Flood assessment using machine learning and its implications for coastal spatial planning in Phu Yen Province, Vietnam

Spatial modeling of flood probability using geo-environmental variables and machine learning models, case study: Tajan watershed, Iran

Over the past few decades, urban expansion has accelerated worldwide. This process can increase future flood risks due to local changes in hydrological conditions and the increased exposure and vulnerability of communities in flood-prone areas. Therefore, assessing the impact of urban expansion on flood susceptibility is an important task that can support local authorities in urban planning and in mitigating flood impacts. The objective of this study was to assess the impact of urban expansion on flood susceptibility in Hanoi using machine learning models: Deep Neural Networks (DNN), Adaptive Boosting (ADB), Extreme Gradient Boosting (XGB), and Random Forest (RF). A total of 1058 flood points and 14 conditioning factors corresponding to 2014 and 2024 were used as input to the models. Statistical indices, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Area Under the Curve (AUC), and Coefficient of Determination (R2) were used to evaluate the performance of the proposed model. The results showed that the DNN model achieved the highest performance in assessing the impact of urban expansion on flood susceptibility (AUC=0.92), followed by XGB (0.91), ADB (0.86), and RF (0.82). During 2014–2024, urban expansion combined with the impacts of climate change has significantly increased the areas susceptible to flooding. In Hanoi, areas in the "high" and "very high" flood-susceptibility categories have been expanding continuously, accounting for about 25% of the total study area. In contrast, the "medium" group has a slight decreasing trend, while the "low" and "very low" areas have narrowed. This shows that urban expansion is increasing the area prone to flooding. The results of this study provide a solid scientific basis, supporting planners and policymakers in identifying limitations in current flood risk adaptation measures and in developing more appropriate spatial and temporal strategies to minimize flood impacts.

A new approach based on biology-inspired metaheuristic algorithms in combination with random forest to enhance the flood susceptibility mapping

The Mekong Basin is the most critical transboundary river basin in Asia. This basin provides an abundant source of fresh water essential for the development of agriculture, domestic consumption, and industry, as well as for the production of hydroelectricity, and it also contributes to ensuring food security worldwide. This region is often subject to floods that cause significant damage to human life, society, and the economy. However, flood risk management challenges in this region are increasingly substantial due to conflicting objectives between several countries and data sharing. This study integrates deep learning with optimization algorithms, namely Grasshopper Optimisation Algorithm (GOA), Adam and Stochastic Gradient Descent (SGD), and open-source datasets to identify the region of probably occurring floods in the Mekong basin, covering Vietnam and Cambodia. Various statistical indices, namely Area Under the Curve (AUC), root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²), were used to evaluate flood susceptibility models. The results show that the proposed models performed well with AUC values above 0.8, specifying that the DNN-Adam model achieved an AUC of 0.98, outperforming DNN-GOA (AUC = 0.89), DNN-SGD (AUC = 0.87), and XGB (AUC = 0.82. Regions with very high flood susceptibility are concentrated in the Mekong Delta of Vietnam and along the Mekong River in Cambodia. The findings of this study are significant in supporting decision-makers or planners in proposing appropriate flood mitigation strategies, planning policies, and strategies, particularly in the Mekong River watershed.

Flooding is the most dangerous of all environmental hazards and leads to significant impacts both on the environment and on human life. Environmental protection and water management can be achieved by modeling flood susceptibility, to inform strategy that will reduce flood damage. The aim of this study is to develop novel hybrid models based on a radial basis functions neural network (RBFNN), arithmetic optimization algorithm (AOA), artificial bee colony (ABC), and ant lion optimizer (ALO), to build flood susceptibility maps in the Quang Tri province of Vietnam. The obtained models were trained and validated using 1511 flood points and 14 conditioning factors. Various statistical indices were used to assess the performance of the models, namely root mean square error, receiver operation characteristics (ROC), area under the receiver operating characteristic curve (AUC), and the coefficient of determination (R2). The comparison analysis highlights that the RBFNN‐ABC model was better than other models with an AUC of 0.98, followed by RBFNN‐AOA, support vector machine, and random forest, all with an AUC value of 0.96, and finally RBFNN‐ALO, with an AUC of 0.95. From the flood susceptibility map, it can be seen that the eastern areas of the study area have high and very high flood susceptibility, that requires local government attention. The approach and results of this study can support national and local authorities, decision‐makers, and other planners in the construction of appropriate strategies to reduce potential damage in the future.

