Our approach employs a strategic environmental assessment (SEA) logic, based on a rigorous triangulation of three complementary sources:

HistoricalESIAs: A critical analysis of ten ESIA reports from major projects (SGO-Sabodala, PMC-Mako, Boto, Sored, Afrigold) covering the 2006-2023 period.IndependentFieldData: Original empirical data from sampling campaigns conducted between 2021 and 2024 at 43 points, covering industrial, artisanal, and community areas.InternationalandNationalNormativeFrameworks: WHO [29], IFC [30], and the Senegalese standard for wastewater discharge (NS 05-061) [31].

The methodological differences and the temporal gap between these sources were explicitly addressed. ESIAs were treated as secondary data requiring prior critical assessment for scientific compatibility and spatial-temporal coherence before being used in the WPPI modeling [32]. This approach of confronting historical ESIA data with contemporary field measurements has previously been applied to soil quality assessment in the same territory [32], establishing methodological continuity in our research program.

2.2.1. Criteria for ESIA Report Selection and Data Extraction

The selection of ESIA reports followed three inclusion criteria: i) availability of quantitative water quality data (pH, heavy metals, cyanides, etc.); ii) explicit documentation of sampling locations and protocols; iii) coverage of the 2006-2023 period. A total of ten ESIA reports met these criteria (Table 1). For each selected report, the following information was systematically extracted: document owner, year of publication, sampling period, water matrix (surface water, groundwater, or effluent), and the specific physico-chemical variables measured. Where available, GPS coordinates of sampling points and analytical methods were also recorded to enable spatial and methodological comparability with field data.

2.2.2. Matching Protocol for Historical vs. Contemporary Data

To ensure valid comparisons between historical ESIA data and contemporary field measurements, we applied four matching criteria:

i) Geographic matching: ESIA sampling points and field sampling points were considered matched if located at the same GPS coordinates or within a radius of 100 m. This threshold was selected based on the positioning accuracy of handheld GPS devices (±3 - 5 m) and the practical constraints of relocating historical sampling points from often imprecise ESIA documentation;

ii) Water matrix matching: Comparisons were restricted to identical water matrices (surface water with surface water; groundwater with groundwater; effluent with effluent);

iii) Seasonal matching: Only dry season data (January-March) were compared, as the vast majority of ESIA reports (100%) were collected during the dry season. This minimizes seasonal variability in contaminant concentrations due to dilution or runoff effects;

iv) Analyte matching: Comparisons were restricted to common analytes—parameters that were measured in both the ESIA and the field campaign (e.g., pH, As, Pb, CN⁻, etc.).

After applying these four criteria, n = 12 matched site-pairs were retained for the historical vs. field comparison. These 12 pairs represent the maximum number of comparable observations across the ten ESIA reports and the 43 field sampling points. The results of this matching procedure are presented in Section 3.1.

Table 1. ESIA reports included in this study: source, period, matrix, extracted variables, and methodological weaknesses.

2.2.3. Methodological Weaknesses of ESIA Data

Our methodological analysis of ESIAs (Figure 2 and Table 1) revealed significant weaknesses, consistent with findings across West Africa [33]:

HeterogeneityofProtocols: Sampling protocols, parameters, and monitoring frequencies varied considerably between projects, limiting inter-project comparability [33].LimitedAssessmentofCumulativeEffects: None of the ESIAs analyzed proposed a cumulative reading of hydric impacts, treating projects as independent entities despite hydrological connectivity [8][11].WeakNormativeAlignment: Most reports rely solely on the national standard NS 05-061, without systematic cross-referencing with WHO or IFC standards. Critical mining contaminants (e.g., cyanides, sulfates) are often absent [12][34].InsufficientTraceability: A lack of precise GPS coordinates, measurement dates, and laboratory details limits the reproducibility of results [32].

These governance gaps are common challenges in the region, mirroring critiques of regulatory practices in other African mining contexts [11][12][35].

