Groundwater is a vital resource, particularly in rural areas of developing nations, where it often serves as the primary source for domestic uses. However, these crucial resources are increasingly threatened by contamination, primarily driven by human activities and complex hydrogeological characteristics [1]. Factors such as shallow groundwater tables, permeable soils, and fractured bedrock can facilitate contaminant transport, overwhelming natural filtration processes [1], a situation further exacerbated by climate change [2]. Effective groundwater protection relies on accurate identification of contamination risk zones within watersheds. Vulnerability maps, generated through the overlay of various physical factors, are a common tool for this purpose [3]-[6]. Key parameters often include topography, lithology, depth to water table, rainfall, land use, conductivity, and vadose zone impact [7]. Among these, lithology and depth to water table are recognized as particularly sensitive factors determining aquifer vulnerability [8]. Despite its importance, previous studies have typically treated lithology as a single, homogeneous factor, often measured solely by hydraulic conductivity. This approach, however, inadequately captures the nuanced contribution of stratified subsurface layers to aquifer vulnerability. Stratified sublayers possess varied textural characteristics and spatial subsurface topologies, leading to diverse abilities to retain and transmit groundwater. These variations arise from differences in weathering processes within the Earth’s crust, influencing vertical flow, lateral flow, and internal storage due to varied flow gradients. As demonstrated in our earlier findings , the variations alter flow convergence zones, a critical factor in aquifer vulnerability mapping and its partitioning. The spatial integration of each sublayer is thus hypothesized to result in a more accurate prediction of aquifer vulnerability. The present study aims to evaluate whether incorporating stratified lithological sublayers into aquifer vulnerability mapping enhances the accuracy of contamination risk assessment. By systematically testing the contribution of subsurface heterogeneity to flow convergence and aquifer susceptibility, the work seeks to establish the potential value of a segregated lithology approach in groundwater protection.

The study was conducted in the Tochi watershed, located in Northern Uganda, as shown in Figure 1. The geology generally consists of sedimentary materials of igneous rock origin. Boreholes are the preferred source of water for most rural communities due to their safety. However, well contamination is being reported [10]. Those that have existed for a very long time have shown elevated biological and physical characteristics when compared to newly constructed ones [11]-[13]. Mixed land use exists within the watershed, with intensive subsistence agriculture practiced on predominantly sandy clay soil. This situation creates the possibility of pollutant transport to the shallow aquifer, necessitating the mapping of vulnerable areas [13].

Figure 1. Map of the study area showing the spatial distribution of boreholes and meteorological stations.

Identification of aquifer vulnerability regions involved the use of factors such as groundwater table, land use, rainfall, and segregated lithology (soil, laterite, saprolite, and granite). Daily point data for production of the rainfall map were obtained from the NASA Power website (https://power.larc.nasa.gov/data-access-viewer/). The data period ranged between the years 1982 to 2023. Data from the Uganda Meteorological Authority for the same period were used to validate the satellite data. Land use data, with a 30-meter resolution, were obtained from the Earth Explorer website (https://earthexplorer.usgs.gov/) for the following paths and rows: path 171 row 058, path 172 row 058, and path 071 row 059. The acquired images were used to derive the land use map. The segregated lithology map was developed from the borehole drill logs in Geomodeller software from our previous study. 52 drill logs were used, with construction dates ranging between March 2003 and December 2016. The spatial distribution of boreholes within the watershed is one borehole per 42 square kilometers. The usable lithological intervals of the borehole in meters were as follows: soil (0 - 4), laterite (0 - 28), saprolite (0.3 - 40), and granite (13 - 65). The drill logs acquired from the districts of Gulu, Omoro, Nwoya, Oyam, and Apac were further used to develop the groundwater table map in ArcGIS.

Aquifer vulnerability mapping in this study used segregated and unsegregated lithologies presented as maps in addition to factors of groundwater table, land use, and rainfall. Segregated lithology was prepared using an implicit geo-modelling approach. The open lowest interval between 0 m and 67 m was used to construct the 3D model structure. The onlap relationship formed the basis for interpolation and segregation of sub-surfaces. The observed lithological depth from borehole logs formed the basis for segregation into layers with varying hydraulic properties and subsurface topology. The approach generated continuous topographic sub-surfaces from which a Digital Elevation Model (DEM) was developed. Reclassification of DEM generated lithological layers with varying vulnerability potential and score rating utilized in the Multicriteria Decision Method-Analytical Hierarchy Process (MCDM-AHP). Unsegregated lithology was generated through digitization of lithological boundaries from the Geology Map of Uganda-2008. The partitioning in the subsurface is assumed to follow the digitized boundaries. The digitized lithological boundaries were overlaid onto DEM, which formed the basis for reclassification into vulnerability potential. The core of our approach was to compare vulnerability classifications derived from two distinct lithological treatments: segregated and unsegregated. The land use map was generated through supervised classification of imagery generated using the Principal Component Analysis (PCA) method. The depth at which first water was struck during drilling operations was considered to represent the saturated zone and was used in this study to approximate groundwater table depth. The concept is supported by the aquifer system within the watershed characterized as semiconfined. Aquifer thickness purifies contaminants in infiltrated water through particle entrapment in void spaces or reaction of contaminants with lithology material through geochemical processes. A thick overburden provides more residence time within the medium for such processes to take place, thereby increasing purification likelihood before reaching the groundwater table. Groundwater table and rainfall maps were generated using kriging interpolation of depth to first water strike and annual rainfall received, respectively. The Groundwater Quality Index (GQI) was used to objectively interpret the results of the vulnerability mapping.

