The study was carried out in the town of Daloa, which is located in central-western Ivory Coast and covers an area of 97.28 km². It is situated between 6˚22' and 6˚29' west longitude and between 6˚49' and 6˚56' north latitude ( Figure 1 ). It is the country’s third most populous city after Abidjan and Bouaké, with an estimated

population of 421,879 ( INS, 2021 ). It has a transitional tropical climate, with a rainy season from March to October and a dry season from November to February ( Kouadio , 2022 ). The Daloa formations belong to the Precambrian basement and there are generally composite aquifers in this environment: alterite aquifers (superficial) and fractured aquifers ( Lachassagne et al., 2011 ). Alterite aquifers are subdivided into two layers ( Figure 2 ): alloterites and isalerites ( Kouassi et al., 2024 ). The city of Daloa is located in a single geological domain, namely migmatites on heterogeneous granitoids.

The majority of hydrogeological parameters control water flow and permit assessment of the vertical vulnerability of groundwater. It is about the physical properties that play a role in the migration and attenuation of a contaminant within the soil-unaerated zone-aerated zone complex. There are many hydrogeological parameters, but we have chosen to focus on:

Well depth: this represents the distance (in metres) from the soil surface to the piezometric surface. It determines the thickness of the unsaturated zone and conditions the pollutant transfer time to the water table ( Patel et al., 2023 ; Fusco et al., 2024 ). For this study, campaigns to measure the water level in 30 wells were carried out in March and November 2022, which mark the start of the rainy and dry seasons, using an OTT 150 m light and sound piezometric probe. In practice, this involved undoing the graduated tape until the tip touched the surface of the water and a sound was emitted before the value was taken. It should be noted that this value is used to calculate the water’s edge.

Annual recharge: this is the main vehicle for transferring pollutants to ground- water. The greater the recharge, the greater the risk of contamination ( Maqsoom et al., 2021 ). In this study, recharge was determined on the basis of the piezometric fluctuation method (Equation 1):

$R=\theta \times \Delta H$ (Equation 1)

with R: total recharge (mm/an); θ: total porosity determined from soil samples taken with an auger around the wells using the saturation method; ΔH: seasonal water level variation.

Hydraulic conductivity of alterites: this is a physical characteristic that represents the ease with which a material allows fluid to be transferred through a connected network.

It controls the propagation speed of the pollutant in the aquifer. In the field, Porchet’s method, based on the principle of constant-load infiltration, was applied using a double-ring infiltrometer.

The principle is to follow the evolution of the water level as a function of time in the inner ring to determine the rate of infiltration at each time step. After a certain time, a steady state is established. The conductivity is then calculated on the basis of equation 2:

$C=\frac{r\times h^{\prime }}{t\times \left(2h+r\right)}$ (Equation 2)

where r: radius of the hole (m) in relation to the inner ring; $h^{\prime }$. height of the percolated water (m); t: elapsed time (s); h: height of the water column (initial value, in m).

Aquifer material: This is the lithology of the saturated zone. This parameter intervenes in the trapping of the pollutant, which can escape due to the absorption capacity of the soil ( Mbuluyo et al., 2017 ). Thus, the data collected from logs and lithological sections of boreholes in the study area give a more or less detailed idea of this parameter. To determine the ‘Aquifer material’ parameter, the layers representative of the saturated zone were considered.

Slope: This parameter indicates whether the water will run off at the surface or infiltrate into the ground. Indeed, the steeper the slope, the greater the runoff and the less the water will infiltrate ( Farah et al., 2021 ). On the other hand, an area with a weak slope favours infiltration and increases the potential for contaminant migration. To determine this, we used a digital terrain model (DTM) that can be downloaded from the Earth Explorer website.

Soil type: Soil is a hydraulic property that controls recharge during the water infiltration process ( Mboudou et al., 2016 ; Maqsoom et al., 2021 ). Its characteristics are of considerable importance in the path of water and therefore in the transfer time of pollutants from the soil surface to water. For this reason, the soil map of Côte d’Ivoire produced by Perraud & De La Souchère (1969) was used, revealing that the soil in Daloa is sandy-clayey.

