Metropolitan Water District of Southern California Case Study

Case Study Contents

The Santa Ana River in Anaheim in Orange County, photo by Derek Neumann / Getty Images

Photo by Derek Neumann / Getty Images

The Metropolitan Water District of Southern California (Metropolitan), a water wholesaler, produces an Integrated Water Resources Plan (IRP) about every five years that lays out a strategy for meeting future water demands through imported supplies and local sources and conservation (see Figure 1). Although the IRPs have always accounted for hydrologic variability, over the past ten years, Metropolitan has worked to account for how uncertainties beyond historical hydrologic variation could affect its investment needs. Beginning with the 2010 IRP [PDF], Metropolitan has used methods for Decisionmaking Under Deep Uncertainty (DMDU) to stress-test its long-term plan for a broader range of uncertainties, including climate change, and inform an adaptive management strategy.

This case study describes an unpublished RAND Corporation study, funded by Metropolitan and completed in early 2020, that used Robust Decision Making (RDM) to evaluate the robustness of Metropolitan’s 2015 IRP [PDF] to a wide range of uncertain future trends and conditions. The study team used these findings to describe a nascent monitoring approach for Metropolitan to inform adaptation of the IRP to future conditions.

This case study provides the reader the opportunity to review the steps of an advanced RDM-style vulnerability analysis and explore a monitoring framework that can guide the adaptation of a long-term robust water management strategy.

Figure 1. Metropolitan Water District of Southern California Service Area

Map of the Metropoliltan Water District of Southern California showing member agencies.

SOURCE: Courtesy Metropolitan Water District of Southern California. Used with permission.


Metropolitan is a water wholesaler that provides water to 26 Southern California water agencies that together serve about 19 million customers. Metropolitan imports water from Northern California via the State Water Project (SWP) and from the Colorado River via the Colorado River Aqueduct (CRA). The City of Los Angeles, which is within the Metropolitan service area, additionally imports water from the Owens Valley in Eastern California via the Los Angeles Aqueduct (LAA). Metropolitan also supports the development of local resources and conservation throughout the region.

Metropolitan’s 2015 IRP presents estimates of future retail water demands by its member agencies based on demographic and economic projections, historical local climate conditions, and expected levels of conservation. It then combines these retail demands with estimates of local supplies (reflecting historical climate variability) to determine the demand over time for water from Metropolitan.

The IRP defines an adaptive management strategy to guide the development of new supplies and conservation to meet its targets. Such targets include (1) support and financial resources for new infrastructure in the Bay-Delta known at the time as the California WaterFix, (2) maintenance of supplies from the Colorado River in the face of significant regional challenges (see the Colorado River Basin Case Study), (3) additional conservation savings, and (4) development and protection of local water supplies. The strategy also describes a set of conceptual Future Supply Actions, which includes actions related to public outreach, legislation and regulation, technical studies and support, and land and resources acquisition.

Following publication of the 2015 IRP, Metropolitan asked RAND researchers to use RDM methods to evaluate the IRP’s vulnerabilities to a broader range of uncertainties, including climate change, demographic changes, and the implementation of improvements to the Bay-Delta system. Figure 2 highlights the key RDM steps from this case study. The RAND study did not formally develop alternative strategies to compare, but instead described a monitoring approach for the IRP’s currently defined investments, which is a key component of an adaptive, robust plan.

Figure 2. Robust Decision Making Steps Used in the Southern California 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 Metropolitan RDM analysis was an internal study, and the decision framing step was performed by a small group of water planners and RAND researchers. The study was designed to evaluate the IRP under a broader set of uncertain futures and perform a vulnerability analysis to identify signposts for adaptation.
  • Evaluate strategies across futures: The study team expanded Metropolitan’s modeling framework to account for additional uncertainties and evaluate the IRP across hundreds of futures. These results were summarized and presented using interactive visualizations.
  • Vulnerability analysis: An RDM-style vulnerability analysis was used to analyze how well the IRP would perform across the results of the hundreds of simulations modeled in the previous step. This process identified three simple decision-relevant scenarios (i.e., descriptions of future conditions that would require significant management changes, investments, or both).
  • New futures and strategies: This study proposes a framework for monitoring future conditions to provide indications for how the IRP might need to address pending vulnerabilities.