In this study, artificial neural network (ANN), frequency ratio (FR) and evidential belief function (EBF) methods were used to prepare the flood susceptibility map. For this purpose, the parameters of ten, slope, land curvature, topographic moisture index, distance from river and geology and type of lands in Haraz watershed in Mazandaran province were performed. Eleven conditioning factors including slope, land curvature, distance to river, river density, elevation, rainfall, stream power index (SPI), topographic wetness index (TWI), lithology, land use and normalized difference vegetation index (NDVI) were used in Haraz watershed in Mazandaran province. In addition, 201 floodplains were located in the area. The points were randomly divided into groups of 141 points (70%) and 60 points (30%) for training and validation, respectively. Furthermore, the probability of flooding for each class of each factor was calculated. Hence, the weights obtained for each class in the Geographic Information System (GIS) were applied in the respective layers, and the flood susceptibility maps of the study area were obtained. Based on the flood susceptibility map, the area was divided into 5 classes with very high, high, medium, low and very low sensitivity. These methods were evaluated by area under the curve (AUC) method. The results indicate that the lower and near elevation to river have a high probability and sensitivity to flooding. The results of the current study showed that the frequency ratio (AUC = 0.97) and evidential belief function (AUC = 0.94) and artificial neural network (AUC = 0.87) methods had the highest accuracy in predicting flood occurrence, respectively. The results suggest that these models can be useful and reliable in predicting flood risk potential, especially in different areas, including urban spaces, due to their high efficiency. Extended Abstract Introduction To prevent, control and control floods and Prevention of possible damages, areas with high flood potential first should be considered and foremost identified and then Identify the factors that produce and create floods. In this regard, the level of flood-prone and flood-prone areas in the country has increased and Many cities, villages, industrial and agricultural facilities and residential areas They are at risk of flooding. In the event of a flood There are many factors involved. Generally Climatic factors, regional factors and human factors play a role in creating floods. Climatic factors can be He pointed to Dry area, heavy rainfall and relatively short continuity. One of the most important factors in the field can be mentioned Geological condition, vegetation, basin area, basin shape and form, basin slope and focal point. Also human intervention in the natural water cycle via Destruction of vegetation in watersheds, Irregular land use, Development of impenetrable levels and the like Increased the likelihood of flooding in various areas. In Sail management, some of these factors are controllable and ‌In design, flood control They need more attention. Due to the increasing trend of floods in the country and the growing negative effects of its occurrence in the northern parts of the country, its necessary to reduce the risk of loss of life, property and environmental risk, Necessary measures should be considered. among the various watersheds in the north of the country, in this study, Haraz watershed has been selected as the study area That The reason for choosing it on the one hand It is located and adjacent to key cities in the north of the country, including The cities of Amol, Mahmoud Abad, Babol, Babolsar, Ghaemshahr, Sari, Pol-e Sefid, Shirgah, Neka, Behshahr, Galugah and Bandar-e-Gaz and also Hundreds of rural points and thousands of hectares of agricultural and garden lands and Part of the road along the Caspian Sea (Rasht to Gorgan) and Parts of the mountainous roads of Amol to Tehran and Ghaemshahr to Tehran in this basin and on the other hand There has been a growing flood in recent years in this geographical area that Numerous social, economic and environmental damages and challenges. So these are the reasons The preparation of a flood susceptibility map in the Haraz watershed makes it even more necessary. according to the above and Description of flood hazards in the northern regions of the country, the questions in this study are: What are the most dangerous parts of Haraz watershed in terms of flood sensitivity? Efficiency of which of the artificial neural network models, Frequency ratio and Is the function of definitive evidence more to prepare a flood susceptibility map in Haraz watershed? Methodology Current research in terms of purpose Is a type of applied research and done by quantitative method. According to the objectives of the research, the required data Has been collected from the relevant organizations and organs (Regional Water Company, Natural Resources Department, etc.) and to analyze this data Used the ArcGIS software. Overall, the research process is as follows First Prepared List of past floods in the study area and so on has been identified Effective parameters in flood occurrence and using Three models of definite evidence function (EBF), frequency ratio (FR) and artificial neural network (ANN), A flood sensitization map of Haraz watershed has been prepared. The following is a review Model Validation Using the ROC curve. Results and discussion The weights obtained in each method, for each class of each factor Applied in Geographic Information System (GIS) and Flood susceptibility maps were prepared for Haraz watershed. Flood susceptibility maps Launched in ArcGIS10.3 software environment in five classes, the sensitivity is very low, low, medium, high and very high. In order to assess the accuracy of the flood prediction map, 60 flood events were used (Experimental data) Related to previous courses and These events have not been entered to predict flood potential in probabilistic models. Given that the area below the curve for the model, the frequency ratio is 0.97 So this model is more efficient Definitive Evidence for Model Function Models (0.93) and The neural network is artificial (0.78). Conclusion The present study is done with the aim of preparing a map of the possibility of floods in the watershed of Haraz and Evaluate the efficiency of frequency-ratio models, the function of definitive evidence, and the artificial neural network in the preparation of flood susceptibility maps. To do this, 201 flood points were recorded and 141 Flood situation for modeling and 60 positions were set aside for model validation. To prepare these maps, the first step is to prepare the factors that affect the occurrence of floods. The findings of this study indicate the accuracy of the probability frequency ratio model in identifying areas with flood susceptibility in Haraz watershed in Mazandaran province. Therefore, the use of probability frequency model It is useful and reliable in assessing the risk of flooding. But since The accuracy of predicting models of definite evidence and artificial neural networks is also acceptable. These methods can also be used, but in general, the frequency ratio has a higher accuracy in predicting flood areas. In the maps produced, Parts with low and low elevation classes Exit area, they have the highest amount of tracking. generally, Areas with low elevation and low slope, they are most likely to be flooded. The predictive results also showed that Slope parameters, height, land curvature, lithology, land type, river distance, river carrying capacity and topographic moisture index are influential on Potential flooding potential and using them is useful in probabilistic models, flood potential assessment. Flood formation mechanism and landslide flooding in the form of spatial analysis, it can be extended to other parts of the watershed. The approach presented in this research in fact, some variables affecting the occurrence of floods have been used Which are very important in the flood risk prediction map in the study area which can be used using the results of these maps, He took appropriate management measures to reduce the damage and casualties caused by the floods. To be careful in predicting flood occurrence It is necessary to use other machine learning models or a combination of these models Which will increase the accuracy of the flood prediction. The above findings, in addition to having practical and operational aspects for management devices and institutions in particular, the Crisis Management Headquarters of the northern provinces of the country, can be used as a suitable template, By researchers and those interested in flood urban crisis management planning. Prepare a hybrid susceptibility map for multiple hazards (Flood, earthquake, drought, etc.) Using hybrid models for the study area and other watersheds of the country Especially in areas with high urban population density. Recommended as a basis for future studies.