Unsaturated zone: this is located between the soil layer and the water table. The nature of this zone is an important parameter, because it influences the speed at which pollutants propagate towards the aquifer. The influence of this zone on the potential pollution of the aquifer depends essentially on either the thickness, lithology or transit time ( Gourcy , 2022 ). For the present study, only lithology was considered. This parameter was determined on the basis of logs and lithological sections from boreholes.

In order to acquire socio-environmental data, it was necessary to carry out surveys of state structures and households. The information sought concerned:

Population density: this refers to the number of individuals per square kilometre. This parameter influences all human activities carried out on the surface of the soil, which are likely to produce pollutants directly or indirectly. For this study, the population data provided by the National Statistics Institute (INS) during the general population census in 2021 were used.

Management of wastewater and household waste: these parameters refer to the type of sanitation system used by the population. For these data, a survey was conducted among households in Daloa in 2021. The questions concerned the type of sanitation system used (septic tanks, refuse bins, nature). The proportions of households discharging wastewater and household waste into the environment were calculated.

Seasonal chemical analysis of well and spring water has enabled chemical parameters to be identified and quantified. It was revealed that heavy metals, particularly chromium, are present in high concentrations, regardless of the season.

The characteristics of chromium have therefore been taken into account in order to better explain all the physico-chemical interactions between the aquifer and the pollutant. These are the degree of adsorption (Ad) of the pollutant by the unsaturated zone and its solubility (So) in water, the information for which comes from documentary research.

This study consists to present the areas of high vulnerability linked to a specific pollutant on a map. The assessment and mapping of areas vulnerable to chromium pollution was made possible by the application of Multi-Criteria Analysis (MCA) based on the AHP method. This technique has proved effective in other studies of groundwater contamination. The map is drawn up in several steps. For this reason, a simplified conceptual model was developed to facilitate the work ( Figure 3 ).

The identification of factors relating to the study of the specific vulnerability of aquifers is conditioned by several criteria such as the hydrogeological criterion, the human criterion and the pollutant criterion. As part of this study, twelve (12) factors were selected ( Table 1 ), namely: depth (D), recharge (R), aquifer material (A), lithology of the unsaturated zone (LUZ), conductivity (C), slope (Sl) and soil (S), which constitute the hydrogeological factors. As for human factors, the population density (PD), the household waste management method (Hw) and the waste water management method (Ww) were retained. The pollutant factors are the degree of adsorption (Ad) of the pollutant in the unsaturated zone and its solubility (So) in water.

All the field data obtained was compiled and interpolated. Qualitative data (aquifer material, slope, unsaturated zone lithology, soil type, solubility and adsorption) were transformed into raster and quantitative values before being interpolated to the entire study area. They were then reclassified and rated. In practice, the data was interpolated in the ArcGIS environment. This was done using the Inverse Distance Weighting (IDW) method, known as the local deterministic method in the ‘Spatial Analyst tools’ module. The values were then reclassified on the basis of the several-author rating system shown in Tables 2-4 . Finally, as the factors were measured on different scales, they had to be standardised. As a result, each factor was associated with an entire scale with a score ranging from 1 to 10 in the ‘Reclassify’ tool.

Given the large amount of data to be processed and the complexity of the problem to be studied, the method developed by Saaty (1977) was used to quantify and rank the criteria and factors characterising vulnerability to pollution. It is a simple method whose backbone is consolidated by mathematical calculations that generate weighting coefficients whose sum is equal to 1. The methodological approach can be summarised in four (4) steps.

Construction of the hierarchy: this has consisted of structuring the problem to be solved into three levels ( Figure 4 ). Level 0 represents the main objective to be achieved. Level 1 corresponds to the three (3) criteria taken into account in this study. Finally, level 2 presents the twelve (12) evaluation factors.