Figure 2. Comparative synthesis of methodological weaknesses of ESIAs by mining project.

Sampling Design and Data Aggregation

A total of 43 sampling points were monitored across four campaigns (2021-2024), covering both dry (January-March) and wet (August-October) seasons. For sites with repeated measurements (n = 12 sites), WPPI values were aggregated using arithmetic means per parameter.

Repeated observations were aggregated as follows: i) arithmetic mean for physico-chemical parameters (pH, conductivity, turbidity, TSS, DO, temperature); ii) arithmetic mean for heavy metal concentrations (As, Pb, Hg, Cd, Cr, Cu, Zn); iii) median for parameters with outliers (identified using the IQR method); iv) WPPI calculated from the resulting aggregated parameter values for each site. This aggregation protocol ensures that each site contributes a single representative WPPI value to the spatial analysis, while preserving seasonal and inter-annual variability in the underlying data (Table 1).

Field data were collected to fill the structural gaps in the ESIAs. Our sampling strategy was designed to capture spatial and seasonal variability [17][36].

SamplingSites: 43 georeferenced points (GPS, WGS84) were selected across industrial zones (Sabodala, Mako), artisanal mining zones (Diabougou, Fadougou), and strategic villages (Khossanto, Madina Bafé) (Figure 3). The complete list of 43 sampling points is provided in AppendixE.MeasuredParameters: We focused on priority physico-chemical parameters critical in a mining context: heavy metals and trace elements (As, Pb, Hg, Zn, Cd, Cr, Cu), basic parameters (pH, temperature, turbidity, electrical conductivity, TSS, dissolved oxygen), and mining-specific pollutants (total and free cyanides, nitrates, sulfates, phosphates, fluorides) [3]-[5].SamplingandAnalyticalProtocols: A harmonized protocol was applied, inspired by the CCME [37], FAO [38], and WHO [29]. Samples were collected in both dry (Jan-Mar) and wet (Aug-Oct) seasons, georeferenced, and analyzed in accredited laboratories using a rigorous chain of custody (Photo 1 and Table 2). Detailed protocols are provided in AppendixA.

We used a weighted hierarchy of three normative frameworks:

Figure 3. Location of sampling sites and Falémé/Gambia hydrological zoning.

Photo 1. Sampling sites and collection protocols.

Table 2. Protocol for sampling, analysis, and modeling of the Water Pollution Pressure Index (WPPI).

WHO (2017): Primary reference for domestic and community uses, providing health-based guidelines for key contaminants [29].IFC(2012): Used to evaluate the compliance of industrial sites with international environmental and social performance standards for effluent discharge [30].NS05-061: Used as the Senegalese national standard for wastewater discharge, serving as a secondary reference for effluent compliance [31].

The thresholds were integrated as normalized scales (0 - 1) to calculate partial and global WPPI scores per site (Table 3). The use of international benchmarks is crucial, as relying solely on national standards can lead to an underestimation of health risks, particularly for vulnerable populations [10][39]. The reference norm (Ni) is selected according to the primary water use context: the WHO guideline is applied for domestic/drinking water uses; the IFC standard is applied for assessing industrial effluent compliance; the national NS 05-061 standard serves as a secondary reference for wastewater discharge.

Table 3. Comparison of WHO, IFC, and NS 05-061 thresholds for key measured parameters.

NS05-061refers to the Senegalese national standard for wastewater discharge[31].

To compare water pressure across different sites, we developed the WPPI as a weighted aggregation tool. Our index is fundamentally different from classical WQIs. While WQIs are primarily designed to assess the overall quality of a water body, the WPPI is designed to assess the pressure this quality exerts on a given community.