RDM Step: Decision Framing

The RAND RDM study convened Metropolitan staff and water managers from outside agencies to develop the scope of the analysis.1 Using the results of this convening and subsequent discussion, the study team framed the analysis as summarized in Table 1.

Table 1. XLRM Matrix for Metropolitan Water District of Southern California Case Study

(X) Uncertainties (L) Management Options and Strategies
  • Demographics and water use rates
  • Climate conditions
  • Groundwater yields
  • Bay-Delta operating conditions
  • Implementation of Bay-Delta conveyance improvements
  • 2015 IRP (including 20 thousand acre-feet [TAF] of Future Supply Actions)
(R) Relationships or Systems Model (M) Performance Metrics
  • Augmented IRP Simulation Model (IRPsim)
  • Net balance (negative implies shortages)

The following subsections describe the key uncertainties, performance metrics, and modeling framework used in the RDM study. Unlike the RDM analyses described in the other case studies, the Metropolitan RDM study evaluated only the current strategy described in the 2015 IRP.

Uncertainties (X)

There are various longer-term trends and factors that could affect Southern California water management needs in the future. The Metropolitan RDM study focused on uncertainty about future demographics that affect demand, climate conditions that affect demand and supplies, groundwater yields, Bay-Delta operating conditions, and the implementation of Bay-Delta conveyance improvements.

Demographics and Water-Use Rates

Changes in Southern California housing patterns and water-use rates will affect how much water is needed in the Metropolitan service area. According to the IRP, the population served by Metropolitan is expected to grow from around 1.9 million people in 2018 to between 22 million people and 23 million people by 2050, but this projection is uncertain. How much water each person uses is also highly uncertain. Per capita water use has declined since 2008 in response to more-efficient new housing, replacement of old water-using devices with more-efficient models, replacement of landscapes, and behavioral response to past droughts. But it is difficult to predict how this trend will evolve going forward. For Metropolitan, assuming high population growth and reversal of declining water use would suggest significantly more water demand in Southern California. Preparing for this future yet experiencing one in which population growth slows and per capita demands continue to decline could lead to significant overinvestments. Conversely, preparing for too little demand would lead to costly future shortages.

The Metropolitan RDM study considered the IRP projections of households and per-household demand along with three other projections developed by Metropolitan for the RDM study—Balanced Growth, High Growth, and Periurban. Visualization 1 shows how total occupied housing and per-household demand combine to yield estimates of water demand. The top figure shows how per-household demand in acre-feet per year (vertical axis) and total occupied housing units (horizontal axis) have varied from 1990 to 2017 (brown line) and Metropolitan’s projections going forward from 2015 to 2050 (blue, green, orange, and purple lines). The lower figure shows the corresponding demand projection over time. This visualization shows that all four projections anticipate significant increases in the number of housing units and declining per-household demand, and these changes all translate into increasing demand. Mouse over the different projected demographic and water use conditions (top) to see how they translate into demand (below).

Visualization 1. Water Demand and Drivers for Historical Period and Four Future Projections

NOTE: AF = acre-feet.

Climate Conditions

Climatic conditions, including temperature and precipitation patterns, affect demand for water in Metropolitan’s service area and supplies from the various source regions. Climatic changes, however, are not predictable. Global climate models (GCMs), for example, project a wide range of temperature and precipitation trends across Metropolitan’s source basins (Upper Colorado River, Northern California, Eastern Sierras, and Southern California). To illustrate this point, Visualization 2 shows projected trends in temperature and precipitation across the different basins that supply Metropolitan. Trends are based on the change in temperature and precipitation from a historical baseline period (1950–2000) to the decade between 2041 and 2050. You can adjust the global emissions scenario (or representative concentration pathway [RCP]) used to drive the GCMs and select different GCMs to see how temperature and precipitation could change through 2050 across the basins.

