Search

In the context of early disease detection, machine learning (ML) has emerged as a vital tool. Feature selection (FS) algorithms play a crucial role in ensuring the accuracy of predictive models by identifying the most influential variables. This study, focusing on a retrospective cohort of 4778 COVID-19 patients from Iran, explores the performance of various FS methods, including filter, embedded, and hybrid approaches, in predicting mortality outcomes. The researchers leveraged 115 routine clinical, laboratory, and demographic features and employed 13 ML models to assess the effectiveness of these FS methods based on classification accuracy, predictive accuracy, and statistical tests. The results indicate that a Hybrid Boruta-VI model combined with the Random Forest algorithm demonstrated superior performance, achieving an accuracy of 0.89, an F1 score of 0.76, and an AUC value of 0.95 on test data. Key variables identified as important predictors of adverse outcomes include age, oxygen saturation levels, albumin levels, neutrophil counts, platelet levels, and markers of kidney function. These findings highlight the potential of advanced FS techniques and ML models in enhancing early disease detection and informing clinical decision-making. [ABSTRACT FROM AUTHOR]

This study examines the effectiveness of various quantile regression (QR) and machine learning (ML) methodologies developed for analyzing the relationship between meteorological parameters and daily reference evapotranspiration (ETref) across diverse climates in Iran spanning from 1987 to 2022. The analyzed models include D-vine copula-based quantile regression (DVQR), multivariate linear quantile regression (MLQR), Bayesian model averaging quantile regression (BMAQR), as well as machine learning algorithms such as extreme learning machine (ELM), random forest (RF), M5 model Tree (M5Tree), least squares support vector regression algorithm (LSSVR), and extreme gradient boosting (XGBoost). Additionally, empirical equations (EEs) such as Baier and Robertson (BARO), Jensen and Haise (JEHA), and Penman (PENM) models were considered. While the EEs demonstrated acceptable performance, the QR and ML models exhibited superior accuracy. Among these, the MLQR model displayed the highest accuracy compared to DVQR and BMAQR models. Moreover, LSSVR, XGBoost, and M5Tree models outperformed ELM and RF models. Notably, LSSVR, XGBoost, and MLQR models exhibited comparable performance (R2 and NSE > 0.92, MBE and RMSE < 0.5, and SI > 0.05) to M5Tree and BMAQR models across all climates. Importantly, these models significantly outperformed EEs, DVQR, ELM, and RF models in all climates. In conclusion, high-dimensional QR and ML models are recommended as promising alternatives for accurately estimating daily ETref in diverse global climate conditions. [ABSTRACT FROM AUTHOR]

Tree decline is a highly complex process and is inherently a function of manifold climatic, physiologic, and anthropogenic factors. Monitoring decline processes and their underlying dynamics primarily entails identifying their location and intensity across different ecosystems, for which airborne and satellite remote sensing approaches offer cost-effective and spatially explicit alternatives to field methods. Consumer-grade unmanned aerial vehicles (UAVs) can barely be used as standalone means for large-area monitoring due to their constrains in spatial and spectral domains. However, they could effectively be integrated alongside satellite data to unlock their information for subsequent upscaling on landscape level. We designed a novel two-step workflow to describe the severity of tree decline by linking UAV-RGB information to space-borne multispectral and digital elevation model (DEM) data over 15 forest sites dominated by Persian oak across the latitudinal gradient of Zagros Forests in western Iran. We display how to 1) leverage UAV as reference data across multiple structurally different Persian oak-dominated sites in semi-arid Zagros mountains of Iran; 2) link UAV, Copernicus DEM, and Sentinel-2 data to retrieve decline information within a model-driven context; and 3) analyze the sensitivity of models by means of a global variance-based sensitivity analysis. Results suggested a high association between UAV and field data on the intensity of decline, which enabled using sampled UAV data as reference to estimate the decline severity using space-borne data by means of semi-parametric generalized additive model (GAM) and non-parametric random forest (RF) approaches. Conclusively, this study provided a baseline for multi-scale analysis of tree decline using budget and partially free data sources, which can be of high scientific and practical assets for monitoring in remote, sparse, mountainous, and continuously degrading forest areas. [ABSTRACT FROM AUTHOR]

Floods in Iran cause a lot of damage in different places every year. The 2019 floods of the Gorgan and Atrak rivers basins in the north of Iran were one of the most destructive events in this country. Therefore, investigating the flood hazard of these areas is very necessary to manage probable future floods. For this purpose, in the present study, the capability of Random Forest (RF) and Support Vector Machine (SVM) algorithms was investigated in combination with Sentinel series and Landsat-8 images to prepare the 2019 flood map. Then, the flood hazard map of these areas was prepared using the new hybrid Fuzzy Best Worse Model-Weighted Multi-Criteria Analysis (FBWM-WMCA) model. According to the results of the FBWM-WMCA model, 38.58%, 50.18%, 11.10%, and 0.14% of the Gorgan river basin and 45.11%, 49.96%, 4.17%, and 0.076% of the Atrak river basin are in high, medium, low, and no hazards, respectively. The highest flood hazard areas in Gorgan and Atrak rivers basins in the north, northwest, west, and east, and south and southwest are mostly at medium flood hazard. Also, the results of RF and SVM algorithms with an overall accuracy of more than 85% for Sentinel-1, Sentinel-2, and Landsat-8 images and 80% for Sentinel-3 images indicate that the flooding is related to the western, southwestern, and northern regions including agricultural, bare lands and built up. According to the obtained results and the efficiency of the FBWM-WMCA model, the Gorgan and Atrak rivers basins need proper planning for flood hazard management. [ABSTRACT FROM AUTHOR]

In recent decades, fires in natural ecosystems, particularly forests and rangelands, have emerged as a significant threat. To address this challenge, our study aims to identify and prioritize forest fire-prone areas while highlighting key environmental and anthropogenic factors contributing to forest fires in Iran's Firouzabad region, Fars province. We compiled a forest fire incident map using data from the Data Center of the Natural Resources Department in Fars province, cross-referenced with field surveys. We examined 80 forest fire sites, randomly divided into a "training dataset" (70%) and a "validation dataset" (30%). We created "Forest Fire Susceptibility" (FFS) maps using GIS-based Bayesian and Random Forest (RF) methodologies, incorporating twelve unique environmental and human-induced variables. The performance of these methodologies was evaluated using the "Area Under the Curve-AUC." RF outperformed the Bayesian model with AUC scores of 0.876 and 0.807, respectively. The RF model identified 37.86% of the area as having a high fire risk, compared to the Bayesian model's estimate of 48.46%. Key factors influencing fire occurrences included elevation, mean annual precipitation, distance to roads, and mean annual temperature. Conversely, variables such as slope direction, topographic wetness index, and slope percent had a lesser impact. Given the presence of at-risk flora and fauna species in the area, our findings provide essential tools for pinpointing high fire susceptibility zones, aiding regional authorities in implementing preventive measures to mitigate fire hazards in forest ecosystems. In conclusion, our methodologies allow for the rapid creation of contemporary fire susceptibility maps based on fresh data. [ABSTRACT FROM AUTHOR]

