Multi-hazards modeling using machine learning algorithms in Southwestern Saudi Arabia

Multi-hazards modeling using machine learning algorithms in Southwestern Saudi Arabia | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Multi-hazards modeling using machine learning algorithms in Southwestern Saudi Arabia Ahmed M. Youssef, Ali M. Mahdi, Mohamed M. Al-Katheri, Soheila Pouyan, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-1554302/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The current study aimed at producing a multi-hazard risk map for Hasher-Fayfa Basin. The Basin is part of Jazan region in the southwestern Saudi Arabia and is distinguished by mountainous terrain. Recently, this area has experienced many extreme natural processes that become natural hazard events when it intersects with human activities (urban areas and infrastructures). This work is mapping the probabilities of three main hazards; landslides, floods, and gully erosion using machine learning algorithms named boosted regression tree (BRT), a generalized linear model (GLM), Flexible discriminant analysis (FDA), random forest (RF), and multivariate discriminant analysis (MDA). Several factors obtained from various sources, including topographical, geological, meteorological, hydrological, and human activities were incorporated to produce the final multi-hazard risk model. Area under the curve (AUC) was applied to identify the best predictive model for each natural hazard type. AUC values between 80 and 90% indicated that the model is very good and above 90% indicated that the model is excellent in predictive capability. Based on the accuracy evaluation, it was found that the FDAFDA model is the most accurate for predicting landslides with AUC values of (92.7%, excellent performance), RF model is the most accurate for predicting floods and erosion with AUC values of (97.2%-excellent and 83.3%-very good performance, respectively). Finally, multi-hazard risk map was prepared by coupling of the mentioned three hazards. Results showed that 33.5% of the total area is safe (no-hazard), whereas 66.5% is characterized by at least a single hazard and 2–32 − 3 hazard combination. Machine learning approaches are useful tools as baselines for management and mitigation processes, based on multi-hazard modeling. It could help planners and decision-makers to manage future human activities and expansions. Multi-hazards Susceptibility Future development Sustainability KSA Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 1. Introduction Managing disasters is challenging due to various kinds of hazards (e.g., natural and anthropogenic hazards) (De Silva and Kawasaki, 2018 ). Natural processes will become natural hazards, when they intersect and have a negative effect on people life (Gill and Malamud, 2017 ; Ward et al., 2020 ). Subsequently, natural hazards are the physical events that pose devastating impacts on communities, damaging critical infrastructure, affecting economic viability, and claiming lives (IPCC, 2007 ). Worldwide, an exponential increase in the human population leads to lateral expansion in development (urbanization, agricultural activities, infrastructures, and life lines), subsequently increasing the exposure and interaction with different types of natural hazards (Weinkle et al., 2018 ). Globally, several areas are vulnerable to events involving several types of hazards that are related to the same area and at the same time (Leonard et al., 2014 ). Natural hazards have impacted people and natural environments in many undeveloped countries more severely than developed countries, causing substantial loss of lives and enormous economic plunge (Iglesias et al., 2021 ). In a simple definition, a hazard refers to the probability of potential damage in certain areas (location or where?) and a specific period (time or when?) due to an event with specific magnitude (how large?) (Shi, 2019). Arid and semi-arid areas around the world are impacted by different natural hazards events, ranging from insignificant influence events to extremely events (IPCC, 2012 ; Tabari 2020 ; Rutgersson et al., 2021 ). These natural hazards frequently occur, threatening everything (modern cities, ruler areas, infrastructures, and life lines) (De Silva and Kawasaki, 2018 ). These disasters can cause great economic damage, disruption of transportation systems, injury, and claiming lives. Mountainous provinces are highly likely among the most disaster-susceptible areas due to their various characteristics such as lithology, tectonics, climate, and hydrology (Baig et al., 2021 ). Risk related to natural hazards is high in middle East countries due to lack of disaster preparedness and public awareness, and inadequate funding support (AlQahtanya and Abubakarb, 2020 ). Saudi Arabia is one of these countries that experienced many natural hazards, yearly (Alyami et al., 2021 ). These events cost lives and bring the economy to a static state. Moreover, not much has been done to map different prone areas related to these natural hazards. Mountainous regions that are inhabited by people and crossed by infrastructures are prone to not only one type of hazard, but they are frequently the scene of multiple disasters that interact together, such as earthquakes, avalanches, landslides, floods, mudslides, ground subsidence, soil erosion, and wildfires (Gill and Malamud, 2014 ; Shah et al., 2018 ). Duncan et al. ( 2016 ) defined multi-hazard as all potential and cascading hazards, in any particular area at a certain period. Most of the time, these hazards can cause severe damage to the various human activities located and intercept with the hazard prone areas by causing fatalities and injuries, damaging their urban and industrial areas, destroying infrastructures (roads, tunnels, and railways), and disrupt lifelines (power lines, water systems, and gas mains) (Bell and Glade, 2004 ). Due to the complexity of natural hazards, many studies deal with individual hazard types as an independent approach (Wastl et al., 2011 ). Also, problems related to these problems (e.g., floods, landslides, and gully erosions) have been evacuated individually by applying various machine learning algorithms-MLAs (Sarkar and Mishra, 2018 ; Ghorbanzadeh et al., 2019 ; Hosseiny et al., 2020 ; Zhou et al., 2021 ). However, Earth system science approach indicates that a significant interaction between different types of hazard components due to the interaction between the component systems (e.g., the lithosphere, atmosphere, hydrosphere, and biosphere). Accordingly, a holistic approach to understand all different hazards in a certain area is a must and can prevent cascading hazards that can be formed by the interaction of different hazard types. A multi-hazard risk evaluation could be significant to control hazard interactions (Komendantova et al., 2014 ). To prevent these natural hazards and protect people and their properties and the country economy from long-term plunging, effective predictive models must develop (Bathrellos et al., 2017 ). These models should base on profound understanding of the most influential and triggering factors that have a significant contribution on these hazards (van Westen and Greiving, 2017 ). Multi-hazard modeling becomes an essential tool in land use development at both regional and national scales (Saunders and Kilvington, 2016 ). These multi-hazard modeling approaches gain good attention recently worldwide based on their ability to consider different types of hazards that could impact the area (Schmidt et al., 2011 ; Skilodimou et al., 2019 ; Lombardo et al., 2020 ). Different approaches were used to conduct multi-hazards modeling e.g., by using two decision-making tools including, sequential Monte Carlo method and decision-making tool (Komendantova et al., 2014 ); using a deterministic equation based (theoretical) and empirical understanding (Bout et al., 2018 ); and by using multi-criteria analysis and GIS (Skilodimou et al., 2019 ). Recently, MLAs have been widely used to predict various hazards based on random forest (RF), and support vector machine-SVM (Nachappa et al., 2020 ), boosted regression tree-BRT-, and generalized additive model-GAM (Ye et al., 2020 ). In this study, no multi-hazard assessment has been done in Saudi Arabia before and the application of machine learning techniques in multi-hazard assessment will be a novel study in the area. In this study, developing a multi-hazard risk map using machine learning techniques for the most impacted natural hazards (landslides, floods, and gully erosion) in this area are crucial for effective landuse management. The current study evaluates machine learning models (MDA, GLM, FDA, BRT, and RF) as effective and accurate models to produce multi-hazard risk map for Hasher-Fayfa Basin that might be used by authorities, developers, and decision-makers. 2. Study Area The Jazan province has an area of 11,671 km 2 , is situated at the southwestern corner of Saudi Arabia, bordering the Red Sea to the west and Asir region to the north (Fig. 1 a). The Jazan region includes the study area that is bounded by latitudes 17°10' to 17°30' N and longitudes 42°58' and 43°13' E. Hasher-Fayfa Basin covers 513 Km 2 . Its elevation ranges from 220 m to 2,340 m above sea level (e.g. Fayfa Mountain up to 2.3 km high). This makes it the one of the highest areas in KSA. The population of the Jazan region is about 1.67 million persons in 2020 ( https://knoema.com/atlas/Saudi-Arabia/Jazan/Total-Population ). The area under study inhabited by many people who live in the mountainous areas as in (Fayfa Mountain, Fig. 1 (b-c)). Since about six decades, Saudi Arabia was discovered the oil fields, many urban centers and infrastructures were established. The Jazan area has been expanded drastically. The area has all the elements that could put it under a high risk of floods, landslides, and erosion that could impact the economic development of the region. These hazards will pose a high risk to people and their property (buildings, highways, roads, infrastructure, and lifelines). The basin is characterized by a rugged landscape with the iconic features of the Hasher and Fayfa mountains. These mountains have steep reliefs (up to 88 ◦ ). Rainfall is the main trigger factor for landslide, flood, and erosion in the area under consideration. The region receives up to 500 mm/year of precipitation, with a high rainfall value of 1,400 mm in 1979 (Hasanean and Almazroui 2015 ). The high rainfall in the region is mainly due to the tropical air masses affecting the Jazan and Asir regions (the southwestern provinces of KSA) and the high altitudes that produce orographic rainfall (Şen et al., 2017 ). Additionally, the area is distinguished by the presence of intense tectonic deformation (folds, faults, shear zones, fractures, and joints). The Red Sea Mountains of Saudi Arabia consists of extrusive and intrusive igneous rocks, and metamorphic rocks. These rocks were formed by the collection of island arcs and closure of the interleaving oceanic arcs 700 − 1,000 Ma (Greenwood, 1982 ; Stoeser and Camp, 1985 ). These mountains were raised up due to the spreading of the Red Sea about thirty million years ago. These Precambrian rocks are unconformably overlain by sedimentary sequences of Paleozoic and Cenozoic age (Agar, 1987 ). 3. Methodology The current work involved five steps (Fig. 2 ): (1) data collection of the various events based on intensive field work, analysis of various civil defense reports, technical reports, and questionnaires with the local population (for the period between 2000 and 2020); (2) determine the various events based on a through literature review; (3) model different hazard types using various machine learning techniques (; (4) model validation and selection of the most appropriate model for each hazard type; and (5) generation of multi-hazard risk map (MHRM) by integrating the high-AUC model for each hazard. 3.1. The inventory of hazard data (multi-hazards inventory) Creating an inventory of susceptible landforms is an essential and crucial stage in hazard susceptibility mapping. The inventory of landslides, gully erosion, and flooding in the watershed of the study area was compiled based on field investigations with the help of GPS instrument (732 sites were investigated through the field work between 2015 and 2018) (Fig. 3 ). National and regional data from various sources, including the Civil Defense, Jazan Region Authority, Ministry of Transportation, Saudi Geological Survey, review of technical reports and scientific publications and the private sector of the Hasher and Fayfa region were collected and reviewed. As some of the vulnerable areas are located in mountainous areas and may have been overlooked during the field survey; so, high-resolution Google Earth images were also used to detect landslides, gully erosion features, and flooded areas. Based on this information, 100 landslide locations, 60 sites for gully erosion, and 70 sites for flood locations were used to establish the multi-hazard inventory map (Fig. 1 a). In this study, we used a hazard and non-hazard locations and machine learning techniques. Accordingly, equal numbers of hazards and non-hazardous locations were randomly selected (Fang et al., 2021 ). A random partitioning method was used to select training and validation sites for each hazard type in the inventory map. In the current work, training datasets (70% of each hazard) were used for model generation and the remaining 30% of hazard locations were used for model validation (Zhao et al., 2018 ; Wang et al., 2020 ). 3.2. Hazard-predictive factors (HPFs) It indicated that hazard-predictive factors (HPFs) are the first step of the modeling process (Rusk et al., 2022 ). Accordingly, intensive review of previous studies was done to identify the most effective factors for each hazard type (Roy and Saha, 2019 ; Razavi-Termeh et al., 2020 ; Wang et al., 2020 ). Raster database layers of different topography, climatology, hydrology, vegetation and landuse/landcover, geology, and anthropogenic were selected and prepared using ArcGIS 10.8 of 5 × 5 m resolution (Table 1 , Fig. 4 ). These factors include 16 HPFs; elevation (Fig. 4 a), slope angle (Fig. 4 b), topographic wetness index (TWI – Eq. 1 ) (Beven and Kirkby, 1979) (Fig. 4 c), plan curvature (Fig. 4 d), aspect (Fig. 4 e), topographic roughness index (TRI - Eq. 2) (Nellemann and Reynolds, 1997) (Fig. 4 f), profile curvature (Fig. 4 g), and valley depth (VD) (Fig. 4 h) were extracted from DEM 5 m resolution that were generated using 1:10,000 topographic map; lithology (Fig. 4 n) and distance to fault (Fig. 4 o) were extracted from a geologic map of 1:250,000 scale obtained from Saudi Geological Survey; rainfall distribution map was generated from 20 rainfall stations (acquired from the MEWA, covering a period from 1965 to 2020) (Fig. 4 i), distance to wadis (Fig. 4 j) and stream density ((Fig. 4 k) were generated from topographic map 1:10,000 scale and converted to distance to wadis and stream density using Euclidian distance tool and density tools in ArcGIS 10.8; distance to road (Fig. 4 p) was extracted from high resolution images (Google Earth, then converted to distance to road using Euclidian distance tool in ArcGIS 10.8, landuse/land cover (LULC) (Fig. 4 m) and normalized differential vegetation index (NDVI) (Fig. 4 l) were generated from sentenil-2 image (10 m resolution) acquired in May 2020, LULC map was generated using maximum likelihood classification method, and finally NDVI was extracted using NDVI equation (Eq. 3 ). Shalaby and Tateishi ( 2007 ) pointed out that the maximum likelihood method is a stable and accurate model for image classification because it can apply multiple statistical features in its work. Table 1 The most influencing indicators for multi-hazard susceptibility mapping. Factor Type Hazard-indicator factors and abbreviation Data type Factor range /Units Hazard type Landslide (LS) Flood (FL) Gully erosion (GE) Topography Elevation (EL) Grid 220-2,340 m Ꚛ Ꚛ Ꚛ Slope (S) Grid 0–88 degree Ꚛ Ꚛ Ꚛ Topographic wetness index (TWI) Grid -1.78–24.74 Ꚛ Ꚛ Ꚛ Plan curvature (PC) Grid -271.5–1452.2 Ꚛ Ꚛ Ꚛ Aspect (A) Grid 9 directions Ꚛ Ꚛ Ꚛ Terrain ruggedness index (TRI) Grid 0–174 Ꚛ Profile curvature (PrC) Grid -1627.8–917.3 Ꚛ Valley depth (VD) Grid 0–261 m Ꚛ Climatology Rainfall (R) Point 383–495 mm Ꚛ Ꚛ Hydrology Distance to wadis (DtW) Polyline 0–2,900 m Ꚛ Ꚛ Stream density (SD) Polyline 0.95–23.7 km/km 2 Ꚛ Ꚛ Vegetation and landuse/land cover Normalized difference vegetation index (NDVI) Grid -0.639–0.675 Ꚛ Ꚛ Ꚛ Land use/Landcover (LULC) Grid 5 classes Ꚛ Ꚛ Ꚛ Geology Lithology (Lth) Polygon 6 groups Ꚛ Ꚛ Ꚛ Distance to fault (DtF) Polyline 0–3,153 m Ꚛ Anthropogenic Distance to road (DtR) Polyline 0–5,780 m Ꚛ Ꚛ All, layers were converted to have 5 m * 5 m pixel size using ArcGIS 10.8. Table (1) shows that different factors used to prepare hazard susceptibility maps; 11 factors used for landslides, 12 factors were used for floods, and 13 factors were used for gully erosion. $$TWI=lin \frac{\text{A}}{\text{tan}B}$$ 1 Where, A is the cumulative basin area (m 2 ), and β is the slope (in degrees) at a point. TRI = \({\text{Y}\left[\sum {({x}_{ij}-{x}_{00})}^{2}\right]}^{1/2}\) (2) Where, xii is the height of each pixel adjacent to the pixel (0, 0). Areas with a slope of 0 have a TRI value of zero, while the roughened areas with steep elevations have positive values of TRI. $$NDVI=\frac{(NIR-R)}{(NIR+R)}$$ 3 Where, NIR is the near infrared band (wave length = 832.8 nm) and R is the red band (wave length = 664.6 nm) of the spectral bands. 3.3. Machine learning algorithms In the current work, five machine learning algorithms (MLAs) were used to generate landslide, flood, and gully erosion susceptibility maps that affect the study area. Various literatures were conducted using various MLAs to overcome the problems related to natural hazards by predicting their susceptibility, such as landslides (Park and Kim, 2019 ; He et al., 2021 ), floods (Janizadeh et al., 2019 ; El-Haddad et al., 2021 ), and gully erosion (Ghorbanzadeh et al., 2020 ; Amare et al., 2021 ; Yang et al., 2021 ). A detailed description of the MLAs that used in this study is given in the following parts: 3.3.1. Random forest Random Forest (RF) is an ensemble classification technique (Breiman, 2001 ) and considered as one of the learning techniques that involves multiple steps, starting with training datasets, followed by bootstrapping, then ensemble of trees, and finally aggregation (classification phase) (Hawryło et al., 2018; Sarker et al., 2019 ). Bootstrapping refers to the parallel training of each decision tree with other training subsets using different available features (Herrera et al., 2019 ). In this phase, each individual decision tree is unique and is used to minimize the overall variance of the RF model. In the last stage, the RF model summarizes the decisions of the individual trees; therefore, the RF model shows good generalization. The RF algorithm tends to outperform most other classification approaches in terms of accuracy, with no problems of overfitting (Pedregosa et al., 2011 ). The RF algorithm is robust to training sample selection and noise in datasets and does not require feature scaling (fits both categorical and continuous values) (Kim et al., 2018). RF can overcome outliers in predictors, automatically handle omitted data and increase the diversity of classification trees. 