Water Supply Planning in the Face of Drought and Ecosystem Flows: Examining the Impact of the Bay-Delta Plan on Bay Area Water Supply

Abstract

In California, recent Bay-Delta Plan legislation attempts to balance water supply and ecosystem protection by requiring 40% of the flow to remain in-stream in the Tuolumne River from February through June. Serious questions remain about what this means for the Bay Area water supply, especially during drought. Our work develops a new approach to analyze how in-stream flow policy coupled with climate change could impact regional water supply over the coming decades. Results show that the new in-stream flow demand would exceed urban water deliveries in a typical year. In wet years, water supply performance is minimally impacted, but in drought, the policy can lead to less water in storage, delayed reservoir recovery, and increased time at critically low storage. Storage impact exceeding 50 000 acre-feet (60 million m³) is anticipated with at least 18% frequency, demonstrating that, climate uncertainty notwithstanding, this impact must be planned for and managed to ensure a reliable future water supply.

Short abstract

Water-stressed communities are struggling to find sustainable solutions to meet their water needs without compromising ecosystem health. Our work advances planning efforts through a new analysis approach examining drought, policy, and climate change impacts to water supply performance.

Introduction

Water Supply in the Tuolumne River System

Throughout the San Francisco Bay Area, 2.7 million customers rely on the Hetch Hetchy Regional Water System (RWS) for water supply.¹ Approximately 85% of this supply comes from the Tuolumne River watershed.² The Tuolumne River also serves as critical habitat for migratory fish including threatened steelhead (Oncorhynchus mykiss) and fish species of special concern, the fall-run Chinook salmon (Oncorhynchus tshawytscha).^3,4 The State of California water code recognizes water supply and ecosystem protection as “co-equal goals”,⁵ yet in practice this has been a difficult and contentious balance to strike.

In 2018 in response to precipitously declining fish populations, the State Water Board adopted amendments to the Water Quality Control Plan for the San Francisco Bay/Sacramento-San Joaquin Delta Estuary (“Bay-Delta Plan”).⁶ The amended Bay-Delta Plan requires 40% of unimpaired flow (with an adaptive range of 30–50%) to be left in-stream for ecosystem use between February 1 and June 30 each year.⁷ Unimpaired flow, also called full natural flow, is the streamflow produced by a watershed, assuming no upstream diversions or storage. The Bay-Delta Plan’s unimpaired flow requirement applies to the Tuolumne River and to two other major tributaries of the San Joaquin River.

The Bay-Delta Plan’s 40% unimpaired flow requirement has been passed but not yet implemented, and there is substantial debate over its impact on water supply reliability, especially in dry years.^2,8,9 Currently, the Tuolumne River retains 21% of unimpaired flow during February through June of a median year and only 11% during the driest quartile of years.¹⁰ This means that significant diversion cutbacks will be required to increase the in-stream flow to the 40% benchmark required by the amended Bay-Delta Plan.

Implications of Water Supply Cutbacks

The San Francisco Public Utilities Commission (SFPUC), which owns and operates the RWS, anticipates that it will be responsible for 51.7% of the diversion cutbacks required by the amended Bay-Delta Plan.² This expectation is based on an existing agreement between San Francisco and the Turlock and Modesto Irrigation Districts¹¹ (“the Districts”), which would jointly contribute the remaining 48.3%. SFPUC argues that, in a severe drought, meeting the unimpaired flow requirement would result in large delivery cutbacks far exceeding the past precedent of 25% rationing mandated during the recent 2012–2016 drought.^2,12−14

Concern over such severe water supply impacts has prompted the SFPUC and the irrigation Districts to seek legal relief¹⁵ and to propose an alternative Tuolumne River Voluntary Agreement emphasizing nonflow interventions including habitat improvements and predation management.¹⁶ Meanwhile, however, SFPUC is grappling with the challenge of planning for water supply reliability under the new Bay-Delta Plan requirements.¹⁷

Beyond the policy impact, SFPUC must also contend with climate change effects. Rising temperature and evapotranspiration will place additional demand on freshwater systems, and increasingly variable precipitation will result in more frequent occurrence of extreme wet and extreme dry years.^18,19 As more precipitation falls as rain instead of snow, snowpack storage is lost, and peak streamflow shifts earlier in the year and further out of alignment with peak summer demand.²⁰

Limitations of Existing Planning Approach

In California, urban water suppliers assess reliability and outline their water shortage response in an Urban Water Management Plan.²¹ This plan features a reliability assessment comparing water supply and demand under three planning scenarios: a single normal year, a single dry year, and a 5-year drought period.²² The reliability assessment considers a future planning window of 20 years. Supply is represented using historical hydrology corresponding to each planning scenario. Future demand is estimated using past deliveries adjusted for anticipated land use, climate, policy, and other influences.

