Next Article in Journal
Petrology of Lancang (Upper Mekong) River Sand
Previous Article in Journal
Upwellings and Mantle Ponding Zones in the Lower Mantle Transition Zone (660–1000 km)
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Geosciences
Volume 15
Issue 11
10.3390/geosciences15110414
geosciences-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Dual-Method Assessment of the Yarmouk Basin’s Groundwater Vulnerability Using SINTACS and Random Forest

by
Ibraheem Hamdan
1,2,*,
Ahmad AlShdaifat
1,
Majed Ibrahim
3,
Abdel Rahman Al-Shabeeb
3,
Rida Al-Adamat
3 and
A’kif Al-Fugara
4
1
Department of Applied Earth and Environmental Sciences, Faculty of Earth and Environmental Sciences, Al al-Bayt University, P.O. Box 130040, Mafraq 25113, Jordan
2
Environment, Water, and Energy Research Center, Al al-Bayt University, P.O. Box 130040, Mafraq 25113, Jordan
3
Department of GIS and RS, Faculty of Earth and Environmental Sciences, Al al-Bayt University, P.O. Box 130040, Mafraq 25113, Jordan
4
Department of Surveying Engineering, Faculty of Engineering, Al al-Bayt University, P.O. Box 130040, Mafraq 25113, Jordan
*
Author to whom correspondence should be addressed.
Geosciences 2025, 15(11), 414; https://doi.org/10.3390/geosciences15110414
Submission received: 16 September 2025 / Revised: 25 October 2025 / Accepted: 27 October 2025 / Published: 30 October 2025
(This article belongs to the Special Issue Advancing Groundwater Sustainability: Integrated Risk Assessment and Management Strategies)
Browse Figures
Versions Notes

Abstract

Water scarcity and increased human pressures are crucial issues facing Jordan. Chemical pollutants significantly influence groundwater quality in Jordan due to increased pollution risks, ease of contamination, and various human activities that release harmful compounds into the groundwater. The Yarmouk River Groundwater Basin (YRB) is one of the main basins in northern Jordan. It is exploited for domestic, drinking, agricultural, and industrial uses. This study assessed the groundwater vulnerability for the YRB through the implementation of a dual-method approach, employing the SINTACS intrinsic groundwater vulnerability model and the Random Forest (RF) machine learning method. The results revealed similarities and differences between the two models. The delineation of low-vulnerability zones was similar, suggesting that the intrinsic hydrogeological characteristics of these areas provide robust natural protection against contamination. In addition, both models suggest that the eastern, northern, and southern parts are areas of ‘high’ and ‘very high’ vulnerability. Subtle differences can be observed, particularly in the precise delineation of boundaries and the fragmentation of vulnerability zones. Generally, the results show that over (47%) and (43%) of the basin area falls into the high- and very high-vulnerability classes, while the very low and low classes make up about (14%) and (15%), based on the SINTACS and RF models, respectively. Using the SINTACS and RF groundwater vulnerability assessments in the YRB provides valuable insights into groundwater susceptibility in this critical area of Jordan. The identified high- and very high-vulnerability areas within YRB highlight the urgent need for protective measures to safeguard this vital groundwater resource for both present and future generations. The SINTACS model proves to be a reliable tool for intrinsic vulnerability assessment in the study area, consistent with its application in other parts of Jordan and similar dry regions.

1. Introduction

Water shortage is a significant socio-economic problem in several countries around the world. Temporal decreases in freshwater availability in Jordan are primarily due to increasing urbanization rates, intensive agricultural activities, climate change, and over-exploitation [1,2,3]. In Jordan, due to limited surface water resources, groundwater (GW) is the main source of water for agricultural, domestic, industrial, and other uses [4]. In recent decades, GW resources in Jordan drastically decreased, placing Jordan as one of the poorest countries in the world in terms of fresh water availability. According to the Jordanian Ministry of Water and Irrigation’s (MWI) national water strategy 2023–2040 [5], the share per capita of freshwater is 60 m3/capita/yr, compared to the international water scarcity line of 500 m3/capita/yr.
The water resource shortage is accompanied by the high population growth rate (1.9%) [6], the developmental growth in the country, refugee influxes, and GW overexploitation [7,8], putting stresses on this limited resource [9,10,11]. Furthermore, chemical contamination has become an important quality hazard for the limited GW resources in Jordan due to natural or anthropogenic activities, urging the need for alternative solutions [12].
Vulnerability assessments, one of the common methods of GW quality monitoring, provides outcomes that can provide information about the possibility of GW contamination which aids in decision making and proper management processes [13,14,15]. The GW vulnerability assessment was first introduced by Margat in 1968 [16] and was further developed to include various assessment models, such as DRASTIC [17], GLA [18], SINTACS [19], COP [20,21], PI [22], and VULK [23].
According to Mimi et al. [24], there is no universal and specific groundwater vulnerability assessment method at present. Banerjee et al. [25] described groundwater vulnerability as a non-absolute characteristic. It is a relative, dimensionless, and non-measurable property indicating zones where groundwater contamination is most likely to occur based on the geological, hydrogeological, and morphological characteristics of the aquifer. Nonetheless, the authors classified the GW vulnerability assessment methods into (1) statistical assessment methods, (2) process-based simulation assessment methods, and (3) overlay and index assessment methods. Goldscheider [26], on the other hand, classified groundwater assessments into (1) intrinsic vulnerability assessment methods and (2) specific groundwater vulnerability assessment methods. The intrinsic groundwater vulnerability methods take into consideration the natural characteristics of the groundwater system to determine the sensitivity of the groundwater to contamination generated by anthropogenic activities [27], while specific groundwater vulnerability methods take into consideration the characteristics of the natural geological and hydrogeological characteristics of the aquifer in combination with the pollutant itself and include simulation modeling to determine the “specific vulnerability” of groundwater for this pollutant [26,28].
Most GW vulnerability index models depend on (1) classification, (2) rating, and (3) the rating of different geomorphological, geological, hydrological, and hydrogeological parameters [29]. They have become widely accepted tools for decision-makers for better planning to protect the limited water resources from pollutants, and ensure that activities that could pose a potential hazard to groundwater resources are appropriately located in areas of low contamination risk [30,31,32].
Recently, researchers have been employing different Machine Learning (ML) models to aid in the interpretation of these traditional assessment methods. Support Vector Machine (SVM) and Random Forest (RF) algorithms are two commonly used algorithms for GW vulnerability assessment [33], GW potential mapping [34], and other hydrological studies [35,36]. RF is an ML method that produces assessment models through averaged repeated predictions [37].
This study aims to assess the GW vulnerability of the YRB through the implementation of a dual-method approach: the SINTACS model [19,38] and the RF algorithm simulated model. This study goes beyond the application of an existing model and provides a comparative framework, demonstrating the potential of AI-driven models to enhance the precision of groundwater vulnerability mapping, especially in complex and data-scarce arid regions like Jordan. In addition, the RF model is capable of providing a more accurate spatial delineation of vulnerability classes to support the outcomes of the traditional model, allowing for complex patterns to be captured and refined vulnerability mapping.

2. Study Area

2.1. Description of the Study Area

The YRB covers ca. 1280 km2 between (32°20′00″ to 32°45′00″ N) and (35°42′00″ to 36°23′00″ E) (Figure 1). A semi-arid Mediterranean climate prevails in the west, while the east of the basin is arid [39]. The elevation varies from 1130 m above sea level (m a.s.l) at the southwestern parts to 150 m below sea level at the northeastern parts, based on a 12.5 m resolution digital elevation model [40]. The average annual rainfall varies from 100 mm/yr in the eastern parts of the basin to 500 mm/yr in the northwestern parts [41].
The YRB receives most of its surface water flow from highlands in the north (Golan Heights in Syria) and the west, the Ajloun mountains [42,43]. According to Goode Et Al. [43], GW levels in YRB are declining dramatically.

