents, dike failure, consequences and understanding the ongoing processes during a flood event are important issues in flood risk management. Computational modelling is a key method in supporting flood risk management and tackling the mentioned challenges.

While several computer-based models for assisting flood risk management exist, typically they concentrate on only one component of the flood risk analysis chain such as rainfall generation, hydrological/hydraulic modelling or damage analysis. They do not merge the other components on one platform which may result in encapsulated conclusions. In recent years the availability of higher detailed data, larger study domains, more computational power and more innovative models paved the way for more effective solutions.

In this work we present ProMaIDes (Protection Measures against Inundation Decision support), an open-source, free software package for risk-based evaluation of flood risk mitigation measures¹. The software package consists of numerous relevant modules for a flood risk analysis in riverine and coastal regions: the HYD-module for a hydrodynamic analysis, the DAM-module for an analysis of consequences (including economical damage, consequences to people and the disruption of critical infrastructure services), the FPL-module for the reliability analysis of dikes and dunes as well as a combining RISK-module and the decision support MADM-module. To support a user-friendly model setup, visualization of input and data results, a connection with the free QGIS-system is established by QGIS-plugins and a PostgreSQL-database as data-management system. A detailed online documentation featuring theory, application and programming is available². A community of users is currently set-up.

In order to give a better understanding and to demonstrate the capabilities of ProMaIDes, the tool itself, but also the modules combined with case studies are shortly presented.

¹ https://promaides.h2.de

² https://promaides.myjetbrains.com/youtrack/articles/PMID-A-7/General

How to cite: Bachmann, D., Schotten, R., and Khosh Bin Ghomash, S.: Introducing ProMaIDes: A State-of-the Science Flood Risk Management Tool, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2696, https://doi.org/10.5194/egusphere-egu22-2696, 2022.

Slope instability of river dikes during floods is often driven by the evolution of groundwater pressures. Despite the temporal nature of high river water levels, pressure heads during floods are often assumed to reflect steady-state seepage conditions, leading to conservative estimates of dike slope safety. Here, we investigate the influence of transient groundwater conditions that result from variable flood wave shapes on probabilistic safety estimates of slope stability. We have sampled a large number of flood waves, aiming to maximize the variability in the flood wave shapes, and used them in a modeling chain consisting of a hydrological model (MODFLOW) and a probabilistic dike slope safety assessment (FORM). We compared the resulting time-dependent probabilistic dike safety for inner (landward) slope and outer (riverward) slope stability with the current flood safety assessment in the Netherlands. This comparison showed that current methods based on steady-state and analytical solutions seem to underestimate dike safety. Other methods, based on a design discharge wave, are more consistent with the multi-flood wave dike reliability, but their error increases at extreme water levels. In line with the temporal component of variable flood water levels, the failure probability also has a strong temporal component. Our results indicate that the highest failure probability always occurs after the river water level peak, with a delay of up to 15 days for both inner slope and outer slope stability. In addition, the uncertainty in the shape of the flood wave can be as important as the uncertainty in the geomechanical material properties for explaining the variation in dike failure probabilities. Therefore, this research strongly suggests that transient-groundwater conditions as a function of variable flood wave shapes should be incorporated in dike safety assessment. As a first step, we recommend further research on the occurrence probability of the most influential waveform characteristics, being the total flood wave volume (for the inner slope) and the total water level decrease after the peak (for the outer slope).

How to cite: van Woerkom, T., van der Krogt, M., and Bierkens, M.: On the incorporation of transient groundwater conditions resulting from variable flood wave shapes in probabilistic slope stability assessments of dikes, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-3099, https://doi.org/10.5194/egusphere-egu22-3099, 2022.

This study presents a forecasting model for pluvial flooding in the city of Zadar, Croatia, where a huge mesoscale convective system recently caused massive pluvial flooding and widespread property damage. Flood forecasting approaches based on hydrologic-hydraulic models require a large set of accurate data to provide reliable simulations. They also require many simulations, which can be computationally expensive and time consuming. Therefore, we are investigating the possibility of using a data-driven approach based on local news reports of pluvial flooding combined with a local high-resolution rain gauge. To this end, we considered two different computational approaches. The first - a conventional one - is based on rainfall threshold curves that define the critical rainfall depth for different time periods above which flooding is likely to occur. The second approach is based on machine learning and a classification problem - predicting whether accumulated rainfall depths over different time periods will lead to pluvial flooding. For the second approach, we considered 10 different methods that belong to five categories of machine learning typically used for classification problems. They are logistic regression, support vector machine, discriminant analysis, decision trees, and nearest neighbours. After a careful analysis, we defined rainfall threshold curves for Zadar that can be used for an early warning system and flood forecasting. We show that some machine learning models can provide slightly more accurate predictions than the threshold curve, with quadratic discriminant analysis being the most successful method for this purpose. Overall, this study shows that flood forecasting based on news reports in the city of Zadar can be a reliable approach. The analysis conducted in this study has laid the foundation for the implementation of an early warning system and pluvial flood forecasting in the Croatian coastal area.