Reducing emissions from deforestation and forest degradation (REDD) is a specific strategy for combating deforestation and forest degradation to alleviate the effects of climate change. In this study, the potential greenhouse gas (GHG) emission reduction resulting from the implementation of a REDD project is estimated. Changes in forest cover throughout the years 1985, 1990, 1995, 2000, 2010, 2015, and 2020 were analyzed using time-series Landsat imagery (TM, ETM + , and OLI) and a random forest algorithm. Multilayer perceptron neural networks were used to model the transition potential of the forest cover, which were then predicted via Markov chain analysis. The change detection analysis revealed two discernible patterns in forest cover dynamics. Between 1985 and 2000, a notable decrease in forest cover was seen, whereas from 2000 to 2020, it significantly increased. The results suggested that the absence of REDD implementation would result in the deforestation of approximately 199,569 hectares of forest cover between 2020 and 2050, leading to the release of 1,995,695 tCO2e of emissions into the atmosphere. However, with the implementation of REDD, these emissions would be reduced to 405,512 tCO2e, effectively preventing the release of 1,590,183 tCO2e of emissions into the upper atmosphere. This study demonstrates that the implementation of REDD projects can be an effective strategy for reducing GHG emissions and mitigating climate change in the Hyrcanian forests. [ABSTRACT FROM AUTHOR]

A machine learning-based approach is applied to simulate and forecast forest fires in the Golestan province in Iran. A dataset for no-fire, medium confidence (MC) fire events, and high confidence (HC) fire events is constructed from MODIS-MOD14A2. Nine climate variables from NASA's FLDAS are used as input variables, and 12 dates and 915 study points are considered. Three machine learning ensemble multi-label classifiers, gradient boosting (GBC), random forest (RFC), and extremely randomized tree (ETC), are used for forest fire simulation for the period 2000 to 2021, and ETC is found to be the most accurate classifier. Future fire projection for the near-future period of 2030 to 2050 is carried out with the ETC model, using CMIP6 EC-Earth3-SSP245 General Circulation Model (GCM) data. It is projected that MC forest fire occurrences will decrease, while HC forest fire occurrences will increase, and that the summer months, especially September, will be the most affected by fire. [ABSTRACT FROM AUTHOR]

This study addresses the challenges associated with emergency department (ED) overcrowding and emphasizes the need for efficient risk stratification tools to identify high-risk patients for early intervention. While several scoring systems, often based on logistic regression (LR) models, have been proposed to indicate patient illness severity, this study aims to compare the predictive performance of ensemble learning (EL) models with LR for in-hospital mortality in the ED. A cross-sectional single-center study was conducted at the ED of Imam Reza Hospital in northeast Iran from March 2016 to March 2017. The study included adult patients with one to three levels of emergency severity index. EL models using Bagging, AdaBoost, random forests (RF), Stacking and extreme gradient boosting (XGB) algorithms, along with an LR model, were constructed. The training and validation visits from the ED were randomly divided into 80% and 20%, respectively. After training the proposed models using tenfold cross-validation, their predictive performance was evaluated. Model performance was compared using the Brier score (BS), The area under the receiver operating characteristics curve (AUROC), The area and precision–recall curve (AUCPR), Hosmer–Lemeshow (H–L) goodness-of-fit test, precision, sensitivity, accuracy, F1-score, and Matthews correlation coefficient (MCC). The study included 2025 unique patients admitted to the hospital's ED, with a total percentage of hospital deaths at approximately 19%. In the training group and the validation group, 274 of 1476 (18.6%) and 152 of 728 (20.8%) patients died during hospitalization, respectively. According to the evaluation of the presented framework, EL models, particularly Bagging, predicted in-hospital mortality with the highest AUROC (0.839, CI (0.802–0.875)) and AUCPR = 0.64 comparable in terms of discrimination power with LR (AUROC (0.826, CI (0.787–0.864)) and AUCPR = 0.61). XGB achieved the highest precision (0.83), sensitivity (0.831), accuracy (0.842), F1-score (0.833), and the highest MCC (0.48). Additionally, the most accurate models in the unbalanced dataset belonged to RF with the lowest BS (0.128). Although all studied models overestimate mortality risk and have insufficient calibration (P > 0.05), stacking demonstrated relatively good agreement between predicted and actual mortality. EL models are not superior to LR in predicting in-hospital mortality in the ED. Both EL and LR models can be considered as screening tools to identify patients at risk of mortality. [ABSTRACT FROM AUTHOR]

Drought is an unpleasant climatic phenomenon that directly impacts the different dimensions of human societies. To pry-awareness and choice optimum management decision, design and development of an integrated approach is needed to control this phenomenon more effectively and provide early warnings. This study used twelve remotely sensed indices of the Moderate Resolution Imaging Spectroradiometer (MODIS) and digital elevation model (DEM) to monitor drought during the 2000–2018 growing season. Standardized Precipitation Index (SPI) with 1 to 12 months time scales was used as reference data. A machine learning approach modulated the relations between thirteen indices and SPI with different time scales. The random forest technique was used to construct a comprehensive drought monitoring model in Ilam Province. Validation data were provided based on relative soil moisture, Standardized Precipitation Evapotranspiration Index (SPEI), and crop yield data. It was observed that random forest produced good applicability (R2 = 0.88) for SPI prediction. In the next step, the Drought Hazard Index (DHI) was generated based on the probability occurrences of drought using the comprehensive drought model made in the previous step. The Drought Vulnerability Index (DVI) was calculated by using 7 socioeconomic indices. Finally, the Drought Risk Index (DRI) was obtained by multiplying DHI and DVI for Ilam province. The result of the DRI map showed that two counties are at very high risk of drought, 4 counties are at high risk, and 4 counties are at moderate and low risk of drought. Overall, the result of our study provides a comprehensive method for the assessment of regional drought. Also based on this model, counties with high vulnerability can be identified to provide timely management programs to help improve the situation. [ABSTRACT FROM AUTHOR]

With increasing levels of air pollution, air quality prediction has attracted more attention. Mathematical models are being developed by researchers to achieve precise predictions. Monitoring and prediction of atmospheric PM2.5 levels, as a predominant pollutant, is essential in emission mitigation programs. In this study, meteorological datasets from 9 years in Isfahan city, a large metropolis of Iran, were applied to predict the PM2.5 levels, using four machine learning algorithms including Artificial Neural |Networks (ANNs), K-Nearest-Neighbors (KNN), Support Vector |Machines (SVMs) and ensembles of classification trees Random Forest (RF). The data from 7 air quality monitoring stations located in Isfahan City were taken into consideration. The Confusion Matrix and Cross-Entropy Loss were used to analyze the performance of classification models. Several parameters, including sensitivity, specificity, accuracy, F1 score, precision, and the area under the curve (AUC), are computed to assess model performance. Finally, by introducing the predicted data for 2020 into ArcGIS software and using the IDW (Inverse Distance Weighting) method, interpolation was conducted for the area of Isfahan city and the pollution map was illustrated for each month of the year. The results showed that, based on the accuracy percentage, the ANN model has a better performance (90.1%) in predicting PM2.5 grades compared to the other models for the applied meteorological dataset, followed by RF (86.1%), SVM (84.6%) and KNN (82.2%) models, respectively. Therefore, ANN modelling provides a feasible procedure for the managerial planning of air pollution control. [ABSTRACT FROM AUTHOR]