2.2. Geological and Hydrogeological Setting

According to the Ministry of Energy and Mineral Resources reports, the geological formations in the basin include (1) the bedded limestone and dolomitic limestone Wadi As Sir formation (A7 aquifer), (2) the massive chalk and fossiliferous to coquinal yellow-gray limestone marly Wadi Umm Ghudran formation (B1 aquitard), (3) the alternating micritic limestone with chert and phosphatic chert, phosphatic limestone with chert, and phosphate Amman silicified limestone and Al-Hisa phosphorite formation (B2 aquifer), (4) the massive gray to white chalk marl Muwaqqar formation (B3 aquitard), (5) the alternating limestone, marl, and chalk with chert Um Rijam formation (B4 aquifer), (6) the chalky and marly limestone Wadi Shallala formation (B5 aquifer), (7) the Basalt (BA), and (8) the recent (Quaternary) fluvial and alluvial sediments (Figure 2) [44,45,46,47,48,49].
Water within the YRB is potentially found in two aquifers: the B4/B5 and the A7/B2, separated by the B3 aquitard. Although the B1 formation acts as an aquitard, according to Hobler et al. [50], BRGM [51], and Hamdan et al. [52], the B1 in northern Jordan is hydraulically connected to the underlying A7 formation and the overlying B2 formation due to the high number of fractures, fissures, and faults within its layers. Therefore, the uppermost unit of the Ajlun Group (A7 formation) and the lower part of the Belqa Group (B2 formation) are considered as one hydrogeological unit, namely, the A7/B2 Aquifer. The basalt in the study area is regarded as a recharge zone for the underlying A7/B2 aquifer. The longitudinal fractures within the basalts facilitate water infiltration into the underlying A7/B2 aquifer during recharge events. The B4/B5 layer plays a relatively important role in the northwestern parts of Jordan (Figure 3). Exploitation of the B4/B5 aquifer for domestic purposes in YRB is almost negligible because the groundwater quality often does not meet drinking water quality standards.

3. Methodology and Data Analysis

3.1. Data Requirement

To apply the SINTACS groundwater vulnerability model, the required data inputs are compiled, spatially processed, and evaluated for the YRB (Table 1). The Map Algebra Raster Calculator tool in ArcMap (V.10.8) (License Number: 10.7.0.10450) was used to process and analyze all data with a grid resolution of (30 × 30) m.

3.2. SINTACS Vulnerability Model

SINTACS is an acronym for the seven parameters that are utilized to determine the final groundwater vulnerability for YRB. These include the following: depth to groundwater (S1), net recharge (I), unsaturated zone (N), soil media (T), aquifer media (A), hydraulic conductivity (C), and topographic slope (S2). The Weighted Linear Combination (WLC) method is used to generate the final SINTACS Groundwater Vulnerability map for the YRB, where specific weights were assigned for each criterion. The final score is calculated by summing the values and multiplying the rate by the weight for each layer. Table 2 lists the SINTACS parameters along with their corresponding rates and weights.

3.3. Random Forest Machine Learning Model

Machine learning methods are among the most common predictive models, because they do not require a predefined statistical counting among variables [57]. Within this study, the Random Forest (RF) model is used together with the SINTACS model to evaluate the groundwater vulnerability within YRB.
Random Forest (RF) is one of the ensemble machine learning algorithms that is widely used for both classification and regression tasks. The trees in this model are built using random resampling along with model training data [58,59]. In the context of groundwater vulnerability evaluation, RF algorithms are increasingly employed due to their ability to handle complicated, non-linear relationships between input parameters and vulnerability, as well as their robustness to noisy data and multicollinearity. The Random Forest technique was used, with the Gini Index being the function to minimize so that each of the decision trees that it consists of has less impurity. The indicator is the proportion of times that a randomly chosen item is mislabeled, and the model would be driven to the most balanced branches [60]; in addition, the RF model was built and executed using the Scikit-learn library for Python (V 2.7) [61].
RF basically relies on building decision trees and the subsequent use of training data. This model is suitable for big datasets where no definition of the relationships amongst the variable is required [58]. This method employs Equation (1) [57]:
h   x , i k ,   k = 1 ,   2 ,   3 ,   4 ,   … … … n
where (h) is the Classification index, ( i k ) are the SINTACS parameters, and (1, 2, …n) are vectors.
The procedure for running the RF model entails the following:
1.
Data requirement: hydrogeological, land use, and topographic parameters are required. These often comprise effective infiltration, groundwater depth, soil type, unsaturated zone characteristics, hydraulic conductivity, aquifer media, topographic gradient, rainfall, and land use/cover. These parameters are typically represented as spatial layers in a geographic information system (GIS). For supervised learning, an RF model requires a key variable representing groundwater vulnerability. This variable can come from existing vulnerability maps (for example, from well-known indices such as SINTACS or DRASTIC, which serve as proxy indicators of vulnerability) or, ideally, from detailed groundwater data and quality information (such as nitrate levels or pesticide presence) that indicate contamination levels. The target variable is often classified into specific vulnerability categories, such as very low, low, medium, extreme, and very high. The primary step involves extracting data points, including grid cells and pixels, from all parameter layers and the final vulnerability layer. These sampling factors create training and testing datasets. It is important to ensure that the training data accurately represents the full range of vulnerable situations across the study area.
2.
Model Training and Validation: The dataset was split into 80% training and 20% testing, ensuring a comprehensive evaluation of the model [62]. The model used 6000 random points for training and testing. In this work, the area under the receiver operating characteristic curve (AUC-ROC) measured the overall performance of the radio frequency (RF) model after training. Performance comparisons often use common metrics for classification methods, including accuracy, precision, recall, and F1 score. Random Forest (RF) model performance was compared with overall performance based on Area Under the Receiver Operating Characteristic Curve (AUC-ROC). The AUC-ROC curve was plotted and attained an extremely high AUC of 0.97, indicating the extremely high ability of the model to discriminate between different groundwater vulnerability classes [63].
3.
Hyperparameter Analysis: To obtain the Random Forest (RF) model to run best and to prevent overfitting, hyperparameter tuning was carried out [64]. The most important parameters were tuned (n_estimators, max_depth, min_samples_split and min_samples_leaf). The optimal set of these hyperparameters was found using grid search with 5-fold cross-validation (GridSearchCV) on the training set. The algorithm iteratively goes through a given grid of parameter settings and measures each setting by finding the performance through cross-validation to prevent overfitting. Usually, hyperparameters are the parameter value settings before learning [65]. The process of hyperparameter tuning is used to identify the optimal set of parameter values so that model performance, accuracy, and generalization abilities can be improved [66].
4.
Vulnerability Mapping: After training and validating the RF model, it predicts groundwater vulnerability for each pixel or grid cell in the study area based on the input parameter values for each cell. These results are displayed in a vulnerability map. The RF model produces a continuous vulnerability index, which is then categorized into discrete vulnerability levels (ranging from very low to very high) using appropriate normalization methods to create the final vulnerability map.

4. Results and Discussion

Groundwater pollution reduces the suitability of abstracted groundwater for drinking purposes and can affect groundwater-dependent ecosystems [67]. Different sources of anthropogenic pollutants can affect groundwater quality. Pollution from agricultural activities is considered a widespread source of groundwater pollution, often including large amounts of nitrate, pesticides, and other chemicals that can affect and pollute groundwater [68].