How to cite: Krvavica, N., Horvat, B., Marinović, I., and Šiljeg, A.: Rainfall threshold curves and machine learning approaches for pluvial flood forecasting based on local news reports in Croatia, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4201, https://doi.org/10.5194/egusphere-egu22-4201, 2022.

Surat is a district that has seen numerous floods and high rainfall over the last two decades. The solution to the problem, and the primary aim of this study, is to construct a storage facility, such as a dam, as part of flood prevention measures. The concept of multi-criteria decision making (MCDM) is now widely employed for everyday real-life challenges. Recent advancements and diverse approaches in geographic information systems (GIS) and remote sensing, along with the MCDM technique, will enable us to make an informed decision about where to build a dam site location model (DSLM). The Analytic Hierarchy Process (AHP) is the most frequently utilised MCDM technique for resolving water-related issues. To produce DSLM, ten thematic layers were considered: precipitation, stream order, geomorphology, geology, LULC, soil, distance to road, elevation, slope, and major fault fracture. Precipitation and stream order were the two most important elements affecting the DSLM. The weights of the thematic map layers were determined using the analytical hierarchy process (AHP) technique. These thematic maps and weights are used to perform overlay analysis, resulting in a suitability map with five classes ranging from high to low suitability. Three main sites have been selected as the best candidates for the construction of a new dam. By implementing this low-cost strategy, we may be able to reduce the amount of effort required in the traditional method of dam site selection while increasing decision-makers' accuracy. Approximately 14% of the Surat district is classified as a very high adaptability area, while 27.2 percent is classified as a high suitability area. This method can be applied all over the world to locate possible dam sites, which can be helpful for flood mitigation measures. In addition to that, the presented approach unveiled the scientific method for flood mitigation measures, which are in immediate demand all over the globe, especially in data-scarce regions.

How to cite: Raaj, S., Pathan, A., Mohseni, U., Patidar, N., jariwala, K., Kachhawa, N., Agnihotri, Dr. P. G., Patel, Dr. D., Prieto, Dr. C., Gandhi, Dr. P., and Đurin, Dr. B.: An Integrated Approach of AHP-GIS Based Dam Site Suitability Mapping - A Noval Approach for Flood Alleviating Measures, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4210, https://doi.org/10.5194/egusphere-egu22-4210, 2022.

Both frequency and severity of urban pluvial floods have been increasing due to rapid urbanization and climate change. Hydrological and two dimensional (2D) hydrodynamic models are still too computationally demanding to be used for real-time applications for large urban areas (i.e. flood management scale). As an alternative, data-driven models could be used for flood susceptibility mapping. This study evaluated and compared the performance of image-based models represented by a convolutional neural network (CNN) and point-based models represented by an artificial neural network (ANN), a random forest (RF) and a support vector machine (SVM) with regard to the spatial resolution of the input data. We also examined model transferability. Eleven variables representing topography, anthropogenic aspects and precipitation were selected to predict flood susceptibility mapping. The results showed that: (1) all models were skilful with a minimum area under the curve AUC = 0.87. (2) The RF models outperformed the other models for all spatial resolutions. (3) The CNN models were superior in terms of transferability. (4) Aspect and elevation were the most important factors for flood susceptibility mapping for image-based and point-based models respectively.

How to cite: Seleem, O., Ayzel, G., Costa Tomaz de Souza, A., Bronstert, A., and Heistermann, M.: Towards Urban Flood Susceptibility Mapping Using Data-Driven Models, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4345, https://doi.org/10.5194/egusphere-egu22-4345, 2022.