Pistachio is one of the most important and valuable orchard products in Iran and some other places in the world. Because it is adaptable to adverse environmental conditions, especially drought and salinity, more land is being used for pistachios. In an increasingly resource constrained world, producers, researchers and policy makers need to clearly identify suitable land to optimize production. The present study modelled the relationship between pistachio yield and soil variables by regression, linear and non-linear (MLR and NMLR), feed forward back propagation artificial neural network (FFBP-ANN), adaptive neural fuzzy inference system (ANFIS) and random forest (RF) models in areas under pistachio production from Iran. For this purpose, 124 pistachio orchards were selected and sampled in Kerman, Fars, Khorasan Razavi, Isfahan and East Azarbaijan provinces of Iran. The results indicated that MLR and NMLR could explain 72 and 77% of the pistachio yield variation, respectively, whereas prediction accuracy increased when the data of pistachio orchards were entered in intelligent models: ANFIS, ANN-based and RF to 86, 92 and 96%, respectively. The RF model was found to be most parsimonious. As the models were highly sensitive to gravel, electrical conductivity, exchangeable sodium, CaCO3, gypsum, and available phosphorus and potassium concentrations, special attention should be paid to the measurement and management of these properties. [ABSTRACT FROM AUTHOR]

Machine learning approaches are increasingly being employed to forecast the key characteristics of strong ground motions, including the challenging classification of Pulse-Like (PL) ground motions. The PL ground motions are characterized by their impulsive nature and have the potential to cause significant damage to structures. The classification of PL ground motions continues to be a significant challenge due to the absence of consensus on their definition and categorization. This paper investigates the potential benefits of several Machine Learning Classifiers (MLCs) algorithms such as decision tree, random forest, logistic regression, naive Bayes, support vector machine, K-nearest neighbor, ensemble model, and artificial neural network models for predicting PL and Non-Pulse-Like (NPL) ground motions. In this regard, a dataset comprising 200 near-fault ground motions records compiled from active tectonic regions like Taiwan, Turkey, Iran, and Japan, was divided into 2 portions, with 75% used to train the model and the remaining 25% used for testing. Plots of performance metrics, confusion matrix, and receiver operating curve indicate that the ensemble classifier outperforms the other classifier with 86.2% accuracy and the lowest misclassification of PL and NPL ground motions. Additionally, the trained MLC has been compared with the existing ground motion classification models to further assess the accuracy of the different classifiers in the present study. [ABSTRACT FROM AUTHOR]

Drought as a natural phenomenon has always been a serious threat to regions with hot and dry climates. One of the major effects of drought is the drop in groundwater level. This paper focused on the SPI (Standardized Precipitation Index) and SWI (Standardized Water-Level Index) to assess meteorological and hydrological drought, respectively. In the first part, we used different time frames of SPI (3, 6, 12, and 24 months) to investigate drought in Yazd, a dry province in the center of Iran for 29 years (1990–2018). Then, in the second part, the relationship between SPI and SWI was investigated in the three aquifers of Yazd by some rain gauge stations and the closest observation wells to them. In addition to using SPI and SWI, we also used different machine learning (ML) algorithms to predict drought conditions including linear model and six non-linear models of K_Nearest_Neighbors, Gradient_Boosting, Decision_Tree, XGBoost, Random_Forest, and Neural_Net. To evaluate the accuracy of the mentioned models, three statistical indicators including Score, RMSE, and MAE were used. Based on the results of the first part, Yazd province has changed from mild wet to mild drought in terms of meteorological drought (the amount of rainfall according to SPI), and this condition can worsen due to climate change. The models used in ML showed that SPI-6 (score ave = 0.977), SPI-3 (score ave = 0.936), SPI-24 (score ave = 0.571), and SPI-12 (score ave = 0.413) indices had the highest accuracy, respectively. The models of Neural_Net (score ave = 0.964–RMSE ave = 0.020–MAE ave = 0.077) and Gradient_Boosting (score ave = 0.551–RMSE ave = 0.124–MAE ave = 0.248) had the highest and lowest accuracy in prediction of the SPI in all four-time scales. Based on the results of the second part, about the SWI, Random_Forest model (score = 0.929–RMSE = 0.052–MAE = 0.150) and model of Neural_Net (score = 0.755–RMSE = 0.235–MAE = 0.456) had the highest and lowest accuracy, respectively. Also, hydrological drought (reduction of the groundwater level) of the region has been much more severe, and according to the low correlation coefficient of average SPI and SWI (R2 = 0.14), we found that the uncontrolled pumping wells, as a main factor than a shortage of rainfall, have aggravated the hydrological drought, and this region is at risk of becoming a more arid region in the future. [ABSTRACT FROM AUTHOR]

In this study, a land-subsidence vulnerability map has been prepared using Machine Learning (ML) models fusing, Random Forest (RF), Support Vector Machine (SVM) and a GIS technique in the Hamadan Province, Iran. The information layers used as input of the ML models in "R" software are comprised of elevation, slope, plan and profile curvatures, slope aspect, Topographic Wetness Index (TWI), Normalized Difference Vegetation Index (NDVI), soil texture, distance from rivers, distance from the fault, geology, land use, and groundwater level drawdown. The accuracy of the results obtained stood at 89% in the SVM, and up to 96% in the RF models. The RF model demonstrates a greater efficiency than the SVM model. To determine each parameter's effect on the land-subsidence of the study area, the Partial Least Squares (PLS) model has been used in the R software. The use of the PLS model shows a greater effect of elevation and groundwater level decline compared to the other parameters on the land-subsidence phenomenon. Finally, the raster vulnerability map in the GIS software was divided into four classes in terms of intensity as 'low,' 'medium,' 'high,' and 'very high' utilizing the natural break method. In the optimal RF model 45% of aquifers were assessed as being low, 23% as moderate, 20% as high, and 12% as very high. The study of the groundwater changing process, using GRACE satellite data in Google Earth Engine environment confirmed a decrease in groundwater level, which has led to land-subsidence in the aquifer. [ABSTRACT FROM AUTHOR]

This study was conducted in the Lorestan Province in the west of Iran with two objectives of identifying major environmental variables in spatial risk modeling and identifying spatial risk patches of livestock predation by the Persian leopard. An ensemble approach of three models of maximum entropy (MaxEnt), generalized boosting model (GBM), and random forest (RF) were applied for spatial risk modeling. Our results revealed that livestock density, distance to villages, forest density, and human population density were the most important variables in spatial risk modeling of livestock predation by the leopard. The center of the study area had the highest probability of livestock predation by the leopard. Ten spatial risk patches of livestock predation by the leopard were identified in the study area. In order to mitigate the revenge killing of the leopards, the findings of this study highlight the imperative of implementing strategies by the Department of Environment (DoE) to effectively accompany the herds entering the wildlife habitats with shepherds and a manageable number of guarding dogs. Accordingly, the identified risk patches in this study deserve considerable attention, especially three primary patches found in the center and southeast of Lorestan Province. [ABSTRACT FROM AUTHOR]

