| 1. |
- Ghayur Sadigh, Armin, et al.
(author)
-
Comparison of optimized data-driven models for landslide susceptibility mapping
- 2023
-
In: Environment, Development and Sustainability. - 1387-585X .- 1573-2975.
-
Journal article (peer-reviewed)abstract
- Locations prone to landslides must be identified and mapped to prevent landslide-related damage and casualties. Machine learning approaches have proven effective for such tasks and have thus been widely applied. However, owing to the rapid development of data-driven approaches, deep learning methods that can exhibit enhanced prediction accuracies have not been fully evaluated. Several researchers have compared different methods without optimizing them, whereas others optimized a single method using different algorithms and compared them. In this study, the performances of different fully optimized methods for landslide susceptibility mapping within the landslide-prone Kermanshah province of Iran were compared. The models, i.e., convolutional neural networks (CNNs), deep neural networks (DNNs), and support vector machine (SVM) frameworks were developed using 14 conditioning factors and a landslide inventory containing 110 historical landslide points. The models were optimized to maximize the area under the receiver operating characteristic curve (AUC), while maintaining their stability. The results showed that the CNN (accuracy = 0.88, root mean square error (RMSE) = 0.37220, and AUC = 0.88) outperformed the DNN (accuracy = 0.79, RMSE = 0.40364, and AUC = 0.82) and SVM (accuracy = 0.80, RMSE = 0.42827, and AUC = 0.80) models using the same testing dataset. Moreover, the CNN model exhibiting the highest robustness among the three models, given its smallest AUC difference between the training and testing datasets. Notably, the dataset used in this study had a low spatial accuracy and limited sample points, and thus, the CNN approach can be considered useful for susceptibility assessment in other landslide-prone regions worldwide, particularly areas with poor data quality and quantity. The most important conditioning factors for all models were rainfall and the distances from roads and drainages.
|
|
| 2. |
- Lei, Xinxiang, et al.
(author)
-
Urban flood modeling using deep-learning approaches in Seoul, South Korea
- 2021
-
In: Journal of Hydrology. - : Elsevier BV. - 0022-1694 .- 1879-2707. ; 601
-
Journal article (peer-reviewed)abstract
- Identification of flood-prone sites in urban environments is necessary, but there is insufficient hydraulic information and time series data on surface runoff. To date, several attempts have been made to apply deep-learning models for flood hazard mapping in urban areas. This study evaluated the capability of convolutional neural network (NNETC) and recurrent neural network (NNETR) models for flood hazard mapping. A flood-inundation inventory (including 295 flooded sites) was used as the response variable and 10 flood-affecting factors were considered as the predictor variables. Flooded sites were then spatially randomly split in a 70:30 ratio for building flood models and for validation purposes. The prediction quality of the models was validated using the area under the receiver operating characteristic curve (AUC) and root mean square error (RMSE). The validation results indicated that prediction performance of the NNETC model (AUC = 84%, RMSE = 0.163) was slightly better than that of the NNETR model (AUC = 82%, RMSE = 0.186). Both models indicated that terrain ruggedness index was the most important predictor, followed by slope and elevation. Although the model output had a relative error of up to 20% (based on AUC), this modeling approach could still be used as a reliable and rapid tool to generate a flood hazard map for urban areas, provided that a flood inundation inventory is available.
|
|
| 3. |
- Panahi, Mahdi, et al.
(author)
-
A Country Wide Evaluation of Sweden's Spatial Flood Modeling With Optimized Convolutional Neural Network Algorithms
- 2023
-
In: Earth's Future. - : American Geophysical Union (AGU). - 2328-4277. ; 11:11
-
Journal article (peer-reviewed)abstract
- Flooding is one of the most serious and frequent natural hazards affecting human life, property, and the environment. This study develops and tests a deep learning approach for large-scale spatial flood modeling, using Convolutional Neural Network (CNN) and optimized versions combined with the Gray Wolf Optimizer (GWO) or the Imperialist Competitive Algorithm (ICA). With Sweden as an application case for nation-wide flood susceptibility mapping, this modeling approach considers ten geo-environmental input factors (slope, elevation, aspect, plan curvature, length of slope, topographic wetness index, distance from river, distance from wetland, rainfall, and land use). The GWO and ICA optimization improves model prediction by 12% and 8%, respectively, compared with the standalone CNN model performance. The results show 40% of the land area, 45% of the railroad, and 43% of the road network of Sweden to have high or very high flood susceptibility. They also show the aspect to have the highest input factor impact on flood susceptibility prediction while, for example, rainfall ranks only seven of the total 10 considered geo-environmental input factors. In general, accurate nation-wide flood susceptibility prediction is essential for guiding flood management and mitigation efforts. This study's approach to such prediction has emerged as well-performing and cost-effective for the case of Sweden, calling for further application and testing in other world regions.
|
|
| 4. |
- Panahi, Mahdi, et al.
