Reply on RC2 DOI Creative Commons
Omar Seleem

Published: Dec. 19, 2022

Abstract. Data-driven models have been recently suggested to surrogate computationally expensive hydrodynamic map flood hazards. However, most studies focused on developing for the same area or precipitation event. It is hence not obvious how transferable are in space. This study evaluates performance of a convolutional neural network (CNN) based U-Net architecture and random forest (RF) algorithm predict water depth, models' transferability space improvement using transfer learning techniques. We used three areas Berlin train, validate test models. The results showed that (1) RF outperformed CNN predictions within training domain, presumable at cost overfitting; (2) had significantly higher potential than generalize beyond domain; (3) could better benefit from technique boost their outside domains

Language: Английский

Particle swarm optimization based LSTM networks for water level forecasting: A case study on Bangladesh river network DOI Creative Commons
Jannatul Ferdous Ruma, Mohammed Sarfaraz Gani Adnan, Ashraf Dewan

et al.

Results in Engineering, Journal Year: 2023, Volume and Issue: 17, P. 100951 - 100951

Published: Feb. 10, 2023

Floods are one of the most catastrophic natural disasters. Water level forecasting is an essential method avoiding floods and disaster preparedness. In recent years, models for predicting water levels have been developed using artificial intelligence techniques like neural network (ANN). It has demonstrated that more advanced sequenced-based deep learning techniques, long short-term memory (LSTM) networks, superior at hydrological data. However, historically, LSTM hyperparameters were based on experience, which typically did not produce best outcomes. The Particle Swarm Optimization (PSO) was utilized to adjust hyperparameter increase capacity learn data sequence characteristics. Utilizing observation from stations along Bangladesh's Brahmaputra, Ganges, Meghna rivers, model estimate flood dynamics. Nash Sutcliffe efficiency (NSE) coefficient, root mean square error (RMSE), MAE used assess model's performance, where PSO-LSTM outperforms ANN, PSO-ANN, in all stations. provides improved prediction accuracy stability improves varying lead times. findings may aid sustainable risk mitigation study region future.

Language: Английский

Citations

58
Connecting hydrological modelling and forecasting from global to local scales: Perspectives from an international joint virtual workshop DOI Creative Commons
Antara Dasgupta, Louise Arnal, Rebecca Emerton

et al.

Journal of Flood Risk Management, Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 17, 2023

Abstract The unprecedented progress in ensemble hydro‐meteorological modelling and forecasting on a range of temporal spatial scales, raises variety new challenges which formed the theme Joint Virtual Workshop, ‘Connecting global to local hydrological forecasting: scientific advances’. Held from 29 June 1 July 2021, this workshop was co‐organised by European Centre for Medium‐Range Weather Forecasts (ECMWF), Copernicus Emergency Management (CEMS) Climate Change (C3S) Services, Hydrological Ensemble Prediction EXperiment (HEPEX), Global Flood Partnership (GFP). This article aims summarise state‐of‐the‐art presented at provide an early career perspective. Recent advances forecasting, reflections use forecasts decision‐making across means minimise barriers communication virtual format are also discussed. Thematic foci included model development skill assessment, uncertainty communication, action, co‐production services incorporation knowledge, Earth observation, data assimilation. Connecting societal needs through effective capacity‐building identified as critical. Multidisciplinary collaborations emerged crucial effectively bring newly developed tools practice.

Language: Английский

Citations

19
An Efficient U-Net Model for Improved Landslide Detection from Satellite Images DOI
Naveen Chandra, Suraj Sawant, Himadri Vaidya

et al.

PFG – Journal of Photogrammetry Remote Sensing and Geoinformation Science, Journal Year: 2023, Volume and Issue: 91(1), P. 13 - 28

Published: Jan. 26, 2023

Language: Английский

Citations

14
Accelerating hydrodynamic simulations of urban drainage systems with physics-guided machine learning DOI Creative Commons
Rocco Palmitessa,

Morten Grum,

Allan P. Engsig‐Karup

et al.

Water Research, Journal Year: 2022, Volume and Issue: 223, P. 118972 - 118972

Published: Aug. 11, 2022

We propose and demonstrate a new approach for fast accurate surrogate modelling of urban drainage system hydraulics based on physics-guided machine learning. The surrogates are trained against limited set simulation results from hydrodynamic (HiFi) model. Our reduces times by one to two orders magnitude compared HiFi It is thus slower than e.g. conceptual hydrological models, but it enables simulations water levels, flows surcharges in all nodes links network largely preserves the level detail provided models. Comparing time series simulated model, R2 values order 0.9 achieved. Surrogate training currently hour. However, they can likely be reduced through application transfer learning graph neural networks. will useful interactive workshops initial design phases systems, as well real applications. In addition, our model formulation generic future research should investigate its simulating other systems.

Language: Английский

Citations

22
Concepts and Misconceptions in Climate Change Risk Assessment: Considerations for Sea Level Rise and Extreme Precipitation Risk DOI Open Access
Efthymia Koliokosta

Journal of Geoscience and Environment Protection, Journal Year: 2025, Volume and Issue: 13(01), P. 178 - 214

Published: Jan. 1, 2025

Language: Английский

Citations

0
Flood Susceptibility Zonation Using Geospatial Frequency Ratio and Artificial Neural Network Techniques within Himalayan Terai Region: A Comparative Exploration DOI
Deepanjan Sen, Swarup Das, Sumon Dey

et al.

Communications in computer and information science, Journal Year: 2025, Volume and Issue: unknown, P. 136 - 148

Published: Jan. 1, 2025

Language: Английский

Citations

0
Unraveling the complexities of urban fluvial flood hydraulics through AI DOI Creative Commons
Md Abdullah Al Mehedi, Virginia Smith, Hossein Hosseiny

et al.