Purpose: Hypercholesterolemia is a major risk factor for a wide range of cardiovascular diseases. Developing countries are more susceptible to hypercholesterolemia and its complications due to the increasing prevalence and the lack of adequate resources for conducting screening and/or prevention programs. Using machine learning techniques to identify factors contributing to hypercholesterolemia and developing predictive models can help early detection of hypercholesterolemia, especially in developing countries. Methods: Data from the nationwide 2016 STEPs study in Iran were used to identify socioeconomic, lifestyle, and metabolic risk factors associated with hypercholesterolemia. Furthermore, the predictive power of the identified risk factors was assessed using five commonly used machine learning algorithms (random forest; gradient boosting; support vector machine; logistic regression; artificial neural network) and 10-fold cross validation in terms of specificity, sensitivity, and the area under the receiver operating characteristic curve. Results: A total of 14,667 individuals were included in this study, of those 12.8% (n = 1879) had (undiagnosed) hypercholesterolemia. Based on multivariate logistic regression analysis the five most important risk factors for hypercholesterolemia were: older age (for the elderly group: OR = 2.243; for the middle-aged group: OR = 1.869), obesity-related factors including high BMI status (morbidly obese: OR = 1.884; obese: OR = 1.499; overweight: OR = 1.426) and AO (OR = 1.339), raised BP (hypertension: OR = 1.729; prehypertension: OR = 1.577), consuming fish once or twice per week (OR = 1.261), and having risky diet (OR = 1.163). Furthermore, all the five hypercholesterolemia prediction models achieved AUC around 0.62, and models based on random forest (AUC = 0.6282; specificity = 65.14%; sensitivity = 60.51%) and gradient boosting (AUC = 0.6263; specificity = 64.11%; sensitivity = 61.15%) had the optimal performance. Conclusion: The study shows that socioeconomic inequalities, unhealthy lifestyle, and metabolic syndrome (including obesity and hypertension) are significant predictors of hypercholesterolemia. Therefore controlling these factors is necessary to reduce the burden of hypercholesterolemia. Furthermore, machine learning algorithms such as random forest and gradient boosting can be employed for hypercholesterolemia screening and its timely diagnosis. Applying deep learning algorithms as well as techniques for handling the class overlap problem seems necessary to improve the performance of the models. [ABSTRACT FROM AUTHOR]

In this study, two models of Random Forest and copula-based simulation were used to evaluate the accuracy and efficiency of the IHACRES rainfall-runoff model in simulating the daily discharge of the Siminehroud River in the south of Lake Urmia basin, Iran. A trivariate copula-based model was created using discharge, rainfall and temperature data on a daily scale in the period 1992–2018. Vine family models and their conditional densities were used to implement the copula-based model. By calibrating the IHACRES model and also selecting the tree sequence in accordance with the data, rainfall-runoff simulations were performed in the study area. The accuracy and efficiency of the studied models were evaluated using RMSE and NSE criteria, and also violin plot and Taylor diagram. The results of comparing the error rate of rainfall-runoff simulation in the study area showed that the vine-based model reduces the RMSE statistics by about 14.5 and 16.5%, respectively, compared to the IHACRES and Random Forest models. According to the presented diagrams, the efficiency and certainty of IHACRES and copula-based simulation models are acceptable. While the Random Forest model does not have acceptable accuracy and efficiency in the study area. The copula-based simulation model has a good performance due to the unique tree sequence as well as involving the marginal distributions fitted to the data. Although the copula-based simulation model has increased the efficiency of the model in simulating the daily discharge by about 5% compared to the IHACRES model, it is not significant compared to the mathematical complexity of the copula-based model. [ABSTRACT FROM AUTHOR]

This study introduces two integrated methods to model the short- to long-term droughts in terms of the Standardized Precipitation Index (SPI). In this regard, the monthly precipitation data from three stations located in different climates of Iran was used. Firstly, the capability of Random Forest (RF) and M5 Pruning tree (M5P) models were assessed for short- to long-term drought forecasting. Then, the impact of data multi-pre-processing on improving the models' efficiency was investigated. Therefore, inputs were serially decomposed via Discrete Wavelet Transform (DWT) and Variational Mode Decomposition (VMD). The most dominate subseries were selected based on their energy values and used as inputs of the RF and M5P. Furthermore, since the drought variable consists of two deterministic and stochastic sections and in the modeling process the stochastic part is not often considered, in the third step, the Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model was applied for determining the random parts of the SPIs. Results showed that the integrated models led to better outcomes compared to the standalone models. The second and third steps increased the modeling accuracy up to 45% and 55%, respectively. Results showed that the distribution range of the Root Mean Square Errors (RMSE) criteria for the SPI-3 modeling decreased from 0.438–0.551 (in using raw data) to 0.184–0.26 (in using the integrated-GARCH model). Results showed that using the integrated-GARCH models, SPIs could be successfully simulated with considering only one input variable. [ABSTRACT FROM AUTHOR]

Permeability is one of the important issues that must be considered in the investigation of dam sites. Determination of this parameter in the boreholes is time-consuming, costly, and in some cases impossible. In this study, the values of Q classification system, Lugeon and joint spacing in five-meter intervals of boreholes in limestone rocks of Bazoft and Khersan II dam sites, in south Iran were determined. Then, the Lugeon number was estimated based on Q classification system and joint spacing in control and trial grouting boreholes by statistical analysis (SA), multilayer perceptron neural network (MPNN) by feed-forward method, support vector regression (SVR), adaptive neuro-fuzzy inference system (ANFIS), and random forest (RF) methods. Results showed that rock mass is categorized in moderate permeability class based on mean Lugeon value and in the good category based on Q classification system. The Q classification system and joint spacing indicated the highest effect on the Lugeon and the depth showed the least effect on the Lugeon. The SA displayed that it is possible to predict Lugeon values by Q and Js with a precision higher than 78% based on the data of both dam sites. According to the criteria, the accuracy of the MPNN (R = 0.91), RF (R = 0.97), ANFIS (R = 0.93) and SA (R = 0.80–0.83) to estimate the Lugeon number was lower than SVR (R = 0.98) method. [ABSTRACT FROM AUTHOR]

Inorganic carbon is the largest source of carbon in terrestrial surface, particularly in arid and semiarid regions, including the Chahardowli Plain in western Iran. Inorganic carbon plays an equal or greater role than organic soil carbon in these areas, although less attention has been paid in quantifying their variability. The objective of this study was to model and map calcium carbonate equivalent (CCE) presenting inorganic carbon in soil using machine learning and digital soil mapping techniques. Chahardowli Plain in foothills of the Zagros Mountains in the southeast of Kurdistan Province in Iran was taken as a case study area. CCE was measured at 0–5, 5–15, 15–30, 30–60, and 60–100 cm depths following GloalSoilMap.net project specifications. A total of 145 samples were collected from 30 soil profiles using the conditional Latin hypercube (cLHS) method of sampling. Relationships between CCE and environmental predictors were modeled using random forest (RF) and decision tree (DT) models. In general, the RF model performed slightly superior than the DT model. The mean value of CCE increased with soil depth, from 3.5% (0–5 cm) to 63.8% (30–60 cm). Remote sensing (RS) variables and terrestrial variables were equally important. The importance of RS variables was higher at the surface than terrestrial variables, and vice versa. The most significant variables were Channel Network Base Level (CNBL) variable and Difference Vegetation Index (DVI) with the same variable importance value (21.1%). In areas affected by river activities, the use of the CNBL and vertical distance to channel networks (VDCN) as variables in digital soil mapping (DSM) could increase the accuracy of soil property prediction maps. The VDCN played a principal role in soil distribution in the study area by affecting the rate of discharge and, thus, erosion and sedimentation. A high percentage of carbonate in parts of the region could exacerbate nutrient deficiencies for most crops and provide information for sustainably managing agricultural activity. [ABSTRACT FROM AUTHOR]