Binary combination: this consisted in comparing the relative importance of all the elements taken two by two to configure reciprocal square matrices based on the model established by Saaty (1980) and on the numerical scale of El Majorni (2002) . Firstly, the various factors were established in a precise order based on actual information on the alterite aquifers in the city of Daloa. The order of classification is as follows: P-Sl-A-R-S-LUZ-C, Ww-Hw-PD and So-Ad. Thereafter, when two factors have the same importance in the study of vulnerability to pollution, the value 1 is given to them. However, if one factor is more important than the other, then it is given the value 5. Otherwise, it is given the opposite value.

Determining logical consistency: this involves calculating the RC consistency ratio. This operation is composed of seven (7) steps:

Normalisation of the original matrix: this step consists of dividing each numerical scale of each parameter following each column by the total for that column. In this matrix, a total is determined for each line.

Calculating the priorities of the parameters according to the lines: this involves dividing each total of the lines by the total of the totals of the lines.

Creating a new matrix by multiplying each priority by the columns of the original matrix. A total is then determined for each line.

Ratio of each line total by each priority.

Calculation of the maximum eigenvalue of the elements compared (λ_max) based on the following formula:

${\lambda }_{\max }=\frac{{\displaystyle \sum \text{RV}}}{n}$ (Equation 3)

With: RV_k, report values and n, number of elements compared.

Determination of the coherence index (CI) by the following formula:

$\text{CI}=\frac{{\lambda }_{\max }-n}{n-1}$ (Equation 4)

Calculation of the coherence ratio. Saaty defined, by experiment, a coherence ratio as the ratio of the coherence index (CI) to the random index (RI) of a matrix of the same dimension. The RI values are 1.35, 0.52 and 0 respectively for the hydrogeological, human and pollutant matrices ( Saaty , 2000 ; Alonso & Lamata, 2006 ).

$\text{CR}=\frac{\text{CI}}{\text{RI}}$ (Equation 5)

The consistency ratio provides information on the consistency of the judgements, because the values assigned can present a certain degree of inconsistency. This ratio is compared to 10% ( Saaty , 2000 ).

If RC ≤ 10%, then the matrix is said to be coherent;

If RC > 10%, then the matrix is not consistent and should be revised.

Calculation of factor weights: the eigenvectors and weighting coefficients by line were determined according to the equations below:

$\text{Ev}=\sqrt[n]{W_i\times \cdots \times W_n}$ (Equation 6)

$\text{WC}=\frac{{\text{Ev}}_i}{{\text{Ev}}_i+\cdots +{\text{Ev}}_n}$ (Equation 7)

with WC: weighting coefficient; Ev: eigenvector for each parameter; W_n: main scores assigned to the parameters.

Complete aggregation of the factors involved multiplying each reclassified factor map by its respective weighting coefficient and adding the results to produce a suitability index on a scale of 1 to 10.

Mena (2000) considers that this method is the only one applicable when there are several criteria which vary continuously in space. By adapting the vulnerability index formula established by Gogu & Dessargue (2000) , we can write:

$\text{SAI}={\displaystyle {\sum }_{i=1}^n{\text{WC}}_i\times {\text{RM}}_i}$ (Equation 8)

with SAI: specific ability index; WC_i: weight of factor i; RM_i: reclassified map of factor i.

This equation was used individually for the three decision criteria. Subsequently, the specific chromium vulnerability index noted (SVI) was obtained according to the following formula:

$\text{SVI}=\text{WC\_Hg}\times \text{SAI\_Hg}+\text{WC\_Hu}\times \text{SAI\_Hu}+\text{WC\_P}\times \text{SAI\_P}$ (Equation 9)

with WC_Hg, WC_Hu and WC_P the weighted coefficient of the different criteria considered.

Next, the equal interval classification method of Aydi (2018) was used to classify the vulnerability to specific pollution into five classes ( Table 5 ).

