Flood susceptibility modelling using advanced ensemble machine learning models

Geoscience Frontiers, Journal Year: 2020, Volume and Issue: 12(3), P. 101075 - 101075

Published: Oct. 5, 2020

Floods are one of nature's most destructive disasters because the immense damage to land, buildings, and human fatalities. It is difficult forecast areas that vulnerable flash flooding due dynamic complex nature floods. Therefore, earlier identification flood susceptible sites can be performed using advanced machine learning models for managing disasters. In this study, we applied assessed two new hybrid ensemble models, namely Dagging Random Subspace (RS) coupled with Artificial Neural Network (ANN), Forest (RF), Support Vector Machine (SVM) which other three state-of-the-art modelling susceptibility maps at Teesta River basin, northern region Bangladesh. The application these includes twelve influencing factors 413 current former points, were transferred in a GIS environment. information gain ratio, multicollinearity diagnostics tests employed determine association between occurrences influential factors. For validation comparison ability predict statistical appraisal measures such as Freidman, Wilcoxon signed-rank, t-paired Receiver Operating Characteristic Curve (ROC) employed. value Area Under (AUC) ROC was above 0.80 all models. modelling, model performs superior, followed by RF, ANN, SVM, RS, then several benchmark approach solution-oriented outcomes outlined paper will assist state local authorities well policy makers reducing flood-related threats also implementation effective mitigation strategies mitigate future damage.

Language: Английский

---

Short-term water quality variable prediction using a hybrid CNN–LSTM deep learning model

Stochastic Environmental Research and Risk Assessment, Journal Year: 2020, Volume and Issue: 34(2), P. 415 - 433

Published: Feb. 1, 2020

Language: Английский

---

A comprehensive review of deep learning applications in hydrology and water resources

Water Science & Technology, Journal Year: 2020, Volume and Issue: 82(12), P. 2635 - 2670

Published: Aug. 5, 2020

Abstract The global volume of digital data is expected to reach 175 zettabytes by 2025. volume, variety and velocity water-related are increasing due large-scale sensor networks increased attention topics such as disaster response, water resources management, climate change. Combined with the growing availability computational popularity deep learning, these transformed into actionable practical knowledge, revolutionizing industry. In this article, a systematic review literature conducted identify existing research that incorporates learning methods in sector, regard monitoring, governance communication resources. study provides comprehensive state-of-the-art approaches used industry for generation, prediction, enhancement, classification tasks, serves guide how utilize available future challenges. Key issues challenges application techniques domain discussed, including ethics technologies decision-making management governance. Finally, we provide recommendations directions models hydrology

Language: Английский

---

Evaluation of deep learning algorithms for national scale landslide susceptibility mapping of Iran

Geoscience Frontiers, Journal Year: 2020, Volume and Issue: 12(2), P. 505 - 519

Published: Aug. 7, 2020

The identification of landslide-prone areas is an essential step in landslide hazard assessment and mitigation landslide-related losses. In this study, we applied two novel deep learning algorithms, the recurrent neural network (RNN) convolutional (CNN), for national-scale susceptibility mapping Iran. We prepared a dataset comprising 4069 historical locations 11 conditioning factors (altitude, slope degree, profile curvature, distance to river, aspect, plan road, fault, rainfall, geology land-sue) construct geospatial database divided data into training testing dataset. then developed RNN CNN algorithms generate maps Iran using calculated receiver operating characteristic (ROC) curve used area under (AUC) quantitative evaluation Better performance both phases was provided by algorithm (AUC ​= ​0.88) than ​0.85). Finally, each province found that 6% 14% land very highly susceptible future events, respectively, with highest Chaharmahal Bakhtiari Province (33.8%). About 31% cities are located high susceptibility. results present study will be useful development strategies.