Visualization 2. Projected Trends in Temperature and Precipitation Across Source Basins

To summarize these changes for the purposes of the analysis, the RAND RDM study team calculated temperature and precipitation trends averaged over each basin for each GCM simulation. Visualization 3 shows the ten-year trends through the 2050s for the set of GCM runs available for each of the four source basins’ respective climate regions. From this visualization, one can see how uncertain the projections are within a single region and across the regions by cycling through the climate regions using the radio buttons in the upper right. Hovering over the symbols shows the climate trends over time.

Visualization 3. Climate Trends for the Selected Source Basins

Groundwater Yields

The IRP assumes constant Southern California groundwater availability for all years based on best estimates of the basins’ sustainable yield. For the RDM analysis, however, the study team considered how Southern California precipitation trends could affect groundwater sustainable yields. The team specified three levels of change—no change, 10-percent change by 2050, and 20-percent change by 2050—and assumed a negative change for cases in which the precipitation trend was negative and a positive change for cases in which the precipitation trend was positive.

Regulations on Bay-Delta Exports

Supplies from the SWP are further influenced by San Francisco Bay-Delta regulations that affect how and when water can be pumped from the Delta into the SWP aqueduct. Two flow conditions—existing conveyance high outflow (ECHO)2 and existing conveyance low outflow (ECLO)3—are evaluated.

Implementation of Bay-Delta Conveyance Improvements

Another major element of the IRP is the implementation of a Bay-Delta conveyance project that would replace Delta water diversions from the southern part of the Delta with a diversion point from the Sacramento River, north of the Delta. At the time of the 2015 IRP and the subsequent RDM study, different proposals for conveyance were under consideration and being evaluated as part of the California WaterFix, a project designed to increase the reliability of exports that was expected to be constructed and implemented by the early 2030s.

Recognizing the uncertainty in implementation of this expensive and contentious project, researchers evaluated futures in which the WaterFix would be implemented in the 2030s and futures in which it would be delayed past the time horizon of the study. The permit application for the WaterFix has since been withdrawn because of both political and scientific concerns, and another proposal is under development.

Performance Metrics (M)

The RDM study evaluates the performance of the IRP for each future year based on the estimated regional net water balance over a set of 91 different hydrologic conditions that reflect historical variability. A positive net balance implies that the region is able to use or store water available to it, and a negative net balance indicates that the region is unable to meet all customer demands (i.e., a shortage). The team considered shortages that would be greater than 10 TAF, 10 percent of the time, to be of concern.

Relationships or Systems Model (R)

To address climate and other uncertainties in Southern California and additional source regions in a consistent way, the RDM study team augmented Metropolitan’s standard planning model (IRPsim) with additional modules that translate deviations in temperature and precipitation for the source regions into changes in supplies from the CRA, SWP, and LAA and account for uncertainties about Bay-Delta conditions. Details about IRPsim can be found in the 2015 IRP Technical Appendix 2011 [PDF].

RDM Step: Evaluate Strategies Across Futures

The RDM study team used the expanded IRPsim to simulate and evaluate the performance of the Southern California water system under many different plausible futures. To do this, the team developed a set of 100 climate futures based on combinations of temperature and precipitation trends for the local service area, CRA watershed, SWP watershed, and LAA watershed from the GCMs described earlier. Visualization 4 shows the temperature and precipitation trends across the four basins for each of the 100 climate futures plus one future representing no change. By selecting one sample in either the temperature plot (top) or the precipitation plot (bottom), the user can see how a single future defines both temperature and precipitation trends. The graph also shows the strong correlation in temperature trends across the four basins and the less-strong correlation among precipitation trends across the same basins.