Understanding the spatial variation of soil properties is essential for monitoring land capabilities as well as the sustainable management of soil resources. The aim of this study was to predict digital soil properties mapping using 23 environmental variables, i.e., terrain attributes and remote sensing (RS) indices, across 1500 km2 of Mashhad plain lands. To achieve this purpose, a total of 180 soil samples (0–10 cm) were taken. The random forest (RF) model combined with ordinary kriging (OK), as well as regression kriging (RK), were applied to relate environmental variables and the studied soil properties. The results revealed that RF-OK was the best model with R2 and RMSE for silt (0.89% and 0.10%), followed by calcium carbonate equivalent (0.88% and 3.30%), clay (0.87% and 2.26%), soil organic carbon (0.86% and 0.24%), sand (0.84% and 4.21%), and pH (0.82% and 5.42%). The RS covariates, including band 5 (B5), modified soil-adjusted vegetation index (MSAVI), difference vegetation index (DVI), band 2 (B2), carbonate rock index (CRI2), gypsum index (GI), and enhanced vegetation index (EVI), and terrain attributes, including topographic wetness index (TWI) and elevation (EL), and topographic position index (TPI), were the most important variables in modeling different soil properties. RF-OK showed the prediction and uncertainty maps related to high precision and low standard deviation in most study areas, which indicate low overfitting and overtraining in modeling processes. In general, the RF-OK model, with low cost and high accuracy, can be applicable to use for predicting different soil properties, as well as spatial information acquired from an effort to maps to managing agriculture in areas at different conditions. Finally, this method can be applied to other regions of similar properties and for similar purposes. [ABSTRACT FROM AUTHOR]

In this study, the performance of continuous autoregressive moving average (CARMA), CARMA-generalized autoregressive conditional heteroscedasticity (CARMA-GARCH), random forest, support vector regression and ant colony optimization (SVR-ACO), and support vector regression and ant lion optimizer (SVR-ALO) models in bivariate simulating of discharge based on the rainfall variables in monthly time scale was evaluated over four sub-basins of Lake Urmia, located in northwestern Iran. The models were assessed in two stages: train and test. The results showed that the CARMA-GARCH hybrid model offered better performance in all cases than the stand-alone CARMA. The improvement percentages of the error rate in the CARMA model compared to the CARMA-GARCH hybrid model in the Mahabad Chai, Nazlu Chai, Siminehrood, and Zola Chai sub-basins were 9, 20, 17, and 6.4%, respectively, in the training phase. Among the models, the hybrid SVR models integrated with ACO and ALO optimization algorithms presented the best performance based on the Taylor diagram and evaluation criteria. Considering the use of ant colony and ant lion optimization algorithms to optimize the support vector regression model's parameters, these models offered the best performance in the study area to simulate the discharge. The improvement percentages of the error rate in the SVR-ACO model compared to the CARMA-GARCH hybrid model in the Mahabad Chai, Nazlu Chai, Siminehrood, and Zola Chai sub-basins were 11, 10, 19, and 21%, respectively, in the training phase. In contrast, the random forest model provided the lowest accuracy and the highest error in discharge simulation. [ABSTRACT FROM AUTHOR]

The present study was conducted to compare generalized linear model (GLM), random forest (RF), and Cubist to produce available phosphorus (AP) and potassium (AK) maps and to identify the covariates that control mineral distribution in Lorestan Province, Iran. To this end, the locations for collecting 173 soil samples were determined through the conditioned Latin hypercube sampling (cLHS) method, at four different land-uses (orchards, paddy fields, agricultural, and abandoned fields). The performance of the models was assessed by coefficient of determination (R2), root-mean-square error (RMSE), and mean absolute error (MAE) indices. The results showed that the RF model fitted better than GLM and Cubist models and could explain 40 and 57% of AP and AK distribution, respectively. The R2, RMSE, and MAE of the RF model were 0.4, 2.81, and 2.43 for predicting AP and equal to 0.57, 143.77, and 116.61 for predicting AK, respectively. The most important predictors selected by the RF model were valley depth and soil-adjusted vegetation index (SAVI) for AP and AK, respectively. The maps showed higher AP and AK content in apricot orchards compared to other land-uses. No difference was observed between AP and AK content on paddy fields, agricultural, and abandoned areas. The higher AP and AK contents were related to orchard management practices, such as failure to dispose of plant residuals and fertilizer consumption. It can be concluded that the orchards (by increasing soil quality) was the best land-use in line with sustainable management for the study area. However, generalizing the results needs more detailed research. [ABSTRACT FROM AUTHOR]

A dust storm is a major environmental problem affecting many arid regions worldwide. The novel contribution of this study is combining indicators extracted from RS- and statistic-based predictive models to spatial mapping of land susceptibility to dust emissions in a very important dust source area in the borders of Iran and Iraq (Khuzestan province in Iran and Al-Basrah and Maysan provinces in Iraq). In this research, remote sensing (RS) techniques and machine learning techniques, including multivariate adaptive regression spline (MARS), random forest (RF), and logistic regression (LR), were used for dust source identification and susceptibility map preparation. To this end, 152 DSA for the period of 2005–2020 were identified in the study area. Of these DSA data, 70% was assigned to the Dust Source Susceptibility Mapping (DSSM) (training dataset) and 30% to model validation. Consequently, six factors (i.e., soil, lithology, slope, normalized vegetation differential index (NDVI), geomorphology, and land use units) were prepared as DSA's independent and effective variables. The results of all three models indicated that land use had the most impact on DSA. The validation results of these models using the test data showed sub-curves of 0.92, 0.86, and 0.76 for the RF, MARS, and LR models, respectively. Also, results showed that the RF model outperformed MARS (AUC = 0.89) and LR (AUC = 0.78) methods. In all three models, high and very high susceptibility classes generally covered a large percentage of the case study. The highest percentage of dust source points was also in this susceptibility category. Overall, the results of this study can be useful for planners and managers to control and reduce the risk of negative dust consequences. [ABSTRACT FROM AUTHOR]

This paper evaluates a comparison between three machine learning algorithms (MLAs), namely support vector machine (SVM), multilayer perceptron artificial neural network (MLP-ANN) and random forest (RF), in landslide susceptibility mapping and addresses a optimization algorithm to optimize the performance of a MLA to yield more accurate and reliable results. A genetic algorithm (GA) approach as a part of evolutionary algorithms was utilized in order to optimize the performance of best model among three utilized MLAs. The study area (Tarom–Khalkhal sub-basin) is located in North-West of Iran with mountainous nature (western part of Alborz Mountains), wherein numerous landslide occurrences were recorded. In this case, fifteen predisposing factors, gathered from aerial images and field surveys, were considered to generate the final landslide susceptibility models. The validation procedure was conducted with taking advantage of confusion matrices for different algorithms. Finally, landslide susceptibility maps were generated and evaluated through receiver operator characteristic (ROC) curves. RF algorithm showed the best performance; therefore, hybridized genetic random forest (GRF) was employed in order to optimize the hyperparameters (number of trees, number splits and depth) of the model, which can affect the performance of model. As a result, GRF has best performance among all mentioned algorithms with AUC = 0.93. As a conclusion, genetic algorithm was found to be suitable in optimizing the performance of machine learning algorithms, which is crucial when it comes to landslide susceptibility mapping. [ABSTRACT FROM AUTHOR]