Coastal areas are directly vulnerable to natural disasters like floods, which causes massive damages to natural resources and human resources. Dam induces floods can be devastating for surrounding low lying areas. Bharuch is a district with substantial industrial growth, and intended human activities were causing an imbalance in natural resources for planning and fulfilling other demands. Floods can be devastating concerning the Bharuch district's social, economic, and environmental perspectives. The proper analysis becomes very important to reduce the impact and find mitigation measuring techniques. I did flood susceptibility mapping using the frequency ratio model for the six sub-districts of the area. The susceptibility of a flood was analysed using the frequency ratio model by considering nine different independent variables (land use/land cover, elevation, slope, topographic wetness index, surface runoff, lithology, distance from the main river, soil texture, river network) through weighted-based bivariate probability values. In total, 151 historical floods were reported. I took locations for this study, from which I used 72 locations for susceptibility mapping. I combined the independent variables and historic flood locations to prepare a frequency ratio database for flood susceptibility mapping. The developed frequency ratio was varied from 0 to 13.2 and reclassified into five flood vulnerability zones, namely, very low (less than 0.99), low (0.99-2.04), moderate (2.04-5.58), high (5.58-13.2) and very high susceptibility (more than 13.2). The flood susceptibility analysis will be a valuable and efficient tool for local government administrators, researchers, and planners to devise flood mitigation plans.

Keywords: Flood Susceptibility · Flood · Frequency Ratio · Vulnerability · Bharuch

How to cite: Jariwala, K., Agnihotri, P., Patel, D., Pathan, A., Mohseni, U., and Patidar, N.: Application of frequency ratio modelling technique for predictive flooded area susceptibility mapping using remote sensing and GIS, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4771, https://doi.org/10.5194/egusphere-egu22-4771, 2022.

Flood is one of the most devastating natural disasters. The damages of flood usually vary with the consideration of different factors (depth, duration, velocity, materials of infrastructures) of flooding. Therefore, flood damage estimation is a complex process. Most of previous studies considered only flood depth in developing flood damage functions for residential houses. However, the consideration of other flood parameters such as flood duration and flood velocity are also crucial to estimate flood damage more reliably. Therefore, this study aimed to consider various flood parameters such as flood depth, flood duration, and flood velocity in development of flood damage functions for residential houses.  In this study, the Teesta River Basin in Bangladesh was chosen as the study area. A detailed household questionnaire survey was conducted in flood-affected areas of Lalmonirhat and Rangpur districts (administrative unit of Bangladesh) to collect data of 2017 and 2019 flood events.  Most of the houses in the surveyed flood-affected areas are composed of mud base and side wall of corrugated iron sheets (called “MC type”). For each house, the questionnaire aimed to identify the flood information (flood depth, flood duration, the qualitative representation of flood velocity), household structure information (area, plinth height, ceiling height), structural damage mechanism and the required amount of material with labor work to repair the damage after each flood event. Using the survey data, we have developed depth-damage functions for MC type of house by considering different flood velocity and flood duration combinations. The newly developed depth-damage functions can generalize thresholds of flood depth, flood velocity and flood duration that are responsible for specific type of structural damages (mud removal from the base, mud removal from the base together with side wall instability, full structure instability) of MC type house. Finally, a grid-based approach through the integration of new depth-damage functions with hydrologic-hydraulic model (RRI) and Nays2DFlood Solver (iRIC software) simulation results has been developed to estimate the total flood damage for MC type houses in flood-affected areas of the Teesta River Basin. This comprehensive method can be easily used to derive the depth-damage functions and estimation of total damage for other types of houses if enough surveyed data can be obtained from the field.

Keywords: Flood damage estimation, Depth-damage function, MC type house, Hydrologic-hydraulic model

How to cite: Haque, S., Ikeuchi, K., Shrestha, B. B., and Minamide, M.: Generalizing flood damage mechanism processes of MC Type houses by developing comprehensive flood damage estimation method for Teesta River Basin, Bangladesh, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-5110, https://doi.org/10.5194/egusphere-egu22-5110, 2022.

In Québec, Canada, extraordinary spring conditions in 2017 and 2019 have provided major incentives for the provincial government to commission the updating of current flood inundation maps. Indeed, some of these maps, dating back as far as the 1980’s, do not adequately reflect actual flood risks. Classical hydrodynamic models, such as HEC-RAS (1D, mixed, or full 2D), are generally used to perform the mapping, but they do require significant expertise, hydrometric data, and high-resolution bathymetric surveys. Given the need for updating flood inundation maps and reducing the associated financial costs (data collection and human resources), there is an emerging demand for simplified conceptual methods. In recent years, several models have been developed to fulfill this need, including the geomatic Height Above the Nearest Drainage (HAND) method which solely relies on a the digital elevation model (DEM).