Using this reliability assessment methodology, SFPUC estimates that meeting Bay-Delta Plan ecological flow requirements in a multiyear drought would result in up to 49% reduction in annual water supply deliveries.^2,17 Meanwhile, the environmental nonprofit Tuolumne River Trust criticizes this conclusion, arguing from its own analysis that such a severe reduction is actually the consequence of management decisions that hold back too much water in storage.²³ This debate highlights important limitations in the planning and communication of urban water reliability, in this instance and in California more broadly. Reporting reliability as a single score communicates a straightforward message but can isolate this message from the role of assumptions and management decisions that influence water supply performance. Because a reduction in RWS deliveries results from proactive rationing of storage reserves rather than from the total depletion of storage, delivery reduction as a metric indicates how aggressively conservation is being implemented but provides less direct information about the system’s health or drought severity.

The effectiveness of the reliability assessment is limited not only by communication but also by certain embedded assumptions. Among these is the assumption of stationary hydrology, i.e., unchanging hydrologic properties, across the historical record and planning period, which is inconsistent with observed and projected climate change effects.^18,19,24 SFPUC attempts to incorporate climate effects by compiling an extended “design drought” that stitches together two severe historical droughts in sequence, representing a more extreme event than has occurred historically.¹⁷ However, even with this adjustment, the multiyear drought scenario is still analyzed outside the context of a continuous hydrologic record, which artificially isolates drought from surrounding wet years and implicitly assumes full reservoir recovery between drought events, which is not guaranteed. There is no indication of a return period for the design drought event; therefore, it is unclear how planning scenarios align with stakeholder preferences around risk and uncertainty. Without additional hydrologic context, the existing assessment approach fails to address important dimensions of reliability, including reliability under less extreme but more common drought scenarios, as well as the influence of management actions in nondrought years.

In another planning effort, SFPUC has partnered with the Water Research Foundation to produce the Long-Term Vulnerability Assessment and Adaptation Plan,²⁵ which models RWS performance to assess vulnerability to climate change and other stressors. The assessment offers an insightful analysis of how key performance metrics respond to stressors alone and jointly and at what point desired water supply performance targets may no longer be met. However, for in-stream flow requirements as a stressor, the assessment evaluates only one performance metric—reliability—which gives a limited picture of overall system health. Another limitation is that, despite finding the performance to be sensitive to demand, the assessment considers demand growth from a baseline of 227 MGD (859 000 m³/d), whereas actual deliveries from the past 5 years (FY 2018–2019 to FY 2022–2023) average less than 200 MGD (757 000 m³/d).²⁶⁻³⁰

Our work aims to overcome these planning limitations through a risk-based analysis focused on what the new Tuolumne River ecological flow requirement means for Bay Area water supply reliability. Prioritizing nonstationarity, uncertainty, and transparency, this approach enhances the planning conversation by evaluating multiple facets of water supply performance and communicating risk profiles rather than condensing the performance into a simple but opaque score.

Methods

Hetch Hetchy Regional Water System

Moving from upstream to downstream, the Tuolumne watershed includes three reservoirs owned by SFPUC (Hetch Hetchy, Cherry, and Eleanor), followed by the Don Pedro Reservoir owned by the Turlock and Modesto Irrigation Districts, followed by the lower Tuolumne channel feeding into the San Joaquin River (Figure 1A). Together, SFPUC’s three reservoirs and its Water Bank account within the Don Pedro Reservoir provide SFPUC with 1 230 800 AF (1518 million m³) of storage capacity within the Tuolumne River portion of its RWS.² Water leaving the three SFPUC-owned reservoirs is either diverted for water supply or released downstream. Of the amount released downstream, a portion is first used for power generation before being returned to the downstream channel.³¹ SFPUC follows a Water First policy whereby water supply outcomes are prioritized over power generation.³²

Figure 1.

