Coupling Quantity, Conservation, and Habitat through WSOne

An Integrated Reservoir Model

By Andrew Walker, PH-SW, CFM, Michael Kastanotis, Cassandra Albrecht, EIT, and William Paulitz, P.E.


The water supply for the City of Peabody, Massachusetts is comprised of several surface water reservoirs that serve a community of approximately 54,000 residents. These reservoirs include three terminal reservoirs from which raw water is withdrawn (Winona Pond, Suntaug Lake, and Spring Pond) and four additional upstream reservoirs (Browns Pond, Middle Basin, and Lower Spring Pond-Fountain Pond).

The primary recharge for two of the city’s three terminal reservoirs (Winona and Suntaug) is withdrawals from the Ipswich River. Additionally, the city has an active connection to the regional Massachusetts Water Resource Authority (MWRA).

In recent years, due to water scarcity and its impact on the Ipswich River and the North Coastal watershed, the city has had to make quick real-time decisions to sustain adequate water to its customers. Most recently, in the midst of a severe drought in 2022, the city’s three terminal reservoirs reached or nearly reached minimum capacity and the city’s only choice was to purchase more water from the MWRA.

City staff recognized this was not an ideal approach and that a more holistic evaluation would be beneficial. They sought an approach that would utilize predictive modeling to make proactive water supply decisions and to confirm the long-term resilience of the public water supply. This article serves to:

  • document the work completed to develop an Integrated Reservoir Model (Model);
  • present the results of the evaluation of the city’s water supply resilience under present day and future climate conditions;
  • present the Model’s use in identifying and evaluating the potential costs and benefits of various capital improvement projects and reservoir management strategies; and
  • present its incorporation into a web-based guidance tool for the city.

Project Goals

Going into this effort, the city had several project goals that guided the Model’s development. Specifically, the project intended to provide the city with:

  1. A baseline understanding of the existing water supply conditions in Peabody and how a variety of environmental factors such as precipitation, temperature, direct surface runoff, and groundwater combine with transfers from the Ipswich River to impact the surface water levels in the city’s water supply reservoirs at different times of year.
  2. An assessment of the resilience and vulnerabilities of the city's raw water supply system to climate-driven changes like precipitation, temperature, and streamflow; to changes in city-wide demand patterns; and to changes in permit requirements.
  3. An analysis of the financial and water supply impacts associated with a number of capital improvement projects, such as upgrading the raw water intakes and/or equipment at the city’s reservoirs or reactivating two inactive groundwater wells, along with several changes to the city’s operations of their water system, such as purchasing water from the MWRA in greater quantities or at different times of year, implementing water conservation measures and/or restrictions, transferring water between reservoirs, and many more.
  4. A proactive tool in the form of a web-based portal called “oneWater” that city staff and treatment plant operators could use daily, based on current reservoir and watershed conditions, to predict future reservoir level forecasts and guide decision-making around water supply management.

City of Peabody Water Supply System

The City of Peabody’s water supply system is organized around two treatment plants: Bob Walsh (Walsh) and Winona.

Stylized depiction of Peabody’s surface drinking water supply system

Stylized depiction of Peabody’s surface drinking water supply system.

Walsh Water Treatment Plant: Raw water is provided to the Walsh Water Treatment Plant (WTP) via intakes in two terminal reservoirs: Spring Pond and Suntaug Lake. Spring Pond is, in turn, supplied with water from three other waterbodies: Browns Pond, Lower Spring Pond, and Fountain Pond. Lower Spring Pond is hydraulically connected to Fountain Pond under all but the lowest water levels. Water leaves the Lower Spring Pond-Fountain Pond system via a stoplog-controlled outlet. The stoplogs are controlled seasonally to prevent high groundwater and surface water impacts to nearby residential areas. Water is also regularly pumped from the Lower Spring Pond-Fountain Pond system upgradient to Spring Pond to replenish that terminal reservoir as water levels in Lower Spring Pond-Fountain Pond allow. Spring Pond is also supplied by a piped connection from Browns Pond. In addition to being supplied by a raw water intake from Spring Pond, the Walsh WTP also receives raw water from Suntaug Lake.