The uncertainty of flash flood makes them highly difficult to predict through conventional models. The physical hydrologic models of flash flood prediction of any large area is very difficult to compute as it requires lot of data and time. Therefore remote sensing data based models (from statistical to machine learning) have become highly popular due to open data access and lesser prediction times. There is a continuous effort to improve the prediction accuracy of these models through introducing new methods. This study is focused on flash flood modeling through novel hybrid machine learning models, which can improve the prediction accuracy. The hybrid machine learning ensemble approaches that combine the three meta-classifiers (Real AdaBoost, Random Subspace, and MultiBoosting) with J48 (a tree-based algorithm that can be used to evaluate the behavior of the attribute vector for any defined number of instances) were used in the Gorganroud River Basin of Iran to assess flood susceptibility (FS). A total of 426 flood positions as dependent variables and a total of 14 flood conditioning factors (FCFs) as independent variables were used to model the FS. Several threshold-dependent and independent statistical tests were applied to verify the performance and predictive capability of these machine learning models, such as the receiver operating characteristic (ROC) curve of the success rate curve (SRC) and prediction rate curve (PRC), efficiency (E), root-mean square-error (RMSE), and true skill statistics (TSS). The valuation of the FCFs was done using AdaBoost, frequency ratio (FR), and Boosted Regression Tree (BRT) models. In the flooding of the study area, altitude, land use/land cover (LU/LC), distance to stream, normalized differential vegetation index (NDVI), and rainfall played important roles. The Random Subspace J48 (RSJ48) ensemble method with an area under the curve (AUC) of 0.931 (SRC), 0.951 (PRC), E of 0.89, sensitivity of 0.87, and TSS of 0.78, has become the most effective ensemble in predicting the FS. The FR technique also showed good performance and reliability for all models. Map removal sensitivity analysis (MRSA) revealed that the FS maps have the highest sensitivity to elevation. Based on the findings of the validation methods, the FS maps prepared using the machine learning ensemble techniques have high robustness and can be used to advise flood management initiatives in flood-prone areas.

