Multi-criteria analysis applied to the selection of drinking water sources in developing countries: a case study of Cali, Colombia

The high variability of water quality in surface water in Colombia, including the level of risk to human health, is threatening its use as a source for drinking water. Colombia is characterized by its abundance of water resources, with its average water yield exceeding by six times the average global water yield and by three times the Latin American water yield (IDEAM 2010). However, the availability of water is constrained by the deterioration of its quality. Progressive deterioration of the surface water is mainly caused by rapid urbanization, in combination with a lack of integration between water management and spatial planning. The problem has been exacerbated by inappropriate land use, poor protection of the river basins, wastewater and stormwater discharges from municipalities, discharges from domestic and industrial wastewater treatment plants, and also mining, deforestation processes and improper management (van der Kerk 2011).

Multi-criteria analysis (MCA) tools have been widely used in decision-making situations to ensure a drinking water supply, in terms of water quantity and quality, where a diverse range of alternatives is available.

MCA is a decision-analysis tool for dealing with complex decision-making problems with more than a single criterion and involving a variety of stakeholders. As described by Lai et al. (2008), there are various decision-support methodologies available for MCA, which can be classified as follows: elementary procedures (e.g. conjunctive and disjunctive methods); single synthesizing criterion approaches (e.g. Multi-Attribute Utility Theory, Multi-Attribute Value Theory, Analytical Hierarchy Process); outranking methods (e.g. ELECTRE, PROMETHEE, REGIME, NAIADE); and goal or reference point methods (e.g. Goal Programming and Compromise Programming).

The selection of the decision-support methodologies depends on the availability of information and its quality (Lai et al. 2008). Jaber & Mohsen (2001) described the development of a decision-support system for the evaluation and selection of potential non-conventional water supply systems in Jordan using the Analytic Hierarchy Process. Domènech et al. (2013) analyzed the compatibility of non-conventional centralized and decentralized water supply technologies in Spain using a social multi-criteria evaluation.

This paper focuses on the development of an MCA framework to assess the alternative that best suits the necessities of the city of Cali, Colombia. That framework is based on the drinking water supply problems suffered there and the external constraints involved in the decision-making structure in developing countries such as Colombia.

A review was carried out in order to pinpoint the most pressing drinking water related issues in Cali. Moreover, additional data collection and analysis were conducted to achieve a better understanding of the magnitude of the contamination of the Cauca River, which is used as the main water source for the water supply to Cali. The pre-proposed alternatives were evaluated objectively in order to define the more suitable solutions to meet the future water demand of the city.

This study is divided into two parts: (1) impact of water quality on Cali's current drinking water system; and (2) the application of the MCA tool. The methodology used in each part is described as follows:

• (1) The impact of water quality on the functioning of the drinking water system for the city of Cali was reduced to the evaluation of the impacts of the current Cauca River water quality on the Puerto Mallarino water treatment plant (WTP) because, as described above, it represents the highest water coverage for the city. This analysis was based on turbidity and dissolved oxygen (DO) because these parameters correspond to the operational thresholds in the WTP. The turbidity indicator is related to extreme sediment loads, and the DO is associated with extreme contamination events. Therefore, when either of the mentioned parameters reaches or surpasses the operational thresholds, the WTP may not treat the entering water due to the uncertainty about contaminants in the source water. Other parameters are monitored at the WTP intake and in a station located 4.5 km upstream from the WTP intake (pH, color and electrical conductivity). However, only turbidity and DO are linked to the operational thresholds.

• (2) Multi-criteria decision analysis. An inventory was made of the alternatives previously proposed by the drinking water company (EMCALI EICE ESP), environmental authorities, consultants and universities. The alternatives studied at a pre-feasibility and feasibility level, which by themselves or in combination comply completely with the water criteria demanded by the city, were compared by applying the MCA tool. The water company has considered the alternatives should be included as promising solutions, therefore in this study additional alternatives were not proposed. The water demand for the city is expected to reach 11.9 m³/s in 2036. The alternatives were evaluated by projecting the costs until the year 2036. However, it must be pointed out that the ‘useful life’ is not the same for all alternatives, since some of them will use existing facilities that may require upgrading. A sensitivity analysis was conducted by applying the MCA under different criteria scenarios.

Cali is the capital of the Valle del Cauca Department, located in the south-western part of Colombia between the Central Mountain Range and the Pacific Ocean. It is the third most important city in the country, with 560.3 km² of municipal area and a population of approximately 2.5 million people. The city has experienced rapid population growth in the past decades due to its geographic location as the principal urban, cultural, and economic centre in south-western Colombia (Universidad del Valle and UNESCO-IHE 2008).

Puerto Mallarino and Río Cauca WTPs mainly supply the Low network. However, during drought periods these WTPs partially support the High and Reforma networks as well (see Figure 1 and Table 1). Surface water from the Cauca River is extracted and treated directly by both WTPs. The water quality and quantity from the Cauca, Cali and Meléndez rivers are variable (PDA 2008; Universidad del Valle and UNESCO-IHE 2008). Therefore, when there is a shortage of water at any WTP, Puerto Mallarino and Río Cauca WTPs must compensate for these deficits (Table 1). La Rivera WTP extracts its water from the Pance River, which has good water quality in terms of its physical, chemical and microbiological parameters and is in agreement with Decree 1594/84 (Colombian regulation for water use and residual liquids).

Puerto Mallarino and Río Cauca WTPs cover 77% of the city, and occasionally the coverage reaches more than 80% of the city: when water flows in Meléndez, when the Cali rivers are too low, and when the Río Cali and La Reforma WTPs are not able to produce a sufficient water flow. Cali thus relies heavily on the Cauca River as its main source of potable water.

At the end of 2009, the clarified water reservoir was put into operation to deal with turbidity and contamination events. The reservoir stores clarified water, which is the water coming from the compact circular sludge blanket and solids contact reactors. This allows the mixing of raw water and water from the reservoir during high turbidity peak events (turbidity >2,500 NTU) and avoids the intake of raw water during contamination events (using only water from the reservoir when DO is in the raw water <2.5 mg/L).

The Río Cauca WTP accounts for approximately 21% of the city's water demand (PDA 2008; Universidad del Valle and UNESCO-IHE 2008). According to PDA (2008) the WTP currently uses a complete train with solids contact reactors, comprised of a lateral intake; 2 km of pipeline to the WTP; activated carbon dosing in an injection line; one grit removal chamber; one raw water pumping station with six pumps; one distribution chamber to reactors; six reactors for rapid mixing where chlorine (pre-chlorination step) and coagulant (mineral polymer of aluminum) are added and where flocculation and particle settling processes occur; 32 rapid sand filters (only sand) at a constant rate and constant head; chlorine disinfection in a contact tank; pH conditioning where Ca(OH)₂ is added; two chlorine contact tanks; and a drinking water pumping station with five pumps to the distribution network. Its intake is located at a few meters downstream of the intake of Puerto Mallarino WTP.

MCA was used to evaluate alternative solutions for covering future water demands in Cali based on selected criteria. The selection of these criteria was made by comparing similar frameworks on urban water management (Balkema et al. 2002; Domènech et al. 2013); however, specific issues regarding the problems of this case study, such as current legal and social acceptance, were also included. Therefore, the following criteria were selected: investment, O&M requirements, sludge management, environmental impact, vulnerability issues, current legal aspects and social acceptance. A preliminary framework was given to a group of six professionals with senior experience in the water supply sector and from different institutions (universities, environmental authorities and independent consultants) in Cali. The group was asked to review the framework and to evaluate each criterion separately, which resulted in a weighting factor according to the given situation. The final criteria selection and weighting was realized as objectively as possible, in agreement with the concept given by the group of professionals who individually provided input. The inventory of alternatives was also given to the group, who then graded the criteria individually, according to the resulting framework (Table 2).

Table 2 displays the framework for the assessment of the alternatives, including the criteria with their respective descriptions, assigned weight and grading scale, which was the result of the professionals’ judgment. Political and stakeholder interests, especially in developing countries such as Colombia, are key factors in influencing decision-making. Obtaining environmental permits for the execution of major projects such as dams can, therefore, take many years, due in part to bureaucratic procedures and obstacles placed by both legally settled communities and the slums that are frequently located on the banks of rivers and other protected zones (iDMC & NRC 2011). Hence, the legal and social acceptance criteria (weighting factor 20) became important in the use of the MCA for this case study. Additionally, O&M and sludge disposal are the main processes leading to costly water production in the city of Cali. The costs are linked to chemical usage, sludge production and its treatment, and energy consumption due to pumping actions (Universidad del Valle and UNESCO-IHE 2008), making O&M of crucial importance to the selection of alternatives for Cali.

Multi-criteria decision analysis, based on the outranking method, was used in order to provide an approach to select a robust and flexible alternative. The PROMETHEE method was used to evaluate the alternatives. PROMETHEE uses the outranking principle to rank options where, as stated by its authors, it is possible to introduce arbitrary numbers for the weights, giving the relative importance of the criteria according to the problem statement. As described by Brans & Mareschal (2005), the preference structure of PROMETHEE is based on pairwise comparisons, where the deviation between the evaluations of the alternatives for a criterion is considered.

In this study, the judgment of alternatives was based on the complete ranking, where the net outranking flow (Φ) was considered, Φ = Φ⁺–Φ⁻. The net outranking flow is the balance between the positive and the negative outranking flows. The positive outranking flow, Φ⁺, expresses how an alternative outranks all others, while the negative outranking flow, Φ^–, expresses how an alternative is outranked by all others. The net flow is built on clear and simple preference information and relies on comparative statements (Brans & Mareschal 2005).

This proposed MCA approach may serve as an example of how this analytical and decision-making process can be used for the selection of suitable alternatives for solving the drinking water problem facing the city of Cali.

There are two main reasons for the deterioration of water quality at the Puerto Mallarino WTP intake (Universidad del Valle and UNESCO-IHE 2008): diffuse and point sources. Diffuse sources are mainly due to the deforestation upstream, which causes higher levels of erosion leading to higher turbidity during heavy rainfall events, which is a general problem for the river basin. Point sources originate, for example, from the South Canal (see Figure 1) and the Navarro disposal site, which discharge raw sewage and other waste materials into the river and is only 11 km upstream of the Puerto Mallarino WTP. Operation in 1985 of the upstream Salvajina Dam also had a significant impact on the river water quality. This dam was constructed to control floods, dilute contaminants in the water and for hydro-power generation. Because of its operation, the hydraulic characteristics of the river were modified (Ramírez et al. 2010), leading to higher concentrations of suspended solids (EMCALI 1987, cited by Pérez-Vidal et al. 2012).

The Cauca River contains extremely high turbidity values (up to 10,000 NTU) according to EMCALI (personal communication, August 21, 2015) during the winter months. Contamination events reduce the DO to levels below 2.5 mg/L, which can indicate a serious risk to human health due to the uncertainty about the types of pollutants carried by the river. Water containing high suspended solids concentrations and water with low DO levels cannot be treated with the current treatment system at Puerto Mallarino WTP, resulting in the closure of the treatment plant during times of such concentrations. Closures leave a large proportion of the population of the city at risk of having no water during these times, with only the 80,000 m³ emergency reservoir (up to 4 hours' supply capacity) available at Puerto Mallarino to bridge these periods.

High turbidity events in the Cauca River lead to the WTP's intake shutdowns, if they are over a certain threshold (2,500 NTU). An analysis of the raw water intake turbidity data at Puerto Mallarino has yielded the following information related to when turbidity peaks occur, how long they last, and how they are distributed.

DO concentrations in the vicinity of Puerto Mallarino typically drop after heavy rainfalls with the increase in organic waste concentrations (CVC & Universidad del Valle 2004). The resulting low DO concentrations also lead to WTP intake shutdowns, as they are an indicator of high pollution peaks from mainly domestic, as well as industrial, wastewater. Table 3 outlines how often and for how long breaches of the critical 2.5 mg/L DO threshold occurred from 2008 to 2014.

EMCALI and CVC have been studying various alternatives at a pre-feasibility and feasibility level in order to improve the city's water supply and thereby eliminate the shutdown incidents and guarantee a viable water supply in the present context of increasing demand. The water sources and technologies studied are listed below, in which the alternatives considered for the analysis are denoted with the prefix ‘A’.

The rapid development of the District of Aguablanca in the early 1990s led to the construction of four deep wells with a total capacity of 600 L/s as a supply option in this area (Figure 1). During the period between May 1992 and February 1993, the national energy crisis required that the wells be used for a few hours a day, leading to intermittent service. Therefore, the population began to store water, which led to the oxidation of the iron and manganese commonly present in the groundwater, and caused a yellowish color in the water that created displeasure and rejection among the consumers. This resulted in the suspension of the use of those deep wells for their water supply (PDA 2008). EMCALI, in 2012, publicly requested the construction of a groundwater treatment system for the removal of iron, manganese, low DO, microorganisms, methane and hydrogen sulfide in order to reinstate the aforementioned wells (Alternative 8, A8) (EMCALI 2012). A plan has been proposed to supply the extracted and treated water to the Low network. In addition, the construction of 14 more deep wells has been considered to increase water production to 2.5 m³/s (Alternative 9, A9). The wells would be constructed from the location of the current wells up to ‘El Hormiguero,’ a community located approximately 7 km upstream of the South Canal discharge.

Currently, EMCALI is enlarging the water storage capacity of the reservoir by constructing an additional 100,000 m³ reservoir (Alternative 10, A10) to increase the bridging period during critical events in the Cauca River. This decision requires an analysis of its impact on water quality, hydraulic performance and water temperature behavior on the current Puerto Mallarino WTP. This last parameter played an important role in the operational aspects since it could generate a mass inversion in the water, altering the functioning of the sludge blanket and solid contact reactors of the WTP.

Since 2009, riverbank filtration (RBF) (Alternative 11, A11) has been considered as an alternative to resolving the water supply problems in the city of Cali, which implies process changes to the existing water treatment scheme in order to help with the variable influent loading conditions. Production wells should extract the water some distance from the water body. As the surface water travels through the sediments, it is expected that many contaminants are being removed (Schubert 2003). RBF has been used in Europe for more than 100 years (Schubert 2002; Tufenkji et al. 2002). In the United States, RBF has been used for over half a century (Ray et al. 2002). Other countries that are implementing RBF for drinking water supply are India, China, and South Korea (Ray 2008; Sandhu et al. 2011).

Currently, the construction of the 100,000 m³ reservoir and reinstatement of the Aguablanca wells are under execution. Therefore, for the alternatives that involve the use of the reservoir and/or the Aguablanca wells, the investment costs of these facilities were included. In addition, O&M and sludge treatment and disposal costs were considered in the analysis. A few of the alternatives described in this study comply with the water demanded by the city (in terms of water quantity). Only three alternatives can offer the total flow of demanded water:

• (A1) Timba basin;

• (A2) Salvajina dam, delivering water to a WTP to be constructed in Pance; and,

• (A3) La Balsa, delivering water to a WTP to be constructed in Pance.

In addition, a smart combination of specific alternatives can also comply with the water demand. Five combinations of alternatives–so-called combos–were defined for the remaining alternatives:

• (C1): A7 + A8 + A10;

• (C2): A6 + A8 + A11;

• (C3): A8 + A9 + A11;

• (C4): A4 + A6 + A8 + A9; and

• (C5): A5 + A6 + A8 + A9

In Table 4 the evaluation results are shown. The higher the net outranking flow (Φ), the better the alternative (Brans & Mareschal 1994). The table shows that combo C3 seems to be the most adequate solution to solve the drinking water supply for Cali with a Φ of 0.55. This combo is followed by combo C2, which obtained a Φ of 0.34, which represents a good option for the city. Salvajina dam and La Balsa (A2 and A3) are also good options, giving 0.05 and 0.01 Φ, respectively. However, these alternatives imply an important modification would need to be undertaken to the current distribution network. In addition, extracting water from La Balsa point would require pumping raw water up approximately 40 m to the WTP, from where it would be distributed by gravity. These restrictions are represented in the low Φ values obtained.

The remaining alternatives/combos obtained a negative Φ, indicating a low viability of the alternatives/combos to solve the drinking water problems in Cali. The combos C1, C4 and C5 obtained the lower Φ values (–0.1, –0.43 and –0.40, respectively. These combos would necessitate the construction of multiple dams. The alternatives that involve the construction of dams are not well perceived by the communities and would require much time to obtain environmental and construction licenses, thereby reducing their potential as a timely source for drinking water. This is observed in the high Φ⁻ values obtained by these alternatives, and therefore in the low values observed for Φ (Table 4). In the case of the Timba basin option (alternative 1), the acquisition of large areas of land and negotiations for large pipeline trajectories until the new WTP is built would be required. The options linked to the 100,000 m³ reservoir may be susceptible to drought-flood-contamination events.

By not considering legal aspects and social acceptance criteria in the assessment, the following results can be highlighted: RBF with deep wells (C3) achieved a 0.67 Φ, with a 0.74 Φ⁺ value, and the Cali River basin with RBF (C2) achieved a 0.34 Φ, with a 0.57 Φ⁺ value. This indicates that the use of RBF continues to be the best alternative, leaving out the social and legal aspects. A1 (Timba basin) obtained a positive Φ value (0.10), significantly improving its position (from 5th to 3rd). However, this alternative has significant potential for legal and social constraints, which could impede its selection as a feasible solution.

The desktop study considered factors such as potential production capacity, potential water quality improvements, costs, environmental impacts, O&M considerations and regulatory considerations. Both the quality and quantity aspects of implementing RBF in Cali showed promising results, but it must be studied further to determine its technical feasibility. In addition, the high concentration of sediments transported by the Cauca River could lead to an important reduction in its water yield.

The drinking water supply system of Cali is vulnerable due to its dependence on WTPs abstracting water from the Cauca River. Therefore, alternatives are being considered and evaluated with an MCA. RBF, in combination with the deep wells, and RBF with the Cali River basin options seem to be ideal alternatives in order to guarantee safe drinking water in the amounts demanded by the city, while also having a lower environmental impact during construction and operation.

MCA has been shown to be a suitable tool for the analysis of problems where political, social and stakeholders’ interests are involved. However, it is subject to individual preferences that can alter the results depending on what the developers favor. Selection of a broad range of criteria, in addition to the weighting process, plays an essential role, since it allows compensating subjective decisions and guidelines proposed by institutions. The participation of experts and stakeholders in this process provided a comprehensive conceptualization of the problem, facilitating the construction of an integrated multiple-criteria framework to assess the selection of the alternatives. This analysis provided transparency for the decision-making process and engaged stakeholders in the search for solutions that best suit the addressed needs.

The authors would like to acknowledge the support of Colciencias for the scholarship of Juan Pablo Gutiérrez Marín. Special thanks to Eng. PhD. Juan Carlos Escobar (Chief of the Drinking Water Production Department, WTP Puerto Mallarino, EMCALI EICE-ESP), Eng. Hester van der Waa, Eng. Annemarij de Groot and Eng. Andrew Warren who supported the completion of this article with important information.