Winona Water Treatment Plant: In contrast to Walsh, the Winona WTP is supplied by a sole terminal reservoir, Winona Pond. Suntaug Lake and Winona Pond are primarily recharged by seasonal withdrawals from the Ipswich River via Middle Basin, in addition to traditional recharge from precipitation, groundwater, and surface runoff from their immediate drainage areas. Middle Basin has no significant storage capacity of its own but receives water pumped from the city’s registered Ipswich River withdrawal point. Ipswich River withdrawals are allowed, via the city’s Water Management Act registration, between December 1 and May 31, under the condition that a minimum of 15.5 cubic feet per second (cfs) are always left in the river.

In addition to the above water supply sources, the city’s active MWRA connection, and emergency connections to the Town of Danvers and City of Salem, there are two inactive groundwater wells that have historically provided the city with raw water. However, use of these wells was discontinued due to both worsening water quality and the fact that water was no longer able to be pumped to the distribution system without additional treatment.

Developing an Integrated Reservoir Model

The project team developed the Model to be able to evaluate inflows and outflows to all seven waterbodies in the city’s water supply system and their resulting water levels. Inputs to the Model include:

  • precipitation
  • evaporation
  • direct surface runoff
  • groundwater fluxes (both in and out)
  • inter-reservoir transfers
  • withdrawals from the Ipswich River
  • dynamic raw water demands that vary with hydrologic conditions (e.g., greater demand during dry summers)
  • raw water offsets due to the use of regional and/or local purchases, the reactivation of groundwater supplies, and the initiation of water conservation measures.

These inputs were modified to evaluate the vulnerability of the reservoir system to changes in climate, permit requirements, and demand trends and to evaluate the impact of changes to the reservoirs, their interconnections, and their operation and management on the reservoir system’s reliability. Figure 1 shows a schematic of the city’s system.

Existing Data

The team gathered publicly available and relevant city-specific data to support development of the Model. Tables 1 and 2 below list the data gathered, including data type, data source, time period, and frequency of measurement. Datasets like precipitation and Ipswich River flow were used to estimate the addition or inflow of water into the various reservoirs while datasets like temperature and raw water withdrawals were used to estimate losses from the reservoirs. Historical reservoir levels and raw water withdrawal data were critical to testing the Model to confirm it produced accurate results before using it to assess the vulnerabilities and potential modifications to the city’s water supply system.

In addition to the publicly available and city-provided data, the team gathered surface water-groundwater gradients to understand the likely range or order of magnitude of potential groundwater contributions to each waterbody, which aided us during the model calibration process. This involved installing a surface water and shallow groundwater monitoring network and a GPS survey to collect location and elevation data as they related to the monitoring network.

The team also installed four piezometer/staff gauge couplets and a staff gauge to provide a surface water and shallow groundwater monitoring network for the collection of reservoir stage and shallow groundwater head.

After constructing each couplet, the team installed pressure transducer/data loggers (PTDLs) in each piezometer to collect water level and temperature data. The project team programmed the PTDLs to record water levels at 15-minute intervals prior to installation, with collection over 10 weeks.

Mass Balance Models

The Model is effectively a series of five interconnected mass balance models of each of the five reservoirs – Winona, Suntaug, Brown’s, Springs, and Lower Spring-Fountain (combined). Note that due to its small size, Middle Basin was not modeled as a pond with storage capacity, but rather as a flow split between Winona and Suntaug. Reservoir mass balance models tabulate all inflows and outflows to and from a reservoir at regular time intervals, calculating the change in storage over each interval and the associated change in the reservoir surface.

Because the Model represents five separate reservoirs, several of which are interconnected, and because model development and calibration relied upon data from a wide range of sources, an important first step to building the Model was to identify a single vertical datum for all five reservoirs. To confirm that all reservoir levels and outlet structure elevations referenced a single datum, such as feet above mean sea level in the North American Vertical Datum of 1988 (NAVD88), the team surveyed reservoir water surfaces within hours of the city taking their usual weekly reservoir level measurements and developed a conversion between the various local datums at each waterbody and NAVD88.