3.2. Multivariate discriminant analysis Multivariate discriminant analysis (MDA) is a trained classification algorithm as an extension of linear discriminant analysis (LDA). It can evaluate the multivariate nonlinear relationships between different classes within each group (Lombardo et al. 2006 ). To calculate the distance to the nearest cluster, the normal distribution (the variability and correlation between variables is uniform) of each class is used (Lombardo et al., 2006 ). It can be derived using Eq. ( 4 ) (Hair et al., 1998). $${D}_{v}={D}_{w1}{V}_{1}+{D}_{w1}{V}_{1}+\dots +{D}_{wn}{V}_{n}$$ 4 \({D}_{v}\) and \({D}_{w1}\) (i = 1,2,3, ..., n) represent the discriminant value and weights, respectively, and \({V}_{1}\) (i = 1,2,3,..., n) are independent indicators. The MDA model was run using “mda” package in R software (Hastie et al., 2017 ). 3.3.3. Generalized linear model The generalized linear model (GLM) is a linear regression model that can quantify and incorporate specific and temporal factors (Goetz et al., 2015 ). The GLM can scale the accuracy and quality of the outcomes by using multiple regression to establish a unique link between the dependent and independent factors (Scott et al., 1991). By identifying the best regression model, multiple events can be predicted (Payne, 2015). It is an adequate model for numeric factors because of its regression characteristic; however, insignificant and correlated factors can reduce its performance, it also, deal with non-normal distribution data (Kalantar et al., 2020 ). The GLM linkage function is used to establish the relationship between the dependent variable and the independent variables (Soch et al., 2017). In this study, the occurrence probability of hazard event Y can be represented by Eq. ( 5 ). By the logistic transformation, the link function g(y i ) is represented by Eq. 6 : $$P=\frac{\text{e}\text{x}\text{p}({c}_{0}+{c}_{1}{X}_{1}+{c}_{2}{X}_{2}+\dots +{c}_{i}{X}_{i})}{1+\text{e}\text{x}\text{p}({c}_{0}+{c}_{1}{X}_{1}+{c}_{2}{X}_{2}+\dots +{c}_{i}{X}_{i})}$$ 5 $$g\left({y}_{i}\right)={c}_{0}+\sum {c}_{i}{x}_{i}+ {\epsilon }_{i}$$ 6 where P is the event “Y” probability; \({c}_{0}\) is the intercept; \({c}_{1}\dots .{c}_{i}\) . .. are logistic regression coefficients, \({X}_{1}\dots .{X}_{i}\) are the independent variables; and the error residual expressed by ε i . In this study, R software was used to construct the GLM model (R Core Team, 2019 ). A simple Gaussian family is determined as the link function for normally distributed response data. The independent indicators were to be included in the model individually, using a smoothing spline based on two degrees of freedom to avoid overfitting (Aertsen et al., 2009). 3.3.4. Flexible discriminant analysis Ramsay and Dalzell (1991) proposed the Flexible discriminant analysis (FDA) as a nonlinear statistical classification method. The crucial concept of FDA is to treat an observed object with functional characteristics as an integral, regardless of the order of the observed values (Wagner-Muns et al., 2018). It can distinguish unsupervised work where each class is divided into subclasses with a unique value (Zou et al., 2019). FDA can provide a framework by coupling various models and methods (regression algorithms, discriminant analysis, and classification techniques) (Hastie and Tibshirani, 1996 ). In this study, the FDA technique was used to create a hazard susceptibility map using the species distribution package in R software (Naimi and Araújo, 2016). 3.3.5 Boosted regression trees BRT is a combination of regression and boosting techniques. It was proposed by Friedman ( 2001 ). The boosting algorithm is used to improve model accuracy, where the residual errors of the existing tree will be used to adjust the new trees (Ye et al., 2009 ). Shin et al. ( 2012 ) pointed out that the BRT model has several advantages because it can be improved by fitting and combining multiple methods and increasing the number of trees, it uses different types of data without conversion or outlier removal, it works when some data are missing, and it can use sophisticated nonlinear relationships to fit the data. The BRT model is defined by Eq. ( 7 ) (Friedman 2001 ). $$F(X;{\left[{B}_{m}{a}_{m}\right]}^{{m}_{o}}={\sum }_{m=0}^{m}{B}_{m}h(X;{a}_{m})$$ 7 Where is \(h(X;m)\) a simple classification function with variable “a” and factors “x.” Factor “m” is representing the step of the algorithm, and is \({B}_{m}\) the weighting coefficient at step m. A new tree will be added to the original model at each iteration step, which leads to reducing of the loss function. The BRT training stage will be ended as soon as the predefined iteration numbers achieved. The BRT model was run using the package "brt" in R software (Ridgeway and Southworth 2013 ). 3.4. HFIs multicollinearity and importance, and model validation Multicollinearity method is used to evaluate variables effectiveness in different model construction (Yoo and Cho 2019 ). In this method, two indices were used, variance inflation factors (VIF) and tolerance (TOL) as shown in equations ( 8 and 9 ): $$\text{T}\text{O}\text{L}=(1-{R}_{j}^{2})$$ 8 $$\text{V}\text{I}\text{F}=\left[\frac{1}{(1-{R}_{j}^{2})}\right]$$ 9 \({R}_{j}^{2}\) Represents the regression coefficient of the explanatory factor J on all other descriptive factors. Studies have shown that a TOL 5 account for multicollinearity issues (Rahman et al., 2019 ). Additionally, the importance of independent indicators is essential. It measures their contribution in each hazard type modeling. Several techniques were used among them the learning vector quantization (LVQ), and random forest algorithm (RF) (Zhang and Xie, 2012 ; Sun et al., 2020 ). In this study, HIFs importance and contribution to landslide, floods, and gully erosion occurrence were assessed using RF. Finally, to evaluate the accuracy of the results, the receiver operating characteristics (ROC) and the area under the curves (AUC) can also be applied to evaluate the MLMs that used to provide valuable information (Bradley 1997 ). Without validated maps, the models and their results would have no scientific significance or if the AUC is less than 0.5, the model is considered a random model (Marzban, 2004). Evaluation of the accuracy of the different machine learning models for each hazard type is a profound step to create the final hazard model (Amare et al., 2021 ). In this study, the performance of all hazard models was determined using the validation dataset of each hazard type. This will provide quantitative measures of the model accuracy. The predictive performance was done using the AUC. AUC value ranges from 0 to 1 (0 to 100%). A range of 1 (100%) represents high performance, while a range of ≤ 0.5 (50%) indicates a poor classification. Evaluation of the accuracies of these models based on (Guzzetti et al., 2005 ). They divided the AUC values into three groups: between 0.75 and 0.8 for an acceptable model, between 0.8 and 0.9for a good susceptibility model, and above 0.9 for an excellent model. 4. Results And Analysis 4.1. Multicollinearity test The multicollinearity analysis of the hazard-predictor factors used in this study for landslide susceptibility (11 factors), flood susceptibility (12 factors), and gully erosion susceptibility (13 factors) is shown in Table (2). Results showed that both tolerance (ToL) and (VIF) values of landslide-predictor factors used in this study to run the different machine learning models were appropriate, ToL values were more than 0.1 (the lower ToL value is 0.20) and the VIF values were less than 5 (the large VIF value is 4.9) for slope factor. The tolerance (ToL) and (VIF) values of flood-predictor factors were appropriate, ToL values were more than 0.1 (the lower ToL value is 0.257) and the VIF values were less than 5 (the large VIF value is 3.898) for elevation factor. For the gully erosion predictor factors, all factors are adequate except the TRI that provide VIF value more than 5. For the rest of these factors the ToL values were more than 0.1 (the lower ToL value is 0.271) and the VIF values were less than 5 (the large VIF value is 3.697) for the slope facto. Accordingly, the results showed that there was no multicollinearity among the selected FCFs, which means that they made a significant contribution to the model construction in this study. 4.2. The importance of hazard-predictors The analysis of landslide hazard-predictor importance is shown in Figure (5a). Our findings indicated that the top predictors that controlled landslide susceptibility are the slope angle (S), followed by land use/land cover (LULC), elevation (EL), the normalized differential vegetation index (NDVI), followed by aspect (As), and distance to wadis (DtW), distance to road (DtR), topographic witness index (TWI) and distance to fault (DtF). Less importance predictor is lithology (Lth). We believed that this is due to the higher slope angle for the mountainous areas and the gravity forces play a great role than material types (Çellek, 2020 ). However, negative importance factor comes to be plan curvature (PC). Gully erosion hazard-predictor importance is shown in Figure (5b). Results showed that the top predictors-controlled gully erosion susceptibility is TWI followed by VD, EL, NDVI, SD, PrC, S, and R. The less important predictors are aspect (AS), TRI, and PC, followed by LULCs, and Lth. The analysis of flood hazard-predictor importance is shown in Figure (5c). Results showed that the top predictors that controlled flood susceptibility belonged to the hydrologic group, which was DtW, followed by equal importance of slope angle (S), elevation (EL), and stream density (SD), then followed by both TWI and rainfall (R), followed by normalized differential vegetation index (NDVI) and plan curvature (PC). Less importance predictors are LULCs, aspect (AS), and lithology (Lth). However, negative importance factor comes to be distance to road (DtR). 4.3. Different susceptibility maps Maps of landslides, flood, and gully erosion susceptibility were generated based on both hazard-predictive factors (independent factors) and actual hazard locations (dependent variable) by applying various types of machine learning models (MLMs). Each susceptibility map was divided into five classes very low, low, moderate, high, and very high based on Natural Breaks Classifier-Jenks (Jenks and Caspall, 1971 ) (Fig. 6 ). Landslide susceptibility maps for the Hasher-Fayfa Basin were created on the basis of RF, FDA, MDA, GLM, and BRT models (Fig. 6 (a-e)); gully erosion susceptibility maps for the Hasher-Fayfa Basin were created according to the RF, FDA, MDA, GLM, and BRT models (Fig. 6 (f-j)); and flood susceptibility maps were created using RF, FDA, and MDA models (Fig. 6 (k-m)). Detailed analysis was conducted for each model. Landslide hazard models indicated that high and very high susceptible regions based on RF and BRT covered more areas (21.2 and 22.5%, respectively) than the other three models (FDA = 17.8%, MDA = 18.6%, and GLM = 17.6%) (Fig. 7 a). Gully erosion models indicated that high and very high susceptible classes in MDA and BRT, covering more areas (37.5 and 34.2%, respectively) than the other three models (RF = 32.8%, FDA = 31.4%, and GLM = 32.6%) (Fig. 7 b). Flood models showed that high and very high susceptible classes in the MDA and FDA, covering more areas (15.0 and 14.1%, respectively) than RF model (RF = 6.5%) (Fig. 7 c). 4.4. Models’ validation Figure 8 (a) and Table (3) show the AUC curves and accuracy values for the landslide’s five ML models. Results show that landslides susceptibility models give high performance where AUC > 0.9 (90%), the AUC values for ML models BRT, RF, GLM, MDA, and FDA are 0.909 (90.9%), 0.918 (91.8%), 0.923 (92.3%), 0.923 (92.3%), and 0.927 (92.7%), respectively. Results show that the most accurate ML model that has a high predictive accuracy for landslide susceptibility is FDA model with AUC value of 92.7%. Figure 8 (b) and Table (3) show the AUC curves and values for the five ML models that were applied to create the gully erosion susceptibility models. Results show that gully erosion susceptibility models give AUC values range between 0.739 and 0.833 (medium to high performance). The ML models MDA, GLM, FDA, BRT and RF give AUC values of 0.739 (73.9%), 0.776 (77.6%), 0.792 (79.2%), 0.827 (82.7%), and 0.833 (83.3%), respectively. Results show that the most accurate ML model that has high predictive accuracy for gully erosion susceptibility is RF model with AUC value of 83.3%. Figure 8 (c) and Table (3) show the AUC curves and values for the three ML models that were used to generate the flood susceptibility maps. Results show that flood susceptibility models give AUC values range between 0.932 and 0.972 (high performance). The ML models MDA, FDA, and RF give AUC values of 0.932 (93.2%), 0.945 (94.5%), and 0.972 (97.2%), respectively. Results show that the most accurate ML algorithm that has high predictive performance for flood susceptibility is RF model with AUC value of 97.2%. 4.5. Multi-hazard risk map To produce multi-hazard risk map for these three hazard types, three steps were applied: 1) hazard susceptibility map to each hazard type (LS, GE, and FL) was created according to the relationship between the independent factors (hazard-predictors) and the dependent factors (hazard locations of landslides, gully erosions, and floods) using various types of machine learning techniques (Fig. 6 ), 2) the ROC-AUC method was used to determine the highest performance models for each hazard type and then integrated to map multiple hazards (Table 2 ). The most accurate models in this study are FDA model for landslides where AUC = 92.7%, RF for gully erosion with AUC = 83.3%, and RF for floods with AUC = 97.2%. These three susceptibility maps of the three natural hazards (landslides, gully erosion, and floods) were combined in ArcGIS 10.8 to produce an integrated multi-hazard risk map. Table 2 Multicollinearity results of hazard-predictor factors. Hazard-causative factors LS FL GE Tolerance VIF Tolerance VIF Tolerance VIF Elevation (EL) 0.290 3.448 0.257 3.898 0.476 2.101 Slope (S) 0.200 4.900 0.316 3.168 0.271 3.697 Topographic wetness index (TWI) 0.307 3.258 0.304 3.290 0.326 3.067 Plan curvature (PC) 0.869 1.151 0.926 1.080 0.446 2.243 Aspect (A) 0.677 1.477 0.897 1.115 0.843 1.186 Terrain ruggedness index (TRI) - - - - 0.020 49.302 Profile curvature (PrC) - - - - 0.523 1.911 Valley depth (VD) - - - - 0.502 1.991 Rainfall (R) - - 0.719 1.391 0.716 1.396 Distance to wadis (DtW) 0.461 2.168 0.398 2.511 - - Stream density (SD) - - 0.723 1.384 0.613 1.632 Normalized difference vegetation index (NDVI) 0.382 2.620 0.460 2.176 0.699 1.430 Land use/Landcover (LULC) 0.660 1.514 0.741 1.350 0.703 1.422 Lithology (Lth) 0.718 1.393 0.586 1.705 0.672 1.488 Distance to fault (DtF) 0.842 1.188 - - - - Distance to road (DtR) 0.767 1.304 0.705 1.418 - - Table 3 AUC values for different machine learning models used in modeling landslide, gully erosion, and flood. Model Type Test Result Variable(s) Area (AUC) Standard Errors Landslides BRT 0.909 0.054 RF 0.918 0.052 GLM 0.923 0.051 MDA 0.923 0.050 FDA 0.927 0.049 Gully Erosion MDA 0.739 0.064 GLM 0.776 0.059 FDA 0.792 0.057 BRT 0.827 0.053 RF 0.833 0.051 Floods MDA 0.932 0.049 FDA 0.945 0.035 RF 0.972 0.000 Results showed that the multi-hazard risk map included eight susceptibility classes (Figs. 9 and 10 ). These classes include; no-hazard or safe areas (No-H covers 33.6% of the study area). However, about 66.4% of the total area are subjected to various types of hazards as follows: landslides (LS = 22.4%), gully erosion (GE = 28.9%), floods (FL = 1.6%), landslides-erosion (LS-GE = 6.5%), floods-landslides (FL-LS = 0.3%), floods-erosion (FL-GE = 6.5%), and flood-landslides-erosion (FL-LS-GE = 0.3%). 5. Discussion Understanding hazard management could be enhanced using different types of modeling approaches (Hill and Minsker, 2010 ). These models could provide planners and policymakers with useful, efficient, and informative results. Across the world, countless literature has individually studied different types of natural hazards (e.g., Amare et al., 2021 ; Msabi and Makonyo, 2021 ). All these studies provide important and useful results. However, most natural hazards do not occur individually due to cascading effects. One hazard may lead to another; therefore, multiple hazards and their linkages, interactions and cascading impacts can provide much understanding of their processes and optimal ideas for averting and minimizing disaster losses and for effective land management (Godschall et al., 2020 ). These maps can provide valuable information that is critical for planning and managing existing and future human activities. Dealing with multi-hazards has shown that the interaction of numerous hazard types can cause a higher risk than the risk of a single hazard type (Liu et al., 2021 ). In the current work, we investigated three hazards in a Hasher-Fayfa Basin (mountainous catchment) in Saudi Arabia. To evaluate disaster events in this area, multi-hazard modeling was performed. To achieve that five machine learning models (RF, FDA, MDA, GLM and BRT) for landslides and gully erosion and three models (RF, FDA and MDA) for floods. Based on (Guzzetti et al., 2005 ) classification, our study shows that the three best models had predictive ability of above 0.8 (good and excellent performances). The FDA model for landslides has an AUC = 92.7% (excellent performance), RF for gully erosion has an AUC = 83.3% (good performance), and RF for flooding has an AUC = 97.2% (excellent performance). The MHM map was produced by coupling the results of the FDA (for landslides) and RF (for gully erosion and flooding) approaches. Our work is in agreement with that of Nachappa et al. ( 2020 ) that used different machine learning approaches in multi-hazard evaluation. Also, our results in choosing RF for multi-hazard map that provides excellent accuracy for landslides AUC = 0.93 and for floods AUC = 0.97, are in agreement with the Nachappa et al. ( 2020 ) in which they applied multi-hazard evaluation use of random forest (RF), which provides a suitable result for both flood (AUC = 0.87) and landslide (AUC = 0.90), and SVM, which provides AUC = 0.87 for flood and AUC = 0.89 for landslides. They constructed the multi-hazard map using RF and SVM and produced an accurate model for planners and managers. The results show that no hazard areas cover approximately 33.5% of the study area. However, 66.5% of the total area is probably affected by one or more hazards. Our findings indicating the application of the optimal MLTs to predict multiple problems provides crucial informative data about their interactions. These relationships are highly correlated with the scale of the indicators used in the analysis and the specific types of hazards in the area. The current study bridged the gap between the different hazards by fully identifying the cascading interactions between these different types of natural hazards. 