Estimating the vulnerability of groundwater resources to diffuse pollution in highlands areas: review of the literature and critical analysis (highlands of Cameroon)

One-third of humanity lives off groundwater for its daily needs, and 99% of planet's accessible freshwater is found in aquifers (FAO 2016). The quality of these waters is declining worldwide, mainly due to human activity, urbanization, land use change, agriculture, and rising living standards (Nguedia Djatsa et al. 2022; Nguedia et al. 2023). In agricultural areas, the problem arises because fertilizers and plant protection products can reach aquifers and then contaminate them for many years. Potential sources of pollution that can damage groundwater in urban areas include construction sites, vehicular traffic, commercial and industrial activities, polluted areas, and waste and wastewater management (Ducommun 2010).

In Africa and Asia, there are still areas where groundwater vulnerable to pollution is consumed. As this susceptibility is linked to many parameters, there is a range of assessment methods (Mba et al. 2023). These can be highly complex and costly, or simplified, each designed to assess vulnerability in environments that may vary in space and time. There is no guarantee that a method developed in a different natural setting will be valid in the geomorphology of high plateaus, mountains, and hills. Because of this, a literature review was the guiding method for clarifying the notion of vulnerability, the state of knowledge of groundwater, the notion of pollution, and methods for estimating groundwater vulnerability. A critical review showed the indispensability of developing models for each environment, such as the uplands.

The world's freshwater reserves represent 3% of the total volume, fresh groundwater accounting for 33% of that (Marsily 2009). The development of human activity in cities in developing countries suggests that groundwater resources are becoming increasingly vulnerable – e.g., in Cameroon's other cities and the country's western highlands. An understanding of intrinsic versus specific vulnerability is important in this context.

According to Vrba & Zaporozec .(1994), vulnerability is an intrinsic property of a groundwater system that depends on its sensitivity to human impacts. Allier et al. (2008) note that vulnerability is the ‘failure of groundwater to protect or defend itself naturally against pollution threats, depending on local hydrogeological conditions’. In other words, vulnerability represents the natural environment characteristics that determine groundwater's sensitivity to pollution. In this study ‘vulnerability’ is always associated with groundwater contamination.

Schnebelen et al. (2002) acknowledge two types of vulnerability:

• Intrinsic vulnerability represents the natural environment's characteristics that determine the groundwater's sensitivity to pollution from human activity.

• Specific vulnerability defines the groundwater's vulnerability to a particular pollutant or group of pollutants, and takes into account the pollutants' properties and their relationships with the various intrinsic vulnerability components.

The distinction between the two types of vulnerability is important, as they are not generally on the same scale.

Groundwater pollution makes it unsuitable for one of its uses or disrupts the aquatic ecosystem. It can have various sources including phytosanitary products (insecticides, herbicides, fungicides), fertilisers (nitrogenous, etc.), industrial or other accidents (leaking tanks, fires, etc., or lost lorry loads), leaks from sewers, uncontrolled waste dumping, etc. (Margat & Monition 1971; Marsily 2009; Behanzin et al. 2021). Pollutants that can reach the water table behave differently as they migrate through the soil and subsoil, and their mobility and persistence in the subsoil are highly variable depending on solubility and ease of biodegradation. These mechanisms depend not only on the nature of the pollutant but also on pH conditions, microbiological activity, humidity, and soil organic matter content (Allier et al. 2008). Taking the nature of the pollutants into account when mapping vulnerability would therefore mean drawing up numbers of maps, taking into account the specific characteristics of the different species.

In fact, the concept of intrinsic vulnerability is used. It is independent of the pollutant and describes the groundwater resource's susceptibility to being affected by diffuse or point source surface pollution. Combined with information on human activity, it can be used to assess the risk of groundwater pollution.

A number of intrinsic vulnerability studies include the studies by Dibi et al. (2013). There is no standard method for assessing specific vulnerability. Kouakou et al. (2022), studying specific vulnerability to nitrate (NO₃) in groundwater in Abidjan, used a combination of a geographic information system (GIS) and multi-criteria analysis, taking the pollutant's properties and relationships with the various intrinsic vulnerability components into account.