To support the tabulation of precipitation, evaporation, and other reservoir inflows and outflows and to relate changes in reservoir storage to changes in water level, the team developed relationships between reservoir level, surface, and storage volume. Generally, these relationships were developed by combining bathymetric mapping from existing sources with LiDAR-based topographic data above typical water levels.

Watershed Inflows and Outflows

Each of the five reservoirs incorporated into the Model have four watershed inflows and outflows in common: precipitation, evaporation, direct runoff, and groundwater contributions.

The volume of precipitation entering each reservoir on any given day was calculated by multiplying the depth of precipitation recorded at the National Weather Service’s (NWS) meteorological gaging station at Logan Airport in Boston by the current surface area of that reservoir.

Similarly, the volume lost from each reservoir due to evaporation was calculated daily based on the estimated depth of evaporation by the current surface area of that reservoir. The depth of evaporation was estimated from daily maximum and minimum temperature data recorded at the Logan Airport gauge using the Papadakis method.

Direct runoff entering each reservoir from its respective drainage area was estimated daily by multiplying the mean daily discharge value recorded by the USGS Ipswich River gauge in nearby South Middleton by the ratio of a reservoir’s drainage area to the USGS gauge’s watershed size of 44.5 square miles.

Groundwater can act as both inflow and outflow to a reservoir depending on the aquifer material, water table elevation, and reservoir level. Not only can groundwater act as both inflow and outflow, but the magnitude of that contribution can change significantly over time. Groundwater fluxes to each reservoir were estimated monthly during the model calibration process and then checked against likely ranges of groundwater flow estimated from groundwater level data recorded with the staff gauge-piezometer network installed in April 2023.

Reservoir Interconnections and Control Rules

In addition to the naturally occurring inflows and outflows to each reservoir, the Model also incorporates the many piped inlets and outlets, including interconnections between the reservoirs and their respective control rules.

Drinking Water Withdrawals

In addition to the inlets and outlets described above, the Model also incorporates drinking water withdrawals to the Winona and Walsh WTPs. Drinking water withdrawals from Winona Pond, Suntaug Lake, and Spring Pond were identified monthly from historical withdrawal data for most years between 2013 and 2020. Those monthly average withdrawals are presented in Table 3.

The team considered earlier data irrelevant due to changes in demand by the city’s water users and in how and when the city changed how and when they supplemented their own raw water with purchases from the MWRA. In addition, raw water withdrawal data from 2017, 2018, 2021, and 2022 were not used since withdrawal rates were significantly impacted by several WTP shutdowns during those years.

Note that while the Model does not represent the physical withdrawal structures explicitly, it does incorporate the physical limitations of those structures by incorporating reservoir failure levels in each terminal reservoir. When the water level in one of the reservoirs drops below its failure elevation, withdrawals from that reservoir to its WTP are assumed to cease.

Model Calibration

Graph depicting historical vs. Simulated Reservoir Storage, 2015-2022, in Winona Pond.

Figure 2: Historical (observed, in gray) vs. Simulated (blue) Reservoir Storage, 2015-2022, in Winona Pond.

The project team calibrated the Model against historical reservoir level observations to verify that appropriate results are produced. While discharge into or out of the reservoirs through man-made structures is rather precisely modeled and most naturally occurring inflows and outflows are based on high quality, long-term data sources, very little groundwater level data exists with which to estimate groundwater inflows or outflows to each reservoir.

As groundwater contributions can represent a significant portion of a reservoir’s daily water balance, this variable was iteratively modified during calibration to maximize agreement between simulated water levels and historical water levels in all five modeled reservoirs that were observed by city staff between 2015 and 2022. Note that during calibration and validation simulations described in this and the following section, the team replaced average monthly drinking water withdrawal rates with historical withdrawal data for the appropriate time period.

Based on the calibration groundwater contributions identified above, the Model is generally able to reliably reproduce historically observed reservoir levels; an example of that agreement is shown in Figure 2. As the figure indicates, the Model generally reproduces historical reservoir levels quite accurately during all seasons and across a wide range of year-to-year conditions. The simulated minimum annual reservoir levels, perhaps the most important output from the Model, are particularly well matched during the 2015-2022 calibration period.

Graph of historical vs. Simulated Reservoir Levels in Suntaug Lake.