6. Conclusions The western and southern parts of Saudi Arabia face various natural hazards. Many mountainous areas are inhabited by people and these areas are vulnerable to several (compound) natural hazards. The delineation of high-risk areas is the critical stage and the most difficult task for all developers, and decision-makers. In the current work, we applied various machine learning techniques (MLTs) to map individual hazards affecting the Hasher - Fayfa basin (landslides, gully erosion, and flooding). The final map of this study is a multi-hazard risk map created by linking the three predominant natural hazards in the area (landslides, gully erosion, and flooding). The multi-hazard risk map is divided into 8 zones, the no hazard zone (covers 20% of the study area). Three single hazard zones namely landslide hazard zone (3%), which is confined to high altitude and steeply sloping areas, gully erosion hazard zone (2%) which is concentrated along gullies and mainly characterized by soil and weathered materials that can be easily eroded by rainfall, and flood hazard zone (4%) which is confined to main wadis and mainly downstream parts as water collects from various tributaries. Three zones include two types of interacting hazards, such as landslide-flood hazard zone (5%), landslide-erosion hazard zone (9%), and flood- erosion hazard zone (10%). Finally, there is a multiple hazard zone affected by the three hazard types landslide - erosion - flood, which covers 10% of the study area. The machine learning methods achieved acceptable accuracy in predicting the different hazard types; so, the multi-hazard map was produced with a high confidence level. Due to population growth and expansion of infrastructure and urban areas to mountainous and floodplains, and to ensure sustainable development, multi-hazard assessments are essential. Our ability to produce multi-hazard map will help planners and decision-makers make concrete management decisions for existing and future developments. This will make communities more resilient and able to act proactively to minimize future damage not only caused by a single hazard type, but also that may result from cascades of hazards or combined hazards. Declarations Authors Contributions Ahmed M. Youssef, Ali M. Mahdi, Mohamed M. Al-Katheri, Soheila Pouyan, Hamid Reza Pourghasemi designed the experiments, ran models, analyzed the results, and wrote and reviewed the manuscript. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this paper. References Agar R (1987) The Najd fault system revisited: a two-way strike-slip orogeny in the Saudi Arabian shield. J Struct Geol 9:41–48 AlQahtanya AM, Abubakarb IR (2020) Public perception and attitudes to disaster risks in a coastal metropolis of Saudi Arabia. Int J Disaster Risk Reduct 44:101422. Doi: 10.1016/j.ijdrr.2019.101422 Alyami A, Dulong CL, Younis MZ, Mansoor S (2021) Disaster Preparedness in the Kingdom of Saudi Arabia: Exploring and Evaluating the Policy, Legislative Organisational Arrangements Particularly During the Hajj Period. Eur J Environ Public Health 5(1):em0053. https://doi.org/10.29333/ejeph/8424 Amare S, Langendoen E, Keesstra S, Ploeg Mvd, Gelagay H, Lemma H, Zee SEATMvd (2021) Susceptibility to Gully Erosion: Applying Random Forest (RF) and Frequency Ratio (FR) Approaches to a Small Catchment in Ethiopia. Water 13:216. https://doi.org/10.3390/w13020216 Baig SU, Rehman MU, Janjua NN (2021) District-level disaster risk and vulnerability in the Northern mountains of Pakistan, Geomatics. Nat Hazards Risk 12(1):2002–2022. DOI: 10.1080/19475705.2021.1944331 Bathrellos GD, Skilodimou HD, Chousianitis K, Youssef AM, Pradhan B (2017) Suitability estimation for urban development using multi-hazard assessment map. Sci Total Environ 575:119–134 Bell R, Glade T (2004) Multi-hazard analysis in natural risk assessments. WIT Trans Ecol Environ 77:1–10 Bout B, Lombardo L, van Westen CJ, Jetten VG (2018) Integration of two-phase solid fluidequations in a catchment model for flashfloods, debris flows and shallow slope failures. Environ Model Softw 105:1–16. DOI. 10.1016/j.envsoft.2018.03.017 Bradley AP (1997) The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recogn 30:1145–1159 Breiman L (2001) Random forests. Mach Learn 45(1):5–32 Çellek S (2020) Effect of the Slope Angle and Its Classification on Landslide. Nat. Hazards Earth Syst. Sci. Discuss. https://doi.org/10.5194/nhess-2020-87 , 2020 De Silva MMGT, Kawasaki A (2018) Socioeconomic Vulnerability to Disaster Risk: A Case Study of Flood and Drought Impact in a Rural Sri Lankan Community. Ecol Econ 152:131–140 Duncan M, Edwards S, Kilburn C, Twigg J, Crowley K (2016) An interrelated hazards approach to anticipating evolving risk GFDRR (Ed.), The Making of a Riskier Future: How Our Decisions Are Shaping Future Disaster Risk, Global Facility for Disaster Reduction and Recovery, Washington, USA, pp. 114–121 El-Haddad BA, Youssef AM, Pourghasemi HR, Pradhan B, El-Shater A, El-Khashab MH (2021) Flood susceptibility prediction using four machine learning techniques and comparison of their performance at Wadi Qena Basin, Egypt. Nat Hazards 105, 83–114 (2021). https://doi.org/10.1007/s11069-020-04296-y Fang Z, Wang Y, Duan G, Peng L (2021) Landslide Susceptibility Mapping Using Rotation Forest Ensemble Technique with Different Decision Trees in the Three Gorges Reservoir Area, China. Remote Sens 13(2):238. https://doi.org/10.3390/rs13020238 Friedman JH (2001) Greedy function approximation: A gradient boosting machine. Ann Stat 29:1189–1232 Ghorbanzadeh O, Blaschke T, Gholamnia K, Meena SR, Tiede D, Aryal J (2019) Evaluation of different machine learning methods and deep-learning convolutional neural networks for landslide detection. Remote Sens 11(2):196. https://doi.org/10.3390/rs11020196 Ghorbanzadeh O, Shahabi H, Mirchooli F, Kamran KV, Lim S, Aryal J, Jarihani B, Blaschke T (2020) Gully erosion susceptibility mapping (GESM) using machine learning methods optimized by the multi–collinearity analysis and K-fold cross-validation. Geomatics Nat Hazards Risk 11(1):1653–1678. DOI: 10.1080/19475705.2020.1810138 Gill JC, Malamud BD (2014) Reviewing and visualizing the interactions of natural hazards. Rev Geophys 52(4):680–722. https://doi.org/10.1002/2013RG000445 Gill JC, Malamud BD (2017) Anthropogenic processes, natural hazards, and interactions in a multi-hazard framework. Earth Sci Rev 166:246–269. https://doi.org/10.1016/j.earscirev.2017.01.002 Godschall S, Smith V, Hubler J, Kremer P (2020) A Decision Process for Optimizing Multi-Hazard Shelter Location Using Global Data. Sustainability 2020, 12, 6252. https://doi.org/10.3390/su12156252 Goetz JN, Brenning A, Petschko H, Leopold P (2015) Evaluating machine learning and statistical prediction techniques for landslide susceptibility modeling. Comput Geosci 81:1–11 Greenwood WR (1982) Late Proterozoic Island-Arc Complexes and Tectonic Belts in the Southern Part of the Arabian Shield. Ministry of Petroleum and Mineral Resources, Riyadh, Saudi Arabia Guzzetti F, Reichenbach P, Cardinali M, Galli M, Ardizzone F (2005) Probablistic Landslide Hazard Assessment at the Basin Scale. Geophys J Roy Astron Soc 72:272–299. https://doi.org/10.1016/j.geomorph.2005.06.002 Hasanean H, Almazroui M (2015) Rainfall: Features and Variations over Saudi Arabia. Rev Clim 3(3):578–626. https://doi.org/10.3390/cli3030578 Hastie T, Tibshiran R, Leisch F, Hornik K, Ripley BD (2017) Mixture and flexible discriminant analysis. (2017). https://cran.r-project.org/web/packages/mda/mda.pdf Hastie T, Tibshirani R (1996) Discriminant Analysis by Gaussian Mixtures. J. R. Stat. Soc. Ser. B 1996, 58, 155–176 He Q, Jiang Z, Wang M, Liu K (2021) Landslide and wildfire susceptibility assessment in southeast asia using ensemble machine learning methods. Remote Sens 13:1572 Herrera VM, Khoshgoftaar TM, Villanustre F, Furht B (2019) Random forest implementation and optimization for Big Data analytics on LexisNexis’s high performance computing cluster platform. J Big Data 6:68. https://doi.org/10.1186/s40537-019-0232-1 Hill DJ, Minsker BS (2010) Anomaly detection in streaming environmental sensor data: A data-driven modeling approach. Environ Model Softw 25:1014–1022 Hosseiny H, Nazari F, Smith V, Nataraj C (2020) A Framework for Modeling Flood Depth Using a Hybrid of Hydraulics and Machine Learning. Sci Rep 10, 8222 (2020). https://doi.org/10.1038/s41598-020-65232-5 Iglesias V, Braswell AE, Rossi MW, Joseph MB, McShane C, Cattau M, Koontz MJ, McGlinchy J, Nagy RC, Balch J, Leyk S, Travis WR (2021) Risky Development: Increasing Exposure to Natural Hazards in the United States. Earth's Future, https://doi.org/10.1029/2020EF001795 IPCC (2007) Climate change 2007. Synthesis report: Contribution of Working Groups I, II, and III to the fourth assessment report of the Intergovernmental Panel on Climate Change. Cambridge University Press. Retrieved from http://www.ipcc.ch/pdf/assessment-report/ar4/wg2/ar4-wg2-ts.pdf IPCC (2012) Managing the risks of extreme events and disasters to advance climate change adaptation: special report of the intergovernmental panel on climate change. The Intergovernmental Panel on Climate Change Janizadeh S, Avand M, Jaafari A, Phong TV, Bayat M, Ahmadisharaf E, Prakash I, Pham BT, Lee S (2019) Prediction Success of Machine Learning Methods for Flash Flood Susceptibility Mapping in the Tafresh Watershed, Iran. Sustainability 2019, 11, 5426. https://doi.org/10.3390/su11195426 Jenks GF, Caspall FC (1971) Error on choroplethic maps: Definition, measurement, reduction Ann. Assoc Am Geogr 61(2):217–244 Kalantar B, Ueda N, Saeidi V, Ahmadi K, Halin AA, Shabani F (2020) Landslide Susceptibility Mapping: Machine and Ensemble Learning Based on Remote Sensing Big Data. Remote Sens. 2020, 12, 1737. https://doi.org/10.3390/rs12111737 Komendantova N, Mrzyglocki R, Mignan A, Khazai B, Wenzel F, Patt A, Fleming K K (2014) Multi-hazard and multi-risk decision-support tools as a part of participatory risk governance: feedback from civil protection stakeholders. Int J Disaster Risk Reduct 8:50–67. DOI: 10.1016/j.ijdrr.2013.12.006 Leonard M, Westra S, Phatak A, Lambert M, van den Hurk B, McInnes K, Risbey J, Schuster S, Jakob D, Stafford-Smith M (2014) A compound event framework for understanding extreme impacts Wiley Interdiscip. Rev Clim Chang 5:113–128. DOI: 10.1002/wcc.252 Liu B, Han X, Qin L, Xu W, Fan J (2021) Multi-hazard risk mapping for coupling of natural and technological hazards. Geomatics Nat Hazards Risk 12(1):2544–2560. DOI: 10.1080/19475705.2021.1969451 Lombardo F, Obach RS, DiCapua FM, Bakken GA, Lu J, Potter DM, Zhang Y (2006) A hybrid mixture discriminant analysis–random forest computational model for the prediction of volume of distribution of drugs in human. J Med Chem 49(7):2262–2267 Lombardo L, Tanyas H, Nicu IC (2020) Spatial modeling of multi-hazard threat to cultural heritage sites. Eng Geol 277:105776 Msabi MM, Makonyo M (2021) Flood susceptibility mapping using GIS and multi-criteria decision analysis: A case of Dodoma region, central Tanzania, Remote Sensing Applications: Society and Environment, 21: 100445. https://doi.org/10.1016/j.rsase.2020.100445 Nachappa TG, Ghorbanzadeh O, Gholamnia K, Blaschke T (2020) Multi-Hazard Exposure Mapping Using Machine Learning for the State of Salzburg. Austria Remote Sens 12:2757. https://doi.org/10.3390/rs12172757 Park S, Kim J (2019) Landslide susceptibility mapping based on random forest and boosted regression tree models, and a comparison of their performance.Appl Sci9:942Return to ref 2019 in article Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V et al (2011) Scikit-learn: machine learning in python. J Mach Learn Res. 2011;12:2825–30 R Core Team (2019) R: A Language and Environment for Statistical Computing. Foundation for Statistical Computing, Vienna: R [Available at https://www.R-project.org/ ] Rahman M, Ningsheng C, Islam MM, Dewan A, Iqbal J, Washakh RMA, Shufeng T (2019) Flood Susceptibility Assessment in Bangladesh Using Machine Learning and Multi-criteria Decision Analysis. Earth Syst Environ 3, 585–601 (2019). https://doi.org/10.1007/s41748-019-00123-y Razavi-Termeh SV, Sadeghi-Niaraki A, Choi S-M (2020) Gully erosion susceptibility mapping using artificial intelligence and statistical models, Geomatics. Nat Hazards Risk 11(1):821–844. DOI: 10.1080/19475705.2020.1753824 Ridgeway G, Southworth MH (2013) RUnit S. Package ‘gbm.’ Viitattu; 10:40 Roy J, Saha S (2019) Landslide susceptibility mapping using knowledge driven statistical models in Darjeeling District, West Bengal, India. Geoenviron Disasters 6:11. https://doi.org/10.1186/s40677-019-0126-8 Rusk J, Maharjan A, Tiwari P, Chen T-HK, Shneiderman S, Turin M, Seto KC (2022) Multi-hazard susceptibility and exposure assessment of the Hindu Kush Himalaya. Sci Total Environ 804:150039. https://doi.org/10.1016/j.scitotenv.2021.150039 Rutgersson A, Kjellström E, Haapala J, Stendel M, Danilovich I, Drews M, Jylhä K, Kujala P, Guo Larsén X, Halsnæs K, Lehtonen I, Luomaranta A, Nilsson E, Olsson T, Särkkä J, Tuomi L, Wasmund N (2021) Natural Hazards and Extreme Events in the Baltic Sea region, Earth Syst. Dynam Discuss. https://doi.org/10.5194/esd-2021-13 Sarkar T, Mishra M (2018) Soil erosion susceptibility mapping with the application of logistic regression and artificial neural network. J Geovis Spat Anal 2(1):8 Sarker IH, Watters P, Kayes ASM (2019) Effectiveness analysis of machine learning classification models for predicting personalized context-aware smartphone usage. J Big Data. 2019;6(1):1–28 Saunders WSA, Kilvington M (2016) Innovative land use planning for natural hazard risk reduction: A consequence-driven approach from New Zealand. Int J Disaster Risk Reduct 18:244–255. https://doi.org/10.1016/j.ijdrr.2016.07.002 Schmidt J, Matcham I, Reese S, King A, Bell R, Henderson R, Smart G, Cousins J, Smith W, Heron D (2011) Quantitative multi-risk analysis for natural hazards: a framework for multi-risk modelling. Nat. Hazards 58, 1169–1192 (2011) Şen Z, Al-Harithy S, As-Sefry S, Almazroui M (2017) Aridity and Risk Calculations in Saudi Arabian Wadis: Wadi Fatimah Case. Earth Syst Environ 1, 26 (2017). https://doi.org/10.1007/s41748-017-0030-x Shah AA, Khwaja S, Shah BA, Reduan Q, Jawi Z (2018) Living With Earthquake and Flood Hazards in Jammu and Kashmir, NW Himalaya. Front Earth Sci 6:179. https://doi.org/10.3389/feart.2018.00179 Shalaby A, Tateishi R (2007) Remote sensing and GIS for mapping and monitoring land cover and land-use changes in the Northwestern coastal zone of Egypt. Appl Geogr. 27 28 – 41 https://doi.org/10.1016/j.apgeog.2006.09.004 Shin Y, Kim T, Cho H, Kang KI (2012) A formwork method selection model based on boosted decision trees in tall building construction. Autom Constr 23:47–54 Skilodimou HD, Bathrellos GD, Chousianitis K, Youssef AM, Pradhan B (2019) Multi-hazard assessment modeling via multi-criteria analysis and GIS: a case study. Environ Earth Sci 78, 47 (2019). https://doi.org/10.1007/s12665-018-8003-4 Solberg AHS (1996) Texture fusion and classification based on flexible discriminant analysis. In Proceedings of the 13th International Conference on Pattern Recognition, Vienna, Austria, 25–29 August; 596–600 Stoeser DB, Camp VE (1985) Pan-African microplate accretion of the Arabian shield. Geol Soc Amer Bull 96:817–826 Sun D, Wen H, Wang D, Xu J (2020) A random forest model of landslide susceptibility mapping based on hyperparameter optimization using Bayes algorithm. Geomorphology 362:107201. https://doi.org/10.1016/j.geomorph.2020.107201 Tabari H (2020) Climate change impact on flood and extreme precipitation increases with water availability. Sci Rep 10:13768. https://doi.org/10.1038/s41598-020-70816-2 Van Westen CJ, Greiving S (2017) Multi-hazard risk assessment and decision making. Environmental Hazards Methodologies for Risk Assessment and Management. IWA publishing, London, UK, pp 31–94 Wang Y, Fang Z, Hong H, Peng L (2020) Flood susceptibility mapping using convolutional neural network frameworks. J Hydrol 582:124482. https://doi.org/10.1016/j.jhydrol.2019.124482 Wang Y, Fang Z, Wang M, Peng L, Hong H (2020) Comparative study of landslide susceptibility mapping with different recurrent neural networks. Comput Geosci 138:104445. doi: 10.1016/j.cageo.2020.104445 Ward PJ, Blauhut V, Bloemendaal N, Daniell JE, de Ruiter MC, Duncan MJ, Emberson R, Jenkins SF, Kirschbaum D, Kunz M, Mohr S, Muis S, Riddell GA, Schäfer A, Stanley T, Veldkamp TIE, Winsemius HC (2020) Review article: Natural hazard risk assessments at the global scale. Nat Hazards Earth Syst Sci 20:1069–1096. https://doi.org/10.5194/nhess-20-1069-2020 Wastl M, Stötter J, Kleindienst H (2011) Avalanche risk assessment for mountain roads: A case study from Iceland. Nat Hazards 56(2):465–480 Weinkle J, Landsea C, Collins D, Musulin R, Crompton RP, Klotzbach PJ, Pielke R (2018) Normalized hurricane damage in the continental United States 1900–2017. Nat Sustain 1(12):808–813. https://doi.org/10.1038/s41893-018-0165-2 Yang A, Wang C, Pang G, Long Y, Wang L, Cruse RM, Yang Q (2021) Gully Erosion Susceptibility Mapping in Highly Complex Terrain Using Machine Learning Models. ISPRS Int J Geo-Information 10(10):680. https://doi.org/10.3390/ijgi10100680 Ye J, Chow J-H, Chen J, Zheng Z (2009) Stochastic gradient boosted distributed decision trees, in Proceedings of the ACM 18th International Conference on Information and Knowledge Management (CIKM '09), 2061–2064, Hong Kong, November 2009 Ye T, Liu W, Mu Q, Zong S, Li Y, Shi P (2020) Quantifying livestock vulnerability to snow disasters in the Tibetan Plateau: Comparing different modeling techniques for prediction. Int J Disaster Risk Reduct 48101578. https://doi.org/10.1016/j.ijdrr.2020.101578 Yoo C, Cho E (2019) Effect of Multicollinearity on the Bivariate Frequency Analysis of Annual Maximum Rainfall Events. Water 11:905. https://doi.org/10.3390/w11050905 Zhang C, Xie Z (2012) Combining object-based texture measures with a neural network for vegetation mapping in the Everglades from hyperspectral imagery. Remote Sens Environ 124:310–320 Zhao G, Pang B, Xu Z, Yue J, Tu T (2018) Mapping flood susceptibility in mountainous areas on a national scale in China. Sci Total Environ 615:1133–1142. doi: 10.1016/j.scitotenv.2017.10.037 Zhou X, Wen H, Zhang Y, Xu J, Zhang W (2021) Landslide susceptibility mapping using hybrid random forest with GeoDetector and RFE for factor optimization. Geosci Front 12(5):101211. https://doi.org/10.1016/j.gsf.2021.101211 Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-1554302","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":106525885,"identity":"4aadd436-6e52-4087-805f-06b69db3781a","order_by":0,"name":"Ahmed M. Youssef","email":"","orcid":"","institution":"Saudi Geological Survey","correspondingAuthor":false,"prefix":"","firstName":"Ahmed","middleName":"M.","lastName":"Youssef","suffix":""},{"id":106525886,"identity":"912e55d2-3b5d-4a5b-a71c-64ae07b19662","order_by":1,"name":"Ali M. Mahdi","email":"","orcid":"","institution":"South Valley University","correspondingAuthor":false,"prefix":"","firstName":"Ali","middleName":"M.","lastName":"Mahdi","suffix":""},{"id":106525887,"identity":"13afb125-0d0a-4a4f-9188-846039e61b9c","order_by":2,"name":"Mohamed M. Al-Katheri","email":"","orcid":"","institution":"Saudi Geological Survey","correspondingAuthor":false,"prefix":"","firstName":"Mohamed","middleName":"M.","lastName":"Al-Katheri","suffix":""},{"id":106525888,"identity":"5c22115a-734a-4dc5-af06-273d2fc70f52","order_by":3,"name":"Soheila Pouyan","email":"","orcid":"","institution":"Shiraz University","correspondingAuthor":false,"prefix":"","firstName":"Soheila","middleName":"","lastName":"Pouyan","suffix":""},{"id":106525889,"identity":"b4445251-14de-4f62-a1d6-82a676463fad","order_by":4,"name":"Hamid Reza Pourghasemi","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0003-2328-2998","institution":"Shiraz University","correspondingAuthor":true,"prefix":"","firstName":"Hamid","middleName":"Reza","lastName":"Pourghasemi","suffix":""}],"badges":[],"createdAt":"2022-04-13 12:39:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-1554302/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-1554302/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":21873335,"identity":"df73ecc9-a94c-4509-8bfd-f394fffe5efc","added_by":"auto","created_at":"2022-05-25 15:15:56","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":150386,"visible":true,"origin":"","legend":"\u003cp\u003eLocation of Hasher-Fayfa Basin in Jazan province, KSA, with training and validation locations of landslide, flood, and gully erosion, b, c) 3D perspective from Google Earth and photograph by the author (Ahmed M. Youssef), respectively, showing the densely populated area in Fayfa Mountain.