2.5.1. General Mathematical Formulation

The WPPI is designed to generate a normalized score between 0 (absence of pressure) and 1 (maximum pressure). First, we calculate a Partial Pressure Index (PPI) for each parameter:

where:

Ci = Measured concentration of parameter i.Ni = Reference norm for parameter i (from Table 3, selected according to use context).Pi = Toxicological weight of parameter i (scale 1 - 5, detailed in Table 4). These weights are expert-derived ordinal scores, constructed specifically for the WPPI. They are informed by established toxicological evidence and regulatory benchmarks, but should not be interpreted as quantitative toxicity coefficients from EPA or WHO sources. Rather, they represent a relative ranking of concern based on the severity and frequency of documented health effects in mining contexts.Fi = Local usage and socio-ecological factor (0.9 - 2.0 in this operationalization, detailed in Table 5).Di = Distance from the contamination source to the village/water point (km).a = Regularization constant = 0.1 km.

2.5.2. Toxicological Weighting Factor (Pi)

The weighting factor Pi (scale 1 - 5) is assigned based on three criteria: environmental toxicity, frequency of normative exceedance in mining contexts, and criticality for human health. These ordinal weights are informed by established toxicological evidence [40], regulatory benchmarks [29][30], and health risk assessments [34][41][42]. The complete scoring grid is presented in Table 4.

Table 4. Toxicological weighting factor (Pi) scoring grid.

2.5.3. Local Usage and Socio-Ecological Factor (Fi)

The Fi factor captures the nature and intensity of the relationship between a community and its water resource. It is a function of the dependency and vulnerability of the community. In this operationalization, Fi ranges from approximately 0.9 to 2.0. The factor is formally defined as:

The complete scoring grid for each component is presented in Table 5.

Table 5. Socio-ecological factor (Fi) component scoring grid.

2.5.4. Normalization of Partial Pressure Indices

To bound each PPI between 0 and 1, we define a maximum reference value (PPI_max). For normalization purposes, a reference exceedance ratio of Ci/Ni = 10 was adopted as a conservative upper-bound scenario for the present dataset. This threshold was selected based on the empirical distribution of our 43 field samples, where the 95th percentile of all measured Ci/Ni ratios was 8.7. The value of 10 thus represents a robust upper bound for >95% of our observations. The influence of this assumption was evaluated through sensitivity analysis using ratios ranging from 5 to 20, as described below.

Using conservative estimates (P_max = 5, F_max = 2.0, D_min = 0.5 km, a = 0.1 km), we obtain:

The normalized PPI for each parameter is then:

2.5.5. Final WPPI Calculation

The WPPI is calculated as the arithmetic mean of the normalized PPIs for all n parameters included in the assessment:

This mean-based approach ensures that the index is bounded between 0 and 1 and is not dependent on the number of parameters measured. The final classification is:

A: 0.00 - 0.20 (Low)B: 0.20 - 0.40 (Moderate)C: 0.40 - 0.60 (High)D: 0.60 - 0.80 (Very High)E: 0.80 - 1.00 (Critical)

2.5.6. Handling of Non-Detects (ND) and Non-Measured (NM) Parameters

For parameters reported as “not detected” (ND), the value was replaced with LOD/2 (half the limit of detection) for WPPI calculations. This approach follows standard practice in environmental chemistry for handling censored data below the analytical detection limit [29][40].

For parameters “not measured” (NM)—i.e., parameters that were not analyzed or reported in the ESIA or field dataset—the parameter was excluded from the WPPI calculation for that specific site. To ensure comparability across sites with different numbers of measured parameters, the final WPPI was renormalized by the number of available parameters:

where n_available is the number of parameters actually measured for that site (i.e., excluding NM parameters). This mean-based renormalization ensures that sites with missing data are not penalized or advantaged relative to sites with complete data.

A sensitivity analysis was conducted to assess the impact of NM parameter exclusion on site rankings (see 2.5.7, point 3, and AppendixC). The analysis confirmed that the exclusion of NM parameters did not alter the ranking of any site by more than one class, validating the robustness of this approach.