Visualization 4. Temperature and Precipitation Trends Across Three Source Basins for 101 Climate Futures

Each climate future, plus the historical climate condition, was then combined with the four demographic scenarios, three assumptions about groundwater sensitivity, two assumptions about Bay-Delta operations, and whether the California WaterFix is successfully implemented to define a set of integrated futures to use in the vulnerability analysis. This yielded a sample size of 4,848 futures (see Figure 3).

Figure 3. Experimental Design and Number of Futures

RDM Step: Vulnerability Analysis

To stress-test Metropolitan’s IRP, the RDM study team evaluated the IRP’s performance across the wide range of plausible futures described earlier and analyzed the results using RDM techniques.

The RDM study considers that the IRP does not achieve its objective of ensuring Southern California water reliability if shortages greater than 10 TAF occur more frequently than 10 percent of the time for a given year. The IRP is vulnerable to a future if such shortage conditions occur sooner than 2035, just after the time that all IRP investments would be complete. In these cases, the IRP would need to adapt to meet these more-difficult conditions.

The performance of the IRP in each of the plausible futures can be summarized by a robustness plot. In Visualization 5, the position of the symbol indicates the assumptions for a subset of the uncertainties. The Xs indicate vulnerable outcomes (i.e., cases in which shortages larger than 10 TAF would occur about 10 percent or more of the time by 2035). The Os indicate nonvulnerable outcomes. This visualization shows that there are specific regions (or combinations of uncertainties) that tend to lead to vulnerable outcomes. For example, most futures corresponding to the High Growth demographic future are vulnerable. For the other demographic futures, vulnerabilities also depend on climate and groundwater conditions. Changing to outcomes corresponding to No WaterFix in the upper-right corner shows that, without the WaterFix, the level of vulnerability is greatly increased.

Visualization 5. Robustness Plot Showing Performance of the IRP Across All Futures

Like in the Colorado River Basin Water Supply and Demand Study (CRBS), the Metropolitan RDM study team used scenario-discovery methods to analyze which uncertain input conditions lead to low reliability by 2035. They first focused on half of the cases: specifically, those in which the WaterFix is successfully implemented. The study team used the PRIM algorithm to evaluate the 565 cases that would lead to vulnerable outcomes of the 2,424 total cases (23 percent). In Visualization 5, click “Yes” on the “Show Vulnerability Scenarios” button to see the results that correspond to each of the three scenarios. Mouse over these results to see specific definitions.

This analysis identified three key IRP-relevant scenarios. The first is simply the condition in which the demographic assumptions are consistent with the High Growth future. In this scenario, 58 percent of cases are vulnerable. The nonvulnerable cases are those in which precipitation increases across all of the supply basins, compensating for the higher demands.

The second IRP-relevant scenario is the most interesting because it pertains to the more likely demographic conditions: the IRP Base Case and Periurban demographic futures (see Table 2). For these cases, the IRP is vulnerable if local service area precipitation trends are even slightly negative and temperatures rise more than about 2.5 degrees. (Precipitation trends in the other three basins—SWP, Upper Colorado River Basin, and LAA source regions—would all need to be significantly wetter to compensate.) In this scenario, 65 percent of the cases would lead to vulnerabilities. This scenario also covers more than half of the vulnerabilities not captured by the High Growth scenario.

Table 2. IRP-Relevant Scenario 2: IRP Base Case and Periurban Demographic Futures

Trends Vulnerability
Precipitation trends (through 2050)
  • Southern California: < –1.7 percent
  • LAA: < +15.2 percent
  • SWP: < +17.3 percent
  • Upper Colorado River Basin: < +12.3 percent
Temperature trends (by 2050)
  • Southern California: > +2.48 degrees Fahrenheit
  • LAA: any
  • SWP: any
  • Upper Colorado River Basin: any

The third IRP-relevant scenario covers those conditions in which demographic growth is low (Table 3). In these cases, the greater the groundwater sensitivity, the greater the vulnerability is. The climate conditions leading to vulnerability are particularly low precipitation for both the Metropolitan service area and the SWP watershed.