\u003c/p\u003e\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Fig1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/7873ef5234bd30b171285125.jpg"},{"id":21873558,"identity":"b9e573e4-cb29-4ef0-9e0a-2e6f4c19d5d6","added_by":"auto","created_at":"2022-05-25 15:20:56","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":161830,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart showing the sequential steps of this study.\u003c/p\u003e\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Fig2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/a2a3ebdcc83c57f83ae9214d.jpg"},{"id":21872707,"identity":"8fb4158a-ecb7-4349-ac4f-80c9f344fead","added_by":"auto","created_at":"2022-05-25 15:10:57","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":130144,"visible":true,"origin":"","legend":"\u003cp\u003ePhotographs showing the three hazards collected between 2008 and 2020 in the study area, landslides (a, b, c); gully erosion (d, e, f); and flash floods (g, h). They were taken by Ahmed M. Youssef (first author).\u003c/p\u003e\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/36a1d205473b49ae14f67a6d.jpg"},{"id":21872706,"identity":"6210c42b-5c64-4255-9472-f9438e101c33","added_by":"auto","created_at":"2022-05-25 15:10:56","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":7425841,"visible":true,"origin":"","legend":"\u003cp\u003ePredictive factors used to construct multi-hazard model: (a) elevation, (b) slope, (c) TWI, (d) plan curvature, (e) aspect, (f) TRI, (g) profile curvature, (h) valley depth, (i) rainfall, (j) distance to wadis, (k) stream density, (l) NDVI, (m) LULC, (n) lithology, (o) distance to fault, and (p) distance to road.\u003c/p\u003e\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Fig4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/704eb584ff9e858f55883e4b.jpg"},{"id":21872704,"identity":"1de28c58-7372-4679-a9ba-93fc6b253ac2","added_by":"auto","created_at":"2022-05-25 15:10:56","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":62677,"visible":true,"origin":"","legend":"\u003cp\u003eHazard-predictive factors importance EL; S; TWI; PC, A, TRI, PrC, VD, R, DtW, SD, NDVI, LULC, Lth, DtF, and DtR) for a) landslide; b) gully erosion; and c) flood.\u003c/p\u003e\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Fig5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/0062f8baf23ac5746dc33c60.jpg"},{"id":21872698,"identity":"4ed520dd-5851-4355-9b1a-69c7a8d153df","added_by":"auto","created_at":"2022-05-25 15:10:56","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":188193,"visible":true,"origin":"","legend":"\u003cp\u003esusceptibility maps for landslide, gully erosion, and flood using MLMs.\u0026nbsp;\u003c/p\u003e","description":"","filename":"Fig6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/400b52b53f1cc710f36a4335.jpg"},{"id":21873338,"identity":"be8fa7cb-8ea6-4722-b0c7-d455c17471d8","added_by":"auto","created_at":"2022-05-25 15:15:56","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":65117,"visible":true,"origin":"","legend":"\u003cp\u003eArea percentage of different hazard classes for (a) landslide, (b) gully erosion, and (c) flood.\u003c/p\u003e","description":"","filename":"Fig7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/614b792a1c82d2cd01cc6f93.jpg"},{"id":21873336,"identity":"229d0f49-4b8c-42db-a7df-2d66cad5135e","added_by":"auto","created_at":"2022-05-25 15:15:56","extension":"jpg","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":67957,"visible":true,"origin":"","legend":"\u003cp\u003eAccuracy assessment of different hazard models for (a) landslides, (b) gully erosion, and (c) floods\u003c/p\u003e","description":"","filename":"Fig8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/7f0e3bd728184c89f6758eca.jpg"},{"id":21872697,"identity":"379f7dd7-f632-4222-926b-90c9aa3dd5c6","added_by":"auto","created_at":"2022-05-25 15:10:56","extension":"jpg","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":99371,"visible":true,"origin":"","legend":"\u003cp\u003eThe multi-hazard risk map was constructed by integrating of highly accurate models of FDA for landslide, RF for gully erosion, and RF for floods.\u003c/p\u003e\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Fig9.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/2db5150c01291b5350756496.jpg"},{"id":21872699,"identity":"a8cfbd7c-5afb-4625-8529-aef537fb3f23","added_by":"auto","created_at":"2022-05-25 15:10:56","extension":"jpg","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":46572,"visible":true,"origin":"","legend":"\u003cp\u003eDifferent types of hazard percentage in the study area. No-H = No hazard, LS = landslide, GE = gully erosion, FL = flood, LS-GE = landslide-erosion, FL-LS = flood-landslide, FL-GE = flood-gully erosion, and FL-LS-GE flood-landslide-gulley erosion.\u003c/p\u003e\u003cp\u003e\u003cbr\u003e\u003c/p\u003e","description":"","filename":"Fig10.jpg","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/ba09cea59fed6f4c8889118c.jpg"},{"id":22861823,"identity":"a51b0316-ba61-4ffc-91e9-75377829d729","added_by":"auto","created_at":"2022-06-21 01:09:27","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1466393,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-1554302/v1/189f4041-e69d-485d-904b-2b2b17816f14.pdf"}],"financialInterests":"","formattedTitle":"Multi-hazards modeling using machine learning algorithms in Southwestern Saudi Arabia","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eManaging disasters is challenging due to various kinds of hazards (e.g., natural and anthropogenic hazards) (De Silva and Kawasaki, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Natural processes will become natural hazards, when they intersect and have a negative effect on people life (Gill and Malamud, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Ward et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Subsequently, natural hazards are the physical events that pose devastating impacts on communities, damaging critical infrastructure, affecting economic viability, and claiming lives (IPCC, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Worldwide, an exponential increase in the human population leads to lateral expansion in development (urbanization, agricultural activities, infrastructures, and life lines), subsequently increasing the exposure and interaction with different types of natural hazards (Weinkle et al., \u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Globally, several areas are vulnerable to events involving several types of hazards that are related to the same area and at the same time (Leonard et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Natural hazards have impacted people and natural environments in many undeveloped countries more severely than developed countries, causing substantial loss of lives and enormous economic plunge (Iglesias et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In a simple definition, a hazard refers to the probability of potential damage in certain areas (location or where?) and a specific period (time or when?) due to an event with specific magnitude (how large?) (Shi, 2019).\u003c/p\u003e \u003cp\u003eArid and semi-arid areas around the world are impacted by different natural hazards events, ranging from insignificant influence events to extremely events (IPCC, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Tabari \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Rutgersson et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These natural hazards frequently occur, threatening everything (modern cities, ruler areas, infrastructures, and life lines) (De Silva and Kawasaki, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). These disasters can cause great economic damage, disruption of transportation systems, injury, and claiming lives. Mountainous provinces are highly likely among the most disaster-susceptible areas due to their various characteristics such as lithology, tectonics, climate, and hydrology (Baig et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Risk related to natural hazards is high in middle East countries due to lack of disaster preparedness and public awareness, and inadequate funding support (AlQahtanya and Abubakarb, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Saudi Arabia is one of these countries that experienced many natural hazards, yearly (Alyami et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These events cost lives and bring the economy to a static state. Moreover, not much has been done to map different prone areas related to these natural hazards.\u003c/p\u003e \u003cp\u003eMountainous regions that are inhabited by people and crossed by infrastructures are prone to not only one type of hazard, but they are frequently the scene of multiple disasters that interact together, such as earthquakes, avalanches, landslides, floods, mudslides, ground subsidence, soil erosion, and wildfires (Gill and Malamud, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Shah et al., \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Duncan et al. (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) defined multi-hazard as all potential and cascading hazards, in any particular area at a certain period. Most of the time, these hazards can cause severe damage to the various human activities located and intercept with the hazard prone areas by causing fatalities and injuries, damaging their urban and industrial areas, destroying infrastructures (roads, tunnels, and railways), and disrupt lifelines (power lines, water systems, and gas mains) (Bell and Glade, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2004\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDue to the complexity of natural hazards, many studies deal with individual hazard types as an independent approach (Wastl et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Also, problems related to these problems (e.g., floods, landslides, and gully erosions) have been evacuated individually by applying various machine learning algorithms-MLAs (Sarkar and Mishra, \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Ghorbanzadeh et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Hosseiny et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Zhou et al., \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). However, Earth system science approach indicates that a significant interaction between different types of hazard components due to the interaction between the component systems (e.g., the lithosphere, atmosphere, hydrosphere, and biosphere). Accordingly, a holistic approach to understand all different hazards in a certain area is a must and can prevent cascading hazards that can be formed by the interaction of different hazard types. A multi-hazard risk evaluation could be significant to control hazard interactions (Komendantova et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). To prevent these natural hazards and protect people and their properties and the country economy from long-term plunging, effective predictive models must develop (Bathrellos et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). These models should base on profound understanding of the most influential and triggering factors that have a significant contribution on these hazards (van Westen and Greiving, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eMulti-hazard modeling becomes an essential tool in land use development at both regional and national scales (Saunders and Kilvington, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). These multi-hazard modeling approaches gain good attention recently worldwide based on their ability to consider different types of hazards that could impact the area (Schmidt et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Skilodimou et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Lombardo et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDifferent approaches were used to conduct multi-hazards modeling e.g., by using two decision-making tools including, sequential Monte Carlo method and decision-making tool (Komendantova et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2014\u003c/span\u003e); using a deterministic equation based (theoretical) and empirical understanding (Bout et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2018\u003c/span\u003e); and by using multi-criteria analysis and GIS (Skilodimou et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Recently, MLAs have been widely used to predict various hazards based on random forest (RF), and support vector machine-SVM (Nachappa et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), boosted regression tree-BRT-, and generalized additive model-GAM (Ye et al., \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn this study, no multi-hazard assessment has been done in Saudi Arabia before and the application of machine learning techniques in multi-hazard assessment will be a novel study in the area. In this study, developing a multi-hazard risk map using machine learning techniques for the most impacted natural hazards (landslides, floods, and gully erosion) in this area are crucial for effective landuse management. The current study evaluates machine learning models (MDA, GLM, FDA, BRT, and RF) as effective and accurate models to produce multi-hazard risk map for Hasher-Fayfa Basin that might be used by authorities, developers, and decision-makers.\u003c/p\u003e"},{"header":"2. Study Area","content":"\u003cp\u003eThe Jazan province has an area of 11,671 km\u003csup\u003e2\u003c/sup\u003e, is situated at the southwestern corner of Saudi Arabia, bordering the Red Sea to the west and Asir region to the north (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). The Jazan region includes the study area that is bounded by latitudes 17\u0026deg;10' to 17\u0026deg;30' N and longitudes 42\u0026deg;58' and 43\u0026deg;13' E. Hasher-Fayfa Basin covers 513 Km\u003csup\u003e2\u003c/sup\u003e. Its elevation ranges from 220 m to 2,340 m above sea level (e.g. Fayfa Mountain up to 2.3 km high). This makes it the one of the highest areas in KSA. The population of the Jazan region is about 1.67\u0026nbsp;million persons in 2020 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://knoema.com/atlas/Saudi-Arabia/Jazan/Total-Population\u003c/span\u003e\u003cspan address=\"https://knoema.com/atlas/Saudi-Arabia/Jazan/Total-Population\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e The area under study inhabited by many people who live in the mountainous areas as in (Fayfa Mountain, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e(b-c)). Since about six decades, Saudi Arabia was discovered the oil fields, many urban centers and infrastructures were established. The Jazan area has been expanded drastically. The area has all the elements that could put it under a high risk of floods, landslides, and erosion that could impact the economic development of the region. These hazards will pose a high risk to people and their property (buildings, highways, roads, infrastructure, and lifelines). The basin is characterized by a rugged landscape with the iconic features of the Hasher and Fayfa mountains. These mountains have steep reliefs (up to 88\u003csup\u003e◦\u003c/sup\u003e). Rainfall is the main trigger factor for landslide, flood, and erosion in the area under consideration. The region receives up to 500 mm/year of precipitation, with a high rainfall value of 1,400 mm in 1979 (Hasanean and Almazroui \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). The high rainfall in the region is mainly due to the tropical air masses affecting the Jazan and Asir regions (the southwestern provinces of KSA) and the high altitudes that produce orographic rainfall (Şen et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAdditionally, the area is distinguished by the presence of intense tectonic deformation (folds, faults, shear zones, fractures, and joints). The Red Sea Mountains of Saudi Arabia consists of extrusive and intrusive igneous rocks, and metamorphic rocks. These rocks were formed by the collection of island arcs and closure of the interleaving oceanic arcs 700\u0026thinsp;\u0026minus;\u0026thinsp;1,000 Ma (Greenwood, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1982\u003c/span\u003e; Stoeser and Camp, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e1985\u003c/span\u003e). These mountains were raised up due to the spreading of the Red Sea about thirty million years ago. These Precambrian rocks are unconformably overlain by sedimentary sequences of Paleozoic and Cenozoic age (Agar, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1987\u003c/span\u003e).\u003c/p\u003e"},{"header":"3. Methodology","content":"\u003cp\u003eThe current work involved five steps (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e): (1) data collection of the various events based on intensive field work, analysis of various civil defense reports, technical reports, and questionnaires with the local population (for the period between 2000 and 2020); (2) determine the various events based on a through literature review; (3) model different hazard types using various machine learning techniques (; (4) model validation and selection of the most appropriate model for each hazard type; and (5) generation of multi-hazard risk map (MHRM) by integrating the high-AUC model for each hazard.