Aquifer contamination is a complex issue driven by various factors, including surface waste, geochemical processes, and their combined effects [15]. Anthropogenic activities and land use patterns are widely recognized for their negative impact on groundwater quality [16][17]. In the Tochi watershed, our Groundwater Quality Index (GQI) analysis indicated that 70.6% of boreholes fall within the very low-risk category, and 22% within the low-risk category, suggesting that the water sources within the period March 2003 to December 2016 were generally safe for use. However, this may not be the case in the current year, 2026, due to the cumulative effect of contaminants in soil arising from anthropogenic activities.

Aquifer depth plays a critical role in natural water purification by providing sufficient residence time for contaminant attenuation during infiltration. However, the presence of shallow groundwater tables (<35 m) in high vulnerability regions, particularly when coupled with intense rainfall, can significantly increase the likelihood of contaminants reaching the groundwater directly [18]. Previous studies in Uganda have shown that most aquifers are semi-confined, with depths typically ranging from 27 m to 35 m [19]. This semi-confined nature renders them susceptible to pollution from external sources [20][21]. Indeed, surface-derived groundwater contamination, including elevated levels of total coliform, nitrate, and sulphate, has been reported in Gulu district within similar semi-confined aquifer depths (12 m to 76 m) [15]. These findings underscore the urgent need for effective intervention strategies to mitigate future groundwater contamination risks.

This study advanced aquifer vulnerability mapping by comparing two methodologies: one incorporating segregated subsurface lithology and another using unsegregated lithology, alongside other influencing factors. The vulnerability maps, classified into high, low, and very low risk, were statistically different between the two methods (P > chi-square = 0 at α = 0.05%). This significant difference highlights the potential impact of lithological detail on vulnerability assessment in addition to the potential contribution of AHP weight and criteria.

Specifically, the segregated lithology approach yielded a statistically significant improvement in delineating high vulnerability zones compared to the unsegregated lithology (p-value = 0.0295). This enhanced performance is likely attributable to the method’s ability to capture the intricate variations in hydraulic properties and flow paths within distinct subsurface layers (soil, laterite, saprolite, granite). The observed differences can further be investigated by focusing on the interplay between lithological details, AHP weight, and criteria for a better conclusion. In spite of the contribution of other factors in vulnerability mapping, the existence of significant elevation differences, which encourage flow convergence zones, is more accurately represented when lithology is disaggregated, allowing for a more precise identification of high-risk areas. For instance, laterite, with its highly variable hydraulic conductivity and dual-porosity characteristics, facilitates faster horizontal water movement along relict bedrock foliation planes and fractures [22]. By segregating such layers, the model better reflects the actual flow dynamics. In contrast, the unsegregated approach, which lumps these layers into broader geological units, may oversimplify the subsurface hydrogeology, leading to less accurate vulnerability predictions.

Within the low vulnerability class, the accuracy of the two methods did not vary significantly (P-value = 0.47). This suggests that for areas with inherently low vulnerability, the level of lithological detail might have a less pronounced impact on the overall assessment. However, the one-sample t-test performed on vulnerability classes for segregated lithology further confirmed that the vulnerability potential matched the associated risk levels derived from the GQI (p-value=0.00103 for high potential and p-value=0.02205 for low potential). The high AUC values (75% for high vulnerability potential and 100% for low vulnerability potential) further affirm the robustness and accuracy of the segregated lithology approach in mapping aquifer vulnerability. This aligns with studies by Zghibi et al. (2016), who emphasized the role of hydrogeological system properties, particularly hydraulic conductivity and aquifer medium, as primary intrinsic factors influencing aquifer vulnerability. By segmenting the aquifer thickness into units with more uniform textural characteristics, our method effectively assessed the contribution of varying subsurface topology gradients to identify potential flow convergence zones, leading to improved vulnerability mapping. [23]

This study successfully compared two methodologies for aquifer vulnerability mapping in the Tochi watershed, utilizing factors such as groundwater table, rainfall, land use, and both segregated (soil, laterite, saprolite, granite) and unsegregated (Amuru group, gneissic granitoids, metagabbro) lithological data. The primary objective was to demonstrate the improved accuracy achieved by integrating segregated lithology compared to a single unsegregated lithology thematic map.