Schematic of the Tuolumne River system with (A) key water supply features and their positions relative to the Bay Area service area, (B) locations of hydrologic data sets, (C) model representation, and (D) description of hydrologic data sets.

Under the existing policy, SFPUC must release a certain flow from its reservoirs to satisfy the senior water rights of the downstream irrigation districts. The Districts are jointly entitled to the first 2416 cubic feet per second (cfs; 68 m³/s) of unimpaired Tuolumne River flow, while SFPUC may divert or store any flow in excess of this amount.^17,33,34 Between April 15 and June 13 annually, the 2416 cfs (68 m³/s) entitlement increases to 4066 cfs (115 m³/s). When unimpaired Tuolumne River flow is less than the full potential entitlement, SFPUC owes the entire unimpaired flow to the Districts. When SFPUC releases more than the entitled flow, the excess volume is credited to its Don Pedro Reservoir Water Bank account, which can later be debited to offset future releases to the Districts.¹¹

Select terms related to the RWS are described below:

• The Districts: Turlock and Modesto Irrigation Districts, which hold senior water rights on the Tuolumne River and own and operate the Don Pedro Reservoir

• Water Bank: SFPUC storage account in the Don Pedro Reservoir, consisting of predelivered water which can offset future releases to the Districts

• Full Natural Flow: streamflow produced by a watershed assuming no upstream diversions or storage; also called unimpaired flow

• Bay-Delta Plan: amended policy not yet implemented that requires 40% of unimpaired flow to be left in-stream for ecosystem use between February 1 and June 30 each year

• Water First Policy: SFPUC manages reservoirs so that power generation does not negatively impact water supply outcomes, consistent with California water law

Hydrologic Data

Historical Observations

This project relies on hydrologic data sets accessed through the USGS Streamgaging Network,³⁵ the California Data Exchange Center,³⁶ and correspondence with SFPUC (Nicholas Waelty, personal communication, February 9, 2022 to March 1, 2023). The hydrologic observations include time series data sets of streamflow, reservoir storage, and evaporation at locations within the Tuolumne watershed. The waterbody, sampling rate, time availability, and data source associated with each data set are detailed in Figure 1D, and measurement locations are illustrated in Figure 1B.

Streamflow Projections

As part of California’s Fourth Climate Change Assessment, Pierce et al.³⁷ model unimpaired Tuolumne River streamflow under many possible climate futures and share resulting streamflow projections through Cal-Adapt.³⁸ The Tuolumne streamflow projections are generated through land surface modeling of bias-corrected and downscaled global climate models. Pierce et al. evaluated 10 global climate models and identified a subset of four that capture the range of climate outcomes: HadGEM2-ES (warm/dry scenario), CNRM-CM5 (cool/wet scenario), CanESM2 (average scenario), and MIROC5 (complement scenario, most unlike the other three). Each of these four models is evaluated at two emissions levels (medium/Representative Concentration Pathway-RCP 4.5 and high/RCP 8.5) to generate a total of eight future streamflow scenarios. Our project uses these eight scenarios to represent uncertain future hydrology for planning the Bay Area water supply.

Demand Projection

This work models future demand with a moderate growth assumption drawing from two sets of projections prepared by SFPUC. In its 2020 Urban Water Management Plan, SFPUC plans for reliability by estimating upper-bound future water demand associated with full realization of regional development goals,³⁹ an approach that in the past has consistently overpredicted future demand.⁴⁰ By contrast, in its 10-year Financial Plan, SFPUC targets a more realistic yet conservative demand projection to inform its budget and operations.³⁹ Because of their respective purposes and biases, neither projection is appropriate for our analysis, which seeks a realistic yet reliability-oriented projection. Instead, our analysis assumes a moderate growth case falling midway between SFPUC’s upper- and lower-bound estimates.