Landslides are natural hazards that can cause catastrophic life losses and damage to infrastructures and communities. In Iran, landslide exposure has been predominantly increasing in the Zagros Mountains, notably along the lifelines, such as road networks. Therefore, this study aimed to investigate the landslide vulnerability of a 6682 km road network in the Chaharmahal and Bakhtiari Province, Iran, using a two-step methodology comprised of: (1) landslide susceptibility mapping using four machine learning methods—boosted regression trees (BRT), multiple discriminant analysis (MDA), multivariate adaptive regression splines (MARS), and random forest (RF); and (2) mapping road exposure to landslides using the analytic hierarchy process (AHP) that computed the weight for four buffer zones (0–50, 50–150, 150–300, and > 300 m) from the road network. The combined results of steps 1 and 2 produced a map of the road network vulnerability to landslides that demonstrated that 9.7 km (13.6%) of the road network was located in the very-high vulnerability class. Specifically, the roads of the Ardal and Kohrang counties have been found to be the most vulnerable to landslide risk. The finding of this study could be useful for decision-makers and civil engineering to better manage road networks in terms of landslide risk and community resilience in the aftermath of major landslides. [ABSTRACT FROM AUTHOR]

The Balangu (Lallemantia iberica) species have a high gastronomical impact in the Middle East and Balkan region. It is widely used in the local food industry, such as confectionery, edible oil, and protein food. In this study, 49 ecotypes were collected from different regions of Iran. 37 agronomic traits were measured during the growing season and at harvest time. To find the correlation between the grain yield per unit area, grain yield per single plant (GYSP), oil percent (OP), and protein percent (PP) with other measured traits, which these were utilized as the labels of different machine learning (ML) procedures including Linear Regression (LR), Support Vector Regression (SVR), Random Forest Regression (RFR), and Gradient Boosting Decision Tree Regression (GBDTR). It was observed that there is a linear relationship between the measured agronomic traits and the considered labels. So, the LR, RFR, and GBDTR models showed the lowest mean absolute error, mean square error, and root mean square error than SVR models and good prediction ability of the test data. Although, the RFR and GBDTR have naturally lower bias than other methods in this study, but the GBDTR scheme is preferred because of the over-fitting shortcoming of the RFR technique. The GBDTR method showed better results rather than the other ML regression methods according to the RMSE 3.302, 0.040, 0.028, and 0.060 for GYUA, GYSP, OP, and PP, respectively. [ABSTRACT FROM AUTHOR]

The aim of this study was to improve our understanding of factors that affect the spatial distribution of wildfire occurrences at the regional scale. We employed the random forest, boosted regression tree, and genetic algorithm rule-set production models to assess the spatial interplay between fire events and climate, topography, and anthropogenic factors in order to characterize wildfire occurrence in the Zagros eco-region of western Iran. We constructed a geospatial database using the historical fires from the period 2007–2020 and topography, climate, and human related factors. The results demonstrated that human activities (i.e., land use and distance from the settlements and roads) contributed 45% to the probability model of wildfire occurrence in the study region. The models ranked the climate factors (rainfall, temperature, and wind effect) as the second most influential drivers of fire occurrences, whereas topographic features (slope, elevation, and aspect) did not significantly influence fire probability in the landscape. Overall model performance was assessed with the area under the receiver operating characteristic (AUROC) method that showed the superior performance of the RF model in the training phase (AUROC = 0.92) and in its ability to predict upcoming fires (AUROC = 0.90). The insights obtained from this research can bring into focus both the locations and the types of suppression policies that are required to alleviate the effects of the upcoming wildfires in the early twenty-first century. [ABSTRACT FROM AUTHOR]

Estimation of reservoir inflow is of particular importance in optimal planning and management of water resources, proper allocation of water to consumption sectors, hydrological studies, etc. This study aimed to estimate monthly inflow (Q) to the Maroon Dam reservoir located in Iran utilizing climatic data such as minimum, maximum, and mean air temperatures (Tmin, Tmax, T), reservoir evaporation (E), and rainfall (R). The impact of any of the mentioned variables was analyzed by the entropy-based pre-processing technique. The results of the pre-processing showed that the rainfall is the most important parameter affecting the reservoir inflow. Therefore, three types of input patterns were taken into consideration consisting the antecedent Q-based, antecedent R-based, and combined antecedent Q and R-based input combinations. To estimate the monthly reservoir inflow, a random forest (RF) was firstly employed as the standalone model. Then, two different types of hybrid models were proposed via coupling the RF on complete ensemble empirical mode decomposition (CEEMD) and wavelet analysis (W) in order to implement the coupled CEEMD-RF and W-RF models. It is worthwhile to mentioning that six mother wavelets were used in developing the hybrid W-RF models. Four error metrics including root mean square error (RMSE), mean absolute error (MAE), Kling-Gupta efficiency (KGE), and Willmott index (WI) were used to assess the accuracy of implemented models. The attained results indicated the superiority of proposed hybrid models over the classic RF for estimating the monthly reservoir inflow. The most precise model during the test phase was W-RF(3) utilizing the Sym(2) as the mother wavelet under a lagged Q-based pattern with error measures of RMSE = 15.011 m3/s, MAE = 10.439 m3/s, KGE = 0.832, WI = 0.773. [ABSTRACT FROM AUTHOR]

In Streamflow prediction the most important triggering/controlling variables are related to climate, physiography, and landscape patterns. This study investigated the effect of different landscape metrics to relate spatial patterns to surface runoff processes and predict monthly streamflow using climatic and physiographic variables for the 42 sub-basins of the Urmia Lake Basin in Iran. We developed an innovative data-driven framework and considered two different modelling approaches i.e., modelling in homogenous clusters (local approach) and modelling in the entire area as an entity (global approach). The results of basin LULC monitoring from the 20-year experimental period display drastic changes in the land use of the basin such as reduction in lake area (48.3%) due to increasing irrigated areas (22.5%), increasing residential areas (14.2%), and decrease in rangeland (6.0%). Streamflow prediction results in the global experiment showed Group Method of Data Handling (GMDH) and Random Forest (RF) with NSE of 0.76 and NRMSE of 6.44% have similar results and outperformed Partial Least Squares regression (PLS), but in clustering experiment GMDH with NSE of 0.88 and NRMSE of 5% shows the highest accuracy and outperformed both RF and PLS. The results confirmed that modelling in homogenous clusters (local prediction) significantly enhanced the performance of prediction. [ABSTRACT FROM AUTHOR]

Soil characteristics depend on its various physical, chemical, and biological parameters. Soil organic matter content (SOMC) is one of the main parameters of soil that affects all of these parameters. Therefore, this study was performed to estimate SOMC of irrigated lands in the Damavand Absard region of Tehran province using terrain attributes and standardized spectral reflectance (ZPC1) through two modeling approaches of random forest (RF) and artificial neural network (ANN). Soil sampling (0–30 cm) was performed at 60 points using random sampling method, and SOMC was measured in the laboratory. The results of SOMC modeling showed that the RF model predicted SOMC more accurately than the ANN model. Modeling SOMC using RF model showed that the R2, RMSE, and MAE values for training data were 0.81, 0.23%, and 0.22%, respectively. Furthermore, the R2, RMSE, and MAE values for test data were 0.78, 0.20%, and 0.21%, respectively. Modeling SOMC stocks by RF model showed that the efficient factors for estimating the SOC stocks were ZPC1, Slope, Normalized height, Relative Slope Position, elevation, Vertical Distance to Channel Network, Standardized height, and Slope length (LS) factor. According to the obtained results, the combination of terrain attributes and ZPC1 data can be used for managing and monitoring the SOMC in the irrigated lands of Iran with sufficient accuracy. Based on the results of this study, we conclude that the farmers can reference these maps to manage the SOMC in the Damavand Absard region due to high importance of this area for irrigated agriculture. [ABSTRACT FROM AUTHOR]