\u003c/p\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e3.1. The inventory of hazard data (multi-hazards inventory)\u003c/h2\u003e \u003cp\u003eCreating an inventory of susceptible landforms is an essential and crucial stage in hazard susceptibility mapping. The inventory of landslides, gully erosion, and flooding in the watershed of the study area was compiled based on field investigations with the help of GPS instrument (732 sites were investigated through the field work between 2015 and 2018) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). National and regional data from various sources, including the Civil Defense, Jazan Region Authority, Ministry of Transportation, Saudi Geological Survey, review of technical reports and scientific publications and the private sector of the Hasher and Fayfa region were collected and reviewed. As some of the vulnerable areas are located in mountainous areas and may have been overlooked during the field survey; so, high-resolution Google Earth images were also used to detect landslides, gully erosion features, and flooded areas. Based on this information, 100 landslide locations, 60 sites for gully erosion, and 70 sites for flood locations were used to establish the multi-hazard inventory map (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). In this study, we used a hazard and non-hazard locations and machine learning techniques. Accordingly, equal numbers of hazards and non-hazardous locations were randomly selected (Fang et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). A random partitioning method was used to select training and validation sites for each hazard type in the inventory map. In the current work, training datasets (70% of each hazard) were used for model generation and the remaining 30% of hazard locations were used for model validation (Zhao et al., \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Hazard-predictive factors (HPFs)\u003c/h2\u003e \u003cp\u003eIt indicated that hazard-predictive factors (HPFs) are the first step of the modeling process (Rusk et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Accordingly, intensive review of previous studies was done to identify the most effective factors for each hazard type (Roy and Saha, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Razavi-Termeh et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Wang et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Raster database layers of different topography, climatology, hydrology, vegetation and landuse/landcover, geology, and anthropogenic were selected and prepared using ArcGIS 10.8 of 5 \u0026times; 5 m resolution (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). These factors include 16 HPFs; elevation (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea), slope angle (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb), topographic wetness index (TWI \u0026ndash; Eq.\u0026nbsp;\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) (Beven and Kirkby, 1979) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec), plan curvature (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed), aspect (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee), topographic roughness index (TRI - Eq.\u0026nbsp;2) (Nellemann and Reynolds, 1997) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ef), profile curvature (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg), and valley depth (VD) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eh) were extracted from DEM 5 m resolution that were generated using 1:10,000 topographic map; lithology (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003en) and distance to fault (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eo) were extracted from a geologic map of 1:250,000 scale obtained from Saudi Geological Survey; rainfall distribution map was generated from 20 rainfall stations (acquired from the MEWA, covering a period from 1965 to 2020) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ei), distance to wadis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ej) and stream density ((Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ek) were generated from topographic map 1:10,000 scale and converted to distance to wadis and stream density using Euclidian distance tool and density tools in ArcGIS 10.8; distance to road (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ep) was extracted from high resolution images (Google Earth, then converted to distance to road using Euclidian distance tool in ArcGIS 10.8, landuse/land cover (LULC) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003em) and normalized differential vegetation index (NDVI) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003el) were generated from sentenil-2 image (10 m resolution) acquired in May 2020, LULC map was generated using maximum likelihood classification method, and finally NDVI was extracted using NDVI equation (Eq.\u0026nbsp;\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Shalaby and Tateishi (\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) pointed out that the maximum likelihood method is a stable and accurate model for image classification because it can apply multiple statistical features in its work.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe most influencing indicators for multi-hazard susceptibility mapping.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eFactor\u003c/p\u003e \u003cp\u003eType\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eHazard-indicator factors and abbreviation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eData type\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eFactor range\u003c/p\u003e \u003cp\u003e/Units\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e \u003cp\u003eHazard type\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eLandslide (LS)\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003eFlood (FL)\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003eGully erosion (GE)\u003c/b\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"7\" rowspan=\"8\"\u003e \u003cp\u003eTopography\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eElevation (EL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e220-2,340 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSlope (S)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u0026ndash;88 degree\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTopographic wetness index (TWI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.78\u0026ndash;24.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePlan curvature (PC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-271.5\u0026ndash;1452.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAspect (A)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9 directions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTerrain ruggedness index (TRI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u0026ndash;174\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProfile curvature (PrC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1627.8\u0026ndash;917.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValley depth (VD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u0026ndash;261 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClimatology\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRainfall (R)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePoint\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e383\u0026ndash;495 mm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eHydrology\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDistance to wadis (DtW)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePolyline\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u0026ndash;2,900 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eStream density (SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePolyline\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.95\u0026ndash;23.7 km/km\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVegetation and landuse/land cover\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNormalized difference vegetation index (NDVI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.639\u0026ndash;0.675\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLand use/Landcover (LULC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGrid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5 classes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eGeology\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLithology (Lth)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePolygon\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6 groups\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDistance to fault (DtF)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePolyline\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u0026ndash;3,153 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnthropogenic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDistance to road (DtR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePolyline\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u0026ndash;5,780 m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eꚚ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAll, layers were converted to have 5 m * 5 m pixel size using ArcGIS 10.8. Table\u0026nbsp;(1) shows that different factors used to prepare hazard susceptibility maps; 11 factors used for landslides, 12 factors were used for floods, and 13 factors were used for gully erosion.\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$TWI=lin \\frac{\\text{A}}{\\text{tan}B}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere, A is the cumulative basin area (m\u003csup\u003e2\u003c/sup\u003e), and β is the slope (in degrees) at a point.\u003c/p\u003e \u003cp\u003eTRI = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\text{Y}\\left[\\sum {({x}_{ij}-{x}_{00})}^{2}\\right]}^{1/2}\\)\u003c/span\u003e\u003c/span\u003e (2)\u003c/p\u003e \u003cp\u003eWhere, xii is the height of each pixel adjacent to the pixel (0, 0). Areas with a slope of 0 have a TRI value of zero, while the roughened areas with steep elevations have positive values of TRI.\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$NDVI=\\frac{(NIR-R)}{(NIR+R)}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere, NIR is the near infrared band (wave length\u0026thinsp;=\u0026thinsp;832.8 nm) and R is the red band (wave length\u0026thinsp;=\u0026thinsp;664.6 nm) of the spectral bands.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Machine learning algorithms\u003c/h2\u003e \u003cp\u003eIn the current work, five machine learning algorithms (MLAs) were used to generate landslide, flood, and gully erosion susceptibility maps that affect the study area. Various literatures were conducted using various MLAs to overcome the problems related to natural hazards by predicting their susceptibility, such as landslides (Park and Kim, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; He et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), floods (Janizadeh et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; El-Haddad et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), and gully erosion (Ghorbanzadeh et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Amare et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Yang et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). A detailed description of the MLAs that used in this study is given in the following parts:\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e3.3.1. Random forest\u003c/h2\u003e \u003cp\u003eRandom Forest (RF) is an ensemble classification technique (Breiman, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2001\u003c/span\u003e) and considered as one of the learning techniques that involves multiple steps, starting with training datasets, followed by bootstrapping, then ensemble of trees, and finally aggregation (classification phase) (Hawryło et al., 2018; Sarker et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Bootstrapping refers to the parallel training of each decision tree with other training subsets using different available features (Herrera et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In this phase, each individual decision tree is unique and is used to minimize the overall variance of the RF model. In the last stage, the RF model summarizes the decisions of the individual trees; therefore, the RF model shows good generalization. The RF algorithm tends to outperform most other classification approaches in terms of accuracy, with no problems of overfitting (Pedregosa et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). The RF algorithm is robust to training sample selection and noise in datasets and does not require feature scaling (fits both categorical and continuous values) (Kim et al., 2018). RF can overcome outliers in predictors, automatically handle omitted data and increase the diversity of classification trees.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Multivariate discriminant analysis\u003c/h2\u003e \u003cp\u003eMultivariate discriminant analysis (MDA) is a trained classification algorithm as an extension of linear discriminant analysis (LDA). It can evaluate the multivariate nonlinear relationships between different classes within each group (Lombardo et al. \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). To calculate the distance to the nearest cluster, the normal distribution (the variability and correlation between variables is uniform) of each class is used (Lombardo et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). It can be derived using Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e4\u003c/span\u003e) (Hair et al., 1998).\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$${D}_{v}={D}_{w1}{V}_{1}+{D}_{w1}{V}_{1}+\\dots +{D}_{wn}{V}_{n}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\({D}_{v}\\)\u003c/span\u003e \u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({D}_{w1}\\)\u003c/span\u003e\u003c/span\u003e (i\u0026thinsp;=\u0026thinsp;1,2,3, ..., n) represent the discriminant value and weights, respectively, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({V}_{1}\\)\u003c/span\u003e\u003c/span\u003e (i\u0026thinsp;=\u0026thinsp;1,2,3,..., n) are independent indicators. The MDA model was run using \u0026ldquo;mda\u0026rdquo; package in R software (Hastie et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e3.3.3. Generalized linear model\u003c/h2\u003e \u003cp\u003eThe generalized linear model (GLM) is a linear regression model that can quantify and incorporate specific and temporal factors (Goetz et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). The GLM can scale the accuracy and quality of the outcomes by using multiple regression to establish a unique link between the dependent and independent factors (Scott et al., 1991). By identifying the best regression model, multiple events can be predicted (Payne, 2015). It is an adequate model for numeric factors because of its regression characteristic; however, insignificant and correlated factors can reduce its performance, it also, deal with non-normal distribution data (Kalantar et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The GLM linkage function is used to establish the relationship between the dependent variable and the independent variables (Soch et al., 2017). In this study, the occurrence probability of hazard event Y can be represented by Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e5\u003c/span\u003e). By the logistic transformation, the link function g(y\u003csub\u003ei\u003c/sub\u003e) is represented by Eq.\u0026nbsp;\u003cspan refid=\"Equ5\" class=\"InternalRef\"\u003e6\u003c/span\u003e:\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$$P=\\frac{\\text{e}\\text{x}\\text{p}({c}_{0}+{c}_{1}{X}_{1}+{c}_{2}{X}_{2}+\\dots +{c}_{i}{X}_{i})}{1+\\text{e}\\text{x}\\text{p}({c}_{0}+{c}_{1}{X}_{1}+{c}_{2}{X}_{2}+\\dots +{c}_{i}{X}_{i})}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e5\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ5\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ5\" name=\"EquationSource\"\u003e\n$$g\\left({y}_{i}\\right)={c}_{0}+\\sum {c}_{i}{x}_{i}+ {\\epsilon }_{i}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e6\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere P is the event \u0026ldquo;Y\u0026rdquo; probability; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({c}_{0}\\)\u003c/span\u003e\u003c/span\u003e is the intercept; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({c}_{1}\\dots .{c}_{i}\\)\u003c/span\u003e\u003c/span\u003e. .. are logistic regression coefficients, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({X}_{1}\\dots .{X}_{i}\\)\u003c/span\u003e\u003c/span\u003e are the independent variables; and the error residual expressed by ε\u003csub\u003ei\u003c/sub\u003e. In this study, R software was used to construct the GLM model (R Core Team, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). A simple Gaussian family is determined as the link function for normally distributed response data. The independent indicators were to be included in the model individually, using a smoothing spline based on two degrees of freedom to avoid overfitting (Aertsen et al., 2009).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e3.3.4. Flexible discriminant analysis\u003c/h2\u003e \u003cp\u003eRamsay and Dalzell (1991) proposed the Flexible discriminant analysis (FDA) as a nonlinear statistical classification method. The crucial concept of FDA is to treat an observed object with functional characteristics as an integral, regardless of the order of the observed values (Wagner-Muns et al., 2018). It can distinguish unsupervised work where each class is divided into subclasses with a unique value (Zou et al., 2019). FDA can provide a framework by coupling various models and methods (regression algorithms, discriminant analysis, and classification techniques) (Hastie and Tibshirani, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e1996\u003c/span\u003e). In this study, the FDA technique was used to create a hazard susceptibility map using the species distribution package in R software (Naimi and Ara\u0026uacute;jo, 2016).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section3\"\u003e \u003ch2\u003e3.3.5 Boosted regression trees\u003c/h2\u003e \u003cp\u003eBRT is a combination of regression and boosting techniques. It was proposed by Friedman (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2001\u003c/span\u003e). The boosting algorithm is used to improve model accuracy, where the residual errors of the existing tree will be used to adjust the new trees (Ye et al., \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Shin et al. (\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) pointed out that the BRT model has several advantages because it can be improved by fitting and combining multiple methods and increasing the number of trees, it uses different types of data without conversion or outlier removal, it works when some data are missing, and it can use sophisticated nonlinear relationships to fit the data. The BRT model is defined by Eq.