To construct the moderate growth future demand projection, this analysis averages upper- and lower-bound annual projections, scales by 0.85 to focus on the 85% supply contribution of the Tuolumne River subsystem, and then distributes the annual demand along typical monthly fractions observed historically. To create an upper-bound annual projection from the 5-year-increment point estimates in SFPUC’s 2020 Urban Water Management Plan, this analysis fits nonlinear least-squares regressions to the retail and wholesale components and sums these two contributions (Figure S1). Wholesale supply to customers outside of San Francisco is guaranteed only up to a 184 MGD (697 000 m³/d) supply assurance,⁴¹ so the future wholesale component is capped at this level. The lower-bound annual projection is represented by the average annual delivery from the past 5 years (WY 2018–2022), since SFPUC’s financial plan projects no growth in water sales.⁴²

Tuolumne Water Supply Model

To investigate policy and climate impacts on the future reliability of Bay Area regional water supply, this project develops a systems model in R of the Tuolumne River and its Hetch Hetchy RWS contributions.⁴³ The model focuses on the fundamental water allocation to inform key planning questions.

The model representation of the RWS is illustrated in Figure 1C. Inside the system boundary is the total storage available to SFPUC, which includes the Cherry, Eleanor, and Hetch Hetchy reservoirs, as well as the Water Bank account. This combined representation is appropriate because, in practice, the storage features are operated collectively to satisfy release and delivery targets.⁴⁴ Key fluxes across the system boundary include inflow, downstream obligations, urban water supply deliveries, and evaporation.

Policy Scenarios

Water supply performance is compared between two policy scenarios: Existing BDP (existing water rights) and Amended BDP (Bay-Delta Plan in addition to existing water rights). The policy scenario determines the amount of flow that SFPUC must release downstream. In the Existing BDP scenario, the downstream flow obligation is defined by eq 1, which represents the Districts’ entitlement.

1

In the Amended BDP scenario, an ecological demand (eq 2) is added that during select months requires SFPUC to contribute 51.7% of the flow needed to meet the 40% unimpaired flow benchmark, compared to typical existing in-stream flow requirements established in the Federal Energy Regulatory Commission license.⁴⁵

2

Combining the new ecological demand with the existing water rights obligation gives SFPUC’s downstream flow obligation under the Amended BDP scenario (eq 3).

3

Reservoir Simulation

Under each policy scenario, reservoir operation is simulated using a monthly time step, which captures intra-annual seasonality while supporting the assumption of negligible conveyance time throughout the system relative to the monthly time step. The simulation is governed by the water balance in eq 4, where S is total SFPUC storage, Q_FNF is the Tuolumne watershed full natural flow, E is total SFPUC evaporation, Q_ds is SFPUC’s downstream flow obligation (including Districts’ entitlement and ecological demand, if applicable), and D is SFPUC’s urban water supply deliveries.

4

Evaporative losses attributed to SFPUC include all evaporation recorded at the Hetch Hetchy, Cherry, and Eleanor reservoirs as well as a fraction of Don Pedro Reservoir evaporation equal to the fraction of Don Pedro storage comprised by the SFPUC Water Bank. Historical evaporation rates recorded at the reservoirs depend on weather conditions and the water surface area of the reservoir at its historical level. These observed rates cannot be used directly in the simulation since the different policy rules result in reservoir storage and therefore surface area different than what occurred historically. Instead, this analysis uses a month-specific evaporation factor representing evaporative loss as a fraction of the total storage volume. This simplification is appropriate because when historical evaporation is normalized by storage level, the month-specific evaporation factor is highly consistent across the historical record (Figure S2).

SFPUC storage capacity within the Tuolumne River system varies across the year. The SFPUC Water Bank contributes 570 000 AF (700 million m³) toward SFPUC’s total 1 230 800 AF (1518 million m³) Tuolumne storage capacity, yet the Water Bank capacity can expand by up to 170 000 AF (210 million m³) between April 28 and October 6 of each year when flood reserve capacity is not needed within the Don Pedro Reservoir.^2,11 Accordingly, SFPUC Tuolumne storage can rise as high as 1 400 800 AF (1728 million m³) during this summer period. In early October when the nominal storage capacity returns to 1 230 800 AF (1518 million m³), the de facto storage capacity is somewhat smaller because a flood reserve is required operationally within SFPUC’s upcountry reservoirs. In practice, this flood reserve depends on year-specific forecasting analysis, though for our analysis a constant flood reserve of 123 080 AF (152 million m³; 10% of total SFPUC capacity) is used to represent effective storage capacity based on historically observed storage levels (Figure S3). The resulting model demonstrates close agreement between historical observations and simulated storage under the Existing BDP policy (Figure S4).