This project aims at expanding upon earlier work carried out with HAND which was designed to compute the required water height to flood any DEM pixel of a watershed. The information provided by HAND along with the application of the Manning equation allow for the construction of a synthetic rating curve for any homogeneous river reach. This methodological approach has been used to come up with first-instance flood inundation mapping of large rivers in conterminous United States with a matching rate reaching 90% when compared to the use of HEC-RAS. However, to our knowledge, this has not been assessed for small rivers, and our goal here is to validate this simplified conceptual approach using two small watersheds (less than 200 km²) in Quebec.

The results of this study show that the ensuing synthetic rating curves for small rivers are consistent with river hydraulics (Froude numbers meeting the subcritical flow requirement behind the use of Manning equation) and in-situ derived rating curves of six hydrometric stations. The results also demonstrate the relevance of this approach when comparing the use of HAND with HEC-RAS 2D for the hydrographic networks of the two watersheds given flows simulated by a semi-distributed hydrological model (i.e., HYDROTEL). For this demonstration, the forcing data include the precipitation and temperature time series of the Canadian precipitation analysis system. Preliminary results indicate good performances (hitting rate above 60%) for the pilot river watersheds which are located in a data-sparse region.

While the preliminary results illustrate the potential to produce first-instance flood inundation mapping solely based on a DEM and simulated streamflows, future work will contribute to the advancement of our understanding of flood risks in poorly-gauged watersheds. HAND-derived inundation mapping will be further analyzed and compared to HEC-RAS-2D applications (i.e., the diffusion-wave equations), although the presence of complex urban infrastructures such as culverts, pipes, or bridges may represent a major challenge for the proposed approach. We believe a modeling continuum based on hydrological modeling and HAND-derived flood inundation mapping will inform and strengthen land management planning and contribute to the elaboration of public safety protocols.

How to cite: El Batti, Z., Foulon, E., Gordon, C., and Rousseau, A.: Development of a loosely coupled geomatic-hydrological modeling approach for flood inundation mapping in small watersheds, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6322, https://doi.org/10.5194/egusphere-egu22-6322, 2022.

Floods, the most frequent and severe of natural disasters worldwide, inflict significant social, environmental and fiscal impacts, including: loss of human life, damage to natural habitats and damage to infrastructure. Flood risk mapping can be used to mitigate these impacts as it provides a holistic approach to identifying flood prone areas by simultaneously considering socioeconomic and environmental indicators. This research compares the performance of two multi-criteria decision making methods, and one Machine Learning (ML) method in the development of flood risk mapping. This approach was first developed and validated for the Don River watershed in the Greater Toronto Area and subsequently extended to several other watersheds across Southern Ontario. Remote sensing data such as Digital Elevation Models and landuse and lancover datasets were used to develop the environmental flood hazard extent, and combined together with socioeconomic indicators, flood risk maps were developed using subjective and objective weighting schemes in a GIS analysis. The subjective maps were produced using the Analytical Hierarchy Process (AHP), the objective maps were produced using the Shannon Entropy method and the ML maps were produced using Artificial Neural Networks. The accuracy of these maps was compared against the floodplain map of the Don River. For a range of flood risk severity, where 1 was very low risk and 5 was very high risk, the AHP maps were superior in identifying areas where flood risk severity was 4 or greater. Conversely, the Entropy maps were superior in identifying areas where flood hazard risk was 5, however the difference in accuracy for both scenarios was marginal between the two methods. The accuracy of the ML maps showed marginal superior performance under both scenarios in comparison to the multi-criteria maps. Additionally, the uncertainty in the combination of flood risk indicators was quantified through a sensitivity analysis focusing on the discretization of the number of classes in each indicator dataset. The outcome of this research provides an accurate and simplified alternative to using hydrological and hydraulic models, especially when insufficient data limits the use of hydrological and hydraulic models. Future research should focus on an optimisation approach to the discretization of classes in indicator datasets.

How to cite: Khalid, R. and Khan, U. T.: A comparison of multi-criteria and machine learning weighting for flood risk assessment in the Southern Ontario, Canada, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6574, https://doi.org/10.5194/egusphere-egu22-6574, 2022.