2.5.7. Sensitivity Analysis

A sensitivity analysis was conducted to assess the robustness of the normalization procedure using our dataset (n = 43 sampling points):

Regularizationconstant(a): Varying a between 0.01 and 0.5 km. The overall site ranking remained stable, with Spearman’s rank correlation coefficients ranging from ρ = 0.81 to 0.98 relative to the baseline configuration (a = 0.10 km), with particularly high stability around the selected value.Referenceexceedanceratio: Varying the assumed Ci_max/Ni ratio from 5 to 20. The final classification of any site did not change by more than one class, confirming the robustness of the normalization procedure.NMparameterexclusion: Varying the number and type of NM parameters excluded from the WPPI calculation (see 2.5.6). No site was reclassified by more than one class, confirming the robustness of the renormalization approach.

The complete sensitivity analysis is provided in AppendixC.

2.5.8. Worked Example: Comparative Demonstration of the WPPI’s Social Integration

To illustrate the core innovation of the WPPI—the integration of social and spatial vulnerability—we present two hypothetical villages (A and B) with identical water contamination levels but contrasting socio-spatial contexts (Table 6).

VillageA: High vulnerability context. Fi = 1.85, Di = 0.5 km.VillageB: Low vulnerability context. Fi = 1.10, Di = 5.0 km.

Table 6. Comparative WPPI calculation for two villages with identical contamination but contrasting socio-spatial contexts.

Normalization(PPI_max=166.7):

Village A: Pb = 0.46, As = 0.28, Hg = 0.31, CN = 0.16 → WPPI = 0.30(ClassB:Moderate)Village B: Pb = 0.03, As = 0.02, Hg = 0.02, CN = 0.01 → WPPI = 0.02(ClassA:Low)

Interpretation: The contamination is identical, but the pressure is dramatically different. Village A, with high dependency and proximity, experiences Moderate pressure, while Village B experiences Low pressure. This demonstrates that the WPPI assesses not just pollution levels but the actual socio-environmental pressure on communities—a critical innovation over classical WQIs.

2.5.9. Source Identification and Proximity Assessment (Di)

Industrial sources were identified from official mining perimeters and project documentation for the following operations: Sabodala Gold Operations (SGO), Petowal Mining Company (PMC-Mako), Boto (IAMGOLD), Sored, and Afrigold (Kolya). Artisanal sources were identified through field surveys conducted during the 2021-2024 campaigns, complemented by satellite imagery analysis (Google Earth Pro, 2021-2024) to map active artisanal mining sites. The complete list of identified sources is provided in AppendixE.

The proximity factor Di was calculated as the shortest straight-line distance (Euclidean distance) from the sampling point to the nearest mining site (industrial or artisanal). For sites with multiple sources within a 5 km radius (n = 23 sites), Di was assigned based on the closest source, consistent with the “nearest source” approach commonly used in spatial risk assessments [8][12].

This straight-line distance approach was adopted for two reasons: i) it provides a consistent and reproducible metric across all sites; ii) detailed hydrological connectivity data (e.g., surface flow pathways, groundwater flow direction) were not uniformly available for all sampling points. Future refinements could incorporate hydrological connectivity to improve the representation of contaminant transport pathways, particularly in the Falémé and Gambia River corridors.

Spatial analysis is crucial for localizing pressures and modeling vulnerabilities [44]. We integrated the WPPI values and sampling point coordinates into QGIS 3.30.

HydrologicalZoning: We established a zoning based on the Falémé and Gambia River watersheds [22].Interpolation: The Inverse Distance Weighting (IDW) method was used to visualize pressure gradients and produce vulnerability maps (Figure 4) [25]. The methodology is detailed in AppendixB.

Figure 4. IDW interpolation of WPPI values.