4.1. SINTACS Model Parameters

  • Depth to Groundwater (Sottosuolo) (S1): Groundwater pollution can affect the exploitable amount of freshwater, leading to water scarcity problems [69]. Depth to groundwater is an important parameter in groundwater vulnerability analysis, as deeper groundwater levels imply a longer travel time and more natural attenuation for pollutants before reaching the groundwater [10,70,71]. In the YRB, the depth to water level ranges from 41 m and 311 m (Figure 4a). This parameter was given the rate value of (1) for depth to GW in the basin, suggesting it is generally deep with slight variations (Table 2).
  • Infiltration Rate (Infiltrazione) (I): The infiltration rate plays a crucial role in groundwater vulnerability assessment tools. It controls the amount of water that penetrates the ground surface and reaches the aquifer. Specifically, it governs the vertical transport of contaminants through the unsaturated zone toward the aquifer. According to Piscopo [72], it also determines the volume of water available for dispersing and diluting contaminants in the vadose and saturated zones. The infiltration rate was calculated based on the assumed recharge percentage for each aquifer in the basin, following procedures outlined by Hobler et al. [50]. The recharge percentages were assumed to be 20% for the A7/B2 aquifer, 6% for the Basalt and B4/B5 aquifers, and 0% for the B3 aquitard, relative to the total rainfall. An intersection between the aquifers map and the rainfall map was applied, and the infiltration rate was calculated accordingly (Figure 4b); the rates were as listed in Table 2 (Figure 5a).
  • Unsaturated Zone (Non Saturo) (N): The unsaturated zone reflects the system’s natural attenuation and purification capacity against pollutants reflected by the physico–chemical interactions between the pollutants and the rocks within this zone [32,52,73]. The lithological characteristics of the outcropping geological formations in the YRB were examined, using the geological maps (scale 1:50,000), to evaluate the effect of the unsaturated zone (Figure 4c). The geological formations (Figure 5b) are listed in Table 2.
  • Soil Media (Tipologia Copertura) (T): Soil texture describes and controls the runoff potential, infiltration capabilities, and permeability rates of the area [74,75]. According to Tsegay et al. [70], soil texture influences the potential dispersion and purification of contaminants and reduces pollutants’ ability to move vertically into the vadose zone. Within the YRB, five soil texture types are present: clay, clay loam, silty clay, silty clay loam, and silty loam (Figure 4d). The rates are specified in Table 2 (Figure 5c).
  • Aquifer Media (Acquifero) (A): Aquifer characteristics describe the process that happens in the phreatic zone, when a contaminant infiltrates the soil and the unsaturated zone and mixes and interacts with groundwater [54]. These processes include dispersion, sorption, dilution, and chemical reactions between the rock and the contaminants [76]. Within the YRB, three aquifers are present: A7/B2, B4/B5, and the Basalt Aquifers (Figure 4e). The assigned rates are listed in Table 2 (Figure 5d).
  • Hydraulic Conductivity (Conducibilità Idraulica) (C): Hydraulic conductivity represents the fluid movement and speed through the interconnected voids, bedding planes, and fractures within the aquifer [10,55,77]. It regulates the contamination level within the aquifer; with increasing aquifer hydraulic conductivity, the rate at which contaminants are transferred within the aquifer increases. According to Hobler et al. [50], the hydraulic conductivity within the study area was 1.73 m/d for the A7/B2 aquifer, 8.6 × 10−5 m/d for the B3 aquitard, 4.3 m/d for the B4/B5 aquifer, and 34.56 m/d for the Basalt aquifer (Figure 4f and Figure 5d).
  • Topographic Slope (Superficie Topografica) (S2): Slope parameter affects the groundwater vulnerability map indirectly compared to other geological and hydrogeological parameters, through its control over both water and pollutant infiltration rates. Areas with low slopes tend to retain water and pollutants for a longer time, allowing for a greater infiltration of water and pollutants. The digital elevation model (12.5 m resolution [40]) was used to generate the slope map of the study area (Figure 4g and Figure 5f). The resulting slope map was classified according to the slope classes presented in Table 2.

4.2. SINTACS Vulnerability Map

After preparing all rate maps for the SINTACS model parameters, each rate map was multiplied by its specific weight (Figure 6, based on Table 2). The Map Algebra “Raster calculator” tool was used in ArcMap (V. 10.8) to calculate the final SINTACS index map by multiplying the rate map by the weight for each parameter using Equation (2) [19]:
I S I N T A C S =   ∑ i = 1 7 R i W i
where (ISINTACS) is the resulting SINTACS index value, (Ri) is the rate for each parameter, and (Wi) is the weight for each parameter. The depth to groundwater (rate × weight) map was assumed to have a value of 5 because, as mentioned above, the rate for the entire area was 1 and the weight was 5.
To better understand the SINTACS values, these values were normalized using Equation (3) to (1 to 100) and then divided into five equal classes.
I n = I x − I m i n I m a x + I m i n × 100
where (In) is the normalized SINTACS value, (Imin) is the minimum SINTACS index value, and (Imax) is the maximum SINTACS index value.
The infiltration rate map (R × W) values, ranging from 4 to 20, indicate the relative impact of infiltration on vulnerability. Higher RW values (e.g., 16, 20) signify areas with high infiltration rates where water, and potentially contaminants dissolved within it, can rapidly move from the surface to groundwater. The unsaturated zone, with (R × W) values ranging from 10 to 45, has a varied protective capacity where elevated values suggest increasing groundwater vulnerability. The outcomes suggest that some localities in the basin offer minimal natural protection, probably due to geological formations [77]. The hydraulic conductivity, with (R × W) values from 3 to 27, implies faster groundwater flow, which may lead to the rapid spread of contaminants, as reflected by the higher values [78]. The slope map shows (R × W) values in the range of 2 to 20, suggesting an influence on surface runoff. Steeper slopes (higher values) generally indicate higher runoff and reduced infiltration, and vice versa. The wide range of slopes in the Yarmouk River Basin indicates diverse hydrological responses to precipitation events [79].
These maps collectively provide a detailed understanding of the hydrogeological and topographical factors influencing groundwater vulnerability in the YRB. Areas exhibiting higher R × W values for parameters such as infiltration rate, unsaturated zones, coarser soil textures, highly permeable aquifer media, and higher hydraulic conductivity are intrinsically more susceptible to contamination. The spatial heterogeneity observed across all six maps underscores the complex interplay of these factors and emphasizes the need for a comprehensive, integrated assessment to accurately delineate groundwater vulnerability zones. These maps serve as the foundational layers upon which the final vulnerability assessment is built, allowing for a nuanced understanding of the contributing factors to overall vulnerability.
The Normalized SINTACS Groundwater Vulnerability Map for the YRB (Figure 7) synthesizes the information from the individual parameter maps into a single, comprehensive assessment. The map categorizes the basin into five distinct vulnerability classes: very low, low, moderate, high, and very high. The map shows that nearly 50% of the entire area is characterized as highly and very highly vulnerable areas, reflecting the unsaturated zone and aquifer rock type, which is mainly limestone. In comparison, it can be observed that Muwaqqar chalk played a significant role in protecting groundwater from pollutants because it is considered an aquitard and a protective layer for groundwater from various pollution sources.
The SINTACS model has been previously implemented to assess GW vulnerability in Jordan, albeit with some modification. Studies reported successful implementation for areas such as Wadi Shueib and the Jordan Valley, with results consistent with the outcomes of the current study. The reasons for zones being identified as having high and very high vulnerability were attributed to factors such as shallow aquifers, fractured bedrock, and intensive agricultural practices [77,80].
The YRB hydrogeological settings are similar to other regions in Jordan. The groundwater vulnerability in the country is influenced by highly permeable aquifer media, varying soil textures, and unsaturated zones [78,79].

Spatial Distribution of Vulnerability

A heterogenous distribution of vulnerability risks is observed in the final YRB map. This spatial variability is a direct reflection of the diverse hydrogeological and topographical conditions discussed in the individual parameter maps. The delineation of these zones is crucial for targeted groundwater protection strategies.
The very low- and low-vulnerability zones, depicted in dark green and light green, respectively, are predominantly found in the central–western parts of the basin. These regions likely possess a combination of favorable intrinsic characteristics that offer significant natural protection to the groundwater. This could be related to the unsaturated zones, less permeable soil textures, deeper water tables, or aquifer media with lower hydraulic conductivity. Such areas are naturally resilient to contamination and may require less intensive management interventions, though continuous monitoring is still essential [81].
The moderate-vulnerability zones, yellow areas on the map, are more widespread and often act as transitional areas between regions of lower and higher vulnerability. Management in these zones could focus on preventative measures and planning to avoid increased vulnerability [82].
The very high- (red) and high-vulnerability (orange) zones, prominent in the southern, northern, and eastern parts of the basin, are characterized by increased susceptibility to groundwater contamination. The widespread nature of these zones suggests the need for urgent protective measures, such as enforced land use policies, managed waste control, and MAR activities [83].

4.3. Random Forest Machine Learning Map

The values for the final vulnerability map using the Random Forest machine learning model ranged between 5 and 25, where the random forest model assumes weights based on the importance of each parameter. To compare the results of the calculated SINTACS model map and the SINTACS map calculated by the Random Forest machine learning model, the resulting map from the Random Forest was also normalized based on Equation (3) to values from 1 and 100. The normalized map for RF was divided and classified into 5 equal classes (Figure 8). The RF vulnerability map exhibits a spatial distribution of vulnerability classes that, at first glance, appears broadly similar to the SINTACS map. There are distinct zones of varying vulnerability levels across the basin. Similarly to the SINTACS map, areas classified as ‘very low’ (dark green) and ‘low’ (light green) vulnerability are concentrated in the central–western part of the basin. This suggests that these regions consistently demonstrate intrinsic protective characteristics, regardless of the modeling approach. The ‘moderate’ vulnerability areas (yellow) are also widely distributed, forming transitional zones. Their extent and location seem comparable to those identified by the SINTACS model. The ‘high’ (orange) and ‘very high’ (red) vulnerability zones are again prominent, particularly in the northern, eastern, and southern parts of the basin. These areas continue to be identified as critical regions with high susceptibility to groundwater contamination. However, a closer visual inspection reveals some subtle differences in the precise delineation and fragmentation of these high-vulnerability zones compared to the SINTACS map. The RF model might show more granular or less continuous patches of high vulnerability, or vice versa, depending on its learning patterns and how it interprets the complex relationships between the input parameters.