This research examined a new method for coupling flood flow modelling with the machine learning (ML)-based land cover detection from the Unmanned Aerial Vehicle (UAV) and satellite river imagery. We examined a 2 km river channel section with a gravel bed in the Kurobe River, Japan. The method used Random Forests (RF) for riverine land cover detection with the satellite images' RGBs and Near InfraRed (NIRs). In the process, the UAV images were used effectively to train the RF in several small portions of the river channel where the types of riverine land cover were precise. Using these UAV images with the corresponding feature values (i.e., RGBs and NIRs) of the satellite images made it possible to create the training data with high accuracy for land cover detection. The results indicated that combining the high- and low-resolution images in the RF could effectively detect waters, gravel/sand, trees, and grasses from the satellite images with a certain degree of accuracy. Its F-measure, consisting of precision and recall rates, had high enough with 0.8. Then, the ML-based land covers were coupled with a flood flow model. In the coupling, the results of the detected riverine land covers were converted into the roughness coefficients of the two-dimensional flood flow analysis. The flood flow simulation could reproduce the velocity field and water surface profile of flood flows with high accuracy. These results strongly suggest the effectiveness of coupling the current flood flow modelling with the ML-based land cover detection for grasping the most vulnerable portions in river flood management.

How to cite: Sato, T., Iwami, S., and Miyamoto, H.: Flood flow modelling coupled with ML-based land cover detection from UAV and satellite river imagery, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6636, https://doi.org/10.5194/egusphere-egu22-6636, 2022.

Flood warning systems are needed to plan mitigation measures and inform response strategies. The extent and dynamics of floods are typically predicted using physics-based hydrological models, which are computationally expensive and data assimilation is difficult. Deep-learning models can overcome these limitations, enabling fast predictions informed by multiple sources of data. Studies show this can be achieved while retaining or improving the level of detail and accuracy previously attainable. We, therefore, propose a deep-learning flood forecasting tool that combines multiple sources of readily available data to quickly generate flood extent maps, which can inform warnings.

We train a neural network with U-NET architecture consisting of encoder and decoder convolutional modules. In the encoder module, features are extracted from the input and the data is downsampled to reduce complexity. Subsequently, the data is upsampled back to the original dimension in the decoder module and each 10 by 10 m pixel of the output image represents a flood prediction. The input to the neural network includes radar rainfall observations, LIDAR topographic scans, soil type and land use maps, groundwater depth simulations and previous inundation maps. All inputs are individually normalized and pre-processed. The rainfall observations are temporally aggregated to various intervals, hydrological features are highlighted in the topographic scans, and soil types and uses are grouped into categories.

The model is trained and evaluated against a set of maps of surface water extent derived from Synthetic Aperture Radar (SAR) satellite observations. The predictions are scored against the target images by computing the critical success index (CSI), which measures the percentage of correct predictions among the total predicted of observed flooded areas. Permanent water bodies and areas where flooding is not captured by the satellite images (e.g. in forests) are masked out during both training and evaluation. The model is trained on a set of flooding events that occurred between 2018 and 2020 within the Jammerbugt Municipality in northern Denmark, which extends for about 850 km². The model is validated on spatially independent data and tested on temporally independent events from the same study area.

The proposed model yielded up to ~60% CSI with the test dataset, which is comparable to existing flood screening approaches. The test data included both fluvial and pluvial flooding as well as observed surface water in coastal areas. Large flooded areas were correctly predicted, while false negatives were frequently obtained for smaller areas. The overall performance of the proposed method is expected to improve by further tuning the model hyperparameters and by treating separately flood processes with different dynamics (e.g. pluvial vs. fluvial vs. coastal). These tradeoffs are compensated by the minimal computational time required to generate predictions once the model has been trained. Also, it is expected that the model can easily be transferred to other locations since it relies on local topographic information. The additional advantage of using a deep-learning approach is the ability to easily integrate alternative and additional data sources, which enables, for example, longer-term flood warnings driven by rainfall forecasts instead of observations.

How to cite: Palmitessa, R., Hjermitslev, O. G., Johansen, H. E., Arnbjerg-Nielsen, K., Bauer‐Gottwein, P., Mikkelsen, P. S., and Löwe, R.: Improved flood predictions by combining satellite observations, topographic information and rainfall spatial data using deep learning, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-8823, https://doi.org/10.5194/egusphere-egu22-8823, 2022.