Figure 3: Historical (Observed) vs. Simulated Reservoir Levels, 2008-2012, in Suntaug Lake.

Model Validation

To confirm that the assumptions made during the calibration process are appropriate over a wider range of hydrologic and meteorologic conditions, the project team validated the Model by simulating a different multi-year period, in this case the 5-year period from 2008 to 2012, with no modifications to the Model and compared model results to observed conditions during. For example, validation results for Suntaug Lake are shown in Figure 3 below.

As shown in the figure above, the Model can reliably reproduce the behavior of Peabody’s water supply reservoirs over a wide range of conditions and seasons. It is therefore considered an effective tool for identifying and understanding the resilience and/or vulnerability of the city’s water supply system to changes in climate, demand patterns, and permit requirements.  It is also useful in evaluating the effectiveness of various solutions to address those vulnerabilities.

Even while being an effective tool, the Model includes potential weaknesses that should be noted and possibly addressed through additional data gathering and analysis in the future.

  1. The groundwater contributions were estimated monthly as part of the model calibration process, but groundwater contributions can vary significantly based on relative surface water and groundwater levels, which can fluctuate year-to-year even during the same month. Additional data may be required to better understand the relationship between surface water and groundwater levels at each reservoir. As groundwater inflows to many of the reservoirs represent such a significant component of many reservoirs’ ability to recharge, updating the Model to incorporate those relationships will allow it to more accurately predict future reservoir levels.
  2. The Model assumes demand is independent of watershed conditions, using the same drinking water withdrawal value from each reservoir on a monthly basis. In reality, hot and dry summers are often associated with increased demand for irrigation. Incorporating statistical correlation of the relationship between recent hydrologic and meteorologic data and future raw water demand may improve the accuracy of model predictions but was beyond the scope of this project.

Understanding Existing Conditions

The project team first used the Model in assessment mode to simulate existing reservoir management and baseline climate conditions. In this simulation, no changes were made to the reservoirs or their operations. Climate conditions were represented by historical daily observations of precipitation, temperature, and Ipswich River flow from 1950 to 2022. Results of this “Business-as-Usual” simulation of the Model are shown graphically in Figures 4, 5, and 6.

In general, these figures show the simulated water level in a reservoir, in terms of how full it is, in a black line that rises and falls seasonally over a 73-year period. Low points are typically associated with late summer and fall. Differences in dry season low points (and winter recovery in the case of some reservoirs) highlight the variability of natural and anthropomorphic inflows and outflows to a reservoir, sometimes resulting in significant differences in reservoir level year-to-year.

Note that the red line in each figure represents a “failure” level, the point below which raw water can no longer be withdrawn to its respective WTP. Below this level, the Model “turns off” raw water withdrawals in that reservoir and instead increases simulated purchases from the MWRA accordingly. The orange line however represents an action or mitigation level, generally equivalent to two feet above the failure level unless otherwise indicated. The action level was used in several alternative reservoir management scenarios discussed later in this paper.

Figure 4: Business-as-Usual Scenario, Simulated Reservoir Capacity in Winona Pond.

Figure 5: Business-as-Usual Scenario, Simulated Reservoir Capacity in Suntaug Lake.

Figure 6: Business-as-Usual Scenario, Simulated Reservoir Capacity in Spring Pond.

As shown in the simulation results depicted above:

  • Winona Pond is simulated to refill every winter-spring season with the help of withdrawals from the Ipswich River via Middle Basin. Winona Pond draws down to between 50% and 35% of its capacity nearly every year over the 73-year simulation period except for one year during the 1960s drought, the regional drought of record, when the reservoir dipped as low as 26.4% of its capacity, very nearly reaching its failure level.
  • Like Winona Pond, Suntaug Lake also refills every winter-spring refill season, generally drawing down during the summer-fall season to between 65% and 55% of capacity. Occasionally, annual minimums will remain above 65% during exceptionally wet years as it did during two years in the 1950s and again in 2021. Suntaug Lake was also simulated to draw down to its failure level near 52% capacity on seven occasions. These failure periods were all quite brief, with the simulated cessation of drinking water withdrawals generally returning the reservoir to a functioning state within weeks. In total, Suntaug Lake was simulated to be in failure less than 0.3% of the time over the 73-year period.
  • In contrast, Spring Pond enters a failure state far more regularly and does not reliably refill every winter-spring season. Spring Pond was simulated to reach its failure level of about 62% of capacity during parts of 46 years over the 73-year continuous simulation. In total, Spring Pond spent approximately 7% of the 73-year simulation period in failure. Model results indicate a strong correlation between Spring Pond not refilling fully over the winter-spring season and reaching a failure state the following summer-fall season. Spring Pond was simulated not to reach full capacity during 25 of the 73 years. As a result, during 23 of those 25 years, Spring Pond failed during the following summer-fall season. In short, Spring Pond is vulnerable to multi-year droughts as it cannot rely on withdrawals from the Ipswich River to replenish it each winter-spring season as Winona Pond and Suntaug Lake can.

Future Climate Model Results

The team used the Model to evaluate how anticipated changes to precipitation and temperature under potential future climate scenarios might impact the reliability and vulnerabilities of Peabody’s reservoir system. Changes to precipitation and average temperature were estimated on a seasonal basis under a 2070 climate (RCP 8.5) based on the updated ResilientMA Viewer generated by Cornell University. Table 4 presents those anticipated changes to precipitation and average temperature.

To approximate a potential 2070 climate, the 1950-2022 precipitation time series used by the Model to estimate the volume of precipitation landing on the reservoirs’ surfaces was modified by the appropriate seasonal percentage. Note that this change did not increase the frequency of precipitation but merely the depth of precipitation that was simulated to fall on wet days.

Similarly, the daily maximum and minimum temperature time series were assumed to increase by a constant value equal to the anticipated seasonal average temperature increases presented above. These increases in daily maximum and minimum temperatures, in turn, resulted in increased evaporation simulated from the modeled reservoirs. The results of this future climate scenario simulation are presented below, in Figures 7, 8, and 9, along with their corresponding Business-as-Usual counterparts.

Figure 7: Business-as-Usual (left) vs. 2070 Climate Scenario (right), Simulated Reservoir Capacity in Winona Pond.

Figure 8: Business-as-Usual (left) vs. 2070 Climate Scenario (right), Simulated Reservoir Capacity in Suntaug Lake.

Figure 9: Business-as-Usual (left) vs. 2070 Climate Scenario (right), Simulated Reservoir Capacity in Spring Pond.

As depicted in these figures, they show no significant changes in reservoir performance resulting from the potential future climate conditions and their associated impact on reservoir recharge rates.

  • Winona Pond is simulated to fail under future climate conditions, whereas it does not under the Business-as-Usual scenario. However, it fails only briefly at the peak of the historic 1960s drought. In fact, it resides in a failure state only 0.006% of the entire 73-year simulation period.
  • Suntaug Lake also experiences a small increase in failure rate, increasing from 0.26% to 0.36%. No new failures are expected to occur; rather the failures that are already simulated to occur increase slightly in duration.
  • Spring Pond, the reservoir most vulnerable to failures, particularly during multi-year droughts, is also expected to be in a failure state more often, increasing from 7.0% to 7.6% of the time.

While not insignificant, these anticipated changes in reservoir reliability suggest that the Peabody reservoir system is not particularly vulnerable to the changes in precipitation and temperature that may occur due to climate change during this century.

Future Demand Model Results

The team also used the Model to evaluate the reservoir system’s resilience against, or vulnerability to, increases in raw water demands. The project team evaluated a 10% increase in raw water withdrawals from Winona Pond, Suntaug Lake, and Spring Pond every day of the year. Figures 10, 11, and 12 compare the results of that simulation to the Business-as-Usual scenario.

Figure 10: Business-as-Usual (left) vs. 10% Increase to Demand Scenario (right), Simulated Reservoir Capacity in Winona Pond.

Figure 11: Business-As-Usual (left) vs. 10% Increase in Demand Scenario (right), Simulated Reservoir Capacity in Suntaug Lake.

Figure 12: Business-as-Usual (left) vs. 10% Increase to Demand Scenario (right), Simulated Reservoir Capacity in Spring Pond.