\u0026nbsp;(\u003cspan refid=\"Equ6\" class=\"InternalRef\"\u003e7\u003c/span\u003e) (Friedman \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2001\u003c/span\u003e).\u003cdiv id=\"Equ6\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ6\" name=\"EquationSource\"\u003e\n$$F(X;{\\left[{B}_{m}{a}_{m}\\right]}^{{m}_{o}}={\\sum }_{m=0}^{m}{B}_{m}h(X;{a}_{m})$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e7\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere is\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(h(X;m)\\)\u003c/span\u003e\u003c/span\u003e a simple classification function with variable \u0026ldquo;a\u0026rdquo; and factors \u0026ldquo;x.\u0026rdquo; Factor \u0026ldquo;m\u0026rdquo; is representing the step of the algorithm, and is \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({B}_{m}\\)\u003c/span\u003e\u003c/span\u003e the weighting coefficient at step m. A new tree will be added to the original model at each iteration step, which leads to reducing of the loss function. The BRT training stage will be ended as soon as the predefined iteration numbers achieved. The BRT model was run using the package \"brt\" in R software (Ridgeway and Southworth \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.4. HFIs multicollinearity and importance, and model validation\u003c/h2\u003e \u003cp\u003eMulticollinearity method is used to evaluate variables effectiveness in different model construction (Yoo and Cho \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In this method, two indices were used, variance inflation factors (VIF) and tolerance (TOL) as shown in equations (\u003cspan refid=\"Equ7\" class=\"InternalRef\"\u003e8\u003c/span\u003e and \u003cspan refid=\"Equ8\" class=\"InternalRef\"\u003e9\u003c/span\u003e):\u003cdiv id=\"Equ7\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ7\" name=\"EquationSource\"\u003e\n$$\\text{T}\\text{O}\\text{L}=(1-{R}_{j}^{2})$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e8\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ8\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ8\" name=\"EquationSource\"\u003e\n$$\\text{V}\\text{I}\\text{F}=\\left[\\frac{1}{(1-{R}_{j}^{2})}\\right]$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e9\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\({R}_{j}^{2}\\)\u003c/span\u003e \u003c/span\u003e Represents the regression coefficient of the explanatory factor J on all other descriptive factors. Studies have shown that a TOL\u0026thinsp;\u0026lt;\u0026thinsp;0.10 and a VIF\u0026thinsp;\u0026gt;\u0026thinsp;5 account for multicollinearity issues (Rahman et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAdditionally, the importance of independent indicators is essential. It measures their contribution in each hazard type modeling. Several techniques were used among them the learning vector quantization (LVQ), and random forest algorithm (RF) (Zhang and Xie, \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Sun et al., \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). In this study, HIFs importance and contribution to landslide, floods, and gully erosion occurrence were assessed using RF.\u003c/p\u003e \u003cp\u003eFinally, to evaluate the accuracy of the results, the receiver operating characteristics (ROC) and the area under the curves (AUC) can also be applied to evaluate the MLMs that used to provide valuable information (Bradley \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). Without validated maps, the models and their results would have no scientific significance or if the AUC is less than 0.5, the model is considered a random model (Marzban, 2004). Evaluation of the accuracy of the different machine learning models for each hazard type is a profound step to create the final hazard model (Amare et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In this study, the performance of all hazard models was determined using the validation dataset of each hazard type. This will provide quantitative measures of the model accuracy. The predictive performance was done using the AUC. AUC value ranges from 0 to 1 (0 to 100%). A range of 1 (100%) represents high performance, while a range of \u0026le;\u0026thinsp;0.5 (50%) indicates a poor classification. Evaluation of the accuracies of these models based on (Guzzetti et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). They divided the AUC values into three groups: between 0.75 and 0.8 for an acceptable model, between 0.8 and 0.9for a good susceptibility model, and above 0.9 for an excellent model.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results And Analysis","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.1. Multicollinearity test\u003c/h2\u003e \u003cp\u003eThe multicollinearity analysis of the hazard-predictor factors used in this study for landslide susceptibility (11 factors), flood susceptibility (12 factors), and gully erosion susceptibility (13 factors) is shown in Table\u0026nbsp;(2). Results showed that both tolerance (ToL) and (VIF) values of landslide-predictor factors used in this study to run the different machine learning models were appropriate, ToL values were more than 0.1 (the lower ToL value is 0.20) and the VIF values were less than 5 (the large VIF value is 4.9) for slope factor. The tolerance (ToL) and (VIF) values of flood-predictor factors were appropriate, ToL values were more than 0.1 (the lower ToL value is 0.257) and the VIF values were less than 5 (the large VIF value is 3.898) for elevation factor. For the gully erosion predictor factors, all factors are adequate except the TRI that provide VIF value more than 5. For the rest of these factors the ToL values were more than 0.1 (the lower ToL value is 0.271) and the VIF values were less than 5 (the large VIF value is 3.697) for the slope facto. Accordingly, the results showed that there was no multicollinearity among the selected FCFs, which means that they made a significant contribution to the model construction in this study.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.2. The importance of hazard-predictors\u003c/h2\u003e \u003cp\u003eThe analysis of landslide hazard-predictor importance is shown in Figure (5a). Our findings indicated that the top predictors that controlled landslide susceptibility are the slope angle (S), followed by land use/land cover (LULC), elevation (EL), the normalized differential vegetation index (NDVI), followed by aspect (As), and distance to wadis (DtW), distance to road (DtR), topographic witness index (TWI) and distance to fault (DtF). Less importance predictor is lithology (Lth). We believed that this is due to the higher slope angle for the mountainous areas and the gravity forces play a great role than material types (\u0026Ccedil;ellek, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). However, negative importance factor comes to be plan curvature (PC).\u003c/p\u003e \u003cp\u003eGully erosion hazard-predictor importance is shown in Figure (5b). Results showed that the top predictors-controlled gully erosion susceptibility is TWI followed by VD, EL, NDVI, SD, PrC, S, and R. The less important predictors are aspect (AS), TRI, and PC, followed by LULCs, and Lth.\u003c/p\u003e \u003cp\u003eThe analysis of flood hazard-predictor importance is shown in Figure (5c). Results showed that the top predictors that controlled flood susceptibility belonged to the hydrologic group, which was DtW, followed by equal importance of slope angle (S), elevation (EL), and stream density (SD), then followed by both TWI and rainfall (R), followed by normalized differential vegetation index (NDVI) and plan curvature (PC). Less importance predictors are LULCs, aspect (AS), and lithology (Lth). However, negative importance factor comes to be distance to road (DtR).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.3. Different susceptibility maps\u003c/h2\u003e \u003cp\u003eMaps of landslides, flood, and gully erosion susceptibility were generated based on both hazard-predictive factors (independent factors) and actual hazard locations (dependent variable) by applying various types of machine learning models (MLMs). Each susceptibility map was divided into five classes very low, low, moderate, high, and very high based on Natural Breaks Classifier-Jenks (Jenks and Caspall, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e1971\u003c/span\u003e) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). Landslide susceptibility maps for the Hasher-Fayfa Basin were created on the basis of RF, FDA, MDA, GLM, and BRT models (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e (a-e)); gully erosion susceptibility maps for the Hasher-Fayfa Basin were created according to the RF, FDA, MDA, GLM, and BRT models (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e (f-j)); and flood susceptibility maps were created using RF, FDA, and MDA models (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e (k-m)).\u003c/p\u003e \u003cp\u003eDetailed analysis was conducted for each model. Landslide hazard models indicated that high and very high susceptible regions based on RF and BRT covered more areas (21.2 and 22.5%, respectively) than the other three models (FDA\u0026thinsp;=\u0026thinsp;17.8%, MDA\u0026thinsp;=\u0026thinsp;18.6%, and GLM\u0026thinsp;=\u0026thinsp;17.6%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea). Gully erosion models indicated that high and very high susceptible classes in MDA and BRT, covering more areas (37.5 and 34.2%, respectively) than the other three models (RF\u0026thinsp;=\u0026thinsp;32.8%, FDA\u0026thinsp;=\u0026thinsp;31.4%, and GLM\u0026thinsp;=\u0026thinsp;32.6%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb). Flood models showed that high and very high susceptible classes in the MDA and FDA, covering more areas (15.0 and 14.1%, respectively) than RF model (RF\u0026thinsp;=\u0026thinsp;6.5%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ec).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.4. Models\u0026rsquo; validation\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e (a) and Table\u0026nbsp;(3) show the AUC curves and accuracy values for the landslide\u0026rsquo;s five ML models. Results show that landslides susceptibility models give high performance where AUC\u0026thinsp;\u0026gt;\u0026thinsp;0.9 (90%), the AUC values for ML models BRT, RF, GLM, MDA, and FDA are 0.909 (90.9%), 0.918 (91.8%), 0.923 (92.3%), 0.923 (92.3%), and 0.927 (92.7%), respectively. Results show that the most accurate ML model that has a high predictive accuracy for landslide susceptibility is FDA model with AUC value of 92.7%. Figure\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e(b) and Table\u0026nbsp;(3) show the AUC curves and values for the five ML models that were applied to create the gully erosion susceptibility models. Results show that gully erosion susceptibility models give AUC values range between 0.739 and 0.833 (medium to high performance). The ML models MDA, GLM, FDA, BRT and RF give AUC values of 0.739 (73.9%), 0.776 (77.6%), 0.792 (79.2%), 0.827 (82.7%), and 0.833 (83.3%), respectively. Results show that the most accurate ML model that has high predictive accuracy for gully erosion susceptibility is RF model with AUC value of 83.3%. Figure\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e(c) and Table\u0026nbsp;(3) show the AUC curves and values for the three ML models that were used to generate the flood susceptibility maps. Results show that flood susceptibility models give AUC values range between 0.932 and 0.972 (high performance). The ML models MDA, FDA, and RF give AUC values of 0.932 (93.2%), 0.945 (94.5%), and 0.972 (97.2%), respectively. Results show that the most accurate ML algorithm that has high predictive performance for flood susceptibility is RF model with AUC value of 97.2%.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e4.5. Multi-hazard risk map\u003c/h2\u003e \u003cp\u003eTo produce multi-hazard risk map for these three hazard types, three steps were applied: 1) hazard susceptibility map to each hazard type (LS, GE, and FL) was created according to the relationship between the independent factors (hazard-predictors) and the dependent factors (hazard locations of landslides, gully erosions, and floods) using various types of machine learning techniques (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e), 2) the ROC-AUC method was used to determine the highest performance models for each hazard type and then integrated to map multiple hazards (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The most accurate models in this study are FDA model for landslides where AUC\u0026thinsp;=\u0026thinsp;92.7%, RF for gully erosion with AUC\u0026thinsp;=\u0026thinsp;83.3%, and RF for floods with AUC\u0026thinsp;=\u0026thinsp;97.2%. These three susceptibility maps of the three natural hazards (landslides, gully erosion, and floods) were combined in ArcGIS 10.8 to produce an integrated multi-hazard risk map.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMulticollinearity results of hazard-predictor factors.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eHazard-causative factors\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eLS\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003eFL\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eGE\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTolerance\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVIF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTolerance\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eVIF\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eTolerance\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eVIF\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eElevation (EL)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.290\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.448\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.257\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.898\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.476\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2.101\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSlope (S)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.316\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.168\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.271\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e3.697\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTopographic wetness index (TWI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.307\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.258\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.304\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.290\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.326\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e3.067\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlan curvature (PC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.869\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.151\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.926\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.080\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.446\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2.243\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAspect (A)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.677\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.477\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.897\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.115\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.186\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTerrain ruggedness index (TRI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e49.302\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProfile curvature (PrC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.523\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.911\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eValley depth (VD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.502\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.991\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRainfall (R)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.719\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.391\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.716\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.396\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistance to wadis (DtW)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.461\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.168\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.398\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.511\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStream density (SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.723\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.384\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.613\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.632\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNormalized difference vegetation index (NDVI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.382\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.620\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.460\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.176\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.699\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.430\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLand use/Landcover (LULC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.660\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.514\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.741\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.350\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.703\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.422\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLithology (Lth)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.718\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.393\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.586\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.705\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.672\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.488\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistance to fault (DtF)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.842\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.188\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistance to road (DtR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.767\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.304\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.705\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.418\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAUC values for different machine learning models used in modeling landslide, gully erosion, and flood.