Japan has a history of major natural disasters, mostly due to its geographical characteristics and topographic features. Major typhoons and floods cause severe damages to lives, properties and important infrastructure, which may increase in future due to climate change. Therefore, sustainable and cost-effective disaster management strategies are of timely requirement, and paddy fields in the river flood plain areas of Japan can be effectively utilized in this regard. After the paddy harvest season, most paddy fields remain unused for a few months and during this time it can work as storage reservoir with minor interventions. During intense rainfall, water can be stored within the paddy field bunds if the drainage outlets are kept closed for some time. Thus, contribution of rainwater to the river can be lessened, resulting river discharge reduction to some extent and protecting important areas from flood damages. The potential of paddy fields in Japan as storage reservoir is not well represented in any research that involves hydrological modelling. This study is performed to assess the impact of using paddy fields for river discharge and inundation reduction, through hydrological model simulation. Two major river basins in Japan, Abukuma river in Fukushima prefecture and Chikuma river in Nagano prefecture are selected as study areas. Paddy field covers 15-20% of watershed areas of these rivers and most of these fields are very close to the main river stream, which indicates their fair potential to store rainwater and contribute to discharge reduction. A global hydrological and water resources model named ‘H08’ is used in this study to simulate river discharge for two scenarios, where one is the control scenario with no storage of water within the paddy field and another is storing rainwater within the exiting or extended paddy bunds. Simulations are performed for 2018 and 2019 to compare the normal flood year and extreme event (a super typhoon occurred in Japan in 2019). Observed and simulated discharge is compared for model calibration and results show better correlation in the upstream section of the rivers. More adjustment of model parameters is still necessary for better calibration. Simulation results show that for 2018, Abukuma river experienced 21-25% decrease in river discharge when water is stored within the conventional 25cm paddy bund. The reduction increased up to 35% when the paddy bunds are assumed to be extended up to 50cm in height. Similar results are observed for Chikuma river basin. For 2019, discharge shows 10-15% decrease for 25cm paddy bunds and around 20% reduction for proposed 50cm bund. With this discharge reduction potential, if paddy field bunds can be extended up to 50cm with a working public-private partnership, where farmers are aware of the advantages of utilizing unused paddy fields as such an effective means of flood management, then this strategy can be considered a sustainable and cost-effective way of disaster management, where the existing land-cover will act as a natural means of storage reservoir. Moreover, this sustainable strategy can be adopted in other countries having similar geographical features as Japan.

How to cite: Saha, D., Oki, K., Yoshida, K., and Kamiya, H.: Improvement of Disaster Management Approaches in Japan Using Paddy Field, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9037, https://doi.org/10.5194/egusphere-egu22-9037, 2022.

Streamflow data based on the observation may be partially missing due to flood or malfunction of the measuring equipment. Here, it is important to complement the missing flow rate with high accuracy for water resource management and flood risk management. Various statistical approaches such as linear regression and multiple regression models have been proposed as methods for complementing missing flow rates. Among the statistical methods, deep learning has been rapidly evolved with the improvement of computational equipment. Then, deep learning methods have achieved remarkable success in various fields. It may indicate that there is a possibility that the missing flow rate can be complemented with high accuracy by using the deep learning method. Therefore, this study implemented deep learning for missing streamflow complementation. In addition, because the network structure of deep learning may have a great influence on estimation accuracy, this study conducted a sensitivity analysis of the network structure. Among the deep learning methods, Bidirectional LSTM (Bi-LSTM) was implemented in this study. Bi-LSTM is a kind of LSTM that can learn long-term dependence of time series data. Bi-LSTM learns data in both forward and backward directions, compared to Unidirectional LSTM which learns data forward directions. As for the input data, both hourly streamflow and precipitation data were used. For model learning and evaluation, missing streamflow data were artificially generated. The results show that Bi-LSTM can complement the flow rate with high accuracy. It also showed the importance of optimizing the network structure.

How to cite: Nagasato, T., Ishida, K., Sakaguchi, D., Amagasaki, M., and Kiyama, M.: Sensitivity analysis of network structure in missing streamflow data complementation using Bidirectional Short-Term Memory network, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10803, https://doi.org/10.5194/egusphere-egu22-10803, 2022.