(author)
-
Large-scale dynamic flood monitoring in an arid-zone floodplain using SAR data and hybrid machine-learning models
- 2022
-
In: Journal of Hydrology. - : Elsevier BV. - 0022-1694 .- 1879-2707. ; 611, s. 128001-
-
Journal article (peer-reviewed)abstract
- Although the growing number of synthetic aperture radar (SAR) satellites has increased their application in flood-extent mapping, predictive models for the analysis of flood dynamics that are independent of sensor characteristics must be developed to fully extract information from SAR images for flood mitigation. This study aimed to develop hybrid machine-learning models for flood mapping in the Ahvaz region, Iran, based on SAR data. Each hybrid model consists of a support vector machine (SVM) algorithm coupled with one of the following metaheuristic optimization procedures: grey wolf optimization (GWO), differential evolution, and the imperialist competitive algorithm. Sentinel-1 acquired SAR images before and during flooding between 20 March and 26 May of 2019. The goodness-of-fit level and predictive capability of each model were scrutinized using overall accuracy, producer accuracy, and user accuracy. The SVM-GWO approach yielded the highest accuracy with overall accuracies of 96.07% and 93.39% in the training and validation steps, respectively. Furthermore, this hybrid model provided the most accurate classification of water-inundation class based on producer accuracy (96.67%) and user accuracy (95.05%). The results highlight that wetland is the last land-use/land-cover type to return to normal conditions due to the many previously dry oxbow lakes that could trap water for a long time. Furthermore, the nine most suitable sites for flood-protection structures (e.g., embankments and levees) were identified based on floodwater distribution analysis. This work describes a robust, data-parsimonious approach that will benefit flood mitigation studies seeking to identify the most suitable locations for embankments based on spatio-temporal flood dynamics.
|
|
| 5. |
- Paryani, Sina, et al.
(author)
-
Hybrid-based approaches for the flood susceptibility prediction of Kermanshah province, Iran
- 2023
-
In: Natural Hazards. - : Springer Science and Business Media LLC. - 0921-030X .- 1573-0840. ; 116:1, s. 837-868
-
Journal article (peer-reviewed)abstract
- This study aims at optimizing the support vector regression (SVR) model using four metaheuristic methods, Harris hawks optimization (HHO), particle swarm optimization (PSO), gray wolf optimizer (GWO), and bat algorithm (BA). The intent is to create a reliable flood susceptibility map (FSM). In this regard, a flood inventory map for 617 flood locations was generated from the Google earth engine (GEE). Four hundred and thirty-two random locations (70%) were used for spatial flood susceptibility modeling, and 185 random locations (30%) were selected for testing hybrid approaches. Based on the available data and literature, the following eleven factors were selected: altitude, slope angle, slope aspect, plan curvature, stream power index (SPI), topographic wetness index (TWI), distance to river, lithology, drainage density, land use, and rainfall. The normalized frequency ratio (NFR) method was used to obtain a weight for each class of each factor. Next, flood susceptibility maps were produced by SVR-HHO, SVR-PSO, SVR-GWO, and SVR-BA hybrid models. The prediction power of hybrid models was assessed using various indicators of sensitivity, specificity, accuracy, kappa coefficient, receiver operating curve (ROC) diagram, mean square error (MSE), and root-mean-square error (RMSE). Validation results indicated the area under the curve (AUC) of 85.8%, 85.7%, 85.5%, and 84.6% for the SVR-HHO, SVR-GWO, SVR-BA, and SVR-PSO hybrid models, respectively. The results from testing phase reveal the best performance of the SVR-HHO model (RMSE = 0.401, MSE = 0.160, sensitivity = 0.822, specificity = 0.800, accuracy = 0.811, and kappa = 0.622). The SVR-PSO model had a poor performance (RMSE = 0.406, MSE = 0.164, sensitivity = 0.827, specificity = 0.773, accuracy = 0.80, and kappa = 0.60). It can be concluded that the map produced by SVR-HHO is a feasible approach for modeling flood susceptibility.
|
|
| 6. |
- Rahmati, Omid, et al.