To affirm the reliability of the specific pollution vulnerability map of the city of Daloa and to appreciate the relevance of each variable on this map, two (2) verification tests were carried out. The first test consisted in determining the margin of error on the final pollution vulnerability map. The second test, which is an empirical validation criterion, consisted of calculating the coincidence rates of high chromium concentrations analysed in well and spring water in Daloa and in highly vulnerable areas. This is one of the approaches (Superimposing existing groundwater chemical data on the map) widely used to empirically validate a vulnerability map.

Calculation of the margin of error: the margin of error permits to appreciate the quality of the information obtained from the vulnerability map. It was calculated on the basis of the equations bellow:

$\Delta \bar{X}=\frac{\sigma }{\sqrt{m}}$ (Equation 10)

$\text{Er}=\frac{{\displaystyle \sum \Delta \bar{X}}}{\text{IvM}}$ (Equation 11)

with $\Delta \bar{X}$: uncertainty on the average index; 𝜎: standard deviation of vulnerability indices; m: number of data considered; IvM: average vulnerability index.

Determination of coincidence rates: chromium concentrations were projected onto the pollution vulnerability map, and the coincidence rate was determined by calculating the ratio of the number of wells whose pollution level was consistent with the vulnerability level over the total number of wells sampled. To do this, 27 wells with chromium values greater than 0.05 mg/L were considered. In validating the vulnerability map, the areas that are really contaminated must correspond to those with the highest vulnerability indices. This verification technique has been adopted by several authors, including Deh (2013) , Gougo (2016) , Kumar et al. (2017) and Kouadio (2019) .

The obtaining of the different thematic maps has involved multiplying each factor map by its weight (Cp).

Thus, Tables 6-9 show the weighting coefficients based on the numerical values assigned (see square matrices) and the consistency ratio. These tables show ratio values between 0% and 0.05%. This shows that the matrices are consistent. On the basis of the order of importance of each factor established, the weighting coefficients decrease from the most important to the least important factors.

Human vulnerability map: Figure 5(a) shows two (2) classes of vulnerability to pollution, namely high and very high, on the human vulnerability map. These two (2) occupy 89.37% and 10.63% respectively. As a reminder, drawing up this map was made possible thanks to data on population density, wastewater management and household waste. Thus, it is clear that the human activities carried out in the area seem to be crucial in terms of the vulnerability of the alterite aquifers.

Pollutant criteria map: The sensitivity map linked to the characteristics of chromium ( Figure 5(b) ) indicates a very high vulnerability class throughout the study area. This indicates that chromium or all non-reactive and soluble pollutants could migrate towards the groundwater of the city of Daloa.

Hydrogeological criterion map: It shows four (4) classes of vulnerability to pollution (low, medium, high and very high) in the study area after superimposing the thematic maps ( Figure 5(c) ). However, the high class predominates, and represents 86.81% of the surface area. Then come the medium, very high and low classes, which occupy 11.64%, 1.52% and 0.03% respectively. The latter are poorly represented on the map. Consequently, this situation indicates that hydrogeological factors would favour the progression of certain pollutants towards the water table if they were present in these areas.

The specific vulnerability map based on the AMC method, obtained by combining the maps, shows three classes of vulnerability (medium, high and very high) ( Figure 6 ). Analysis of this map shows that the medium and very high vulnerability classes hardly exist in the study area (0.79% and 1.68% of the total surface area of the study area); instead, it is dominated by the high class (97.53%). In this case, a very close eye must be kept on any intense anthropogenic activities that tend to pollute groundwater.

The validity of this map was made possible by the margin of error and the coincidence rates of high chromium concentrations. The margin of error obtained on the map is 0.68% ( Table 10 ), showing that the information on the map is reliable. This error value was obtained by considering only those factors for which the data came from the field missions. In terms of uncertainties, the depth factor has the highest value (0.08).

As for the coincidence rate, the distribution of chromium concentrations on the vulnerability map reveals that 78.26% of chromium concentrations > 0.05 mg/L coincide with areas of high vulnerability ( Figure 7 ).