Abstract. In this paper, Radial Basis Function (RBF) Neural Network and Logistic Regression (LR) models were proposed for hazard prediction of landslides in a part of the Semirom area (Iran) to compare their accuracy and performance. For this purpose, a spatial database of the study area was prepared that consists of 68 landslide locations and 11 influencing information layers including slope, aspect, profile curvature, plan curvature, distance from faults, distance from roads, distance from residential regions, distance from rivers, land use, lithology and rainfall. Landslide hazard maps were prepared for the study area by applying the proposed algorithms. Performance of the models was assessed using the Receiver Operating Characteristic (ROC) curve and area under the ROC curve (AUC). The coefficient of determination (R2), the root mean square error (RMSE), and the Normal Root Mean Square Error (NRMSE) were calculated for proposed methods. The outcomes showed that the RBF Neural Network has the highest R2 (0.8224), in comparison to that of the LR model (0.5365). Also, the ROC plots, RMSEs and NRMSEs showed that the proposed RBF Neural Network is much better than the LR model. Consequently, it can be concluded that the RBF Neural Network is the best regression model in this study and it can be considered as a capable method for landslide hazard mapping in landslide-susceptible areas.

Flood damage is becoming increasingly severe in the context of climate change and changes in land use. Assessing the effects of these changes on floods is important, to help decision-makers and local authorities understand the causes of worsening floods and propose appropriate measures. The objective of this study was to evaluate the effects of climate and land use change on flood susceptibility in Thua Thien Hue province, Vietnam, using machine learning techniques (support vector machine (SVM) and random forest (RF)) and remote sensing. The machine learning models used a flood inventory including 1,864 flood locations and 11 conditional factors in 2017 and 2021, as the input data. The predictive capacity of the proposed models was assessed using the area under the curve (AUC), the root mean square error (RMSE), and the mean absolute error (MAE). Both proposed models were successful, with AUC values exceeding 0.95 in predicting the effects of climate and land use change on flood susceptibility. The RF model, with AUC = 0.98, outperformed the SVM model (AUC = 0.97). The areas most susceptible to flooding increased between 2017 and 2021 due to increased built-up area.