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel Type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTest Result Variable(s)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eArea\u003c/p\u003e \u003cp\u003e(AUC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStandard Errors\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e\u003cb\u003eLandslides\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBRT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.054\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.918\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.052\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGLM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.923\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.051\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.923\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.050\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.927\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.049\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003e\u003cb\u003eGully\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003eErosion\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.739\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.064\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGLM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.776\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.059\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.792\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.057\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBRT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.827\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.053\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.833\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.051\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e\u003cb\u003eFloods\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.932\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.049\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.035\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.972\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eResults showed that the multi-hazard risk map included eight susceptibility classes (Figs.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e and \u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003e). These classes include; no-hazard or safe areas (No-H covers 33.6% of the study area). However, about 66.4% of the total area are subjected to various types of hazards as follows: landslides (LS\u0026thinsp;=\u0026thinsp;22.4%), gully erosion (GE\u0026thinsp;=\u0026thinsp;28.9%), floods (FL\u0026thinsp;=\u0026thinsp;1.6%), landslides-erosion (LS-GE\u0026thinsp;=\u0026thinsp;6.5%), floods-landslides (FL-LS\u0026thinsp;=\u0026thinsp;0.3%), floods-erosion (FL-GE\u0026thinsp;=\u0026thinsp;6.5%), and flood-landslides-erosion (FL-LS-GE\u0026thinsp;=\u0026thinsp;0.3%).\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eUnderstanding hazard management could be enhanced using different types of modeling approaches (Hill and Minsker, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). These models could provide planners and policymakers with useful, efficient, and informative results. Across the world, countless literature has individually studied different types of natural hazards (e.g., Amare et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Msabi and Makonyo, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). All these studies provide important and useful results. However, most natural hazards do not occur individually due to cascading effects. One hazard may lead to another; therefore, multiple hazards and their linkages, interactions and cascading impacts can provide much understanding of their processes and optimal ideas for averting and minimizing disaster losses and for effective land management (Godschall et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). These maps can provide valuable information that is critical for planning and managing existing and future human activities. Dealing with multi-hazards has shown that the interaction of numerous hazard types can cause a higher risk than the risk of a single hazard type (Liu et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn the current work, we investigated three hazards in a Hasher-Fayfa Basin (mountainous catchment) in Saudi Arabia. To evaluate disaster events in this area, multi-hazard modeling was performed. To achieve that five machine learning models (RF, FDA, MDA, GLM and BRT) for landslides and gully erosion and three models (RF, FDA and MDA) for floods. Based on (Guzzetti et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) classification, our study shows that the three best models had predictive ability of above 0.8 (good and excellent performances). The FDA model for landslides has an AUC\u0026thinsp;=\u0026thinsp;92.7% (excellent performance), RF for gully erosion has an AUC\u0026thinsp;=\u0026thinsp;83.3% (good performance), and RF for flooding has an AUC\u0026thinsp;=\u0026thinsp;97.2% (excellent performance).\u003c/p\u003e \u003cp\u003eThe MHM map was produced by coupling the results of the FDA (for landslides) and RF (for gully erosion and flooding) approaches. Our work is in agreement with that of Nachappa et al. (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) that used different machine learning approaches in multi-hazard evaluation. Also, our results in choosing RF for multi-hazard map that provides excellent accuracy for landslides AUC\u0026thinsp;=\u0026thinsp;0.93 and for floods AUC\u0026thinsp;=\u0026thinsp;0.97, are in agreement with the Nachappa et al. (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) in which they applied multi-hazard evaluation use of random forest (RF), which provides a suitable result for both flood (AUC\u0026thinsp;=\u0026thinsp;0.87) and landslide (AUC\u0026thinsp;=\u0026thinsp;0.90), and SVM, which provides AUC\u0026thinsp;=\u0026thinsp;0.87 for flood and AUC\u0026thinsp;=\u0026thinsp;0.89 for landslides. They constructed the multi-hazard map using RF and SVM and produced an accurate model for planners and managers. The results show that no hazard areas cover approximately 33.5% of the study area. However, 66.5% of the total area is probably affected by one or more hazards.\u003c/p\u003e \u003cp\u003eOur findings indicating the application of the optimal MLTs to predict multiple problems provides crucial informative data about their interactions. These relationships are highly correlated with the scale of the indicators used in the analysis and the specific types of hazards in the area. The current study bridged the gap between the different hazards by fully identifying the cascading interactions between these different types of natural hazards.\u003c/p\u003e"},{"header":"6. Conclusions","content":"\u003cp\u003eThe western and southern parts of Saudi Arabia face various natural hazards. Many mountainous areas are inhabited by people and these areas are vulnerable to several (compound) natural hazards. The delineation of high-risk areas is the critical stage and the most difficult task for all developers, and decision-makers. In the current work, we applied various machine learning techniques (MLTs) to map individual hazards affecting the Hasher - Fayfa basin (landslides, gully erosion, and flooding). The final map of this study is a multi-hazard risk map created by linking the three predominant natural hazards in the area (landslides, gully erosion, and flooding).\u003c/p\u003e \u003cp\u003eThe multi-hazard risk map is divided into 8 zones, the no hazard zone (covers 20% of the study area). Three single hazard zones namely landslide hazard zone (3%), which is confined to high altitude and steeply sloping areas, gully erosion hazard zone (2%) which is concentrated along gullies and mainly characterized by soil and weathered materials that can be easily eroded by rainfall, and flood hazard zone (4%) which is confined to main wadis and mainly downstream parts as water collects from various tributaries. Three zones include two types of interacting hazards, such as landslide-flood hazard zone (5%), landslide-erosion hazard zone (9%), and flood- erosion hazard zone (10%). Finally, there is a multiple hazard zone affected by the three hazard types landslide - erosion - flood, which covers 10% of the study area.\u003c/p\u003e \u003cp\u003eThe machine learning methods achieved acceptable accuracy in predicting the different hazard types; so, the multi-hazard map was produced with a high confidence level. Due to population growth and expansion of infrastructure and urban areas to mountainous and floodplains, and to ensure sustainable development, multi-hazard assessments are essential. Our ability to produce multi-hazard map will help planners and decision-makers make concrete management decisions for existing and future developments. This will make communities more resilient and able to act proactively to minimize future damage not only caused by a single hazard type, but also that may result from cascades of hazards or combined hazards.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthors Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAhmed M. Youssef, Ali M. Mahdi, Mohamed M. Al-Katheri, Soheila Pouyan, Hamid Reza Pourghasemi designed\u0026nbsp;the experiments,\u0026nbsp;ran models, analyzed\u0026nbsp;the results, and wrote and reviewed the manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of competing interest\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have influenced the work reported in this paper.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAgar R (1987) The Najd fault system revisited: a two-way strike-slip orogeny in the Saudi Arabian shield. J Struct Geol 9:41\u0026ndash;48\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlQahtanya AM, Abubakarb IR (2020) Public perception and attitudes to disaster risks in a coastal metropolis of Saudi Arabia. Int J Disaster Risk Reduct 44:101422. Doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ijdrr.2019.101422\u003c/span\u003e\u003cspan address=\"10.1016/j.ijdrr.2019.101422\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlyami A, Dulong CL, Younis MZ, Mansoor S (2021) Disaster Preparedness in the Kingdom of Saudi Arabia: Exploring and Evaluating the Policy, Legislative Organisational Arrangements Particularly During the Hajj Period. Eur J Environ Public Health 5(1):em0053. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.29333/ejeph/8424\u003c/span\u003e\u003cspan address=\"10.29333/ejeph/8424\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmare S, Langendoen E, Keesstra S, Ploeg Mvd, Gelagay H, Lemma H, Zee SEATMvd (2021) Susceptibility to Gully Erosion: Applying Random Forest (RF) and Frequency Ratio (FR) Approaches to a Small Catchment in Ethiopia. Water 13:216. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/w13020216\u003c/span\u003e\u003cspan address=\"10.3390/w13020216\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaig SU, Rehman MU, Janjua NN (2021) District-level disaster risk and vulnerability in the Northern mountains of Pakistan, Geomatics. Nat Hazards Risk 12(1):2002\u0026ndash;2022. DOI: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/19475705.2021.1944331\u003c/span\u003e\u003cspan address=\"10.1080/19475705.2021.1944331\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBathrellos GD, Skilodimou HD, Chousianitis K, Youssef AM, Pradhan B (2017) Suitability estimation for urban development using multi-hazard assessment map. Sci Total Environ 575:119\u0026ndash;134\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBell R, Glade T (2004) Multi-hazard analysis in natural risk assessments. WIT Trans Ecol Environ 77:1\u0026ndash;10\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBout B, Lombardo L, van Westen CJ, Jetten VG (2018) Integration of two-phase solid fluidequations in a catchment model for flashfloods, debris flows and shallow slope failures. Environ Model Softw 105:1\u0026ndash;16. DOI. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.envsoft.2018.03.017\u003c/span\u003e\u003cspan address=\"10.1016/j.envsoft.2018.03.017\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBradley AP (1997) The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recogn 30:1145\u0026ndash;1159\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreiman L (2001) Random forests. Mach Learn 45(1):5\u0026ndash;32\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u0026Ccedil;ellek S (2020) Effect of the Slope Angle and Its Classification on Landslide. Nat. Hazards Earth Syst. Sci. Discuss. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5194/nhess-2020-87\u003c/span\u003e\u003cspan address=\"10.5194/nhess-2020-87\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, 2020\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Silva MMGT, Kawasaki A (2018) Socioeconomic Vulnerability to Disaster Risk: A Case Study of Flood and Drought Impact in a Rural Sri Lankan Community. Ecol Econ 152:131\u0026ndash;140\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuncan M, Edwards S, Kilburn C, Twigg J, Crowley K (2016) An interrelated hazards approach to anticipating evolving risk GFDRR (Ed.), The Making of a Riskier Future: How Our Decisions Are Shaping Future Disaster Risk, Global Facility for Disaster Reduction and Recovery, Washington, USA, pp.\u0026nbsp;114\u0026ndash;121\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEl-Haddad BA, Youssef AM, Pourghasemi HR, Pradhan B, El-Shater A, El-Khashab MH (2021) Flood susceptibility prediction using four machine learning techniques and comparison of their performance at Wadi Qena Basin, Egypt. Nat Hazards 105, 83\u0026ndash;114 (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11069-020-04296-y\u003c/span\u003e\u003cspan address=\"10.1007/s11069-020-04296-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFang Z, Wang Y, Duan G, Peng L (2021) Landslide Susceptibility Mapping Using Rotation Forest Ensemble Technique with Different Decision Trees in the Three Gorges Reservoir Area, China. Remote Sens 13(2):238. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/rs13020238\u003c/span\u003e\u003cspan address=\"10.3390/rs13020238\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFriedman JH (2001) Greedy function approximation: A gradient boosting machine. Ann Stat 29:1189\u0026ndash;1232\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGhorbanzadeh O, Blaschke T, Gholamnia K, Meena SR, Tiede D, Aryal J (2019) Evaluation of different machine learning methods and deep-learning convolutional neural networks for landslide detection. Remote Sens 11(2):196. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/rs11020196\u003c/span\u003e\u003cspan address=\"10.3390/rs11020196\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGhorbanzadeh O, Shahabi H, Mirchooli F, Kamran KV, Lim S, Aryal J, Jarihani B, Blaschke T (2020) Gully erosion susceptibility mapping (GESM) using machine learning methods optimized by the multi\u0026ndash;collinearity analysis and K-fold cross-validation. Geomatics Nat Hazards Risk 11(1):1653\u0026ndash;1678. DOI: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/19475705.2020.1810138\u003c/span\u003e\u003cspan address=\"10.1080/19475705.2020.1810138\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGill JC, Malamud BD (2014) Reviewing and visualizing the interactions of natural hazards. Rev Geophys 52(4):680\u0026ndash;722. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/2013RG000445\u003c/span\u003e\u003cspan address=\"10.1002/2013RG000445\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGill JC, Malamud BD (2017) Anthropogenic processes, natural hazards, and interactions in a multi-hazard framework. Earth Sci Rev 166:246\u0026ndash;269. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.earscirev.2017.01.002\u003c/span\u003e\u003cspan address=\"10.1016/j.earscirev.2017.01.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGodschall S, Smith V, Hubler J, Kremer P (2020) A Decision Process for Optimizing Multi-Hazard Shelter Location Using Global Data. Sustainability 2020, 12, 6252. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/su12156252\u003c/span\u003e\u003cspan address=\"10.3390/su12156252\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoetz JN, Brenning A, Petschko H, Leopold P (2015) Evaluating machine learning and statistical prediction techniques for landslide susceptibility modeling. Comput Geosci 81:1\u0026ndash;11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreenwood WR (1982) Late Proterozoic Island-Arc Complexes and Tectonic Belts in the Southern Part of the Arabian Shield. Ministry of Petroleum and Mineral Resources, Riyadh, Saudi Arabia\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuzzetti F, Reichenbach P, Cardinali M, Galli M, Ardizzone F (2005) Probablistic Landslide Hazard Assessment at the Basin Scale. Geophys J Roy Astron Soc 72:272\u0026ndash;299. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geomorph.2005.06.002\u003c/span\u003e\u003cspan address=\"10.1016/j.geomorph.2005.06.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHasanean H, Almazroui M (2015) Rainfall: Features and Variations over Saudi Arabia. Rev Clim 3(3):578\u0026ndash;626. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/cli3030578\u003c/span\u003e\u003cspan address=\"10.3390/cli3030578\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHastie T, Tibshiran R, Leisch F, Hornik K, Ripley BD (2017) Mixture and flexible discriminant analysis. (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://cran.r-project.org/web/packages/mda/mda.pdf\u003c/span\u003e\u003cspan address=\"https://cran.r-project.org/web/packages/mda/mda.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHastie T, Tibshirani R (1996) Discriminant Analysis by Gaussian Mixtures. J. R. Stat. Soc. Ser. B 1996, 58, 155\u0026ndash;176\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe Q, Jiang Z, Wang M, Liu K (2021) Landslide and wildfire susceptibility assessment in southeast asia using ensemble machine learning methods. Remote Sens 13:1572\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHerrera VM, Khoshgoftaar TM, Villanustre F, Furht B (2019) Random forest implementation and optimization for Big Data analytics on LexisNexis\u0026rsquo;s high performance computing cluster platform. J Big Data 6:68. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40537-019-0232-1\u003c/span\u003e\u003cspan address=\"10.1186/s40537-019-0232-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHill DJ, Minsker BS (2010) Anomaly detection in streaming environmental sensor data: A data-driven modeling approach. Environ Model Softw 25:1014\u0026ndash;1022\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHosseiny H, Nazari F, Smith V, Nataraj C (2020) A Framework for Modeling Flood Depth Using a Hybrid of Hydraulics and Machine Learning. Sci Rep 10, 8222 (2020). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-020-65232-5\u003c/span\u003e\u003cspan address=\"10.1038/s41598-020-65232-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIglesias V, Braswell AE, Rossi MW, Joseph MB, McShane C, Cattau M, Koontz MJ, McGlinchy J, Nagy RC, Balch J, Leyk S, Travis WR (2021) Risky Development: Increasing Exposure to Natural Hazards in the United States. Earth's Future, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1029/2020EF001795\u003c/span\u003e\u003cspan address=\"10.1029/2020EF001795\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIPCC (2007) Climate change 2007. Synthesis report: Contribution of Working Groups I, II, and III to the fourth assessment report of the Intergovernmental Panel on Climate Change. Cambridge University Press. Retrieved from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.ipcc.ch/pdf/assessment-report/ar4/wg2/ar4-wg2-ts.pdf\u003c/span\u003e\u003cspan address=\"http://www.ipcc.ch/pdf/assessment-report/ar4/wg2/ar4-wg2-ts.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIPCC (2012) Managing the risks of extreme events and disasters to advance climate change adaptation: special report of the intergovernmental panel on climate change. The Intergovernmental Panel on Climate Change\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJanizadeh S, Avand M, Jaafari A, Phong TV, Bayat M, Ahmadisharaf E, Prakash I, Pham BT, Lee S (2019) Prediction Success of Machine Learning Methods for Flash Flood Susceptibility Mapping in the Tafresh Watershed, Iran. Sustainability 2019, 11, 5426. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/su11195426\u003c/span\u003e\u003cspan address=\"10.3390/su11195426\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJenks GF, Caspall FC (1971) Error on choroplethic maps: Definition, measurement, reduction Ann. Assoc Am Geogr 61(2):217\u0026ndash;244\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalantar B, Ueda N, Saeidi V, Ahmadi K, Halin AA, Shabani F (2020) Landslide Susceptibility Mapping: Machine and Ensemble Learning Based on Remote Sensing Big Data. Remote Sens. 2020, 12, 1737. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/rs12111737\u003c/span\u003e\u003cspan address=\"10.3390/rs12111737\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKomendantova N, Mrzyglocki R, Mignan A, Khazai B, Wenzel F, Patt A, Fleming K K (2014) Multi-hazard and multi-risk decision-support tools as a part of participatory risk governance: feedback from civil protection stakeholders. Int J Disaster Risk Reduct 8:50\u0026ndash;67. DOI: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ijdrr.2013.12.006\u003c/span\u003e\u003cspan address=\"10.1016/j.ijdrr.2013.12.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeonard M, Westra S, Phatak A, Lambert M, van den Hurk B, McInnes K, Risbey J, Schuster S, Jakob D, Stafford-Smith M (2014) A compound event framework for understanding extreme impacts Wiley Interdiscip. Rev Clim Chang 5:113\u0026ndash;128. DOI:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/wcc.252\u003c/span\u003e\u003cspan address=\"10.1002/wcc.252\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu B, Han X, Qin L, Xu W, Fan J (2021) Multi-hazard risk mapping for coupling of natural and technological hazards. Geomatics Nat Hazards Risk 12(1):2544\u0026ndash;2560. DOI: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/19475705.2021.1969451\u003c/span\u003e\u003cspan address=\"10.1080/19475705.2021.1969451\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLombardo F, Obach RS, DiCapua FM, Bakken GA, Lu J, Potter DM, Zhang Y (2006) A hybrid mixture discriminant analysis\u0026ndash;random forest computational model for the prediction of volume of distribution of drugs in human. J Med Chem 49(7):2262\u0026ndash;2267\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLombardo L, Tanyas H, Nicu IC (2020) Spatial modeling of multi-hazard threat to cultural heritage sites. Eng Geol 277:105776\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMsabi MM, Makonyo M (2021) Flood susceptibility mapping using GIS and multi-criteria decision analysis: A case of Dodoma region, central Tanzania, Remote Sensing Applications: Society and Environment, 21: 100445. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rsase.2020.100445\u003c/span\u003e\u003cspan address=\"10.1016/j.rsase.2020.100445\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNachappa TG, Ghorbanzadeh O, Gholamnia K, Blaschke T (2020) Multi-Hazard Exposure Mapping Using Machine Learning for the State of Salzburg. Austria Remote Sens 12:2757. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/rs12172757\u003c/span\u003e\u003cspan address=\"10.3390/rs12172757\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark S, Kim J (2019) Landslide susceptibility mapping based on random forest and boosted regression tree models, and a comparison of their performance.Appl Sci9:942Return to ref 2019 in article\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V et al (2011) Scikit-learn: machine learning in python. J Mach Learn Res. 2011;12:2825\u0026ndash;30\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR Core Team (2019) R: A Language and Environment for Statistical Computing. Foundation for Statistical Computing, Vienna: R [Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.R-project.org/\u003c/span\u003e\u003cspan address=\"https://www.R-project.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e]\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahman M, Ningsheng C, Islam MM, Dewan A, Iqbal J, Washakh RMA, Shufeng T (2019) Flood Susceptibility Assessment in Bangladesh Using Machine Learning and Multi-criteria Decision Analysis. Earth Syst Environ 3, 585\u0026ndash;601 (2019). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s41748-019-00123-y\u003c/span\u003e\u003cspan address=\"10.1007/s41748-019-00123-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRazavi-Termeh SV, Sadeghi-Niaraki A, Choi S-M (2020) Gully erosion susceptibility mapping using artificial intelligence and statistical models, Geomatics. Nat Hazards Risk 11(1):821\u0026ndash;844. DOI: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/19475705.2020.1753824\u003c/span\u003e\u003cspan address=\"10.1080/19475705.2020.1753824\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRidgeway G, Southworth MH (2013) RUnit S. Package \u0026lsquo;gbm.\u0026rsquo; Viitattu; 10:40\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoy J, Saha S (2019) Landslide susceptibility mapping using knowledge driven statistical models in Darjeeling District, West Bengal, India. Geoenviron Disasters 6:11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40677-019-0126-8\u003c/span\u003e\u003cspan address=\"10.1186/s40677-019-0126-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRusk J, Maharjan A, Tiwari P, Chen T-HK, Shneiderman S, Turin M, Seto KC (2022) Multi-hazard susceptibility and exposure assessment of the Hindu Kush Himalaya. Sci Total Environ 804:150039. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2021.150039\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2021.150039\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRutgersson A, Kjellstr\u0026ouml;m E, Haapala J, Stendel M, Danilovich I, Drews M, Jylh\u0026auml; K, Kujala P, Guo Lars\u0026eacute;n X, Halsn\u0026aelig;s K, Lehtonen I, Luomaranta A, Nilsson E, Olsson T, S\u0026auml;rkk\u0026auml; J, Tuomi L, Wasmund N (2021) Natural Hazards and Extreme Events in the Baltic Sea region, Earth Syst. Dynam Discuss. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5194/esd-2021-13\u003c/span\u003e\u003cspan address=\"10.5194/esd-2021-13\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSarkar T, Mishra M (2018) Soil erosion susceptibility mapping with the application of logistic regression and artificial neural network. J Geovis Spat Anal 2(1):8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSarker IH, Watters P, Kayes ASM (2019) Effectiveness analysis of machine learning classification models for predicting personalized context-aware smartphone usage. J Big Data. 2019;6(1):1\u0026ndash;28\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaunders WSA, Kilvington M (2016) Innovative land use planning for natural hazard risk reduction: A consequence-driven approach from New Zealand. Int J Disaster Risk Reduct 18:244\u0026ndash;255. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ijdrr.2016.07.002\u003c/span\u003e\u003cspan address=\"10.1016/j.ijdrr.2016.07.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchmidt J, Matcham I, Reese S, King A, Bell R, Henderson R, Smart G, Cousins J, Smith W, Heron D (2011) Quantitative multi-risk analysis for natural hazards: a framework for multi-risk modelling. Nat. Hazards 58, 1169\u0026ndash;1192 (2011)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eŞen Z, Al-Harithy S, As-Sefry S, Almazroui M (2017) Aridity and Risk Calculations in Saudi Arabian Wadis: Wadi Fatimah Case. Earth Syst Environ 1, 26 (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s41748-017-0030-x\u003c/span\u003e\u003cspan address=\"10.1007/s41748-017-0030-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShah AA, Khwaja S, Shah BA, Reduan Q, Jawi Z (2018) Living With Earthquake and Flood Hazards in Jammu and Kashmir, NW Himalaya. Front Earth Sci 6:179. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/feart.2018.00179\u003c/span\u003e\u003cspan address=\"10.3389/feart.2018.00179\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShalaby A, Tateishi R (2007) Remote sensing and GIS for mapping and monitoring land cover and land-use changes in the Northwestern coastal zone of Egypt. Appl Geogr. 27 28 \u0026ndash; 41 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apgeog.2006.09.004\u003c/span\u003e\u003cspan address=\"10.1016/j.apgeog.2006.09.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShin Y, Kim T, Cho H, Kang KI (2012) A formwork method selection model based on boosted decision trees in tall building construction. Autom Constr 23:47\u0026ndash;54\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSkilodimou HD, Bathrellos GD, Chousianitis K, Youssef AM, Pradhan B (2019) Multi-hazard assessment modeling via multi-criteria analysis and GIS: a case study. Environ Earth Sci 78, 47 (2019). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12665-018-8003-4\u003c/span\u003e\u003cspan address=\"10.1007/s12665-018-8003-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSolberg AHS (1996) Texture fusion and classification based on flexible discriminant analysis. In Proceedings of the 13th International Conference on Pattern Recognition, Vienna, Austria, 25\u0026ndash;29 August; 596\u0026ndash;600\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStoeser DB, Camp VE (1985) Pan-African microplate accretion of the Arabian shield. Geol Soc Amer Bull 96:817\u0026ndash;826\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun D, Wen H, Wang D, Xu J (2020) A random forest model of landslide susceptibility mapping based on hyperparameter optimization using Bayes algorithm. Geomorphology 362:107201. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.geomorph.2020.107201\u003c/span\u003e\u003cspan address=\"10.1016/j.geomorph.2020.107201\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTabari H (2020) Climate change impact on flood and extreme precipitation increases with water availability. Sci Rep 10:13768. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-020-70816-2\u003c/span\u003e\u003cspan address=\"10.1038/s41598-020-70816-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Westen CJ, Greiving S (2017) Multi-hazard risk assessment and decision making. Environmental Hazards Methodologies for Risk Assessment and Management. IWA publishing, London, UK, pp 31\u0026ndash;94\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Y, Fang Z, Hong H, Peng L (2020) Flood susceptibility mapping using convolutional neural network frameworks. J Hydrol 582:124482. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jhydrol.2019.124482\u003c/span\u003e\u003cspan address=\"10.1016/j.jhydrol.2019.124482\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Y, Fang Z, Wang M, Peng L, Hong H (2020) Comparative study of landslide susceptibility mapping with different recurrent neural networks. Comput Geosci 138:104445. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cageo.2020.104445\u003c/span\u003e\u003cspan address=\"10.1016/j.cageo.2020.104445\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWard PJ, Blauhut V, Bloemendaal N, Daniell JE, de Ruiter MC, Duncan MJ, Emberson R, Jenkins SF, Kirschbaum D, Kunz M, Mohr S, Muis S, Riddell GA, Sch\u0026auml;fer A, Stanley T, Veldkamp TIE, Winsemius HC (2020) Review article: Natural hazard risk assessments at the global scale. Nat Hazards Earth Syst Sci 20:1069\u0026ndash;1096. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5194/nhess-20-1069-2020\u003c/span\u003e\u003cspan address=\"10.5194/nhess-20-1069-2020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWastl M, St\u0026ouml;tter J, Kleindienst H (2011) Avalanche risk assessment for mountain roads: A case study from Iceland. Nat Hazards 56(2):465\u0026ndash;480\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeinkle J, Landsea C, Collins D, Musulin R, Crompton RP, Klotzbach PJ, Pielke R (2018) Normalized hurricane damage in the continental United States 1900\u0026ndash;2017. Nat Sustain 1(12):808\u0026ndash;813. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41893-018-0165-2\u003c/span\u003e\u003cspan address=\"10.1038/s41893-018-0165-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang A, Wang C, Pang G, Long Y, Wang L, Cruse RM, Yang Q (2021) Gully Erosion Susceptibility Mapping in Highly Complex Terrain Using Machine Learning Models. ISPRS Int J Geo-Information 10(10):680. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijgi10100680\u003c/span\u003e\u003cspan address=\"10.3390/ijgi10100680\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYe J, Chow J-H, Chen J, Zheng Z (2009) Stochastic gradient boosted distributed decision trees, in Proceedings of the ACM 18th International Conference on Information and Knowledge Management (CIKM '09), 2061\u0026ndash;2064, Hong Kong, November 2009\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYe T, Liu W, Mu Q, Zong S, Li Y, Shi P (2020) Quantifying livestock vulnerability to snow disasters in the Tibetan Plateau: Comparing different modeling techniques for prediction. Int J Disaster Risk Reduct 48101578. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ijdrr.2020.101578\u003c/span\u003e\u003cspan address=\"10.1016/j.ijdrr.2020.101578\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYoo C, Cho E (2019) Effect of Multicollinearity on the Bivariate Frequency Analysis of Annual Maximum Rainfall Events. Water 11:905. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/w11050905\u003c/span\u003e\u003cspan address=\"10.3390/w11050905\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang C, Xie Z (2012) Combining object-based texture measures with a neural network for vegetation mapping in the Everglades from hyperspectral imagery. Remote Sens Environ 124:310\u0026ndash;320\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao G, Pang B, Xu Z, Yue J, Tu T (2018) Mapping flood susceptibility in mountainous areas on a national scale in China. Sci Total Environ 615:1133\u0026ndash;1142. doi: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.scitotenv.2017.10.037\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2017.10.037\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou X, Wen H, Zhang Y, Xu J, Zhang W (2021) Landslide susceptibility mapping using hybrid random forest with GeoDetector and RFE for factor optimization. Geosci Front 12(5):101211. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.gsf.2021.101211\u003c/span\u003e\u003cspan address=\"10.1016/j.gsf.2021.101211\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Multi-hazards, Susceptibility, Future development, Sustainability, KSA","lastPublishedDoi":"10.21203/rs.3.rs-1554302/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-1554302/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe current study aimed at producing a multi-hazard risk map for Hasher-Fayfa Basin. The Basin is part of Jazan region in the southwestern Saudi Arabia and is distinguished by mountainous terrain. Recently, this area has experienced many extreme natural processes that become natural hazard events when it intersects with human activities (urban areas and infrastructures). This work is mapping the probabilities of three main hazards; landslides, floods, and gully erosion using machine learning algorithms named boosted regression tree (BRT), a generalized linear model (GLM), Flexible discriminant analysis (FDA), random forest (RF), and multivariate discriminant analysis (MDA). Several factors obtained from various sources, including topographical, geological, meteorological, hydrological, and human activities were incorporated to produce the final multi-hazard risk model. Area under the curve (AUC) was applied to identify the best predictive model for each natural hazard type. AUC values between 80 and 90% indicated that the model is very good and above 90% indicated that the model is excellent in predictive capability. Based on the accuracy evaluation, it was found that the FDAFDA model is the most accurate for predicting landslides with AUC values of (92.7%, excellent performance), RF model is the most accurate for predicting floods and erosion with AUC values of (97.2%-excellent and 83.3%-very good performance, respectively). Finally, multi-hazard risk map was prepared by coupling of the mentioned three hazards. Results showed that 33.5% of the total area is safe (no-hazard), whereas 66.5% is characterized by at least a single hazard and 2\u0026ndash;32\u0026thinsp;\u0026minus;\u0026thinsp;3 hazard combination. Machine learning approaches are useful tools as baselines for management and mitigation processes, based on multi-hazard modeling. It could help planners and decision-makers to manage future human activities and expansions.\u003c/p\u003e","manuscriptTitle":"Multi-hazards modeling using machine learning algorithms in Southwestern Saudi Arabia","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2022-05-25 15:10:54","doi":"10.21203/rs.3.rs-1554302/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9ce5f584-55ed-41e8-82eb-fe5448f79354","owner":[],"postedDate":"May 25th, 2022","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2022-06-21T01:09:17+00:00","versionOfRecord":[],"versionCreatedAt":"2022-05-25 15:10:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-1554302","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-1554302","identity":"rs-1554302","version":["v1"]},"buildId":"_2-kVJe1T_tPrBINL-cwx","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}