This research analyzes water supply performance through both retrospective and prospective lenses. The retrospective analysis covers a 29-year period encompassing water years 1994–2022, inclusive, using known historical hydrology to understand the water supply effect if the amended Bay-Delta Plan had been in place historically. The prospective analysis looks forward over a 30-year period of water years 2023–2052 and assesses the policy impact under climate change uncertainty by considering several possible hydrologic futures. Together, these analyses describe the possible water supply impacts of the amended Bay-Delta Plan policy and help advance planning discussions working toward a more sustainable Bay Area water future.

Results

System Demand from 40% Unimpaired Flow

Across the 29-year retrospective analysis window, the average full natural flow in the Tuolumne River watershed was 1 930 000 acre-feet per year (AFY; 2380 million m³/y). Had the amended Bay-Delta Plan been in place historically, its 40% unimpaired flow requirement would have placed an average annual demand of 260 000 AFY (320 million m³/y) on the RWS, distributed between February and June (Figure 2). This ecological demand is larger and more variable than the amount diverted for urban water supply, which averaged 220 000 AFY (270 million m³/y). Both these demands are small relative to the irrigation Districts’ water rights entitlement averaging 1 040 000 AFY (1280 million m³/y).

Figure 2.

RWS demands resulting from the 40% unimpaired flow requirement applied during the 29-year retrospective simulation. (A) Annual full natural flow (FNF) availability and supply allocation. (B) Monthly distribution of Bay-Delta Plan ecological demand on the RWS, which applies February–June only. The box-and-whisker plot shows the median and interquartile range of the data, with whiskers extending to the minimum and maximum values within 1.5× the interquartile range and outliers shown as points.

Month-to-month flow allocation within the Tuolumne system varies dramatically by season and year type. Figure 3 shows how the full natural flow is allocated among system demands during the severe 2012–2016 drought and recovery. System demands include the Districts’ entitlement, Bay-Delta Plan ecological flows, urban supply deliveries, evaporation, amounts added to storage, and finally uncaptured amounts which are released downstream in excess of other downstream obligations due to lack of available storage capacity.

Figure 3.

RWS flow allocation by system demand under (A) existing BDP and (B) amended BDP policy scenarios. Shaded amounts below the black line are supplied by available flow, while shaded amounts above the black line are supplied by depleting storage reserves. Gray uncaptured amounts above the black line represent the storage drawdown as the seasonal flood reserve takes effect.

The analysis shows that ecological demand impacts the RWS differently depending on the state of the system. In drought when the Districts’ entitlement comprises all or nearly all of the full natural flow (e.g., 2014 in Figure 3B), the ecological demand must be supplied from storage, and the system is left with less water than it would otherwise have had. By contrast, in wet years with enough full natural flow to meet all demands and top off storage (e.g., 2017 in Figure 3B), the ecological flow released downstream simply replaces a portion of otherwise uncaptured flow and the system is not left with less water than it would otherwise have had. In moderate years (e.g., 2016 in Figure 3B), the ecological demand can be supplied by available full natural flow, but this reduces the amount added to storage, which may be more or less significant, depending on the current storage level and subsequent hydrology.

The monthly flow available to SFPUC is represented by the gap, if any, between the green shaded region and the black line (i.e., any full natural flow in excess of the Districts’ entitlement). Figure 3 shows that this SFPUC flow availability is virtually nonexistent in some years (e.g., 2014–2015), while in others (e.g., 2017) it dwarfs the size of the Districts’ entitlement. By comparison, the Districts’ entitlement is relatively consistent year to year. This results in a risk profile wherein SFPUC is more vulnerable to water shortage in dry years yet better poised to take advantage of excess flow in wet years.

