A spatial–temporal graph deep learning model for urban flood nowcasting leveraging heterogeneous community features

Farahmand, Hamed; Xu, Yuanchang; Mostafavi, Ali

doi:10.1038/s41598-023-32548-x

Cited by 25 publications

(7 citation statements)

References 82 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In general, the inundation depths are well-known spatial and temporal variables 4 , 22 , 44 ; thus, the 2D flood simulation should be carried out subject to the corrections of the inundation depths in time and space. The aforementioned relationships of estimating inundation depth as the IOT-based grids could persist in the temporal correlation of the gridded rainfall-induced inundation depths.…”

Section: Methodsmentioning

confidence: 99%

Modeling rainfall-induced 2D inundation simulation based on the ANN-derived models with precipitation and water-level measurements at roadside IoT sensors

2023

Sci Rep

View full text Add to dashboard Cite

This study aims to develop a smart model for carrying out two-dimensional (2D) inundation simulation by estimating the gridded inundation depths via the ANN-derived models (ANN_GA-SA_MTF), named SM_EID_2D model. Within the SM_EID_2D model, the rainfall-induced inundation depths at the IoT sensors (i.e., IOT-based grids) are first estimated to be then used in the estimation of inundation depths at the ungauged grids (VIOT-based grids), the resulting flood extents and spatial distribution of inundation of what could be achieved. To facilitate the reliability of the proposed SM_EID_2D model in the 2D inundation simulation, a considerable number of rainfall-induced flood events are generated as the training datasets by coupling the hydrodynamic numerical model (SOBEK) with the simulated gridded rainstorms. To proceed with the model validation and application, the Miaoli City of North Taiwan is selected as the study area, and the associated hydrological and geographical data are adopted in the generation of the training datasets. The results from the model validation indicate that the proposed SM_EID_2D model could provide the gridded inundation-depth hydrographs with a low bias (about 0.02 m) and a high fitness to the validated data (nearly 0.7); also, the spatial distribution of inundated and non-inundated grids as well as the induced flooding extent provided could be well emulated by the proposed SM_EID_2D model under acceptable reliability (0.7). The proposed SM_EID_2D model is also advantageous for the 2D inundation simulation in the real-time delineated subbasins by assembling the emulated inundation depths at the specific grids.

show abstract

Section: Methodsmentioning

confidence: 99%

Modeling rainfall-induced 2D inundation simulation based on the ANN-derived models with precipitation and water-level measurements at roadside IoT sensors

2023

Sci Rep

View full text Add to dashboard Cite

show abstract

“…The weights of the edges represent the degree of similarity in flood vulnerability of neighboring regions. The distance between two regions is the main determinant for assessing the degree of similarity in flood vulnerability [58]. In addition to physical distance, some static features should be considered in combination.…”

Section: Adjacent Matrix Of the Networkmentioning

confidence: 99%

A Dual-Layer Complex Network-Based Quantitative Flood Vulnerability Assessment Method of Transportation Systems

Ding,

Wang,

2024

Land

View full text Add to dashboard Cite

Evaluating the vulnerability of urban transportation systems to flood disasters can provide scientific support for urban disaster prevention and mitigation. Current methods for assessing the flood vulnerability of urban roads often overlook the internal relationships within the complex spatial composition of road networks and surface structures. In this study, based on the theory of complex networks, a dual-layer network assessment model is established for evaluating the flood vulnerability of urban transportation systems by coupling basic geographic data with road network vector data. Unlike traditional methods, this model considers the complex relationship between road network structures and ground surfaces, uncovering a correlation between road network structure and road flood vulnerability. By utilizing this model, the flood vulnerability of road networks in Shenzhen, as well as the city’s spatial flood vulnerability, are quantitatively assessed. Based on the quantitative results, we create maps illustrating the distribution of road and spatial flood vulnerability in Shenzhen. The study results reflect that roads highly vulnerable to flooding are mainly located in the central urban area of the southwest, with the flood vulnerability spatially concentrated primarily in the northern and western regions. Using data from government reports, news stories, and other sources over the past five years, we compile recorded instances of urban waterlogging. The quantitative results of the model are consistent with the distribution trend in recorded waterlogging points, indicating that the model’s outcomes are authentic and reliable.

show abstract

“…In a different take, Sun et al [41] demonstrate how physics-based connectivity could be useful for PUB. Finally, Farahmand et al [42] utilize a similar neural network architecture to this study and utilize attention-based graph neural networks.…”

Section: Graph Neural Network (Gnns) In Hydrologymentioning

confidence: 99%

Spatial Downscaling of Streamflow Data with Attention Based Spatio-Temporal Graph Convolutional Networks

Sit

Demiray

Demir

2023

Preprint

View full text Add to dashboard Cite

Accurate streamflow data is vital for various climate modeling applications, including flood forecasting. However, many streams lack sufficient monitoring due to the high operational costs involved. To address this issue and promote enhanced disaster preparedness, management, and response, our study introduces a neural network-based method for estimating historical hourly streamflow in two spatial downscaling scenarios. The method targets two types of ungauged locations: (1) those without sensors in sparsely gauged river networks, and (2) those that previously had a streamflow sensor, but the gauge is no longer available. For both cases, we propose the ScaleGNN, a graph neural network architecture. We evaluate the performance of ScaleGNN against a Long Short-Term Memory (LSTM) baseline and spatial persistence in estimating discharge values over a 36-hour period. Our findings indicate that ScaleGNN surpasses spatial persistence in the first scenario, while both neural network approaches demonstrate their effectiveness compared to spatial persistence in the second scenario.

show abstract

A spatial–temporal graph deep learning model for urban flood nowcasting leveraging heterogeneous community features

Cited by 25 publications

References 82 publications

Modeling rainfall-induced 2D inundation simulation based on the ANN-derived models with precipitation and water-level measurements at roadside IoT sensors

Modeling rainfall-induced 2D inundation simulation based on the ANN-derived models with precipitation and water-level measurements at roadside IoT sensors

A Dual-Layer Complex Network-Based Quantitative Flood Vulnerability Assessment Method of Transportation Systems

Spatial Downscaling of Streamflow Data with Attention Based Spatio-Temporal Graph Convolutional Networks

Contact Info

Product

Resources

About