4.4. Comparison Between the SINTACS Map and the RF Machine Learning Model Map

While both the SINTACS and RF maps generally agree on the broad spatial patterns of vulnerability, there appear to be some differences in the finer details of the classification boundaries. The RF model, being a data-driven approach, might capture more complex, non-linear relationships between the input parameters and vulnerability, potentially leading to a more nuanced or localized classification. For instance, some areas that were uniformly classified in the SINTACS map might show more variability in the RF map, or vice versa. These differences could be attributed to the distinct methodologies: SINTACS relies on expert-assigned weights and ratings, while RF learns these relationships directly from the data. This preliminary visual comparison suggests that while the overall trends are consistent, the RF model might offer a refined or alternative perspective on the spatial distribution of groundwater vulnerability in the YRB.
Specific similarities are observed for low-vulnerability zones. Both models identify the central western parts of the basin as having ‘very low’ and ‘low’ groundwater vulnerability. This suggests that the intrinsic hydrogeological characteristics of this region, such as potentially deeper water tables, the unsaturated zones, or less permeable aquifer media, provide robust natural protection against contamination, which is recognized by both expert-driven and data-driven approaches [84].
Similarly, both models suggest that the eastern, northern, and southern parts are areas of ‘high’ and ‘very high’ vulnerability. This consistency indicates that these regions possess inherent hydrogeological conditions (e.g., shallow aquifers, highly permeable formations, high infiltration rates) that make them intrinsically susceptible to pollution, a fact that is robustly captured by both methodologies [85]. These similarities underscore the fundamental hydrogeological controls on groundwater vulnerability in the Yarmouk River Basin. Regardless of the modeling approach, certain areas are inherently more or less vulnerable due to their natural characteristics.
On the other hand, subtle differences can be observed between the SINTACS and RF maps, particularly in the precise delineation of boundaries and the fragmentation of vulnerability zones. Both maps were used to calculate the areas for the different vulnerability classes to observe any differences in their delineation (Table 3). The SINTACS model relies on assigning weights and ratings to each parameter based on expert knowledge and predefined ranges. This approach is transparent and easily interpretable [86]. The ML RF model primarily learns the input parameters relationships and produces the vulnerability assessment directly from the trained data. Thus, these models resolve complex relationships that are not detected using traditional models, leading to more accurate delineation of the resultant zones [87].
Indeed, studies have reported that RF models produce more accurate groundwater vulnerability assessment, compared to traditional models [35,88]. Thus, the combination of these models could be viewed as a complementary approach rather than being mutually exclusive [88,89]. Furthermore, the application of RF in arid environments like Jordan is evident in other regions [84].
While the SINTACS and RF models provide broadly similar spatial patterns of groundwater vulnerability in the Yarmouk River Basin, the RF model likely offers a more nuanced and data-driven delineation of these zones. The literature supports the idea that machine learning models like RF can enhance the accuracy of vulnerability assessments, particularly by capturing complex relationships. However, the choice of model often depends on data availability, the desired level of interpretability, and the specific objectives of the assessment. For comprehensive groundwater management, integrating insights from both traditional and advanced modeling approaches can provide a more robust and reliable understanding of groundwater vulnerability.

4.5. Model Validation

The ML models demonstrate high accuracy, as shown in the most important area below the AUC curve on the ROC plot (Figure 9). In the ROC curve, the X-axis represents False Positives (FP), while the Y-axis shows True Positives (TP). TP indicates the proportion of correctly identified positive cases, whereas FP indicates the proportion of incorrectly predicted positives. AUC values range from 0.5 to 1. Whereas values close to 0.5 indicate poor performance, values near 1 suggest better accuracy [57]. The results showed that the AUC scores for each method are as follows: RS (97%) and SINTACS (84%).

4.6. Anthropogenic Pressures and Vulnerability

The literature consistently emphasizes the role of anthropogenic activities in exacerbating groundwater vulnerability in Jordan. Rapid population growth, coupled with expanding agricultural and industrial sectors, leads to increased pollutant loads from sources such as untreated wastewater, agricultural runoff (pesticides, fertilizers), and industrial effluents [81]. The high- and very high-vulnerability zones identified in the YRB are likely to be areas where these anthropogenic pressures coincide with intrinsic hydrogeological susceptibility, leading to a heightened risk of contamination. This necessitates not only intrinsic vulnerability mapping but also a comprehensive risk assessment that integrates hazard mapping (sources of pollution) with vulnerability [31,90]. LULC plays an important role in groundwater vulnerability. It controls the water and pollutant recharge rates and patterns towards aquifers. Additionally, urban and industrial areas can significantly increase groundwater vulnerability to pollutants due to high concentrations of pollutants in wastewater and industrial water. The illegal dumping of industrial water in wadis or leakage from households’ septic tanks allows pollutants to percolate more directly.
A spatial join between the SINTACS groundwater vulnerability map and the LULC map for the YRB was performed. Table 4 displays the percentage of area coverage for LULC within each vulnerability class.

4.7. Limitations and a Proposal for Future Work

The current study provides insights into the groundwater vulnerability of the YRB using the SINTACS intrinsic groundwater vulnerability model and RF machine learning model. Within this study, several limitations should be acknowledged. First, the analysis primarily relied on intrinsic vulnerability parameters without incorporating direct groundwater quality measurements, especially those indicative of potential contamination. The absence of hydro-chemical data limits the ability to validate the predicted results from both models against observed contamination patterns.
Additionally, uncertainties related to the spatial resolution and accuracy of input datasets may influence the delineation of groundwater vulnerability zones. Within this study, the grid data had a spatial resolution of 30 m × 30 m. Higher-resolution data may enhance the model results.
Future research should therefore integrate hydro-chemical analyses of the YRB to better validate and refine the results of SINTACS and RF models. Incorporating historical quality data with a temporal monitoring of water quality from the wells would also enhance the understanding of changes in groundwater quality in YRB. Moreover, future work could benefit from coupling intrinsic vulnerability models with process-based or flow models to account for contaminant-specific behaviors and to improve predictive accuracy. Moreover, integrating the socio-economic factors and land use change scenarios with the results from SINTACS and RF vulnerability assessment models could further support sustainable groundwater management and policy development in the YRB.

5. Conclusions

The results of this study provide significant information for effective groundwater protection strategies and management in the YRB. Furthermore, these results are in agreement with the other literature promoting integrated water resource management in Jordan through scientific research, policy implementation, and community engagement.
In conclusion, the SINTACS and RF assessment models provide contribute to the understanding of groundwater susceptibility in this GW basin in Jordan. The results highlight consistent patterns of vulnerability across Jordan due to hydrogeological settings and anthropogenic activities. The outcomes point toward the need for targeted measures for land use planning, pollution control, GW monitoring, and the importance of public awareness. These measure, if put in place, could aid in maintaining the scarce water resources in Jordan.

Author Contributions

Conceptualization, I.H., A.A. and M.I.; methodology, I.H., A.A. and M.I.; software, I.H. and M.I.; validation, I.H., A.A., M.I. and A.A.-F.; formal analysis, I.H., A.A. and R.A.-A.; investigation, I.H. and A.A.; resources, I.H.; data curation, I.H., A.A. and M.I.; writing—original draft preparation, I.H., A.A., M.I., A.R.A.-S., R.A.-A. and A.A.-F.; writing—review and editing, I.H., A.A., M.I., A.R.A.-S., R.A.-A. and A.A.-F.; visualization, I.H.; supervision, I.H. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