The most dangerous landslide disasters always cause serious economic losses and human deaths. The contribution of this work is to present an integrated landslide modelling framework, in which an adaptive neuro-fuzzy inference system (ANFIS) is combined with the two optimization algorithms of whale optimization algorithm (WOA) and grey wolf optimizer (GWO) at Anyuan County, China. It means that WOA and GWO are used as two meta-heuristic algorithms to improve the prediction performance of the ANFIS-based methods. In addition, the step-wise weight assessment ratio analysis (SWARA) method is used to obtain the initial weight of each class of landslide influencing factors. To validate the effectiveness of the proposed framework, 315 landslide events in history were selected for our experiments and were randomly divided into the training and verification sets. To perform landslide susceptibility mapping, fifteen geological, hydrological, geomorphological, land cover, and other factors are considered for the modelling construction. The landslide susceptibility maps by SWARA, SWARA-ANFIS, SWARA-ANFIS-PSO, SWARA-ANFIS-WOA, and SWARA-ANFIS-GWO models are assessed using the measures of the receiver operating characteristic (ROC) curve and root-mean-square error (RMSE). The experiments demonstrated that the obtained results of modelling process from the SWARA to the SAWRA-ANFIS-GWO model were more accurate and that the proposed methods have satisfactory prediction ability. Specifically, prediction accuracy by area under the curve (AUC) of SWARA, SWARA-ANFIS, SWARA-ANFIS-PSO, SWARA-ANFIS-GWO, and SWARA-ANFIS-WOA models were 0.831, 0.831, 0.850, 0.856, and 0.869, respectively. Due to adaptability and usability, the proposed prediction methods can be applied to other areas for landslide management and mitigation as well as prevention throughout the world.

Flooding is a very common worldwide natural hazard causing large-scale casualties every year; Iran is not immune to this thread as well. Comprehensive flood susceptibility mapping is very important to reduce losses of lives and properties. Thus, the aim of this study is to map susceptibility to flooding by different bivariate statistical methods including Shannon's entropy (SE), statistical index (SI), and weighting factor (Wf). In this regard, model performance evaluation is also carried out in Haraz Watershed, Mazandaran Province, Iran. In the first step, 211 flood locations were identified by the documentary sources and field inventories, of which 70% (151 positions) were used for flood susceptibility modeling and 30% (60 positions) for evaluation and verification of the model. In the second step, ten influential factors in flooding were chosen, namely slope angle, plan curvature, altitude, topographic wetness index (TWI), stream power index (SPI), distance from river, rainfall, geology, land use, and normalized difference vegetation index (NDVI). In the next step, flood susceptibility maps were prepared by these four methods in ArcGIS. As the last step, receiver operating characteristic (ROC) curve was drawn and the area under the curve (AUC) was calculated for quantitative assessment of each model. The results showed that the best model to estimate the susceptibility to flooding in Haraz Watershed was SI model with the prediction and success rates of 99.71 and 98.72%, respectively, followed by Wf and SE models with the AUC values of 98.1 and 96.57% for the success rate, and 97.6 and 92.42% for the prediction rate, respectively. In the SI and Wf models, the highest and lowest important parameters were the distance from river and geology. Flood susceptibility maps are informative for managers and decision makers in Haraz Watershed in order to contemplate measures to reduce human and financial losses.