(author)
-
Capability and robustness of novel hybridized models used for drought hazard modeling in southeast Queensland, Australia
- 2020
-
In: Science of the Total Environment. - : Elsevier BV. - 0048-9697 .- 1879-1026. ; 718
-
Journal article (peer-reviewed)abstract
- Widespread detrimental and long-lasting droughts are having catastrophic impacts around the globe. Researchers, organizations, and policy makers need to work together to obtain precise information, enabling timely and accurate decision making to mitigate drought impacts. In this study, a spatial modeling approach based on an adaptive neuro-fuzzy inference system (ANFIS) and several metaheuristic optimizations (ANFIS-BA, ANFIS-GA, ANFIS-ICA, ANFIS-PSO) was developed to predict the spatial occurrence of drought in a region in southeastern Queensland, Australia. In this approach, data describing the distribution of eight drought-contributing factors were prepared for input into the models to serve as independent variables. Relative departures of rainfall (RDR) and relative departures of soil moisture (RDSM) were analyzed to identify locations where drought conditions have occurred. The set of locations in the study area identified as having experienced drought conditions was randomly divided into two groups, 70% were used for training and 30% for validation. The models employed these data to generate maps that predict the locations that would be expected to experience drought. The prediction accuracy of the model-produced drought maps was scrutinized with two evaluation metrics: area under the receiver operating characteristic curve (AUC) and root mean square error (RMSE). The results demonstrate that the hybridized models (ANFIS-BA (AUC(mean) = 83.7%, RMSEmean = 0.236), ANFIS-GA (AUC(mean) = 81.62%, RMSEmean = 0.247), ANFIS-ICA (AUC(mean) = 82.12%, RMSEmean = 0.247), and ANFIS-PSO (AUC(mean) = 81.42%, RMSEmean = 0.255)) yield better predictive performance than the standalone ANFIS model (AUC(mean) = 71.8%, RMSEmean = 0.344). Furthermore, sensitivity analyses indicated that plantavailable water capacity, the percentage of soil comprised of sand, and mean annual precipitation were the most important predictors of drought hazard. The versatility of the new approach for spatial drought modeling and the capacity of ANFIS model hybridization to improve model performance suggests great potential to assist decision makers in their formulations of drought risk, recovery, and response management, and in the development of contingency plans.
|
|
| 7. |
- Rezaie, Fatemeh, et al.
(author)
-
Improving landslide susceptibility mapping using integration of ResU-Net technique and optimized machine learning algorithms
- 2023
-
In: Remote Sensing of Soil and Land Surface Processes: Monitoring, Mapping, and Modeling. - : Elsevier BV. ; , s. 419-438
-
Book chapter (other academic/artistic)abstract
- Landslides are the most common natural disasters in mountainous areas that follow major seismic events, volcanic activity, melting snow, or prolonged and intense rainfalls and cause severe disruptions to ecosystems, economies, and societies worldwide. Therefore, minimizing their negative effects through landslide-susceptibility assessment is essential. In this study, the standard support vector regression (SVR) integrated with the gray wolf optimizer (GWO) and particle swarm optimization (PSO) algorithms were used to map landslide-prone areas. The landslide inventory map was automatically generated using a pixel-based technique based on residual U-Net algorithm from the Sentinel-2 data. In total, 4900 landslide samples were identified and divided randomly into two groups, creating training (70%) and testing (30%) datasets. In addition, nine factors that affect landslides were selected to construct a model using each algorithm. Finally, the performance of the models (SVR, SVR-GWO, and SVR-PSO) were validated and compared using the area under the receiver operating characteristic curve. The findings showed that the hybrid SVR-GWO model performed better than the standard model and is recommended for landslide susceptibility assessment due to its accuracy and efficiency.
|
|
| 8. |
- Bravo, L, et al.
(author)
-
- 2021
-
swepub:Mat__t
|
|
| 9. |
- Tabiri, S, et al.
(author)
-
- 2021
-
swepub:Mat__t
|
|