All data used for this manuscript are provided in the text. The GIS data used for the modeling and the RF model data will be available upon request.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Alqadi, M.; Al Dwairi, A.; Merchán-Rivera, P.; Chiogna, G. Presentation of DeMa (Decision Support Software and Database for Wellfield Management) and Its Application for the Wadi Al Arab Wellfield. Water 2023, 15, 331. [Google Scholar] [CrossRef]
  2. Polemio, M.; Voudouris, K. Groundwater resources management: Reconciling demand, high quality resources and sustainability. Water 2022, 14, 2107. [Google Scholar] [CrossRef]
  3. UN-Water; UNESCO. The United Nations World Water Development Report 2022—Making the Invisible Visible; UN: Paris, France, 2022. [Google Scholar]
  4. Almomani, T.; Al Shraydeh, S.; Shakhatreh, H.; Alroud, R.; Ali, B.; Obayat, A.; Atyeh, A.; Mohammad, A.; Alta’ani, A.; Sa’aydeh, H.; et al. Water Year Book 2016/2017; Ministry of Water and Irrigation: Amman, Jordan, 2018; Available online: https://www.deutsche-rohstoffagentur.de/EN/Themen/Wasser/Projekte/abgeschlossen/TZ/Jordanien/Was_tz_gwrm_fig1_g_en.html (accessed on 17 July 2025).
  5. MWI (Ministry of Water and Irrigation). National Water Strategy 2023–2040; Ministry of Water and Irrigation: Amman, Jordan, 2023. Available online: https://www.mwi.gov.jo/EBV4.0/Root_Storage/AR/EB_List_Page/national_water_strategy_2023-2040.pdf (accessed on 17 July 2025).
  6. Ananbeh, F. DOS 2024, Jordan Department of Statistics. Available online: https://dosweb.dos.gov.jo/ar/news_07_2024/ (accessed on 17 July 2025).
  7. Aljaradin, M.; Bashitialshaaer, R.; Alitawi, H.; Amaireh, M. Water quality, availability and potential of geothermal energy utilization, Afra Water, Jordan. Sustain. Resour. Manag. J. 2017, 1, 17–25. [Google Scholar]
  8. Margane, A.; Alqadi, M.; Alkurdi, O. Updating the Groundwater Contour Map of the A7/B2 Aquifer in North Jordan; Ministry of Water and Irrigation (MWI): Amman, Jordan, 2015; Available online: https://www.deutsche-rohstoffagentur.de/EN/Themen/Wasser/Projekte/abgeschlossen/TZ/Jordanien/syr_report.pdf?__blob=publicationFile&v=2 (accessed on 17 July 2025).
  9. Brückner, F.; Bahls, R.; Alqadi, M.; Lindenmaier, F.; Hamdan, I.; Alhiyari, M.; Atieh, A. Causes and consequences of long-term groundwater overabstraction in Jordan. Hydrogeol. J. 2021, 29, 2789–2802. [Google Scholar] [CrossRef]
  10. Hamdan, I.; Al-Rawabdeh, A.; Al Hseinat, M. Groundwater Vulnerability Assessment for the Corridor Wellfield Using DRASTIC and Modified DRASTIC Models: A Case Study of Eastern Jordan. Open J. Geol. 2020, 10, 991–1008. [Google Scholar] [CrossRef]
  11. Al Naber, M.; Molle, F. Controlling groundwater over abstraction: State policies vs local practices in the Jordan highlands. Water Policy 2017, 19, 692–708. [Google Scholar] [CrossRef]
  12. Muneer, M.; Khan, M.A.; Shinwari, F.U.; Ahmed, I.; Siyar, S.M.; Alshehri, F.; Shahab, M. Assessment of groundwater intrinsic vulnerability using GIS-based DRASTIC method in district Karak, Khyber Pakhtunkhwa, Pakistan. Front. Water 2025, 7, 1540703. [Google Scholar] [CrossRef]
  13. Pronk, M.; Goldscheider, N.; Zopfi, J.; Zwahlen, F. Percolation and particle transport in the unsaturated zone of a karst aquifer. Groundwater 2009, 47, 361–369. [Google Scholar] [CrossRef]
  14. Poulain, A.; Watlet, A.; Kaufmann, O.; Van Camp, M.; Jourde, H.; Mazzilli, N.; Rochez, G.; Deleu, R.; Quinif, Y.; Hallet, V. Assessment of groundwater recharge processes through karst vadose zone by cave percolation monitoring. Hydrol. Process. 2018, 32, 2069–2083. [Google Scholar] [CrossRef]
  15. Lakshminarayanan, B.; Ramasamy, S.; Anuthaman, S.N.; Karuppanan, S. New DRASTIC framework for groundwater vulnerability assessment: Bivariate and multi-criteria decision-making approach coupled with metaheuristic algorithm. Environ. Sci. Pollut. Res. 2022, 29, 4474–4496. [Google Scholar] [CrossRef]
  16. Margat, J. Vulnerabilite des Nappes d’eau Souterraine a la Pollution; BRGM-Publication: Orleans, France, 1968. [Google Scholar]
  17. Aller, L.; Bennett, T.; Lehr, J.; Petty, R.; Hackett, G. DRASTIC: A Standardized System for Evaluating GroundWater Pollution Potential Using Hydrogeologic Setting; EPA/600/2-87/035; EPA (Environmental Protection Agency): Washington, DC, USA, 1987. [Google Scholar]
  18. Hoelting, B.; Haertlé, T.; Hohberger, K.-H.; Nachtigall, K.H.; Villinger, E.; Weinzierl, W. Wrobel; J.-P. Konzept zur Ermittlung der Schutzfunktion der Grundwasserüberdeckung. Geol. Jahrb. Reihe C 1995 Band C 1995, 63, 5–24. [Google Scholar]
  19. Civita, M.; De Maio, M. SINTACS: Un Systema Parametrico per la Valutazione e la Cartografia Della Vulnerabilità Degli Acquiferi All’inquinamento (Methodologia e Automatizzazione); Pitagora Editrice: Bologna, Italy, 1997. [Google Scholar]
  20. Vias, J.M.; Andreo, B.; Perles, M.J.; Carrasco, F.; Vadillo, I.; Jime’nez, P. Preliminary proposal of a method for vulnerability mapping in carbonate aquifers. In Karst and Environment; Carrasco, F., Durán, J.J., Andreo, B., Eds.; Patronato de la Cueva de Nerja: Nerja, Spain, 2002; pp. 75–83. [Google Scholar]
  21. Vias, J.M.; Andreo, B.; Perles, M.; Carrasco, F.; Vadillo, I.; Jime’nez, P. Proposed method for groundwater vulnerability mapping in carbonate (karstic) aquifers: The COP method-application in two pilot sites in southern Spain. Hydrogeol. J. 2006, 14, 912–925. [Google Scholar] [CrossRef]
  22. Goldscheider, N.; Klute, M.; Sturm, S.; Hotzl, H. The PI Method—A GIS-Based Approach to Mapping Groundwater Vulnerability with Special Consideration of Karst Aquifers. Z. Für Angew. Geol. 2000, 46, 157–166. [Google Scholar]
  23. Jeannin, P.Y.; Cornaton, F.; Zwahlen, F.; Perrochet, P. VULK: A tool for intrinsic vulnerability assessment and validation. In Proceedings of the 7th Conference on Limestone Hydrology and Fissured Media, Besancon, France, 20–22 September 2001; Volume 13, pp. 185–188. [Google Scholar]
  24. Mimi, Z.; Mahmoud, N.; Madi, M. Modified DRASTIC Assessment for Intrinsic Vulnerability Mapping of Karst Aquifers: A Case Study. Environ. Earth Sci. 2011, 66, 447–456. [Google Scholar] [CrossRef]
  25. Banerjee, A.; Creedon, L.; Jones, N.; Gill, L.; Gharbia, S. Dynamic groundwater contamination vulnerability assessment techniques: A systematic review. Hydrology 2023, 10, 182. [Google Scholar] [CrossRef]
  26. Goldscheider, N. The Concept of Groundwater Vulnerability. In COST Action 620—Vulnerability and Risk Mapping for the Protection of Carbonate (Karst) Aquifers, Final Report; Zwahlen, F., Ed.; European Commission: Brussels, Belgium, 2003; pp. 5–9. [Google Scholar]
  27. Wachniew, P.; Zurek, A.J.; Stumpp, C.; Gemitzi, A.; Gargini, A.; Filippini, M.; Rozanski, K.; Meeks, J.; Kværner, J.; Witczak, S. Toward operational methods for the assessment of intrinsic groundwater vulnerability: A review. Crit. Rev. Environ. Sci. Technol. 2016, 46, 827–884. [Google Scholar] [CrossRef]
  28. Fannakh, A.; Farsang, A. DRASTIC, GOD, and SI approaches for assessing groundwater vulnerability to pollution: A review. Environ. Sci. Eur. 2022, 34, 77. [Google Scholar] [CrossRef]