DataHeterogeneity: We acknowledge the uneven frequency, quality, and temporal coverage of ESIA and field data [32][33].LogisticalConstraints: Access to certain sites (e.g., along the Falémé) was difficult, potentially affecting spatial representativity [8][17].SubjectivityinWeighting: While our factors are based on best practices and literature, the final selection and weighting of sub-parameters for Fi and Pi involve a degree of expert judgment. The scoring grids presented in Table 4 and Table5 are designed to maximize transparency and reproducibility. Further empirical calibration and sensitivity analysis are recommended [22][23][25].CausalAttribution: Our study design, based on observational data, allows us to identify associations and spatial-temporal trends. However, without source tracing (e.g., isotopic fingerprinting), we cannot definitively attribute observed pollution to specific mining sources. Our discussion of potential causes is therefore framed as plausible interpretations rather than definitive conclusions.

The cross-analysis of ESIA and field data reveals notable discrepancies (Table 7).

Table 7. Comparative analysis of historical baseline/ESIA data and contemporary field measurements (matched site-pairs, n = 12).

Note 1: Values in bold indicate an increase in concentration between the historical ESIA and the contemporary field measurement. ND = Not Detected; NM = Not Measured/Not Reported in the ESIA. Note 2: For ASM (artisanal small-scale mining) sites without formal ESIA reports, baseline values were derived from the earliest available environmental monitoring data or pre-operational surveys in the area (identified as “ASM baseline”). Note 3: Hg concentrations at Gora (SGO) Field measured at 0.0021 mg/L in this sampling campaign; a separate sampling point yielded 0.010 mg/L in a different campaign, reflecting spatial variability within the site (see Section 3.1). Note 4: All 12 matched pairs met the four matching criteria defined in Section 2.2.2: i) geographic matching (same location or within 100 m); ii) water matrix matching (identical matrices); iii) seasonal matching (dry season only); iv) analyte matching (common analytes).

Across the sites assessed, we observed higher concentrations of heavy metals and cyanides in the 2021-2024 field measurements compared to the historical ESIA data. These values frequently exceeded the applicable reference values selected according to water-use context (WHO, IFC, or NS 05-061).

These observations suggest an intensification of mining activities, challenges in effluent treatment, and/or ongoing pollution from ASM. However, due to the heterogeneous nature of the historical data, we frame this as a descriptive trend rather than a statistically significant change [9][18][45].

By aggregating the data per site using the WPPI, we can quantify the temporal dynamics (Table 8, Figure 5).

Table 8. Evolution of normalized WPPI values per site.

Our results show a consistent descriptive increase in WPPI values across the sites for which historical and current data were available, suggesting increasing water pressure over the study period. All sites now have WPPI scores in the “Very High” or “Critical” categories.

Figure 5. Comparison of historical ESIA WPPI and current field-measured WPPI per mining site.

Our detailed analysis of ESIA reports (Table 9) revealed critical omissions and inconsistencies.

Table 9. Cross-evaluation of ESIA data vs. field observations.

The spatial interpolation of WPPI values (Table 10) allowed us to identify several high-pressure zones.

Table 10. Territorial classification of water pressures in the Kédougou mining zone.

We observed a positive association between the density of mining activities and the WPPI value (Figure 6). “Density of mining activities” was defined as the number of active mining sites (industrial and ASM) within a 5 km radius. Using our field data (n = 43 sampling points), the Pearson correlation coefficient between mining activity density and WPPI was r = 0.943 (p < 0.001, 95% CI: 0.91 - 0.96). The complete dataset of 43 sampling points used for this correlation is provided in AppendixE, including site identifiers, category, WPPI values, and number of mining sources within 5 km. This full enumeration allows independent verification of the reported correlation. Zones with a high number of co-located activities exhibited the highest WPPI scores, supporting the presence of cumulative pressure patterns that are poorly captured in conventional project-level ESIAs [1][8][11].

Additional analyses of inter-parameter correlations, multi-source cumulative effects, and individual pollutant contributions to the WPPI are presented in Appendix D. These complementary results provide further insight into the relationships between contaminant co-occurrence, mining activity density, and cumulative water pollution pressures in the study area (see Appendix D).