Table 3. IRP-Relevant Scenario 3: Balanced Growth Demographic Future

Trends Vulnerability
Precipitation trends (through 2050)
  • Southern California: < –22 percent
  • LAA: any
  • SWP: < –11.9 percent
  • Upper Colorado River Basin: any
Temperature trends (by 2050)
  • Southern California: any
  • LAA: any
  • SWP: any
  • Upper Colorado River Basin: any

The identified IRP-relevant scenarios—and particularly the second—strike a balance between coverage (the number of vulnerable cases described) and density (how much the decision-relevant scenario describes vulnerable as opposed to nonvulnerable cases). Analysts could arrive at a different balance between coverage and density based on preferences related to how strictly vulnerabilities should be defined.

To summarize, the vulnerability analysis determined that the Metropolitan IRP would meet its goals under its specified assumptions but would be vulnerable to the following four conditions:

  • high growth (IRP-relevant scenario 1)—conditions of high population growth and slower declines in per-household water use lead to vulnerable conditions across all future climate conditions
  • moderate demographic growth, with declining precipitation and warming trends (IRP-relevant scenario 2)—conditions of moderate population growth and continued declines in per-household water use rates would be vulnerable if declining trends in precipitation, coupled with warming, were experienced in Southern and Northern California, unless they were offset by much wetter trends in the other source basins
  • balanced demographic growth, with significantly declining precipitation and warming trends (IRP-relevant scenario 3), leads to vulnerabilities unless offset by much wetter trends in the other source basins.

The study team also evaluated outcomes for scenarios in which the WaterFix is not implemented. In such cases, the range of uncertainties in which the IRP would have low reliability for all three decision-relevant scenarios increases. This can be seen in Visualization 5 by selecting the “No WaterFix” option.

RDM Step: New Futures and Strategies—Developing a Monitoring Strategy

Although the Metropolitan RDM study did not specifically define an alternative strategy to the IRP, it used the results from the vulnerability analysis to describe a possible monitoring approach to guide its adaptive management process.

Specifically, the study team presented an approach for how Metropolitan could use insights from the vulnerability analysis to develop a framework for monitoring demographic and climate conditions to anticipate the conditions that would require IRP augmentation. We illustrate the framework by applying it to the second IRP-relevant vulnerability scenario (moderate demographic growth, with declining precipitation and warming trends). The framework incorporates three elements, which we describe in Table 4.

Table 4. Elements in the Framework for Monitoring Demographic and Climate Conditions Requiring IRP Augmentation

Monitoring Element Information for IRP
A. Observations of conditions that are uncertain over time Historical trends of temperature and precipitation for each water source basin (including Southern California)
B. Plausible projections of future conditions Projections of future temperature and precipitation trends used in this study to define plausible futures
C. Scenarios describing vulnerabilities For this illustration, we use the IRP-relevant scenario 2

Visualization 6 brings these elements together. The first tab shows historical precipitation (top) and temperature (bottom) data for the user-selected basin. A linear trend is fit through these data and projected through 2055. The second tab shows the same information but also includes the projected precipitation and temperature in 2055 based on the GCM projections. Hover over the symbols to see the specifics of each GCM projection. The third through sixth tabs indicate which 2055 conditions would contribute to a vulnerability for each of the four basins. They are preset to the thresholds defined for vulnerability 2 (see Table 2). By using the slider bars on the right, the user can adjust these vulnerabilities. These visualizations enable a comparison between the current trend derived from the historical data and indicated by the dotted line and the future trends that would contribute to a vulnerability. The user can also plot an additional ten years of hypothetical observations to see how future hypothetical data could change the calculated trends.

Visualization 6. Basic Framework for Monitoring Vulnerable Conditions for Southern California Water Management

By examining each of the basins in turn, one can see that the recent precipitation and temperature trends across all basins are consistent with vulnerability 2, which applies when demographics are consistent with the IRP Base Case or Periurban scenarios. This does not guarantee that the vulnerability will come to pass. For example, if the precipitation pattern from the 1990s in Southern California were repeated over the next ten years, the precipitation trends would be less consistent with vulnerability 2—indicating that the IRP would be less likely to need to adapt.