Simulated Storage Levels

Simulating water supply operations retrospectively results in the storage level series shown in Figure 4. Visible in the figure is the major 2012–2016 drought, as well as the shorter 2007–2009 drought.⁴⁶ The simulation shows that, under the same hydrology, the amended Bay-Delta Plan policy can result in faster depletion, slower recovery, and decreased levels of RWS storage. During the 2012–2016 drought, for example, the Amended BDP scenario experiences a minimum storage level 330 000 AF (410 million m³) lower than that of the Existing BDP scenario. Making up this deficit through conservation alone would require a further 35% reduction in urban supply maintained over four years, on top of the conservation that happened historically during this period. Historical conservation built to a 25% cutback mandate in effect for just under one year, which was itself an unprecedented achievement yet is dwarfed by the larger effort that would be required to offset the effects of the Amended BDP policy. Delayed recovery of storage levels is seen in 2009–2010 and 2016 as reservoirs fill 1–2 years later in the Amended BDP scenario. During the late 1990s, simulated storage remains similarly high under either policy scenario, which is consistent with the State Water Board’s expectation of minimal water supply impact in wet years.⁴⁷

Figure 4.

Comparison of SFPUC storage levels simulated retrospectively under existing BDP and amended BDP policy scenarios.

The flow allocation in Figure 3 works in tandem with Figure 4 to help explain the simulated storage behavior and the variable impact of the ecological demand. During the 2012–2016 drought when the Amended BDP scenario produces a critically low storage level 330 000 AF (410 million m³) below that of the Existing BDP scenario, Figure 3 shows that this difference is attributed not to a single dramatic stressor but instead to a modest ecological demand persisting and accumulating over several years. Following the drought, the Amended BDP scenario storage recovers 1 year later than the Existing BDP storage, partly due to the ecological demand reducing what can be added to storage in 2016 (Figure 3) and partly due to the more critically low storage level from which recovery begins in the Amended BDP scenario (Figure 4). By 2017, the Amended BDP and Existing BDP storage series match (Figure 4) as storage fills to capacity, and ecological demand replaces otherwise uncaptured flow (Figure 3). Thus, after dramatically felt effects in drought, the visible impact of the Amended BDP policy fades from view as wet weather returns.

Water Supply Performance

Future climate uncertainty is represented by eight hydrologic scenarios, and simulating water supply under these scenarios produces eight pairs of Amended BDP and Existing BDP storage level series (Figure S5). Unlike the known retrospective hydrology, these future scenarios capture statistical behavior rather than predict conditions for any specific year. For this reason and to represent future uncertainty more concisely, the prospective impacts from climate and policy are best analyzed through aggregate performance metrics rather than individual time series analysis.

One aggregate performance measure is a cumulative distribution function of absolute storage, which describes how much time the system spends at or below any given storage level. Keeping policy fixed, each hydrologic scenario produces a different cumulative distribution of absolute storage (Figure S6); together, these eight curves define a zone of future performance with climate uncertainty for that policy scenario. Figure 5 compares absolute storage between Amended BDP and Existing BDP policy scenarios for a 30-year prospective planning window.

Figure 5.

Cumulative frequency distribution of absolute storage under prospective amended BDP and existing BDP policy scenarios. More favorable water supply outcomes lie to the right (indicating a greater storage level maintained for a given fraction of time) and below (indicating less time spent below a given storage level). Regions of steeper, more vertical orientation indicate that storage spends a larger fraction of time at the corresponding storage level, whereas a flatter, more horizontal orientation indicates that storage spends a smaller fraction of time at the corresponding storage level.

Figure 5 answers two types of questions about prospective storage with climate uncertainty. First, for any given percentile of performance, what storage level is anticipated? Second, for any storage level of interest, how often does storage fall below this level? The figure answers these questions for any user-defined threshold or percentile so that decision makers can visualize the entire performance profile and access the information they find most valuable. Highlighted in the figure are two sample answers. With the Existing BDP policy, median storage is expected to be approximately 1.1 million AF (1400 million m³), which is the effective capacity during most of the year. By contrast, with the Amended BDP policy, median storage may be as high as 1.1 million AF (1400 million m³) or as low as 920 000 AF (1100 million m³), depending on climate conditions. Using 500 000 AF (600 million m³) as one possible threshold for critically low storage, the figure shows that storage might be expected to fall below this threshold for 0–5% of months with Existing BDP policy but for 0–7% of months with Amended BDP policy.