Aquifers are hydrologic and hydrodynamic systems characterized by their geometry, area, and depth, as well as the intrinsic properties of the rocks that comprise them, such as lithology, porosity, permeability, fracturing, and homogeneity (Gilli et al. 2008). The availability of water in wells between seasons is a subject of debate and disagreement. According to convention, a well should have a positive flow during the rainy season and dry up during the dry season (Barthel et al. 2021). However, studies have shown that unconfined aquifers can receive lateral recharge of greater magnitude than vertical recharge during the dry season, allowing wells to have a good flow compared to the rainy season, as observed by Huang et al. (2021). This observation has also been made by Nguedia Djatsa et al. (2022) in Bafoussam, where water is relatively plentiful during the dry season compared to the rainy season. This is because the soil horizons rest on ignimbritic formations, rich in clay, in the unsaturated zone, which do not favour rapid well recharge, but rather slow recharge that reaches the water table in the dry season. Thus, a well in an unconfined aquifer can be fed either by micro-fractures that drain water, contributing to well recharge, or by low drainage of soils from the watershed.

Pollution is often associated with irresponsible and unsustainable practices (Nguedia et al. 2023). Companies and individuals often prioritize their short-term economic interests at the expense of the environment, resulting in excessive exploitation of natural resources, greenhouse gas emissions, the production of toxic waste, and natural habitat destruction. Pollution can also be critical because of its potential impact on human health. Exposure to atmospheric pollutants such as fine particles, nitrogen oxides, and volatile organic compounds can lead to respiratory problems, cardiovascular diseases, and even premature deaths. Chemical pollution can, equally, contaminate drinking water sources, with adverse effects on human health (Raimi et al. 2022; Tontsa et al. 2023).

Numerous methods for determining groundwater vulnerability have been developed (Gogu & Dassargues 2000). According to Liggett & Talwar (2009), the methods can be divided into three main categories:

• Parametric, with superimposable indices (empirical)

• Physical modelling (numerical mathematics)

• Statistical (predictive)

The choice for producing a vulnerability map depends on the degree of complexity required (Table 1).

Parametric methods enable vulnerability indices to be calculated quickly and empirically. They are used frequently and based on a selection of attributes and physical parameters representative of vulnerability (unsaturated zone thickness, type of soil, etc.), which are discretized and classified by the decreasing interval of supposed effect on contamination attenuation (hence increase in vulnerability) (Gogu & Dassargues 2000). There are three sub-systems:

• Matrices – based on a limited number of rigorously selected physical parameters presented in a matrix form. The vulnerability index is estimated quickly and simply, using a multi-parameter table system.

• Combined classified parameters – the physical parameters deemed necessary for estimating vulnerability are distributed over a scale of their estimated influence on attenuation of the potential contamination. The classes are combined using a defined calculation scheme to give a final vulnerability index.

• Point-counts – similar to the combined classified parameter system, with a weighting factor for each parameter, according to its effectiveness in mitigating potential contamination. Final index values are obtained by the weighted parameter value sum.

Index mapping methods with criteria weighting (point count system models) are relevant because they take the relative importance of each criterion into account linked to the water table's vulnerability. They are also widely recognized and used (Vrba & Zaporozec 1994; Gogu & Dassargues 2000; Murat 2000; Doerfliger et al. 2009; Mohamed et al. 2012; Mfonka et al. 2019; Chakraborty et al. 2022). The most widely recognized include PaPRIKa, EPIK, RISK, DISCO, GOD, DRASTIC, SINTACS, AVI, and SI, and assessment commonly takes the criteria described in Table 2 into account (Bézèlgues et al. 2002).

The most widely recognized methods for estimating groundwater vulnerability depend on the environment in which the aquifer stands.

The first method is dedicated to karstic aquifers (Doerfliger et al. 2009), and it is based on four criteria: epikarst, protective cover (soil), infiltration conditions, and karst network. It was developed to delimit water catchment protection perimeters in accordance with Swiss legislation. An advantage is that it reflects the sensitivity of groundwater to any type of contaminant. It has been used in the Swiss Jura (Doerfliger & Zwahlen 1998).

This was inspired by EPIK because of its karstic specificity. It is based on five criteria: aquifer rock, infiltration, soil, karstification, and epikarst (Petelet-Giraud et al. 2000).

This is a multi-criteria method for mapping karstic aquifers' intrinsic vulnerability, and its targets are the resource and catchment. Developed specifically for karstic aquifers, it cannot be applied indiscriminately to all carbonate aquifers (Doerfliger et al. 2009). If the carbonate aquifer does not function as karst, it is considered a fractured aquifer and the method does not apply (Doerfliger et al. 2009).