Vulnerability 3, however, considers demographic conditions that lead to lower water demand in Southern California (see Table 3). As an exercise, the user can adjust the thresholds to be consistent with this vulnerability, which applies to the Balanced Growth demographic projections. In this case, the precipitation trends in the SWP watershed are projected to remain out of the vulnerability range. Therefore, Metropolitan might conclude that a powerful hedge against the climate trends would be to help shape future demographic conditions so that water demand is lower than projected in the IRP. Figure 4, which is reproducible using Visualization 6, shows this result.

Figure 4. Precipitation Records, Current Trend, and Vulnerability 2 for the State Water Project Source Basin

Summary and Conclusions

In this case study, we describe a climate vulnerability analysis of Metropolitan’s IRP. Using RDM tools that are similar to those used for the CRBS, we define conditions under which the IRP would not meet Metropolitan goals and would need to be modified to adapt to those conditions. These conditions are the IRP-relevant scenarios. We call them decision-relevant because they are relevant for Metropolitan’s decisionmaking regarding the IRP. Second, we describe a framework for monitoring for those conditions and assessing whether such adaptations should be implemented. This information can help prepare Metropolitan for potential adjustments it will need to make and qualitatively indicates how likely those adjustments might be.

The analysis identified several key conditions to which Metropolitan would need to adapt. The monitoring framework shows that historical demographic, temperature, and precipitation trends from the past 40 years are consistent with one of these IRP-relevant scenarios. If these trends continue, Metropolitan will need to adjust its IRP in the coming decades. However, if demographic patterns lead to lower demands because of higher water use efficiency or slower increase in the number of households, then current climate trends might not trigger the need for adaptations.

The framework for monitoring decision-relevant conditions is a novel but natural extension of the traditional RDM process. It helps make strategies agile and responsive to the future as it unfolds. Such agility to changing conditions is an important way to inject robustness into a strategy.

The monitoring approach described in this case study echoes techniques used in other DMDU methods, such as Assumption-Based Planning (ABP). ABP is a qualitative method for assessing conditions under which a plan is vulnerable and identifying signposts that would signal that those conditions are indeed coming to pass.

The next case study on Monterrey, Mexico, builds on this work and takes it further by developing a fully adaptive, robust strategy. In particular, the case study uses optimization to identify no-regrets, near-term policies and defines adaptive pathways along with signposts to guide future policy changes.

Case study authors: David Groves and James Syme

Acknowledgments: The authors would like to thank Brandon Goshi, Jenny McCarthy, Mike Ti, and Jennifer Nevills of Metropolitan for data and information for this case study and for their helpful review and suggestions.


  1. Participants were David Groves, Robert Lempert, and David Catt of RAND; Brandon Goshi, Jennifer Nevills, Jenny McCarthy, Mike Ti, Carlos Carrillo, and David Sumi of Metropolitan; Laurna Kaatz of Denver Water; and Kavita Heyn of Portland Water. (Return to text)

  2. ECHO (Existing Conveyance High Outflow) and ECLO (Existing Conveyance Low Outflow) refer to two scenarios of future environmental regulations developed by the California Department of Water Resources that describe how much freshwater flow from the Sacramento and San Joaquin Rivers must pass through the Delta to the Pacific Ocean. High outflow reduces available freshwater for export to Southern California, whereas low outflow requirements support more water exports.(Return to text)

  3. ECHO (Existing Conveyance High Outflow) and ECLO (Existing Conveyance Low Outflow) refer to two scenarios of future environmental regulations developed by the California Department of Water Resources that describe how much freshwater flow from the Sacramento and San Joaquin Rivers must pass through the Delta to the Pacific Ocean. High outflow reduces available freshwater for export to Southern California, whereas low outflow requirements support more water exports.(Return to text)