Figure 5 shows that, prospectively, under the Amended BDP policy scenario, the RWS is expected to spend larger fractions of time at lower storage volumes. Steeply rising curves at the two seasonal capacity levels indicate that storage can often be maintained at capacity. With the Existing BDP policy, storage is expected to be full during at least 42% of months, even under a severe climate scenario; with the Amended BDP, this number drops to 25%. The red zone describing Amended BDP outcomes is broader than the blue zone describing Existing BDP outcomes. Since both zones are defined by the same set of hydrologic scenarios, the broader Amended BDP zone indicates greater sensitivity of water supply performance to hydrology, and thus increased vulnerability to climate change. The overlapping zone of Amended BDP and Existing BDP outcomes indicates that the Amended BDP policy may have the effect of exacerbating climate change effects on water supply performance, since Amended BDP policy with milder climate could produce a similar storage distribution as Existing BDP policy with more severe climate. This analysis agrees with findings presented in the Long Term Vulnerability Assessment performed by the Water Research Foundation, which finds that climate change amplifies the reliability impacts of other stressors including demand growth and in-stream flow requirements.²⁵

Storage is a valuable indicator of water supply performance because it emphasizes the reserves available to meet water supply needs rather than emphasizing the current intensity of conservation measures, which is more directly manipulated by system managers. In drought, for instance, storage levels will necessarily fall whenever demands exceed available water. By contrast, a variable like urban water deliveries depends more heavily on management decisions of when to initiate conservation measures of what magnitude. This simulation assumes that all demands are fully met whenever sufficient water is available within the system, though in practice these deliveries would eventually be curtailed to extend the life of the storage reserves. The target and timing of such curtailments are management decisions that depend not only on system conditions but also on other factors such as meteorological forecasts, risk preferences, and political will; however, significant curtailments would certainly be expected when storage falls to extremely low levels. Thus, these times of extremely low storage are months when system performance breaks down and cannot meet all demands (unlike earlier, voluntary conservation that occurs when the system could meet full demand, but prudent managers choose not to do so).

Focusing further on the policy impact to water supply outcomes, another useful performance metric is storage level impact, which describes the month-to-month deficit of Amended BDP storage relative to Existing BDP storage. In months when storage is full and excess water is available, the Amended BDP policy results in the same storage as the Existing BDP policy, and the storage level impact is zero. By contrast, in months when the Amended BDP ecological demand diverts water that would otherwise go to other uses, the Amended BDP storage is less than the Existing BDP storage, and the storage level impact is greater than zero. Figure 6 shows the exceedance plot of the storage level impact for a 30-year prospective planning window. Each curve corresponds to a particular hydrologic scenario, and together the curves define a region of performance with climate uncertainty.

Figure 6.

Exceedance plot of prospective storage level impact from the amended BDP. Individual curves correspond to the eight hydrologic scenarios considered: cool, warm, average, and complement models at medium/RCP 4.5 and high/RCP 8.5 emissions levels.

Like the absolute storage analysis, Figure 6 also answers two types of questions about the storage level impact with climate uncertainty. First, for any given percentile of performance, what storage level impact is anticipated? Second, for any impact level of interest, how often does storage impact exceed this amount? The figure answers these questions for any user-defined threshold or percentile so that decision makers can visualize the entire performance profile and access the information they find most valuable. Highlighted in the figure are two sample answers. On the larger end of anticipated impact, the 90th percentile storage impact may be as small as 80 000 AF (100 million m³) or as large as 370 000 AF (460 million m³), depending on climate. Using 50 000 AF (60 million m³) as one possible threshold for significant impact, the figure shows that the Amended BDP policy may produce significant impact in 18–64% of months.

The Amended BDP storage impact is expected to be the smallest under a cool/wet climate scenario and the greatest under the complement scenario. Interestingly, the severity of the storage impact does not follow directly from the severity of the climate scenario. Figure 6 shows that the average and warm/dry scenarios with medium emissions produce quite similar impact distributions, and for a given climate model, the high-emission scenario does not necessarily result in a more severe impact than the medium-emissions scenario. These results underscore the importance of considering the risk profile and vulnerabilities of the water system, beyond general hydrologic characteristics of climate change.