Language: Английский

---

A spatially explicit deep learning neural network model for the prediction of landslide susceptibility

CATENA, Journal Year: 2020, Volume and Issue: 188, P. 104451 - 104451

Published: Jan. 8, 2020

Language: Английский

---

GIS-based comparative assessment of flood susceptibility mapping using hybrid multi-criteria decision-making approach, naïve Bayes tree, bivariate statistics and logistic regression: A case of Topľa basin, Slovakia

Ecological Indicators, Journal Year: 2020, Volume and Issue: 117, P. 106620 - 106620

Published: June 21, 2020

Flood is a devastating natural hazard that may cause damage to the environment infrastructure, and society. Hence, identifying susceptible areas flood an important task for every country prevent such dangerous consequences. The present study developed framework flood-prone of Topľa river basin, Slovakia using geographic information system (GIS), multi-criteria decision making approach (MCDMA), bivariate statistics (Frequency Ratio (FR), Statistical Index (SI)) machine learning (Naïve Bayes Tree (NBT), Logistic Regression (LR)). To reach goal, different physical-geographical factors (criteria) were integrated mapped. access relationship interdependences among criteria, decision-making trial evaluation laboratory (DEMATEL) analytic network process (ANP) used. Based on experts' decisions, DEMATEL-ANP model was used compute relative weights criteria GIS-based linear combination performed derive susceptibility index. Separately, index computation through NBT-FR NBT-SI hybrid models assumed, in first stage, estimation weight each class/category conditioning factor SI FR integration these values NBT algorithm. application LR stand-alone required calculation by analysing their spatial relation with location historical events. revealed very high classes covered between 20% 47% area, respectively. validation results, past points, highlighted most performant Area Under ROC curve higher than 0.97, accuracy 0.922 value HSS 0.844. presented methodological identification can serve as alternative updating preliminary risk assessment based EU Floods Directive.

Language: Английский

---

Flood susceptibility mapping using convolutional neural network frameworks

Journal of Hydrology, Journal Year: 2019, Volume and Issue: 582, P. 124482 - 124482

Published: Dec. 18, 2019

Language: Английский

---

Prediction of groundwater quality using efficient machine learning technique

Chemosphere, Journal Year: 2021, Volume and Issue: 276, P. 130265 - 130265

Published: March 18, 2021

Language: Английский

---

Flash-Flood Susceptibility Assessment Using Multi-Criteria Decision Making and Machine Learning Supported by Remote Sensing and GIS Techniques

Remote Sensing, Journal Year: 2019, Volume and Issue: 12(1), P. 106 - 106

Published: Dec. 27, 2019

Concerning the significant increase in negative effects of flash-floods worldwide, main goal this research is to evaluate power Analytical Hierarchy Process (AHP), fi (kNN), K-Star (KS) algorithms and their ensembles flash-flood susceptibility mapping. To train two stand-alone models ensembles, for first stage, areas affected past by torrential phenomena are identified using remote sensing techniques. Approximately 70% these used as a training data set along with 10 predictors. It should be remarked that techniques play crucial role obtaining eight out conditioning factors. The predictive capability predictors evaluated through Information Gain Ratio (IGR) method. As expected, slope angle results factor highest capability. application AHP model implies construction ten pair-wise comparison matrices calculating normalized weights each predictor. computed input kNN–AHP KS–AHP ensemble Flash-Flood Potential Index (FFPI). FFPI also determined kNN KS models. performance statistical metrics (i.e., sensitivity, specificity accuracy) while validation done constructing Receiver Operating Characteristics (ROC) Curve Area Under (AUC) values density pixels within classes. Overall, best obtained model.

Language: Английский

---

Predicting flood susceptibility using LSTM neural networks

Journal of Hydrology, Journal Year: 2020, Volume and Issue: 594, P. 125734 - 125734

Published: Nov. 8, 2020

Identifying floods and producing flood susceptibility maps are crucial steps for decision-makers to prevent manage disasters. Plenty of studies have used machine learning models produce reliable maps. Nevertheless, most research ignores the importance developing appropriate feature engineering methods. In this study, we propose a local spatial sequential long short-term memory neural network (LSS-LSTM) prediction in Shangyou County, China. The three main contributions study summarized below. First all, it is new perspective use deep technique LSTM prediction. Second, integrate an method with predict susceptibility. Third, implement two optimization techniques data augmentation batch normalization further improve performance proposed method. LSS-LSTM can not only capture attribution information conditioning factors data, but also has powerful modelling capabilities deal relationship floods. experimental results demonstrate that achieves satisfactory (93.75% 0.965) terms accuracy area under receiver operating characteristic (ROC) curve.

Language: Английский

---