Flash-flood susceptibility mapping based on XGBoost, random forest and boosted regression trees

Geocarto International, Год журнала: 2021, Номер 37(19), С. 5479 - 5496

Опубликована: Апрель 23, 2021

Historical exploration of flash flood events and producing flash-flood susceptibility maps are crucial steps for decision makers in disaster management. In this article, classification regression tree (CART) methodology its ensemble models random forest (RF), boosted trees (BRT) extreme gradient boosting (XGBoost) were implemented to create a map the Bâsca Chiojdului River Basin, one areas Romania that is constantly exposed floods. The torrential including 962 delineated from orthophotomaps field observations. Furthermore, set conditioning forces explain floods was constructed which included aspect, land use cover (LULC), hydrological soil groups lithology, slope, topographic wetness index (TWI), position (TPI), profile curvature, convergence stream power (SPI). All indicated slope as most important factor triggering occurrence. highest area under curve (AUC) achieved by RF model (AUC = 0.956), followed BRT 0.899), XGBoost 0.892) CART 0.868), respectively. results showed central part river basin, covers approximately 30% study area, more susceptible flooding.

Язык: Английский

---

Deep learning methods for flood mapping: a review of existing applications and future research directions

Hydrology and earth system sciences, Год журнала: 2022, Номер 26(16), С. 4345 - 4378

Опубликована: Авг. 25, 2022

Abstract. Deep learning techniques have been increasingly used in flood management to overcome the limitations of accurate, yet slow, numerical models and improve results traditional methods for mapping. In this paper, we review 58 recent publications outline state art field, identify knowledge gaps, propose future research directions. The focuses on type deep various mapping applications, types considered, spatial scale studied events, data model development. show that based convolutional layers are usually more as they leverage inductive biases better process characteristics flooding events. Models fully connected layers, instead, provide accurate when coupled with other statistical models. showed increased accuracy compared approaches speed methods. While there exist several applications susceptibility, inundation, hazard mapping, work is needed understand how can assist real-time warning during an emergency it be employed estimate risk. A major challenge lies developing generalize unseen case studies. Furthermore, all reviewed their outputs deterministic, limited considerations uncertainties outcomes probabilistic predictions. authors argue these identified gaps addressed by exploiting fundamental advancements or taking inspiration from developments applied areas. graph neural networks operators arbitrarily structured thus should capable generalizing across different studies could account complex interactions natural built environment. Physics-based preserve underlying physical equations resulting reliable speed-up alternatives Similarly, resorting Gaussian processes Bayesian networks.

Язык: Английский

---

Flood susceptibility modeling in Teesta River basin, Bangladesh using novel ensembles of bagging algorithms

Stochastic Environmental Research and Risk Assessment, Год журнала: 2020, Номер 34(12), С. 2277 - 2300

Опубликована: Сен. 4, 2020

Язык: Английский

---

Depth prediction of urban flood under different rainfall return periods based on deep learning and data warehouse

The Science of The Total Environment, Год журнала: 2020, Номер 716, С. 137077 - 137077

Опубликована: Фев. 1, 2020

Язык: Английский

---

Performance Comparison of an LSTM-based Deep Learning Model versus Conventional Machine Learning Algorithms for Streamflow Forecasting

Water Resources Management, Год журнала: 2021, Номер 35(12), С. 4167 - 4187

Опубликована: Авг. 16, 2021

Язык: Английский

---

A Novel Ensemble Approach for Landslide Susceptibility Mapping (LSM) in Darjeeling and Kalimpong Districts, West Bengal, India

Remote Sensing, Год журнала: 2019, Номер 11(23), С. 2866 - 2866

Опубликована: Дек. 2, 2019

Landslides are among the most harmful natural hazards for human beings. This study aims to delineate landslide hazard zones in Darjeeling and Kalimpong districts of West Bengal, India using a novel ensemble approach combining weight-of-evidence (WofE) support vector machine (SVM) techniques with remote sensing datasets geographic information systems (GIS). The area currently faces severe problems, causing fatalities losses property. In present study, inventory database was prepared Google Earth imagery, field investigation carried out global positioning system (GPS). Of 326 landslides inventory, 98 (30%) were used validation, 228 (70%) modeling purposes. conditioning factors elevation, rainfall, slope, aspect, geomorphology, geology, soil texture, land use/land cover (LULC), normalized differential vegetation index (NDVI), topographic wetness (TWI), sediment transportation (STI), stream power (SPI), seismic zone maps as independent variables process. SVM ensembled prepare susceptibility (LSMs) help (RS) data geographical then classified into four classes; namely, low, medium, high, very high occurrence, breaks classification methods GIS environment. produced by these models an 630 km2 (WofE& RBF-SVM), 474 Linear-SVM), 501km2 Polynomial-SVM), 498 Sigmoid-SVM), respectively, total 3914 km2. results our validated receiver operating characteristic (ROC) curve quality sum (Qs) methods. under (AUC) values WofE& RBF-SVM, WofE & Linear-SVM, Polynomial-SVM, Sigmoid-SVM 87%, 90%, 88%, 85%, which indicates they good identifying zones. As per both validation methods, Linear-SVM model is more accurate than other models. obtained from this new can provide proper significant decision-makers policy planners landslide-prone areas districts.

Язык: Английский

---

Data‐driven flood emulation: Speeding up urban flood predictions by deep convolutional neural networks

Journal of Flood Risk Management, Год журнала: 2020, Номер 14(1)

Опубликована: Дек. 7, 2020

Abstract Computational complexity has been the bottleneck for applying physically based simulations in large urban areas with high spatial resolution efficient and systematic flooding analyses risk assessment. To overcome issue of long computational time accelerate prediction process, this paper proposes that maximum water depth can be considered an image‐to‐image translation problem which rasters are generated using information learned from data instead by conducting simulations. The proposed data‐driven pluvial flood approach is on a deep convolutional neural network trained simulation obtained three catchments 18 hyetographs. Multiple tests to assess accuracy validity were conducted both design real results show networks use only 0.5% compared models, promising generalizability. also potentially applied different but relevant problems, including analysis flood‐safe layout planning.

Язык: Английский

---

A review on applications of urban flood models in flood mitigation strategies

Natural Hazards, Год журнала: 2021, Номер 108(1), С. 31 - 62

Опубликована: Март 30, 2021

Язык: Английский

---

Assessment of long and short-term flood risk using the multi-criteria analysis model with the AHP-Entropy method in Poyang Lake basin

International Journal of Disaster Risk Reduction, Год журнала: 2022, Номер 75, С. 102968 - 102968

Опубликована: Апрель 17, 2022

Язык: Английский

---

From local to regional compound flood mapping with deep learning and data fusion techniques

The Science of The Total Environment, Год журнала: 2021, Номер 782, С. 146927 - 146927

Опубликована: Апрель 6, 2021

Язык: Английский

---