Energy expenditure constitutes a significant portion of total operational costs in greenhouse crop production. Developing accurate energy consumption prediction models presents crucial theoretical foundations for optimizing the environmental control strategies aimed at energy efficiency enhancement. This study focuses on steel-frame solar greenhouses without back slopes in Xinjiang’s Tianshan North Slope region. A physical model was established using thermodynamic equilibrium analysis, elucidating the energy exchange mechanisms between internal and external environments. Key parameters, including outdoor temperature and solar radiation, were identified as primary input variables through systematic energy flow characterization. Building upon this theoretical framework, we developed an enhanced prediction model (WOA-ELM) by integrating the Whale Optimization Algorithm (WOA) with an Extreme Learning Machine (ELM). The WOA’s global optimization capabilities were employed to refine the connection weights between input-hidden layers and optimize hidden neuron thresholds. Comparative evaluations against conventional artificial neural networks (ANNs), radial basis function neural networks (RBFNN), and baseline ELM models were conducted under diverse meteorological conditions. Experimental results demonstrate the superior performance of WOA-ELM across multiple metrics. Under overcast conditions, the model achieved a root mean square error (RMSE) of 0.423, coefficient of determination (R2) of 0.93, and mean absolute error (MAE) of 0.252. In clear weather scenarios, performance further improved with RMSE = 0.27, R2 = 0.96, and MAE = 0.063. The comprehensive evaluation ranked model effectiveness as WOA-ELM > ELM > BP > RBF. These findings substantiate that the hybrid WOA-ELM architecture, combining physical mechanism interpretation with intelligent parameter optimization, delivers enhanced prediction accuracy across varying weather patterns. This research provides valuable insights for energy load management in backslope-less steel-frame greenhouses, offering theoretical guidance for thermal environment regulation and sustainable operation.

Accurate estimation of reference evapotranspiration (ETo) plays a vital role in irrigation and water resource planning. The Penman–Monteith method recommended by the Food and Agriculture Organization (FAO PM56) is widely used and considered a standard to calculate ETo. However, FAO PM56 cannot be used with limited meteorological variables, so it is compulsory to choose an alternative model for ETo estimation, which requires fewer variables. This study built ten machine learning (ML) models based on multi-function, neural network, and tree-based structure against the FAO PM56 method. For this purpose, time series temperature data on a monthly scale are only used to train ML models. The developed ML models were applied to estimate ETo at different test stations and the obtained results were compared with the FAO PM56 method to verify and validate their performance in ETo estimation for the selected stations. In addition, multiple statistical indicators, including root-mean-square error (RMSE), coefficient of determination (R2), mean absolute error (MAE), Nash–Sutcliffe efficiency (NSE), and correlation coefficient (r) were calculated to compare the performance of each ML model on ETo estimation. Among the applied ML models, the ETo tree boost (TB) ML model outperformed the other ML models in estimating ETo in diverse climatic conditions based on statistical indicators (R2, NSE, r, RMSE, and MAE). Moreover, the observed R2, NSE, and r were the highest for the TB ML model, while RMSE and MAE were found to be the lowest at the study sites compared to other applied ML models. Lastly, ETo point data yielded from the TB ML model was used in an interpolation process to create monthly and annual ETo maps. Based on the ETo maps, this study suggests mainly a focus on areas with high ETo values and proper irrigation scheduling of crops to ensure water sustainability.

Machine learning models for predicting rock fracture toughness at different temperature conditions

In this study, a radial basis function (RBF) neural network with three-layer feed forward architecture was developed to effectively predict the viscosity ratio of different ethylene glycol/water based nanofluids. A total of 216 experimental data involving CuO, TiO2, SiO2, and SiC nanoparticles were collected from the published literature to train and test the RBF neural network. The parameters including temperature, nanoparticle properties (size, volume fraction, and density), and viscosity of the base fluid were selected as the input variables of the RBF neural network. The investigations demonstrated that the viscosity ratio predicted by the RBF neural network agreed well with the experimental data. The root mean squared error (RMSE), mean absolute percentage error (MAPE), sum of squared error (SSE), and statistical coefficient of multiple determination (R2) were respectively 0.04615, 2.12738%, 0.46007, and 0.99925 for the total samples when the Spread was 0.3. In addition, the RBF neural network had a better ability for predicting the viscosity ratio of nanofluids than the typical Batchelor model and Chen model, and the prediction performance of RBF neural networks were affected by the size of the data set.