Evaluating average annual loss (AAL) is an essential component of assessing and minimizing future flood risk. A robust method for quantifying flood AAL is needed to provide valuable information for stakeholder decision-making. Several recent studies suggest that the numerical integration method can provide meaningful AAL estimates since this technique includes the full loss‐exceedance probability of flood. While past research focuses on applying the numerical integration method on a single, one-family residential house, calculations across space are necessary for assessing community vulnerability. This research develops a computational framework in MATLAB for integrating across the full loss-exceedance probability curve through space to evaluate flood AAL for multiple single-family homes, including loss to the structure, content, and time spent in refurbishing it (i.e., use), over a case-study census block in Jefferson Parish, Louisiana, USA. To further inform flood mitigation planning, the AAL is also calculated for one, two, three, and four feet of freeboard and separately for each owner-occupant type of residence (i.e., homeowner, landlord, and tenant). Although previous studies provided essential information related to the structure and content loss for one type for ownership-occupant type (homeowner), the wider scope of this study allows for consideration of the use loss and the remaining ownership-occupant types. Results show that individual building AAL varies within a community because of its building attributes. Besides, results highlight the difference of AALs by owner-occupant type, with homeowners generally bearing the highest total AAL and tenants incurring the lowest total AALs. A simple elevation of only one foot can decrease the AAL by as much as 90 percent. A sensitivity analysis underscores the importance of using the exact values of the base flood elevation (BFE) compared to rounding to the nearest integer and excluding damage lower than first flood height (FFH) in the depth-damage functions (DDFs). Expanding the application of the numerical integration method to a broad-scale study area may enhance validity and accuracy as compensating errors are likely to make bulk estimates more reasonable, which might augment its utility at the community level. In general, such techniques improve resilience to flood, the costliest natural hazard, by assisting in better understanding of risk with and without mitigation efforts. 

How to cite: Mostafiz, R. B., Assi, A. A., Friedland, C., Rohli, R., and Rahim, M. A.: A Numerically-integrated Approach for Residential Flood Loss Estimation at the Community Level, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10827, https://doi.org/10.5194/egusphere-egu22-10827, 2022.

In recent years, disasters are more frequent and enormous due to global warming. In the field of hydrology, high-precision rainfall-runoff modeling is required. Recently, deep learning has been applied to rainfall-runoff modeling and shows high accuracy. It is also expected that the accuracy will be improved by using ensemble learning for deep learning. This study tried to improve the accuracy of river flow estimation by performing ensemble learning for deep learning. Stacking was used as the ensemble learning method. For deep learning, LSTM, CNN, and MLP was used and compared. XGBoost was used as the learning device used for ensemble learning. The target area was the Tedori River basin in Ishikawa Prefecture, Japan. In deep learning, the input data were daily average precipitation and temperature. In deep learning and ensemble learning, the target data was the daily average river flow. RMSE was used as the evaluation index. As a result, the accuracy was the highest after ensemble learning when using LSTM. It shows that the selection of the learning device is important for ensemble learning.

How to cite: Sakaguchi, D., Ishida, K., Nagasato, T., Amagasaki, M., and Kiyama, M.: Improvement of river flow estimation accuracy using ensemble learning stacking, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10891, https://doi.org/10.5194/egusphere-egu22-10891, 2022.

Average annual loss (AAL) is used as the basis for the evaluation of risk mitigation measures.  However, the current AAL implementations in flood risk assessment have several shortcomings. For instance, results generated using Riemann trapezoids for the available return periods of a site are typically gross approximations, especially when damage changes rapidly with depth. Monte Carlo simulations offer improvements in precision but at the expense of being computationally intensive. The log-linear method that extrapolates losses to higher return periods and performs piece-wise Riemann sum with these limited return periods can fail to capture the non-linear flood behavior. This paper presents an improved implementation that quantifies AAL at the micro-scale level including the full range of loss‐exceedance probabilities. To demonstrate the methodology, the financial benefit of increasing the lowest floor elevation for a one-story single-family residence is assessed. Several depth-damage functions (DDFs) are selected and compared to examine the variability in AAL results related to the DDF choice. Results demonstrate the need for an AAL estimate that includes the full loss‐exceedance probabilities. Results also highlight the need to assess flood risk at the micro-scale level for a more localized and accurate assessment, whereupon the estimate can be expanded to broader-scale risk estimations with a higher degree of accuracy. The more realistic AAL estimates results could encourage homeowners and communities to take action and support government decision-makers by investing in flood mitigation and considering building code changes.

How to cite: Rahim, M. A., Gnan, E. S., Friedland, C. J., Mostafiz, R. B., and Rohli, R. V.: An Improved Micro Scale Average Annual Flood Loss Implementation Approach , EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10940, https://doi.org/10.5194/egusphere-egu22-10940, 2022.