The aim of DISCO (discontinuities – protective cover) is to define protection zones, taking environmental heterogeneity into account. Three parameters are necessary to assess pollutant transport from any point in the catchment area: the ‘discontinuities’; the ‘protective cover’; and the ‘runoff’, which covers surface runoff phenomena before infiltration (slope runoff, and permanent or temporary watercourses). It has been used to estimate groundwater vulnerability in fissured terrain (Mardhel et al. 2005).

DRASTIC was devised in the 1980s by the National Water-Well Association in response to a request from the Environmental Protection Agency (Newell et al. 2023) to enable intrinsic vulnerability estimation in relation to hydrogeological conditions in the United States. DRASTIC is based on seven parameters, each of which is assigned a multiplying factor (Dp) (Table 3).

DRASTIC is most widely used in Africa, particularly in Cameroon. Ake et al. (2010) used it to study intrinsic vulnerability to nitrate pollution (NO₃) in the loose ground of the Bonoua aquifer (south-east of Côte d'Ivoire), where three vulnerability classes (medium, high, and very high) were identified. It has also been applied to assess vulnerability in fissured terrain. In the western Cameroon highlands, DRASTIC has been used in Dschang and Foumban by Mba et al. (2019) and Mfonka et al. (2019). Although designed for loose formations in agricultural areas, the method has had to be applied in towns, as in the works of Yu et al. (2022), Gintamo (2022), Saravanan et al. (2023).

This is an empirical method devised in 1987 by Foster. Its name comes from the acronym: Groundwater occurrence, Overall aquifer class (aquifer characteristics in terms of lithology and porosity), Depth to water table. Like DRASTIC, it has been used in a number of studies (Djenadi & Djerroud 2018).

Analysing aquifer vulnerability to saline intrusion is a special case of both specific and intrinsic vulnerability. Both salinization and intrinsic vulnerability criteria must be taken into account. In this sense, it is a specific vulnerability. In terms of index mapping, GALDIT is dedicated to coastal aquifer vulnerability to saline intrusion (Doerfliger 2011). GALDIT uses six parameters:

• Type of aquifer: confined or unconfined

• Depth to water table below sea level

• Aquifer hydraulic conductivity

• Distance from shore

• Impact of existing saline intrusion in the area

• Aquifer thickness

The method is used to assess groundwater vulnerability in coastal areas, as in the study by Amarni et al. (2020) who worked in coastal Cherchell aquifer, Central Algeria.

Susceptibility index (SI) is a specific vertical vulnerability assessment method, developed by Ribeiro Aydi (2018), which takes into account agricultural pollutants, particularly nitrate and pesticides. Five parameters are used, the first four parameters are identical to those used in DRASTIC – D: depth to water table; R: effective aquifer recharge; A: lithology; T: topographic slope. The fifth parameter is land use (OS – Table 4), with a score of 0–100 assigned to each land use class. SI has been used by several authors as it gives the specific vulnerability via OS (Ake et al. 2010; Attafi 2018).

SINTACS was developed in the 1990s (Petelet-Giraud et al. 2000) and is a version of DRASTIC adapted to Mediterranean conditions. The method is specific, compared to DRASTIC, because it suggests five vulnerability scenarios, from ‘normal’ to ‘severe’, relating to different aquifer types.

Some authors combine methods to enable comparison to determine the most suitable. DRASTIC is considered complex in most cases but is thought suitable compared to other conventional groundwater vulnerability assessment methods – e.g., Momejian et al. (2019). Recently, Luo et al. (2023) estimated groundwater vulnerability by comparing versions of DRASTIC and GALDIT adapted using hierarchical process analysis (AHP).

The methods are often modified, too. DRASTIC is generally the most suitable, given its complexity and the slightly larger number of parameters used. For example, Bouchnan (2015) used the modified F-DRASTIC method to estimate fractured aquifer vulnerability in Morocco, introducing a fracturing index ‘F’ as the eighth parameter. Sener & Davraz (2013) used an AHP multi-criteria analysis to adapt DRASTIC to assess groundwater vulnerability in the Egirdir basin of Lake Isparta, Turkey. This made it possible to obtain new weights and add two parameters – lineaments and land cover – to DRASTIC. Other authors have relied on modifying the DRASTIC's factor weightings using the AHP method. Examples include the studies by Saravanan et al. (2023) in Cuddalore, India, and by Saravanan et al. (2023) assessing the vulnerability of unconfined groundwater in Thoothukudi, Tamil Nadu, India.