  29. Bordbar, M.; Rezaie, F.; Bateni, S.M.; Jun, C.; Kim, D.; Busico, G.; Moghaddam, H.K.; Paryani, S.; Panahi, M.; Valipour, M. Global review of modification, optimization, and improvement models for aquifer vulnerability assessment in the era of climate change. Curr. Clim. Change Rep. 2023, 9, 45–67. [Google Scholar] [CrossRef]
  30. Yu, H.; Wu, Q.; Zeng, Y.; Zheng, L.; Xu, L.; Liu, S.; Wang, D. Integrated variable weight model and improved DRASTIC model for groundwater vulnerability assessment in a shallow porous aquifer. J. Hydrol. 2022, 608, 127538. [Google Scholar] [CrossRef]
  31. Khazaa’lah, M.; Talozi, S.; Hamdan, I. Assessment of groundwater vulnerability using GIS-based COP model in the northern governorates of Jordan. Model. Earth Syst. Environ. 2023, 9, 19–40. [Google Scholar] [CrossRef]
  32. Awawdeh, M.M.; Jaradat, R.A. Evaluation of aquifers vulnerability to contamination in the Yarmouk River basin, Jordan, based on DRASTIC method. Arab. J. Geosci. 2010, 3, 273–282. [Google Scholar] [CrossRef]
  33. Rahmati, O.; Pourghasemi, H.R.; Melesse, A.M. Application of GIS-based data driven random forest and maximum entropy models for groundwater potential mapping: A case study at Mehran Region, Iran. Catena 2016, 137, 360–372. [Google Scholar] [CrossRef]
  34. Naghibi, S.A.; Ahmadi, K.; Daneshi, A. Application of Support Vector Machine, Random Forest, and Genetic Algorithm Optimized Random Forest Models in Groundwater Potential Mapping. Water Resour. Manag. 2017, 31, 2761–2775. [Google Scholar] [CrossRef]
  35. Baalousha, H.M. Machine Learning Approaches for Groundwater Vulnerability Assessment in Arid Environments: Enhancing DRASTIC with ANN and Random Forest. Groundw. Sustain. Dev. 2025, 28, 101290. [Google Scholar] [CrossRef]
  36. Sajedi-Hosseini, F.; Malekian, A.; Choubin, B.; Rahmati, O.; Cipullo, S.; Coulon, F.; Pradhan, B. A novel machine learning-based approach for the risk assessment of nitrate groundwater contamination. Sci. Total Environ. 2018, 644, 954–962. [Google Scholar] [CrossRef]
  37. Breiman, L. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar] [CrossRef]
  38. Civita, M. Groundwater Vulnerability Maps: A Review. In Proceedings of the IX Symposium on Pesticide Chemistry, Mobility and Degradation of Xenobiotics, Piacenza, Italy, 11–13 October 1993; Edizioni G. Biagini: Lucca, Italia; pp. 587–631. [Google Scholar]
  39. Al-Bakri, J.T.; Shawash, S.; Ghanim, A.; Abdelkhaleq, R. Geospatial techniques for improved water management in Jordan. Water 2016, 8, 132. [Google Scholar] [CrossRef]
  40. ASF. ALOS PALSAR—Digital Elevation Model (12.5 m Spatial Resolution). 2024. Available online: https://asf.alaska.edu/data-sets/derived-data-sets/alos-palsar-rtc/alos-palsar-radiometric-terrain-correction (accessed on 15 May 2025).
  41. MWI-BGR (Ministry of Water and Irrigation; Bundesanstalt für Geowissenschaften und Rohstoffe). Groundwater Resource Assessment of Jordan 2017. 2019. Available online: https://www.bgr.bund.de/EN/Themen/Grundwasser/Downloads/gw_resource_assessment_jordan_pdf.pdf?__blob=publicationFile&v=2 (accessed on 15 May 2025).
  42. Abu-Zreig, M.; Hani, L. Assessment of the SWAT model in simulating watersheds in arid regions: Case study of the Yarmouk River Basin (Jordan). Open Geosci. 2021, 13, 377–389. [Google Scholar] [CrossRef]
  43. Goode, D.J.; Senior, L.A.; Subah, A.; Jaber, A. Groundwater-Level Trends and Forecasts, and Salinity Trends, in the Azraq, Dead Sea, Hammad, Jordan Side Valleys, Yarmouk, and Zarqa Groundwater Basins, Jordan (No. 2013-1061); US Geological Survey: Reston, VA, USA, 2013. [Google Scholar]
  44. Abdelhamid, G. Geological Map of Jarash [Map]; Jordan Ministry of Energy and Mineral Resources—Natural Resources Authority (Geology Directorate): Amman, Jordan, 1993; National Mapping Project, Sheet No. (3154-I). [Google Scholar]
  45. Smadi, A. Geological Map of Al Mafraq [Map]; Jordan Ministry of Energy and Mineral Resources—Natural Resources Authority (Geology Directorate): Amman, Jordan, 1997; National Mapping Project, Sheet No. (3254-IV). [Google Scholar]
  46. Moh’d, B. Geological Map of Irbid [Map]; Jordan Ministry of Energy and Mineral Resources—Natural Resources Authority (Geology Directorate): Amman, Jordan, 1997; National Mapping Project, sheet No. (3155-II). [Google Scholar]
  47. Gharaibeh, A. Geological Map of Ar Ramtha [Map]; Jordan Ministry of Energy and Mineral Resources—Natural Resources Authority (Geology Directorate): Amman, Jordan, 2003; National Mapping Project, Sheet No. (3254-I). [Google Scholar]
  48. Moh’d, B. Geological Map of Ash Shuna Ash Shamaliyya [Map]; Jordan Ministry of Energy and Mineral Resources—Natural Resources Authority (Geology Directorate): Amman, Jordan, 2000; National Mapping Project, Sheet No. (3155-III). [Google Scholar]
  49. Soboh, A. Geological Map of Ar Ramtha [Map]; Jordan Ministry of Energy and Mineral Resources—Natural Resources Authority (Geology Directorate): Amman, Jordan, 2013; National Mapping Project, Sheet No. (3255-II). [Google Scholar]
  50. Hobler, M.; Margane, A.; Almomani, M.; Subah, A. Groundwater Resources of Northern Jordan; Vol. 4 Contributions to the hydrogeology of northern Jordan; Federal Institute for Geosciences and Natural Resources (BGR) and Ministry of Water and Irrigation (MWI): Amman, Jordan, 2001; Available online: https://www.deutsche-rohstoffagentur.de/EN/Themen/Wasser/Projekte/abgeschlossen/TZ/Jordanien/gw_res_management_fb_en.html?nn=1546392 (accessed on 15 May 2025).
  51. BRGM. Jordan Aquifers Modelling Project 3D Geological Model—Intermediary Report No 1; Unpublished Report; BRGM: Amman, Jordan, 2010. [Google Scholar]
  52. Hamdan, I.; Margane, A.; Ptak, T.; Wiegand, B.; Sauter, M. Groundwater vulnerability assessment for the karst aquifer of Tanour and Rasoun springs catchment area (NW-Jordan) using COP and EPIK intrinsic methods. Environ. Earth Sci. 2016, 75, 1474. [Google Scholar] [CrossRef]
  53. MoA (Ministry of Agriculture). National Soil Map and Land Use Project. The Soil of Jordan, Level 2: Semi Detailed Studies 1994. Volume 2: Main Report, Hunting Technical Services in Association with Soil Survey and Land Research Centre, Jordan; Unpublished Report; MoA (Ministry of Agriculture): Amman, Jordan, 1994. [Google Scholar]
  54. Majandang, J.; Sarapirome, S. Groundwater vulnerability assessment and sensitivity analysis in Nong Rua, Khon Kaen, Thailand, using a GIS-based SINTACS model. Env. Earth Sci. 2013, 68, 2025–2039. [Google Scholar] [CrossRef]
  55. Sahu, I.; Prasad, A.; Ahmad, I. Groundwater vulnerability assessment using SINTACS model and GIS: A case study in Raipur city. In IOP Conference Series: Earth and Environmental Science; IOP Publishing: Bristol, UK, 2022; Volume 982, p. 012070. [Google Scholar]
  56. Soyaslan, I.I. Assessment of groundwater vulnerability using SINTACS and modified-SINTACS methods in Burdur Çine Basin (Turkiye). Earth Sci. Res. J. 2025, 29, 201–210. [Google Scholar] [CrossRef]
  57. Band, S.S.; Janizadeh, S.; Chandra Pal, S.; Saha, A.; Chakrabortty, R.; Melesse, A.M.; Mosavi, A. Flash flood susceptibility modeling using new approaches of hybrid and ensemble tree-based machine learning algorithms. Remote Sens. 2020, 12, 3568. [Google Scholar] [CrossRef]
  58. Probst, P.; Wright, M.N.; Boulesteix, A.L. Hyperparameters and tuning strategies for random forest. Wiley Interdiscip. Rev. Data Min. Knowl. Discov. 2019, 9, e1301. [Google Scholar] [CrossRef]