Discussion

This analysis shows that, although the precise impact of the Bay-Delta Plan unimpaired flow requirement is nonconstant in time and contingent on climate, the policy would result in faster depletion and slower recovery of RWS storage during drought and bordering periods when water shortages are most keenly felt. In an extreme case among the prospective hydrologic scenarios considered, this impact could result in a situation in which the system runs out of water and cannot meet all demands. Far more commonly, the impact of the Amended BDP policy would look like earlier conservation response and increased time spent at lower storage levels, where meeting present demands must be balanced more aggressively against reserving stored water for future use.

Because future climate is uncertain, prospective water supply performance also carries uncertainty. This analysis illustrates ranges of performance which may seem imprecise, but which still provide very useful insight to inform decision making. For example, a large (>50 000 AF; >60 million m³) Amended BDP storage impact is anticipated in 18–64% of months. Narrowing this range requires making assumptions about the type of climate future that merely obscures rather than eliminates the uncertainty that surrounds the policy impact. With the range as it stands, water supply planners can recognize that, climate uncertainty notwithstanding, a large storage impact is neither rare nor ubiquitous but something that must be planned for and managed as part of a reliable future water supply.

Drought resiliency planning is not a one-size-fits-all approach. In working to maintain a sustainable water supply amidst climate and policy stressors, planning efforts should recognize and address the distinct risk profile that characterizes the RWS and its vulnerability to these stressors. Because SFPUC may divert any streamflow beyond the Districts’ entitlement, the risk profile of SFPUC looks different from that of the Districts, who receive the first cut but are not guaranteed elevated wet-weather flows. SFPUC is uniquely poised to take advantage of large wet-year flows to aid in dry-year management. This means that storage assets and wet-year management are highly relevant to drought planning. Working within this risk profile, taking fuller advantage of wet-year flows could help balance the heightened vulnerability of being allowed no flow diversions during the driest conditions. This underscores the importance of a planning approach that considers drought in the context of a continuous hydrologic record, which is an element missing from the current Reliability Assessment planning approach, yet aligns well with other efforts in the state including the Department of Water Resource’s Flood-MAR program promoting managed aquifer recharge of excess flows.⁴⁸ SFPUC can improve its ability to take advantage of wet-year flows through projects and partnerships that expand the system storage capacity. This might look like farm-scale recharge basins constructed in partnership with Central Valley irrigation districts,⁴⁹ or like the proposed Calaveras Reservoir Expansion Project currently being considered as part of SFPUC’s Draft Alternative Water Supply Plan.⁵⁰

Along with a continuous hydrologic record, our project incorporates climate-modified hydrology, multiple multiyear drought realizations, and richer decision support information to address gaps in the current planning approach. The visualizations of water supply performance are structured around user input so that planners can evaluate multiple cases and access results specific to parameters of interest. This might entail comparing performance among climate scenarios, anticipating how often storage falls below a user-defined threshold, or describing storage with uncertainty at a user-defined percentile. Planners can access these results by applying their own priorities and assumptions, rather than questioning what is already embedded, so that discussions can build on a shared understanding of the system behavior. Managing the Tuolumne River to balance coequal goals of water supply and ecosystem health is a challenging, high-stakes task. This study advances the conversation by quantitatively assessing the water supply impact of the Bay-Delta Plan ecological flow requirement, which is a fundamental step toward balancing this impact commensurate with ecological outcomes.

Looking forward, a key area for future study is to integrate coping strategies into the analysis to better understand the opportunities available to mitigate the water supply impact of the amended Bay-Delta Plan policy. Also needed is the development of risk-based performance standards for sustainable water supply and quantitative streamflow targets for ecosystem health, along with continued research evaluating the climate change impact to future hydrology and predicting future water system demands.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.3c07398.

• Observed and projected demand, observed and modeled evaporation, effective storage capacity, agreement between observed and simulated storage, storage level time series by hydrologic and policy scenario, storage level distribution by hydrologic and policy scenario (PDF)

Author Present Address

^∥ Department of Environmental and Occupational Health Sciences, University of Washington, Seattle, Washington 98195, United States

Author Contributions

Conceptualization, B.C.G., A.R.S., R.H., S.F., and R.G.L.; formal analysis, B.C.G. and R.G.L.; writing—original draft preparation, B.C.G.; writing—review and editing, B.C.G., A.R.S., R.H., S.F., and R.G.L.; visualization, B.C.G.; supervision, R.G.L.; funding acquisition, B.C.G. and R.G.L.

The authors declare no competing financial interest.