The results demonstrate the value of the WPPI as a tool that potentially goes beyond

Figure 6. Relationship between the number of mining sources (industrial and artisanal) within a 5 km radius and WPPI values (n = 43). (A) Scatter plot with linear regression. Points are color-coded by site category (Industrial: green, n = 12; Artisanal: orange, n = 14; Village exposed: blue, n = 12; Reference: grey, n = 5). The red line represents the linear regression: y = 0.121x + 0.250 (Pearson r = 0.943, R² = 0.889, p < 0.001, 95% CI for r: 0.91 - 0.96). (B) Mean WPPI (± standard deviation) by number of mining sources. (C) Boxplots of WPPI distribution by site category. (D) Residual plot of the linear regression model, showing no clear systematic pattern (mean residual ≈ 0.000; SD = 0.059), supporting the linearity assumption.

classical water quality indices. Classical WQIs, while useful for characterizing water status [15][17][23], often fail to capture the socio-environmental pressure experienced by dependent communities. By integrating the Fi and Di factors, the WPPI provides a more socially and territorially relevant assessment. It translates physicochemical data into a measure of socio-environmental pressure, showing that a site with a moderate WQI but high human dependency (Fi) and proximity (Di) can exert a higher pressure than a more polluted but remote site. This approach aligns with the core principles of SEA, which emphasize the need to consider not just biophysical impacts but also their social and economic consequences [7][26][27]. This is a critical advancement for applied research in resource-constrained and socially vulnerable settings [13][46].

Despite the methodological limitations stemming from the heterogeneous ESIA data, the identified trends are consistent across multiple independent sources and sites. The consistent descriptive trend toward higher heavy-metal concentrations and WPPI values points toward a potentially chronic issue rather than isolated incidents. This is consistent with observations from other mining regions in sub-Saharan Africa, where inadequate monitoring and enforcement have led to progressive environmental degradation [10][12][13][45]. However, we emphasize that this is an observational trend, and definitive causal attribution to specific mining sources would require further investigation, such as isotopic fingerprinting or receptor modeling.

Our findings expose governance gaps in water resource management in an extractive context (Table 11). These implementation and monitoring limitations are common challenges in developing nations, creating a gap between regulatory frameworks and on-the-ground reality [11]-[13][33][35].

Table 11. Proposed roles of stakeholders and avenues for improvement in a mining context.

ESIA Implementation Gaps: ESIAs are often perceived as procedural requirements for obtaining permits rather than as tools for genuine environmental prevention [32].Limited Independent Oversight: Pollution from ASM, which directly affects community water sources, often lacks adequate independent monitoring and regulation [10][13].Weak Integration: There is a clear absence of an integrated system linking environmental data with land-use planning and community grievances [7][27].

This situation is compounded by a lack of data transparency and limited technical capacity for rigorous environmental monitoring, as highlighted by several authors [32][33][44]. The establishment of a multi-stakeholder governance system that includes local communities is essential for addressing these gaps [11][12].

To address these challenges, we propose:

1)StrengtheningESIARegulations: Mandate a minimum parameter list (including heavy metals and cyanides), require multi-seasonal monitoring, and enforce public disclosure of results [12][35].

2)EstablishingLocalWaterObservatories: Create independent bodies involving universities, local authorities, state services, and mining companies for participatory monitoring [12][44].

3)DevelopinganEarlyWarningSystem(EWS): Based on the WPPI, integrate it with an open-data GIS platform to guide local decisions [24].

4)ImplementingStrategicZoning: Develop a strategic hydrological zoning that intersects mining pressure, usage vulnerability, and population density to prioritize interventions [27][47].

5)FurtherExternalValidationoftheWPPI: Conduct comprehensive external validation across independent datasets and contrasting mining contexts, complemented by expanded uncertainty and sensitivity analyses.