  59. Han, S.; Kim, H.; Lee, Y.S. Double random forest. Mach. Learn. 2020, 109, 1569–1586. [Google Scholar] [CrossRef]
  60. Xia, H.; Akay, Y.M.; Akay, M. Selecting relevant genes from microarray datasets using a Random Forest Model. IEEE Access 2021, 9, 97813–97821. [Google Scholar] [CrossRef]
  61. Pantic, I.; Paunovic Pantic, J. Artificial Intelligence in Chromatin Analysis: A Random Forest Model Enhanced by Fractal and Wavelet Features. Fractal Fract. 2024, 8, 490. [Google Scholar] [CrossRef]
  62. Elkhrachy, I. Flash flood water depth estimation using SAR images, digital elevation models, and machine learning algorithms. Remote Sens. 2022, 14, 440. [Google Scholar] [CrossRef]
  63. Heman-Ackah, S.M.; Blue, R.; Quimby, A.E.; Abdallah, H.; Sweeney, E.M.; Chauhan, D.; Hwa, T.; Brant, J.; Ruckenstein, M.J.; Bigelow, D.C.; et al. A multi-institutional machine learning algorithm for prognosticating facial nerve injury following microsurgical resection of vestibular schwannoma. Sci. Rep. 2024, 14, 12963. [Google Scholar] [CrossRef]
  64. Fadhila, R.N.; Ulinnuha, N.; Yuliati, D. Implementation of Random Forest Method with Information Gain Selection and Hyperparameter Tuning for Alzheimer’s Disease Classification. Lontar Komput. J. Ilm. Teknol. Inf. 2025, 16, 1–13. [Google Scholar] [CrossRef]
  65. Putri, T.A.E.; Widiharih, T.; Santoso, R. Penerapan Tuning Hyperparameter Randomsearchcv Pada Adaptive Boosting Untuk Prediksi Kelangsungan Hidup Pasien Gagal Jantung. J. Gaussian 2023, 11, 397–406. [Google Scholar] [CrossRef]
  66. Alibrahim, H.; Ludwig, S.A. Hyperparameter optimization: Comparing genetic algorithm against grid search and bayesian optimization. In Proceedings of the 2021 IEEE Congress on Evolutionary Computation (CEC), Kraków, Poland, 28 June–1 July 2021; pp. 1551–1559. [Google Scholar]
  67. Ravindiran, G.; Rajamanickam, S.; Sivarethinamohan, S.; Karupaiya Sathaiah, B.; Ravindran, G.; Muniasamy, S.K.; Hayder, G.A. Review of the Status, Effects, Prevention, and Remediation of Groundwater Contamination for Sustainable Environment. Water 2023, 15, 3662. [Google Scholar] [CrossRef]
  68. Nakagawa, K.; Amano, H.; Persson, M.; Berndtsson, R. Spatiotemporal variation of nitrate concentrations in soil and groundwater of an intensely polluted agricultural area. Sci. Rep. 2021, 11, 2598. [Google Scholar] [CrossRef] [PubMed]
  69. Siarkos, I.; Sevastas, S.; Mallios, Z.; Theodossiou, N.; Ifadis, I. Investigating groundwater vulnerability variation under future abstraction scenarios to estimate optimal pumping reduction rates. J. Hydrol. 2021, 598, 126297. [Google Scholar] [CrossRef]
  70. Tsegay, T.; Birhanu, B.; Azagegn, T.; Hailu, B.T.; Ayenew, T. Assessing groundwater vulnerability to pollution in a rapidly urbanizing river basin using a modified drastic land use–lineament density method. Geol. J. 2024, 59, 3278–3295. [Google Scholar] [CrossRef]
  71. Ahmadi, M.M.; Mahdavirad, H.; Bakhtiari, B. Multi-criteria analysis of site selection for groundwater recharge with treated municipal wastewater. Water Sci. Technol. 2017, 76, 909–919. [Google Scholar] [CrossRef] [PubMed]
  72. Piscopo, G. Groundwater Vulnerability Map, Explanatory Notes, Castlereagh Catchment; Centre for Natural Resources, NSW Department of Land and Water Conservation: Sydney, Australia, 2001. [Google Scholar]
  73. Rivett, M.O.; Wealthall, G.P.; Dearden, R.A.; McAlary, T.A. Review of unsaturated-zone transport and attenuation of volatile organic compound (VOC) plumes leached from shallow source zones. J. Contam. Hydrol. 2011, 123, 130–156. [Google Scholar] [CrossRef]
  74. Vishwakarma, A.; Goswami, A.; Pradhan, B. Prioritization of sites for Managed Aquifer Recharge in a semi-arid environment in western India using GIS-Based multicriteria evaluation strategy. Groundw. Sustain. Dev. 2021, 12, 100501. [Google Scholar] [CrossRef]
  75. Talozi, S.; Alharahsheh, T.; Hamdan, I. Selecting suitable sites for groundwater recharge in Jordan using the spreading techniques via the integration of multi-criteria decision analysis and geographic information system tools. Groundw. Sustain. Dev. 2023, 22, 100948. [Google Scholar] [CrossRef]
  76. Civita, M.; De Maio, M. Assessing and mapping groundwater vulnerability to contamination: The Italian ‘‘combined’’ approach. Geofísica Int. 2004, 43, 513–532. [Google Scholar] [CrossRef]
  77. Al Kuisi, M.; El-Naqa, A.; Hammouri, N. Vulnerability mapping of shallow groundwater aquifer using SINTACS model in the Jordan Valley area, Jordan. Environ. Geol. 2006, 50, 651–667. [Google Scholar] [CrossRef]
  78. Al Kuisi, M.; Mashal, K.; Al-Qinna, M.; Hamad, A.A. Groundwater vulnerability and hazard mapping in an arid region: Case study, Amman-Zarqa Basin (AZB)-Jordan. J. Water Resour. Prot. 2014, 6, 442. [Google Scholar] [CrossRef]
  79. Margane, A.; Hobler, M.; Subah, A. Mapping of Groundwater Vulnerability and Hazards to Groundwater in the Irbid Area, N Jordan. Z. Für Angew. Geol. 1999, 45, 4. [Google Scholar]
  80. Awawdeh, M.; Al-Kharabsheh, N.; Obeidat, M.; Awawdeh, M. Groundwater vulnerability assessment using modified SINTACS model in Wadi Shueib, Jordan. Ann. GIS 2020, 26, 377–394. [Google Scholar] [CrossRef]
  81. El-Naqa, A.; Hammouri, N.; Kuisi, M. GIS-based evaluation of groundwater vulnerability in the Russeifa area, Jordan. Rev. Mex. De Cienc. Geológicas 2006, 23, 333–342. [Google Scholar]
  82. Noghreyan, A.; Mohamadvali Samani, J.; Zarei, H. Comparison of the SINTACS aquifer vulnerability model to nitrate with three-dimensional numerical model (case study of Varamin Plain Aquifer). Iran. J. Soil Water Res. 2022, 53, 103–116. Available online: https://ijswr.ut.ac.ir/article_87269_en.html?lang=en (accessed on 15 May 2025).
  83. Assefi, M.; Zarei, H.; Radmanesh, F. Improvement of SINTACS method using analytical hierarchy process in geographic information system environment to evaluate aquifer vulnerability (case study of Varamin Plain Aquifer). Irrig. Water Eng. 2015, 5, 109–125. [Google Scholar]
  84. Elzain, H.E.; Chung, S.Y.; Venkatramanan, S.; Selvam, S. Novel machine learning algorithms to predict the groundwater vulnerability index to nitrate pollution at two levels of modeling. Chemosphere 2023, 310, 136859. [Google Scholar] [CrossRef] [PubMed]
  85. Fu, J.; Le, X.C. Improving groundwater vulnerability assessment using machine learning. J. Environ. Sci. 2024, 153, 6–9. [Google Scholar] [CrossRef] [PubMed]
  86. Abba, S.I.; Yassin, M.A.; Jibril, M.M.; Tawabini, B.; Al-Ansari, N. Nitrate concentrations tracking from multi-aquifer groundwater vulnerability zones: Insight from machine learning and spatial mapping. Process Saf. Environ. Prot. 2024, 182, 110–125. [Google Scholar] [CrossRef]
  87. Idir, S.; Khemmoudj, K.; Bouali, H. Comparative analysis of groundwater vulnerability in the Oued Agrioun watershed, northeast Algeria: DRASTIC, GOD, and SINTACS approaches with ROC AUC Validation. Stud. Eng. Sci. 2024, 5, e11736. [Google Scholar] [CrossRef]
  88. Norouzi, H.; Moghaddam, A.A.; Celico, F.; Shiri, J. Assessment of groundwater vulnerability using genetic algorithm and random forest methods (case study: Miandoab plain, NW of Iran). Environ. Sci. Pollut. Res. 2021, 28, 15004–15018. [Google Scholar] [CrossRef]