Climate change has caused a rise in temperature extremes, particularly heatwaves, in recent decades. Physical-empirical models are developed in this study using two classical machine learning algorithms, namely decision tree (DT) and random forests, and a novel hybrid technique consists of Ada-Boost Regression and decision tree (ABR-DT) for forecasting annual heatwave days (HWDs) of Iran from synoptic predictors. The daily temperature data of Princeton Meteorological Forcing were extracted for 27 points to estimate the annual number of HWDs, and the National Centers for Environmental Prediction reanalysis data were used as predictors. The major synoptic variables were extracted for four pressure levels (e.g., 300, 500, 850, and 1000 hPa) and three monthly time lags. The Principal Component Analysis was employed to reduce the diverse predictors and their features to the most optimal structure. The grid point-based performance evaluation revealed the superiority of ABR-DT, which showed a correlation coefficient (CC) of 0.860 and meant absolute error (MAE) of 6.929, using only specific humidity and wind component as predictors. The spatial performance indices over eight different climate regions of Iran also showed the better performance of ABR-DT, which improved the CC and MAE of its two alternatives up to 185 and 19%. The study identified the optimal parameter combination as the predictors of heatwaves by examining the effects of numerous weather components. The results proved the proposed hybrid forecasting approach's effectiveness in predicting heatwave days, a devastating hazard, for many regions. [ABSTRACT FROM AUTHOR]

Iranian plateau is a seismically active region. Within this region, northwestern Iran is of high importance. Before proper planning for mitigating the earthquake-induced hazards can be achieved, it is necessary to identify high-risk areas in terms of susceptibility to earthquakes. In this study, landslide susceptibility in Miandoab Country was modeled using the so-called random forest algorithm (RFA) in MATLAB based on records acquired at 67 earthquake hotspots considering 9 factors affecting the earthquake occurrence (i.e., height, slope, direction, distance from fault, distance from river, distance from road, land use, geology, and precipitation). Predictive power of the model and validity of its results were evaluated using relative operating characteristic (ROC) curve and area under the curve (AUC). The assessment results showed very good accuracy of the model (0.97). It was further found that digital height layer, geology, and distance from fault impose the largest contributions into earthquake potential. The results also showed that 53%, 8.3%, and 38.4% of the study area were classified as being at low risk, moderate risk, and high risk of earthquake. Among other climatic elements, the precipitation exhibits the largest fluctuations; we proceeded to evaluate precipitation trends in the study area for a statistical period of 30 years. This was practiced by implementing Mann–Kendall nonparametric test in MATLAB. This subject-matter is especially important in Iran where the annual precipitation level is as low as 250 mm. The results showed that the precipitation follows an increasing trend in the region. [ABSTRACT FROM AUTHOR]

This study assesses a plausible correlation between a dust intrusion episode and a daily increase in COVID-19 cases. A surge in COVID-19 cases was observed a few days after a Middle East Dust (MED) event that peaked on 25th April 2020 in southwest Iran. To investigate potential causal factors for the spike in number of cases, cross-correlations between daily combined aerosol optical depths (AODs) and confirmed cases were computed for Khuzestan, Iran. Additionally, atmospheric stability data time series were assessed by covering before, during, and after dust intrusion, producing four statistically clustered distinct city groups. Groups 1 and 2 had different peak lag times of 10 and 4-5 days, respectively. Since there were statistically significant associations between AOD levels and confirmed cases in both groups, dust incursion may have increased population susceptibility to COVID-19 disease. Group 3 was utilized as a control group with neither a significant level of dust incursion during the episodic period nor any significant associations. Group 4 cities, which experienced high dust incursion levels, showed no significant correlation with confirmed case count increases. Random Forest Analysis assessed the influence of wind speed and AOD, showing relative importance of 0.31 and 0.23 on the daily increase percent of confirmed cases, respectively. This study may serve as a reference for better understanding and predicting factors affecting COVID-19 transmission and diffusion routes, focusing on the role of MED intrusions. [ABSTRACT FROM AUTHOR]

The preliminary responses on stability assessment are very effective in discontinuous rock slope stabilisations, which geo-mechanical empirical classifications/approaches can be evaluated. Although there are several geomechanical classifications for rock mass characteristics, they still have some complexities and shortcomings. In this paper, correlations based on two flexible classification systems, namely slope mass rating (SMR) and Barton's Qslope, has been discussed as a connection between failure (i.e., safety factor, reliability) and stability conditions (i.e., failure mechanism, support system) which were performed in 300 road/railway slope cases from 12 provinces in Iran. The empirical SMR-Qslope relationship has been used for sedimentary rocks defined lithologically as limestone, marlstone, sandstone, and claystone. The method was used to provide the empirical link to the primary design for discontinuous rock slopes. To this end, after field investigations and the necessary geo-engineering data gathered on the studied slopes. The information used to provide the SMR and Qslope retirements prepare SMR-Qslope relation for Iranian data. The artificial intelligence techniques including k-nearest neighbours, support vector machine, Gaussian process, decision tree, random forest, multilayer perception, and Naïve-Bayes classifiers have been used for detailed classifications. The accurate correlations provided have been implemented and revised in the Python programming language. According to the learning model results, SMR-Qslope equation for Iranian data has been developed as SMR = 6.699 ln (Qslope) + 58.99 / R2 = 0.62 with 0.95 (95%) accuracy. The evaluated results were verified and controlled by Jorda-Bordehore et al. (Stability assessment of rock slopes using empirical approaches: comparison between slope mass rating and Q-slope, 2018) and Maion (Proposta de correlação entre os índices SMR e Q-slope , 2019) empirical relationships as a basic table. By comparing these studies' results, it can be stated that the present study has improved the accuracy of SMR-Qslope relationship. [ABSTRACT FROM AUTHOR]

Regarding the ability of data mining algorithms for post-processing the output of climate models, and on the other hand, the successful application of multi-model ensemble approaches in climate forecasts, in this paper, some important data mining algorithms are evaluated for the monthly forecast of precipitation over Iran. For this purpose, four European climate models, from DWD, ECMWF, CMCC and Meteo-France, with six lead times, are used to be post-processed by applying four different algorithms including artificial neural networks, support vector regression, decision tree and random forests. Based on the proposed approach, 72 different models are provided for 12 months, each month with six lead times. The approach is applied for the monthly forecast of precipitation over Iran. According to the results, the neural network and random forest methods performed better than the decision tree and the support vector machine. This advantage preserved for all months of the year. Also, the proposed multi-model approach outperformed any of the individual European models. [ABSTRACT FROM AUTHOR]

Multiple machine learning (ML) and deep learning (DL) models are evaluated and their prediction performance compared in classifying five wellbore fluid-loss classes from a 20-well drilling dataset (Azadegan oil field, Iran). That dataset includes 65,376 data records with seventeen drilling variables. The dataset fluid-loss classes are heavily imbalanced (> 95% of data records belong to the less significant loss classes 1 and 2; only 0.05% of the data records belong to the complete-loss class 5). Class imbalance and the lack of high correlations between the drilling variables and fluid-loss classes pose challenges for ML/DL models. Tree-based and data matching ML algorithms outperform DL and regression-based ML algorithms in predicting the fluid-loss classes. Random forest (RF), after training and testing, makes only 35 prediction errors for all data records. Consideration of precision recall and F1-scores and expanded confusion matrices show that the RF model provides the best predictions for fluid-loss classes 1 to 3, but that for class 4 Adaboost (ADA) and class 5 decision tree (DT) outperform RF. This suggests that an ensemble of the fast to execute RF, ADA and DT models may be the best way to practically achieve reliable wellbore fluid-loss predictions. DL models underperform several ML models evaluated and are particularly poor at predicting the least represented classes 4 and 5. The DL models also require much longer execution times than the ML models, making them less attractive for field operations that require prompt information regarding rapid real-time decision responses to pending class-4 and class-5 fluid-loss events. [ABSTRACT FROM AUTHOR]

Assessing the role of machine learning (ML) models concerning environmental predictors on spatial variation of soil organic carbon stocks (SOCS) in arid rangelands is very necessary. This study was conducted to explore the variability of surface SOCS in rangeland in the west of Iran using ML approaches. A number of 33 environmental predictors derived from Sentinel-2B and DEM were used. The optimal soil sampling (n = 80) position was determined by Latin hypercube sampling (cLHS) method. Robust and popular random Forest (RF), cubist (CB) along with random forest-ordinary kriging (RF-OK), and cubist-ordinary kriging (CB-OK) hybrid ML models were applied to the prediction of SOCS. Ten-fold CV was implemented for modeling performance and uncertainty map. According to data analysis, the maximum, minimum, and average values of SOCS are 44.50, 10.50, and 20.50 (ton. ha−1) at the surface depth (0–30 cm), respectively. In general, normalized and standardized height covariates had a higher effect related to other predictors. On the other hand, two remote sensing (RS) indices, including salinity ratio (salinity) and GNDVI index, had a better impact on SOCS variability. The external validation of model performance indicated that RF-OK with (R2 = 0.75, RMSE = 6.33 ton. ha−1) with the high and low uncertainty range (3.33–9.50 ton. ha−1) was the outperformed ML model in compare with other models as RF (R2 = 0.65, RMSE = 7.38 ton. ha−1), CB-OK (R2 = 0.56, RMSE = 9.22 ton. ha−1), and CB (R2 = 0.33, RMSE = 10.42 ton. ha−1). In general, the hybrid models improved the accuracy of RF and CB with increased 0.11 until 0.23 of R2, and 1.05 to 1.2 (ton. ha−1) decreased RMSE of model's prediction. Hence, we conclude that the topographic attributes (especially normalized and standardized height) were the most critical factors in controlling surface SOCS in arid rangelands when combining with robust RF ML model, and optimized soil sampling methods like RF-cLHS can prepare acceptable soil properties maps. [ABSTRACT FROM AUTHOR]

Monitoring temporal variation of streamflow is necessary for many water resources management plans, yet, such practices are constrained by the absence or paucity of data in many rivers around the world. Using a permanent river in the north of Iran as a test site, a machine learning framework was proposed to model the streamflow data in the three periods of growing seasons based on tree-rings and vessel features of the Zelkova carpinifolia species. First, full-disc samples were taken from 30 trees near the river, and the samples went through preprocessing, cross-dating, standardization, and time series analysis. Two machine learning algorithms, namely random forest (RF) and extreme gradient boosting (XGB), were used to model the relationships between dendrochronology variables (tree-rings and vessel features in the three periods of growing seasons) and the corresponding streamflow rates. The performance of each model was evaluated using statistical coefficients [coefficient of determination (R-squared), Nash–Sutcliffe efficiency (NSE), and root-mean-square error (NRMSE)]. Findings demonstrate that consideration should be given to the XGB model in streamflow modeling given its apparent enhanced performance (R-squared: 0.87; NSE: 0.81; and NRMSE: 0.43) over the RF model (R-squared: 0.82; NSE: 0.71; and NRMSE: 0.52). Furthermore, the results showed that the models perform better in modeling the normal and low flows compared to extremely high flows. Finally, the tested models were used to reconstruct the temporal streamflow during the past decades (1970–1981). [ABSTRACT FROM AUTHOR]

Knowledge of environmental factors controlling soil organic carbon (SOC) stocks can help predict spatial distribution SOC stocks. So, this study was carried out to select the best environmental factors to model and estimate the spatial distribution of SOC stocks in northwestern Iran. Soil sampling was performed at 210 points by multiple conditioned Latin Hypercube method (cLHm) and SOC stocks were measured. Also, environmental factors, including terrain attributes, moisture index, and normalized difference vegetation index (NDVI), were calculated. SOC stocks were modeled using random forest (RF) and partial least squares regression (PLSR) models. Modeling SOC stocks by RF model showed that the efficient factors for estimating the SOC stocks were slope height (slph), terrain surface texture (texture), standardized height (standh), elevation, relative slope position (rsp), and normalized height (normalh). Also, the PLSR model selected standardized height (standh), relative slope position (rsp), slope, and channel network base level (chnl base) to model SOC stocks. In both RF and PLSR methods, the standh and rsp factors were suitable parameters for estimating the SOC stocks. Predicting the spatial distribution of SOC stocks using environmental factors showed that the R2 values for RF and PLSR models were 0.81 and 0.40, respectively. The result of this study showed that in areas with complex land features, terrain attributes can be good predictors for estimating SOC stocks. These predictors allow more accurate estimates of SOC stocks and contribute considerably to the effective application of land management strategies in arid and semiarid area. [ABSTRACT FROM AUTHOR]

Identifying areas prone to flooding is a key step in flood risk management. The purpose of this study is to develop and present a novel flood susceptibility model based on Bayesian Additive Regression Tree (BART) methodology. The predictive performance of the new model is assessed via comparison with the Naïve Bayes (NB) and Random Forest (RF) based methods that were previously published in the literature. All models were tested on a real case study based in the Kan watershed in Iran. The following fifteen climatic and geo-environmental variables were used as inputs into all flood susceptibility models: altitude, aspect, slope, plan curvature, profile curvature, drainage density, distance from river distance from road, stream power index (SPI), topographic wetness index (TPI), topographic position index (TPI), curve number (CN), land use, lithology and rainfall. Based on the existing flood field survey and other information available for the analyzed area, a total of 118 flood locations were identified as potentially prone to flooding. The data available were divided into two groups with 70% used for training and 30% for validation of all models. The receiver operating characteristic (ROC) curve parameters were used to evaluate the predictive accuracy of the new and existing models. Based on the area under curve (AUC) the new BART (86%) model outperformed the NB (80%) and RF (85%) models. Regarding the importance of input variables, the results obtained showed that the location's altitude and distance from the river are the most important variables for assessing flooding susceptibility. [ABSTRACT FROM AUTHOR]