Scientific Reports, Journal Year: 2022, Volume and Issue: 12(1)

Published: Nov. 4, 2022

As urbanization increases across the globe, urban flooding is an ever-pressing concern. Urban fluvial systems are highly complex, depending on a myriad of interacting variables. Numerous hydraulic models available for analyzing flooding; however, meeting demand high spatial extension and finer discretization solving physics-based numerical equations computationally expensive. Computational efforts increase drastically with in model dimension resolution, preventing current solutions from fully realizing data revolution. In this research, we demonstrate effectiveness artificial intelligence (AI), particular, machine learning (ML) methods including emerging deep (DL) to quantify considering lower part Darby Creek, PA, USA. Training datasets comprise multiple geographic features (e.g., coordinates, elevation, water depth, flooded locations, discharge, average slope, impervious area within contributing region, downstream distance stormwater outfalls dams). ML Classifiers such as logistic regression (LR), decision tree (DT), support vector (SVM), K-nearest neighbors (KNN) used identify locations. A Deep neural network (DNN)-based depth. The values evaluation matrices indicate satisfactory performance both classifiers DNN (F-1 scores- 0.975, 0.991, 0.892, 0.855 binary classifiers; root mean squared error- 0.027 regression). addition, blocked K-folds Cross Validation (CV) detecting locations showed accuracy 0.899, which validates generalize unseen area. This approach significant step towards resolving complexities large multi-dimensional dataset efficient manner.

Language: Английский

Citations

14
A Machine Learning-Based Surrogate Model for the Identification of Risk Zones Due to Off-Stream Reservoir Failure DOI Open Access
Nathalia Silva‐Cancino, Fernando Salazar, Marcos Sanz‐Ramos

et al.

Water, Journal Year: 2022, Volume and Issue: 14(15), P. 2416 - 2416

Published: Aug. 4, 2022

Approximately 70,000 Spanish off-stream reservoirs, many of them irrigation ponds, need to be evaluated in terms their potential hazard comply with the new national Regulation Hydraulic Public Domain. This requires a great engineering effort evaluate different scenarios two-dimensional hydraulic models, for which owners lack necessary resources. work presents simplified methodology based on machine learning identify risk zones at any point vicinity an reservoir without elaborate and run full models. A predictive model random forest was created from datasets including results synthetic cases computed automatic tool numerical software Iber. Once fitted, provided estimate considering physical characteristics structure, surrounding terrain vulnerable locations. Two approaches were compared balancing dataset: minority oversampling undersampling. Results adjusted undersampling technique showed useful estimation zones. On real application test method achieved 91% accuracy.

Language: Английский

Citations

12
Integrated assessment of flood susceptibility and exposure rate in the lower Niger Basin, Onitsha, Southeastern Nigeria DOI Creative Commons

Ani D. Chinedu,

Nkiruka M. Ezebube,

Smart Uchegbu

et al.

Frontiers in Earth Science, Journal Year: 2024, Volume and Issue: 12

Published: June 17, 2024

Background Various methods have been utilized to investigate and mitigate flood occurrences, yet there is a paucity of literature on factors, such as soil compositions, that contribute persistent flooding in river basins like the Lower Niger catchment, specifically at Onitsha. Furthermore, study seeks furnish essential geospatial data concerning vulnerability, risks, exposure rates Catchment area, situated Onitsha, southeastern Nigeria. Materials Soil samples were collected from 10 specific locations identified through GPS ground-truthing techniques. Additionally, satellite imagery Landsat Enhanced Thematic Mapper (ETM +) was utilized, with supervised classification employed extract feature classes. Analysis operations conducted using IDRISI software, resulting creation digital elevation models (DEMs), susceptibility maps, flood-risk zones. Results revealed predominant composition area comprises sandy (84.8%), silt (8.1%), clayey (7.1%) soils. Utilizing these characteristics alongside relevant aerial data, determined various scales delineate most flood-vulnerable zones basin. It found certain areas, accommodating population exceeding 79,426 across 2,926.2 ha, particularly susceptible flooding. Notably, major markets Bridgehead, Textile, Biafra highly susceptible, varying degrees risk. The prevalence soil, which facilitates increased rainwater infiltration but also prone rapid saturation runoff, likely contributes heightened areas. Conclusion Geospatial analysis employing remote sensing indicates high lower River Basin around Urgent mitigation efforts are imperative, necessitating establishment zoned areas equipped effective drainage systems safeguard vulnerable populations.

Language: Английский

Citations

2
Unsupervised Color Based Flood Segmentation in UAV Imagery DOI Open Access
Georgios Simantiris, Costas Panagiotakis

Published: April 16, 2024

We propose a novel unsupervised semantic segmentation method for fast and accurate flood area detection utilizing color images acquired from Unmanned Aerial Vehicles (UAVs). To the best of our knowledge, this is first fully in captured by UAVs, without need pre-disaster images. The proposed framework addresses problem based on parameter-free calculated masks image analysis techniques. First, algorithm gradually excludes areas classified as non-flood over each component LAB colorspace, well an RGB vegetation index detected edges original image. Unsupervised techniques, such distance transform, are then applied, producing probability map location flooded areas. Finally, obtained applying hysteresis thresholding segmentation. tested compared with variations, other supervised methods two public datasets, consisting 953 total, yielding high-performance results, 87.4% 80.9% overall accuracy F1-Score, respectively. results computational efficiency show that it suitable board data execution decision-making during UAVs flight.

Language: Английский

Citations

1