  89. Chakrabortty, R.; Roy, P.; Chowdhuri, I.; Pal, S.C. Groundwater Vulnerability Assessment Using Random Forest Approach in a Water-Stressed Paddy Cultivated Region of West Bengal, India. In Groundwater Geochemistry; Madhav, S., Singh, P., Eds.; Wiley: Hoboken, NJ, USA, 2021. [Google Scholar] [CrossRef]
  90. Al-Adamat, R.A.N.; Foster, I.D.L.; Baban, S.M.J. Groundwater vulnerability and risk mapping for the Basaltic aquifer of the Azraq basin of Jordan using GIS, remote sensing and DRASTIC. Appl. Geogr. 2003, 23, 103–124. [Google Scholar] [CrossRef]
Geosciences 15 00414 g001
Figure 1. Overview map of the study area (based on ALOS PALSAR DEM 12.5 m resolution [40]).
Figure 1. Overview map of the study area (based on ALOS PALSAR DEM 12.5 m resolution [40]).
Geosciences 15 00414 g001
Geosciences 15 00414 g002
Figure 2. Geological map of the study area (based on [44,45,46,47,48,49]).
Figure 2. Geological map of the study area (based on [44,45,46,47,48,49]).
Geosciences 15 00414 g002
Geosciences 15 00414 g003
Figure 3. Hydrogeological map of the study area (modified after [41]).
Figure 3. Hydrogeological map of the study area (modified after [41]).
Geosciences 15 00414 g003
Geosciences 15 00414 g004
Figure 4. Thematic maps of YRB: (a) depth to groundwater, (b) infiltration rate, (c) unsaturated zone, (d) soil texture, (e) aquifer media, (f) hydraulic conductivity, and (g) slope (for references see Table 1).
Figure 4. Thematic maps of YRB: (a) depth to groundwater, (b) infiltration rate, (c) unsaturated zone, (d) soil texture, (e) aquifer media, (f) hydraulic conductivity, and (g) slope (for references see Table 1).
Geosciences 15 00414 g004
Geosciences 15 00414 g005
Figure 5. Rates map for the SINTACS parameter of YRB: (a) infiltration rate, (b) unsaturated zone, (c) soil texture, (d) aquifer media, (e) hydraulic conductivity, and (f) slope.
Figure 5. Rates map for the SINTACS parameter of YRB: (a) infiltration rate, (b) unsaturated zone, (c) soil texture, (d) aquifer media, (e) hydraulic conductivity, and (f) slope.
Geosciences 15 00414 g005
Geosciences 15 00414 g006
Figure 6. Rate × weight map for the SINTACS parameter of YRB: (a) infiltration rate, (b) unsaturated zone, (c) soil texture, (d) aquifer media, (e) hydraulic conductivity, and (f) slope.
Figure 6. Rate × weight map for the SINTACS parameter of YRB: (a) infiltration rate, (b) unsaturated zone, (c) soil texture, (d) aquifer media, (e) hydraulic conductivity, and (f) slope.
Geosciences 15 00414 g006
Geosciences 15 00414 g007
Figure 7. Normalized SINTACS groundwater vulnerability map for YRB.
Figure 7. Normalized SINTACS groundwater vulnerability map for YRB.
Geosciences 15 00414 g007
Geosciences 15 00414 g008
Figure 8. Normalized vulnerability map using the Random Forest (RF) machine learning model for YRB.
Figure 8. Normalized vulnerability map using the Random Forest (RF) machine learning model for YRB.
Geosciences 15 00414 g008
Geosciences 15 00414 g009
Figure 9. ROC Curve between RF model and SINTACS (black dotted line represent the performance of a random classifier with no predictive power).
Figure 9. ROC Curve between RF model and SINTACS (black dotted line represent the performance of a random classifier with no predictive power).
Geosciences 15 00414 g009
Table 1. Information required for applying the SINTACS model.
Table 1. Information required for applying the SINTACS model.
ParameterData/Parameter Values Needed and Their Source
Depth to groundwater (S1)Based on data from MWI and BGR [41]
Net recharge (I)Calculated based on the estimated recharge percentage from literature reviews [51] for rainfall amount based on the aquifer media.
Unsaturated zone (N)Based on geological maps (shape file) scale 1:50,000 [44,45,46,47,48,49].
Soil media (T)Soil units map with texture description (after [53].
Aquifer media (A)Outcropping hydrogeological units [41].
Hydraulic conductivity (C)Values from the literature reviews [50].
Topographic Slope (S2)Derived based on ALOS PALSAR digital elevation model (12.5 m spatial resolution) [40].
Table 2. Rate and weight for SINTACS model parameters (based on [19,54,55,56]).
Table 2. Rate and weight for SINTACS model parameters (based on [19,54,55,56]).
ParameterWeightCategoryRatingsCategoryRatings
Depth to Groundwater (S1) (m)50–101030–403
10–207>401
20–305
Infiltration rate (I) (mm/year)40–251110–1356
25–402135–1657
40–653165–1908
65–904190–2359
90–1105
Unsaturated zone (N) (Class)5Coarse alluvial deposits6–9Sandstone, Conglomerate5–8
Karstified limestone8–10Fissured volcanic rocks5–10
Fractured limestone4–8Marl, clay stone1–3
Fissured dolomite2–5Clay, silt, peat1–2
Medium fine alluvial Deposits3–6Pyrotoclastic rock2–5
Sand complex4–7Fissured metamorphic rocks2–6
Soil media (T) (Texture)3Clay1Sandy clay loam5
Silty clay2Sandy loam6
Clay loam3Loamy sand8
Silty clay loam3.5Sand8.5
Silty loam4No soil cover10
Loam4.5
Aquifer media (A) (Class)4Coarse alluvial deposits4–9Sandstone, conglomerate4–9
Karstified limestone9–10Fissured volcanic rocks8–10
Fractured limestone6–9Marl, clay stone1–3
Fissured dolomite4–7Clay, silt, peat1–3
Medium fine alluvial deposits6–8Pyrotoclastic rock4–8
Sand complex7–9Fissured metamorphic rocks2–5
Hydraulic conductivity (C) (m/day)3<0.115–156–7
0.1–1315–508–9
1–55>5010
Topographic Slope (S2) (%)20–21013–155
3–4916–184
5–6819–213
7–9722–252
10–126>251
Table 3. Comparison between SINTACS and RF models.
Table 3. Comparison between SINTACS and RF models.
Vulnerability ClassSINTACS ModelSINTACS Map Using a Random Forest Machine Learning Model
Area (Km2)(%)Area (Km2)(%)
Very low105.398.3790.37.2
Low74.55.92101.158
Moderate479.2638.07522.7141.5
High271.4321.56238.0518.9
Very high328.4126.09306.7824.4
Table 4. Area coverage (%) of LULC for each vulnerability class.
Table 4. Area coverage (%) of LULC for each vulnerability class.
Vulnerability ClassAgricultureBare SoilForestRocksUrbanWater
Very low3.7323.966--0.6710.005
Low0.5934.9830.001-0.3470.005
Moderate14.33818.4400.0170.0035.3250.004
High3.01815.5350.0840.0132.8720.009
Very high1.83021.4580.1790.0752.4910.005
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Hamdan, I.; AlShdaifat, A.; Ibrahim, M.; Al-Shabeeb, A.R.; Al-Adamat, R.; Al-Fugara, A. A Dual-Method Assessment of the Yarmouk Basin’s Groundwater Vulnerability Using SINTACS and Random Forest. Geosciences 2025, 15, 414. https://doi.org/10.3390/geosciences15110414

AMA Style

Hamdan I, AlShdaifat A, Ibrahim M, Al-Shabeeb AR, Al-Adamat R, Al-Fugara A. A Dual-Method Assessment of the Yarmouk Basin’s Groundwater Vulnerability Using SINTACS and Random Forest. Geosciences. 2025; 15(11):414. https://doi.org/10.3390/geosciences15110414

Chicago/Turabian Style

Hamdan, Ibraheem, Ahmad AlShdaifat, Majed Ibrahim, Abdel Rahman Al-Shabeeb, Rida Al-Adamat, and A’kif Al-Fugara. 2025. "A Dual-Method Assessment of the Yarmouk Basin’s Groundwater Vulnerability Using SINTACS and Random Forest" Geosciences 15, no. 11: 414. https://doi.org/10.3390/geosciences15110414

APA Style

Hamdan, I., AlShdaifat, A., Ibrahim, M., Al-Shabeeb, A. R., Al-Adamat, R., & Al-Fugara, A. (2025). A Dual-Method Assessment of the Yarmouk Basin’s Groundwater Vulnerability Using SINTACS and Random Forest. Geosciences, 15(11), 414. https://doi.org/10.3390/geosciences15110414

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop