Monterrey, Mexico, Case Study

Case Study Contents

A surburban street in Cerro de la Silla, Monterrey, Mexico. Photo by Ana / Adobe Stock

Photo by Ana / Adobe Stock

In this case study, we describe how Robust Decision Making (RDM) was used to develop a robust water management strategy for Monterrey, Mexico. In particular, we describe a new extension of RDM that combines optimization with scenario discovery to develop a robust strategy that

  1. identifies no-regret actions that Monterrey can undertake in the near term
  2. defines pathways to implementing future actions that have, for now, been deferred
  3. describes signposts for future conditions that would help stakeholders choose among these.

This effort has had a significant impact on Monterrey’s water management, moving the region away from a high-cost, risky transfer project toward integrated water resource management over the coming decades.


Planners in growing metropolitan regions in Latin America and across the developing world face many challenges in developing water resources management strategies that will support current and future needs. These issues are particularly relevant in Monterrey, Mexico, the country’s third-largest metropolitan area (see Figure 1). Located in Mexico’s northern state of Nuevo León, the Monterrey Metropolitan Area (MMA) is the economic capital of northern Mexico and an important player in the economic industrial cluster across Mexico’s border with the United States. But Monterrey’s economic success has also led to increased environmental and social challenges, including security and future equity issues related to water (Sisto et al., 2016), which could affect new investment, human capital development, and social welfare in the region (Castro, 2004).

Figure 1. Monterrey, Mexico, Study Area

SOURCE: Molina-Perez et al., 2019, p. 2 from Sisto et al., 2016.

In particular, water security has come to the forefront in public discussions of regional challenges (Bennett, 1995; Sisto et al., 2016). Monterrey’s current system relies on the La Boca, El Cuchillo, and Cerro Prieto surface reservoirs, which provide 70 percent of the current supply. The remaining 30 percent of the supply is provided by groundwater systems, including the MMA, Mina, Buenos Aires, and Santiago well fields.

Since 2010, the Nuevo León government has been exploring the possibility of expanding Monterrey’s water infrastructure to support both population and industrial growth. In particular, the development of the Pánuco Aqueduct project, a 370 km–long water conveyance facility from Veracruz, a less developed state in the south of Mexico, was proposed and promoted by the previous administration (2009–2015). This project was the source of significant controversy. Partly in response to this controversy, the water policy community of Monterrey decided to develop the region’s first long-term water plan (the Monterrey Water Plan [MWP]) in 2016. To support this effort, the Fondo de Agua Metropolitano de Monterrey (FAMM) sponsored a study, led by RAND, that used RDM to create a full analytical framework to support water infrastructure master planning for Monterrey. The study evaluated the vulnerabilities of Monterrey’s water management system to future climate and technological change and demand uncertainty to develop an adaptive strategy designed to minimize vulnerabilities at an acceptable cost.

This case study focuses on how RDM can help researchers develop a robust, adaptive strategy—the final output of a complete RDM analysis. We describe an approach that uses optimization to identify no-regrets, near-term policies and then define adaptive pathways and signposts to guide future policy changes using scenario discovery (see Figure 2).

Figure 2. Robust Decision Making Steps Used in Monterrey, Mexico, Case Study

  1. Decision Framing
  2. Evaluate Strategies Across Futures
  3. Vulnerability Analysis
  4. Tradeoff Analysis
  5. New Futures and Strategies

RDM Steps in This Case Study

  • Decision framing: The researchers held three meetings with a variety of stakeholders to discuss the study’s purpose; identify the key uncertainties, potential water management options, and performance metrics; and discuss data and modeling needs.
  • Evaluate strategies across futures: An integrated assessment model that projects future demand and supply and identifies optimal investment pathways under different uncertain futures was developed and used to evaluate the performance of the Monterrey water system under uncertain climate and demand futures.
  • Vulnerability analysis: The simulation outcomes were analyzed, revealing significant vulnerabilities under all future conditions. This informed the development of new adaptive strategies as part of a process combining the trade-off analysis and new futures and strategies steps.
  • Trade-off analysis: In this study, stakeholders compared cost and reliability trade-offs among different near-term portfolios, one of which formed the basis of an adaptive strategy.
  • New futures and strategies: The study team developed new adaptive strategies by defining near-term portfolios, optimizing for future conditions, and defining triggers for new investments.
  • Robust strategies: The study team defined a framework for planners to select a single robust, adaptive strategy comprising near-term investments, conditions to monitor, thresholds, and deferred investments.

RDM Step: Decision Framing

A series of three workshops, held at Tecnológico de Monterrey, provided an opportunity for a diverse group of stakeholders to discuss what the key aspects of the analysis should be. Participants included state and federal agencies, private companies, nongovernmental organizations, academic institutions, and the public water utility company in Nuevo León. In addition to these workshops, the study team held individual meetings with interested stakeholders in Nuevo León to further expand and refine the scope. The scope resulting from the stakeholder engagements is summarized in Table 1 using an XLRM matrix.

The study team also briefed the stakeholders and FAMM’s technical committee on key findings throughout the process, including the fundamental trade-offs and the performance of the robust, adaptive plans identified. Their feedback led to the selection of a final robust, adaptive strategy for the long term, which is described in the MWP.

Table 1. XLRM Matrix for Monterrey, Mexico, Case Study

(X) Uncertainties (L) Management Options and Strategies
  • Water demand
  • Surface supplies
  • Groundwater availability
  • Desalination costs
  • Infrastructure projects
    • Surface
    • Groundwater
    • Conjunctive management
    • Desalination
  • Water-demand management
  • New water tarrifsa
(R) Relationships or Systems Model (M) Performance Metricsa
  • Integrated assessment model
    • RiverWare model
    • Water-demand econometric model
    • Hydrological models
    • Portfolio optimization model
  • Investment costs
  • Operation costs
  • Reliability

a For brevity, this case study does not describe an analysis of new water tariffs, as was done in the RAND study.

Uncertainties (X)

Water Demand

Future water demand in Monterrey is deeply uncertain because the main drivers of demand are uncertain. Drivers include economic and demographic trends, technological and behavior change, and climate conditions. For this study, the team developed a model of demand that considered water tariffs, sociodemographic characteristics, and climatic and regional conditions. It was calibrated using a comprehensive data set describing water consumption at the household level. This data set was made up of the full set of water utility company consumer-demand records from January 2012 to December 2015.

Surface Supplies

Future water supply from river inflows into Monterrey is also deeply uncertain because of potential impacts from climate change, which could cause conditions to be drier or wetter than historical conditions. Furthermore, the hydrological dynamics of the water basins and rivers on which the city’s supply depends are not completely understood or represented in available models.

To address this challenge, the study team developed 54 different hydrological projections that reflect different variations in streamflow across six sub-basins (i.e., discrete bins with variations from –20- to +20-percent change) and different timing of historical variability, which essentially cycles the 1992 to 2004 historical dry period through the future simulation period.

Groundwater Availability

Monterrey currently derives 30 percent of its supply from groundwater, and several future water management options under consideration rely on groundwater resources. However, stakeholders noted that little is known about the geological composition, storage capacities, or hydrological dynamics of the aquifers from which different groundwater alternatives would be developed. To address the uncertainty in the potential availability of Monterrey’s groundwater resources, the study team considered a wide range of yields, between 70 percent and 130 percent of the current estimate, which is in line with estimated recommendations made by local groundwater experts.

Desalination Costs

Given Monterrey's proximity to the Gulf of Mexico, desalination of seawater is an option for expanding Monterrey’s water supply. Desalination is generally more expensive than conventional surface or groundwater supplies, owing to high energy costs, but there are circumstances in which the cost and reliability attributes can be favorable as part of a portfolio of water management options. Furthermore, as costs decline with growing technological efficiency, desalination projects become increasingly favorable. Thus, trends in desalination costs—e.g., whether costs continue to decline or whether costs plateau—have immense strategic importance for long-term water planning in Monterrey.

The study team sampled uniformly across a range of desalination costs with the most-optimistic cost-reduction scenarios assuming that the desalination project currently under consideration will be 30 percent less expensive by 2027, while the most-pessimistic scenarios assume that this project will be up to 5 percent more expensive in the future.

Management Options and Strategies (L)

Many different options were discussed and are reflected in the analysis. In addition to the controversial Pánuco Aqueduct option, other more-local options were evaluated, including some new surface reservoirs, desalination, groundwater projects, and efficiency. Some options mentioned in the workshops, such as improving filtration of groundwater reservoirs, were not evaluated, because of insufficient information and tools to estimate the potential costs and effects on groundwater supplies. Figure 3 shows the projects and estimated costs.

Figure 3: Projects Included in the Study

NOTE: Length of bar indicates total investment cost, and width of bar is proportional to average project yield. cms = cubic meters per second.

Performance Metrics (M)

The study team identified two key performance measures: reliability of meeting demands and costs of new investments. Reliability, which was estimated as the frequency of monthly supply meeting monthly demand, was reported in the near term (2016–2026), intermediate term (2027–2038), and long term (2039–2050). Costs were calculated by adding the up-front, or capital, cost of an option to the first 15 years of annual operational costs brought to net present value using a 5-percent discount rate. The two cost components were brought to present value because stakeholders in Monterrey were interested in securing funds for building new water infrastructure and supporting its optimal performance.

Relationships or Systems Model (R)

To estimate how the Monterrey water management system would perform in the future with and without an augmented water management strategy, the team developed an integrated assessment model that combines several models within an optimization framework to define optimal water management adaptation strategies. Two models are used to estimate demand and hydrology, and these estimates are combined with information about the water management options and groundwater and desalination assumptions. The water demand and supply balance over time is estimated using a water management model. An optimization routine identifies the adaptation strategy that minimizes investment costs while meeting the reliability performance threshold set by stakeholders. This process is done for each plausible future, as shown in Figure 4.

Figure 4. Integrated Assessment Model

SOURCE: Adapted from Molia-Perez et al., 2019, p. 26.

This modeling framework estimated the optimal management strategy (or sequence of investments) for three decision periods for each specific combination of future demand, hydrological conditions (i.e., surface-water flows), groundwater-availability assumptions, and desalination costs over time. In each future instance (i.e., combination of conditions), the models derived the project portfolio that would be optimal for conditions occurring during the first period; then, based on this decision, the models estimated the optimal expansion of Monterrey’s water infrastructure for the second decision period and then for the third decision period. RDM methods were used to identify the most robust strategy, according to stakeholders’ preferences.

RDM Step: Evaluate Strategies Across Futures

To evaluate strategies across futures, the study team defined a set of plausible futures for the current water management system that reflects the uncertainties described earlier. This experimental design was constructed by first combining each of the 12 water-demand projections with the 54 hydrological projections for a full factorial design of 648 futures. Next, for each of these 648 futures, random values of groundwater assumptions and desalination costs were combined (ranging between 70 percent and 130 percent of the current estimate) to yield a 648-element design that varies demand, hydrology, and groundwater assumptions. These futures were used to represent uncertainty in the following steps of the RDM analysis.1

RDM Step: Vulnerability Analysis

The study team first evaluated the current management strategy across the 648 futures. Next, each simulation was classified in terms of its reliability—cases with reliability below 97 percent were identified as unacceptable. Visualization 1 shows these outcomes across the three uncertainties—water demand, groundwater availability, and surface water inflows—and time. These results show not only that all cases are vulnerable but also that reliability gets very low in future years as demand increases.

Visualization 1. Reliability of Current Management System Across Uncertainty Dimensions and Time

NOTE: GW = groundwater.

Because all of the futures evaluated require augmentation to the current plan, specific decision-relevant scenarios or vulnerabilities were not defined. Instead, the analysis moved straight to the trade-off analysis and new futures and strategies steps to identify which options would best improve performance across all futures.

RDM Steps: Identify New Strategies by Evaluating Trade-Offs of Optional Portfolios

The gold-standard outcome of an RDM study is a robust, adaptive strategy. Such a strategy includes decisions or choices that would be implemented in the near term based on stakeholder deliberations and then defines adaptive pathways that trigger additional choices as the uncertain future unfolds. To identify robust adaptive strategies for Monterrey, the study team focused on determining which of the 15 options (noted in Figure 3) would be required in most of the futures and thus be good candidates for inclusion. Next, individual sequences (or pathways) of additional projects were defined, along with the conditions that would trigger the additional projects. Together, the near-term projects and project pathways describe an actionable, robust strategy.

Optimal Near-Term Portfolios for Each Future

For each of the futures evaluated, an optimal combination of individual options—or portfolios—was identified by iteratively comparing different combinations of options using an optimization routine and selecting the portfolio that would increase reliability to the 97-percent threshold at the least cost. Visualization 2 shows the near-term vulnerability map, but, in this case, it indicates whether the selected option is included in the optimal near-term portfolio for the selected project. For example, one can see that the controversial Pánuco Aqueduct is included in very few plausible futures and only under very high future demand. Other options, such as Monterrey Country Groundwater, are included in almost all cases. The user can hover over the symbols to see the other projects that are included in the optimal, near-term portfolio for the specific future.

Visualization 2. Near-Term Inclusion of Selected Project Across Uncertainty Dimensions

Low-Regret Near-Term Portfolios

Ideally, there would be a portfolio that would be identified as best in the first period across all of the futures shown. However, depending on how the future unfolds, different near-term options would be optimal. Therefore, the study team analyzed the level of correlation across all projects included in the near-term optimal portfolios and identified nine different combinations of projects that were implemented together across many of the plausible futures. These are called the near-term portfolios.

Figure 5 shows which projects are included in each of the nine near-term portfolios. No near-term portfolio includes the two largest and most-expensive projects: desalination in Matamoros and the Pánuco Aqueduct. Instead, all but one portfolio include multiple smaller projects. The exception is P6, which includes the large Cuchillo II Dam project and excludes all other projects.

Figure 5. Projects Included in the Near-Term Portfolios

NOTE: Columns indicate individual low-regret portfolios (P1–P9), and rows indicate the individual projects. Shaded squares indicate which projects are included in each portfolio. Unshaded circles indicate projects that are not included. The symbols are sized by project yield. WW = wastewater.

The identified portfolios represent different ways to improve the system in the near term to prepare for future conditions. As expected, some portfolios will perform better in some futures than in others. To represent these trade-offs, the study team used the concept of regret to express the difference in terms of costs and reliability in the near term for a selected portfolio and the best portfolio for a given future. Regret is often used in RDM studies when comparing different strategies to represent the trade-offs in terms of what could be affected by the different strategies. Visualization 3 shows the cost and reliability regret for the user-selected portfolio across the same uncertainty dimensions shown in Visualization 1. The coloring indicates the reliability and cost regret across the futures. For example, compare the regret results for P4, which is one of the more-expensive near-term portfolios, with those of P7, one of the least-expensive portfolios. P4 shows much lower reliability regret, but it also shows high cost regret for the lower-demand futures, which implies that on many futures P4 overinvests in new capacity. On the other hand, P7 shows low cost regret but high reliability regret for the higher-demand and lower-inflow futures, which indicates that on many futures P7 underinvests in new capacity.

Visualization 3. Reliability and Cost Regret of Portfolios Across Uncertainty Dimensions

Adaptive, Long-Term Strategies

For each of the nine near-term portfolios, the study team developed an adaptive strategy consisting of a set of decision points, based on how conditions could evolve over time, and corresponding near- and long-term options. They used an algorithm called C5.0 to identify the future conditions that would require additional options and created a decision tree for each near-term portfolio. Figure 6 illustrates roughly how the C5.0 classification process works. The algorithm first identifies the optimal portfolio for the different combinations of the uncertainties (see the left side of Figure 6). It then classifies different regions of the uncertainty space according to the optimal portfolios, using splits. It uses the information about the splits to define a decision tree. For example, the common near-term options are implemented now. Then, if uncertainty A is greater than split 1 and uncertainty B is less than split 2, then additional projects are selected according to portfolio X (or the green squares in Figure 6). More information can be found in the study report.

Figure 6. The C5.0 Classification Process

SOURCE: Molina-Perez, et al. 2019, pp. 52, 53.

To aid Monterrey water planners in selecting a portfolio, the team summarized cost regret and reliability regret for the nine near-term portfolios on a single graph (see Visualization 4). Each dashed line traces out the cost and reliability regret trade-off across all of the portfolios for a specific aggregation of future results. For example, the blue-green line and blue-green shapes indicate the mean cost and reliability regret, whereas the red line and red shapes indicate the regret value below which 90 percent of the cases result—or the 90th-percentile result. Click on the shape or color legend items to highlight specific results, and hover over any point to see the near-term projects in the portfolio. Visualization 4 shows how portfolios P4 and P7 represent the extremes of the reliability-cost trade-off.

Visualization 4. Cost and Reliability Regret for the Nine Near-Term Portfolios

NOTE: USD = U.S. dollars.

For each near-term portfolio, a unique decision tree is defined that specifies the conditions that should trigger additional investments. Visualization 5 shows a decision tree for the user-specified near-term portfolio—initially set to P1. On the far left is the near-term node (circle). Hover over the node to see the list of near-term projects, totaling $387 million. To the right of the figure is a set of signposts. Hover over the symbols to see the restriction, or click on the signpost shape in the legend to highlight all of the signposts. For P1, the initial signposts specify different demand levels. If demand is less than 14.6 cubic meters per second (cms), then the strategy follows the top branch. If demand is greater than 17.7 cms, then the strategy follows the lowest branch. Other demand amounts lead to the two intermediate branches. The second and fourth branches also include splits based on groundwater availability. After the signposts, the intermediate-term contingent projects are defined. Hover over the gray squares to see the additional projects, if any, that are indicated for implementation, along with their costs. The top branch includes no additional projects, whereas the lower branches, which correspond to higher demand levels, all include additional projects. There is one more set of signposts before the final contingent projects are defined at the end node. The sizes of the nodes are proportional to the size of cumulative investments, and the colors of the lines indicate the average reliability for the corresponding futures.

By changing the initial portfolio from P1 to another portfolio, a different tree is displayed. Notice that for P4, for example, the final set of projects ranges from ten small projects, costing $583 million, to all 15 projects, including the Pánuco Aqueduct, costing $4.6 billion. The Pánuco Aqueduct project is included only (1) if demand exceeds 19.6 cms by 2026 (lowest branch) or (2) if demand is between 18.6 cms and 19.6 cms in 2026 and groundwater availability is less than 90 percent of expected amounts by 2026 (see the branch that is second from the bottom).

Visualization 5. Interactive Representation of Selected Adaptive Strategy

The analysis lastly evaluated the ultimate reliability in the long term across the uncertainties for the different adaptive strategies. Visualization 6 shows the vulnerability map for the user-selected strategy. As can be seen, there are important near-term trade-offs in reliability across futures. For example, for P1, reliability will remain low for many high water demand cases until the intermediate term, when additional projects are implemented. However, all of the strategies achieve a fairly high level of reliability across all of the futures by the end of the period. Hover over the symbols to see the actual project implementation and cost for the given future.

Visualization 6. Reliability of Selected Adaptive Strategy Across Uncertainty Dimensions and Time

Robust, Adaptive Strategy for Monterrey

Each portfolio defines an adaptive strategy—a set of near-term projects that Monterrey would implement as soon as possible and a set of signposts to guide the implementation of additional projects in the intermediate term and long term if conditions warrant them. Monterrey water planners and stakeholders weighed the cost and reliability results, along with other concerns, for each of the portfolios. After deliberations, they indicated a preference for portfolio P2.

Specifically, the P2 strategy involves the current system being augmented using a broadly diversified set of projects, including new surface-water sources, groundwater sources, and increased efficiency in the network. By deferring decisions on the larger, more-expensive options, such as the Pánuco Aqueduct, costs are relatively low in the near term—U.S. $439 million.

In the following periods, the need for additional projects would be based on future conditions. If demand growth in the 2027–2038 period proves minimal (i.e., below 14.6 cms), then there would be no need to further expand Monterrey’s water infrastructure. If Monterrey’s water demand grows only modestly—less than 15.76 cms in the intermediate term—then the desalination plant would be needed. Intermediate demand trends and future groundwater availability would call for different additional projects. But the large, controversial Pánuco Aqueduct would be needed only if (1) demand exceeds 19.6 cms by 2026 (lowest branch), (2) if demand exceeds 21.0 by 2039 (sixth branch down from the top), or (3) if demand is between 18.6 cms and 19.6 cms in 2026 and groundwater availability is less than 83 percent of expected amounts by 2038 (see the branch that is third from the bottom).


Like in many still-developing regions, where water demands are fast outstripping supply, a focus has been on increasing supply as a presumed sure-shot way of meeting users’ needs. In Monterrey, the predominant policy view was to expand current supply by importing water from another basin. However, like in many regions, the basin had several other options related to managing operations, creating new local supplies, and managing demands. Until recently, these actions have been overlooked in part because it was hard to quantify and trade off their costs and benefits and to incorporate them into a coherent strategy. This case study showed how advanced RDM techniques created a robust, adaptive strategy for Monterrey that takes full advantage of the options available in the basin. In particular, it analyzed a slate of broadly diversified alternatives and

  1. identified no-regret, near-term actions that kept costs low
  2. identified different contingency plans for the intermediate and long terms
  3. identified future conditions (i.e., signposts) that would signal the selection of one contingency action over another.

The result is that Monterrey has avoided a high-cost and risky basin-transfer project in favor of a lower-cost, no-regret strategy that will help it be more responsive to the future. In sum, water managers in Monterrey can be more assured that they are prepared for the future, even though they might not know exactly what that future will be.

Case study authors: David Groves and Edmundo Molina-Perez

Acknowledgments: The case study authors would like to thank the original funder of the RAND study, Fondo de Agua Metropolitano de Monterrey (FAMM), and the additional authors of the source study—Steven W. Popper, Aldo I. Ramirez, and Rodrigo Crespo-Elizondo.


  1. In Molina-Perez et al., 2019, a simpler experimental design was used for the vulnerability analysis. We focus on the second design for simplicity. (Return to text)