Some authors find DRASTIC complex and select its important factors to create a modified version with reduced parameters. An example is the DRIST method (Drias & Tubal 2015), based solely on parameters relating to recharge, the soil (slope and lithologic nature), and the unsaturated zone (nature and thickness). The authors consider that only these parameters are involved in pollution transmission to the water table, and assign the same weights, scores, and classes to them as those defined for DRASTIC.

Vulnerability mapping is a decision-making tool for land use planning and water resource protection, and presupposes the validity of the maps available. Vulnerability assessment parameters, however, tend to vary over time, and this approach is also limited by the temporal variability of other human activities, such as irrigation and exploitation. Consequently, the study of vulnerability dynamics as affected by the parameters' temporal variability is also important, enabling vulnerability evolution to be monitored, while identifying the key factors for groundwater resource management and protection.

The study area – the Cameroon Highlands – is not in a karstic zone, so methods like PaPRIKa, EPIK, and RISK, developed for karstic environments (Doerfliger et al. 2009), are not suitable. DISCO (discontinuities – protective cover) aims to define protection zones and characterise the vulnerability of a highly heterogeneous fissured discontinuous aquifer, taking into account the geological and hydrogeological factors conditioning the fissured aquifer's functioning (Mardhel et al. 2005). For example, Dibi et al. (2013) working in Côte d'Ivoire on granular aquifers used DRASTIC and SINTACS, each with seven vulnerability parameters. (An alterite comprises superficial, usually unconsolidated, material formed in situ by the physico-chemical alteration of previous lithologies, without notable pedological transformations.)

For this study, the margins of error for the map veracity rates were about 17 and 18% for DRASTIC and SINTACS, respectively. The values obtained make it possible to revise the parameter scores and weights in DRASTIC and SINTACS.

Saida (2013) working in the Mitidja (northern Algeria) on groundwater vulnerability to nitrate showed that by applying the standard DRASTIC method coupled with GIS, areas with high nitrate levels were superimposed on areas of high vulnerability. Unlike the study area – low hills flattened by erosion, and flat-bottomed valleys through which permanent and intermittent watercourses flow (Regnoult 1986) –the Mitidja is a vast plain for which the standard DRASTIC method should be modified, as topographic effects on vulnerability will not be as representative because of the non-uniform upland geomorphology.

The comparative study of methods for estimating the intrinsic vulnerability of granular aquifers to pollution in the Laurentian Piedmont by Murat (2000) demonstrates that DRASTIC was not designed for all environments, as it was developed in the context of nitrate pollution in the United States. Each environment has its own properties, which is why Ducommun (2010) specifies that adapted intrinsic vulnerability for urban areas should spatialize the impact of urban elements on the intrinsic vulnerability index values. A map of anthropogenic processes affecting recharge and a spatialization of risks in urban areas should also be produced. Mohamed et al. (2012) evaluated groundwater vulnerability to pollution in Meknes, Morocco, based on indexing and environmental factor weighting (DRASTIC and GOD) and showed that DRASTIC is better for assessing groundwater vulnerability to pollution. Since Meknes is a densely populated city, the method used does not take land use into account. Mfonka et al. (2019) studied the Foubam aquifers on the Western Noun Plain using piezometric monitoring coupled with DRASTIC and showed, through statistical tests, that the water table's depth and the unsaturated zone's impact both have significant impacts on estimating vulnerability.

Over time, authors like Saranya et al. (2021) and Saravanan et al. (2023) have learned to understand and adapt the weights and scores of DRASTIC's factors to their study contexts, using the AHP method. The problems of factor redundancy persist, however, making the model more complex and costly. Another, previous approach for adequate groundwater vulnerability estimation is that authors change (increase/reduce) the number of parameters, as well as adapting the model's weights. Adding factors makes the model more complex with an increase in criteria redundancy, making it more costly. Those who reduce the criteria do not usually add new ones – e.g., Sener & Davraz (2013) and Gintamo (2022) – who opt for simplified models that reflect their natural frameworks' specificities better. Indeed, any work based on DRASTIC but outside its context would be limited by the fact that the factors remain those of the natural framework for which it was developed. It would be appropriate for each study, in fact, to develop a groundwater vulnerability estimating model that takes the intrinsic factors of the natural framework into account.

The study's aim was to present the notions of vulnerability and pollution, as well as methods for estimating groundwater vulnerability. Numerous methods for estimating vulnerability exist that do not fit in with all natural settings – e.g., upland areas. Critical analysis of existing methods and their applicability to uplands revealed that the spatio-temporal variability of the criteria generally considered is a factor in their lack of validity in all types of environments, specifically uplands.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.