Our analysis, validated against the Groundwater Quality Index (GQI) of boreholes, unequivocally demonstrated that the two methods produced statistically different vulnerability maps. Crucially, the segregated lithology approach resulted in a significantly better prediction within the high vulnerability map categories (p-value = 0.0295). This improvement is partly attributed to the enhanced refinement in subsurface lithology, achieved by separating geological units into autonomous layers during the mapping process, which more accurately reflects complex hydrogeological conditions. Furthermore, the variations within the GQI were effectively partitioned according to the vulnerability classes, reinforcing the contribution of segregated lithology to improved accuracy (p-values of 0.00103 for high and 0.02205 for low vulnerability classes, respectively).

While the GQI currently indicates that the majority of water sources in the watershed are safe for use for the period March 2003 to December 2016 (22% low risk and 70.6% very low risk), the prevalence of semi-confined aquifer systems and the projected population growth in Uganda pose significant future risks. The existing mixed land use patterns further necessitate proactive management interventions to prevent future aquifer pollution. Therefore, the adoption of methodologies that incorporate detailed subsurface stratification, such as the segregated lithology approach presented here, is crucial for developing more accurate vulnerability maps and ensuring sustainable groundwater resource management.

To further enhance the accuracy and applicability of aquifer vulnerability models, the following areas warrant future investigation:

1)Production of Subsurface models: More borehole wells should be used in the development of the subsurface model to reduce interpolation error for better identification of flow convergence zones.

2)Incorporation of Fissures and Fractures: The current segregated lithology model assumes a continuous subsurface, with limited explicit attention given to fissures, particularly within the granite layer. Future studies should explore advanced modeling techniques to incorporate the spatial distribution and hydraulic properties of such features, as they can significantly influence groundwater flow and contaminant transport, especially in fractured bedrock environments. The coincidence of strike-slip faults with the lowest elevations in this study, creating flow convergence zones, highlights the importance of these structural features, which may vary across different watersheds.

3)AHP weights, criteria, and segregated lithology:Further research is needed to quantify the individual contribution of AHP weights, criteria, and segregated lithology to aquifer vulnerability mapping.

4)Advanced Land Surface Modeling:Given the shallow depth to the groundwater table in some locations and the semi-confined nature of the aquifer systems, detailed land surface modeling should be considered. This would involve integrating more dynamic representations of infiltration, runoff, and pollutant loading from various land use types. Such models could provide a more comprehensive understanding of contaminant pathways and aid in developing preventative measures to avert future contamination of the aquifer systems, particularly in light of the mixed land use within the watershed.

The authors extend their sincere gratitude to the district water officers of Gulu, Oyam, Omoro, Nwoya, and Apac for providing the borehole drill logs essential for developing the segregated lithology maps. Appreciation is also extended to the Ministry of Energy and Mineral Development for making available the Geology Map of Uganda, which was instrumental in this study.

Conceptualization, Joachim Bongomin and Martine Nyeko; methodology, Joachim Bongomin and Martine Nyeko; software, Joachim Bongomin and Martine Nyeko; validation, Joachim Bongomin, Martine Nyeko, and Denis Nono; formal analysis, Joachim Bongomin, Martine Nyeko, Denis Nono, and Byakatonda Jimmy; investigation, Joachim Bongomin, Martine Nyeko, Denis Nono, Byakatonda Jimmy, Geoffrey Openy, Richard Louis Labeja; resources, Joachim Bongomin, Martine Nyeko, Denis Nono, Byakatonda Jimmy, Geoffrey Openy, Richard Louis Labeja; data curation, Joachim Bongomin and Martine Nyeko; writing—original draft preparation, Joachim Bongomin, Martine Nyeko, and Denis Nono; writing—review and editing, Joachim Bongomin, Martine Nyeko, Denis Nono, Byakatonda Jimmy, Geoffrey Openy, Richard Louis Labeja; visualization, Joachim Bongomin, Martine Nyeko, Denis Nono, Byakatonda Jimmy, Geoffrey Openy, Richard Louis Labeja; supervision, Joachim Bongomin, Martine Nyeko, Denis Nono, Byakatonda Jimmy, Geoffrey Openy, Richard Louis Labeja; project administration, Joachim Bongomin, Martine Nyeko, Denis Nono, Byakatonda Jimmy, Geoffrey Openy, Richard Louis Labeja; funding acquisition, Joachim Bongomin. All authors have read and agreed to the published version of the manuscript.