As Figures 10 through 12 show visually, in contrast to anticipated changes in water levels due to climate change shown in Figures 7 through 9, an increase in raw water demand of only 10% from each reservoir would have significant impacts on system resilience. Winona Pond and Suntaug Lake would continue to refill every winter given the resilience of the Ipswich River withdrawals, but both reservoirs would fail significantly more often.

  • Winona Pond never fails under Business-as-Usual conditions. It is simulated to fail during parts of 16 years during the 73-year simulation, residing in a state of failure 0.4% of the time.
  • Similarly, Suntaug Lake is simulated to experience a significant increase in both the frequency and duration of failure, with the Model indicating it would be in a state of failure about 1.5% of the time compared to less than 0.3% of the time under Business-as-Usual conditions.
  • Spring Pond would also experience significantly more and longer failures, failing most years (10 years in a row, in fact) during the 1960s and 70s. Spring Pond’s failure rate is expected to increase from 7.0% under Business-as-Usual conditions to 11.9% if raw water demands were increased by only 10%. These results suggest that all of the reservoirs in the city’s water supply system are far more vulnerable to increased demand than to potential changes in precipitation, temperature, or Ipswich River flow regimes and minimum flow permit requirements.

Summary of System Vulnerabilities

The results obtained from the Model, some of which are presented above, highlight several general findings regarding the city’s reservoir system vulnerability:

  • The system is vulnerable to increases in raw water demand; however, it is not particularly vulnerable to climate change (i.e., changes in precipitation and temperature) or to changes in Ipswich River flow requirements.
  • With regard to specific reservoirs, Winona Pond has been shown to be particularly resilient, with failure not indicated for any year from 1950-2022 under baseline or future climate conditions.
  • Suntaug Lake is also relatively resilient, failing only once during the entire 72-year period.
  • In contrast, Spring Pond is particularly vulnerable, failing many years over the same time period, particularly during multi-year droughts.

Identifying these vulnerabilities informed the development of reservoir management strategies that sought to improve the reservoir system’s resiliency.


The project team used the Model, in conjunction with other qualitative and quantitative analyses and feedback from a series of workshops with city staff, the Ipswich River Watershed Association, and the Metropolitan Area Planning Commission, to evaluate dozens of alternative scenarios to a Business-as-Usual approach to the city’s raw water supply system. Some scenarios consisted of one or more capital improvement projects, like modifications to intake structures or upsizing of pumps. Other scenarios consisted of changes to reservoir and WTP operations, like adjustments to the raw water blend from Suntaug and Springs Pond waters at Walsh WTP or changes to Middle Basin controls to prioritize refilling Suntaug or Winona to a greater degree. Others include increasing MWRA purchases as reservoir levels dipped below an “action” level. Still other scenarios looked at the impacts of long-term water conservation measures or outdoor watering restrictions.

Analyzing the results of each model simulation, with a particular eye towards reservoir resilience (failure rate), helps identify the most effective actions the city can take to confirm its long-term water supply resilience. Below are the project team’s recommendations for the strategies the city should pursue, largely based on model results and input from city representatives.

Recommendation #1:  Enact stricter outdoor watering restrictions.

Perhaps the most substantial way the city can improve its reservoir resilience is top enact stricter outdoor watering restrictions regularly and for an extended season beyond just the highest demand months of July, August, and September. If the city seeks to obtain demand reductions above 5% and potentially up to 20%, well-enforced outdoor watering restrictions are critical. In lieu of a complete outdoor watering ban, the most efficient restrictions are typically those that limit outdoor watering to 2-3 days a week and to certain hours during the day, such as early morning or late evening. Outdoor watering restrictions would place less of a burden on the city’s reservoirs while also limiting the city’s high demand “peak” reliance on the MWRA. Since the city is not allowed to withdraw from the Ipswich River from June through November, that period offers the greatest potential for implementing water restrictions.

As outdoor watering restrictions effectively improve reservoir resilience and come at no capital cost, the project team recommend that they be incorporated in combination with any of the following scenarios that the city decides to pursue.

Recommendation #2:  Pursue lowering the Suntaug Lake and Spring Pond intake structures.

Throughout discussions with city staff, it was continuously stressed that although Peabody is reliant upon the MWRA as a supplemental source year-round, it only desires to purchase water from the MWRA when absolutely necessary. Therefore, while the scenarios that simulate increasing MWRA purchases year-round drastically reduce reservoir failure rates, they are less desirable and unlikely to gain traction. In contrast, throttling back the city’s own reservoir withdrawals and increasing MWRA purchases as the reservoirs approach failure level is an extension of how the city currently operates, albeit in a more proactive way.

Typically, the city recognizes its need for additional MWRA supply based on demand trends and real-time reservoir levels, but by its nature this is reactive. The project team recommend the city set specific reservoir levels (e.g., four feet above failure) at which they increase MWRA purchases by set percentages (e.g., 25 or 50%) to more proactively, significantly, and confidently improve its reservoir resilience.

The team recommend the city pursue extension (and possible lowering) of the Suntaug Lake intake alongside lowering the Spring Pond intake. As discussed, there are hundreds of millions of gallons of additional capacity in each of these reservoirs, some of which could be utilized. Even in its current state, the Suntaug Lake intake needs to be extended to allow it to be fully operational; otherwise, the failure level is higher than it should be given its intake elevation. Lowering the Spring Pond intake has the greatest contribution to improving reservoir resilience, particularly because in the Business-as-Usual scenario, Spring Pond fails the most often. Lowering the Suntaug Lake intake would allow the city to modify how water is shared between and withdrawn from Suntaug Lake and Spring Pond to great effect.

Recommendation #3:  Consider adjusting the blending at the Walsh WTP seasonally.

Finally, the team recommend the city consider adjusting the blending at the Walsh WTP seasonally to withdraw more from Suntaug Lake. Adjustments to the blending ratio of these sources would be highly effective if done along with lowering the intake in Suntaug Lake. Even with the current infrastructure, reservoir resilience could be significantly improved under some conditions if the blending ratio were adjusted to withdraw more from Suntaug Lake and less from Spring Pond during the fall season. Model simulations indicated that Spring Pond, under Business-as-Usual conditions, spends more time in a failure state than Winona Pond or Suntaug Lake. Therefore, if the city were to withdraw more from Suntaug Lake during the fall season when Spring Pond is most likely to fail and Suntaug Lake has more of a buffer, it could have a beneficial impact on reservoir resilience. This exact scenario was incorporated into the oneWater portal to support the city’s decision-making on when and under what reservoir conditions various scenarios might be most beneficial.

Figure 13: Schematic of the different modules available through the oneWater web application. This project included the Watershed & Aquifer module and predictive Reservoir Health module.

Figure 14: oneWater Home screen showing the locations of the City’s boundaries, its reservoirs, and publicly available gage locations.

In order to allow city staff to continue to use the Model to evaluate several different reservoir management strategies in real time based on current reservoir and watershed conditions, the project team incorporated the Model into an innovative web application known as “oneWater,” depicted in Figure 13. The app allows the user to run the model on-demand by inputting reservoir levels with a corresponding date and, if desired, anticipated short-term withdrawal rates. The portal is hosted on Microsoft Azure and utilizes Azure compute services to perform the model runs. The compute services auto-scale to meet the demand of running the model and keep overall costs to a minimum.

The Home screen, shown in Figure 14, features an interactive map pinpointing essential project locations such as reservoirs, weather stations, groundwater monitoring stations, and stream gauge stations. This map offers a geographical context and helps users visually associate data with the locations from which they are derived.

The Watershed Data screen, shown in Figure 15, displays detailed graphs and data from local NWS stations, USGS groundwater monitoring stations, and USGS stream gauge stations that were pinpointed on the Home screen. This real-time data visualization helps users understand trends, monitor changes, and make data-driven decisions.

Figure 15: oneWater Watershed Data screen, consolidating publicly available data from multiple locations in one location in a customized, easy-to-understand format.

The Start New Model Run screen, shown in Figure 16, is where users can input reservoir levels with a corresponding date. These user-input reservoir levels allow the user to run the Model based on present conditions. Alternatively, the user can opt to run the Model using the default reservoir levels for the selected date, which provides the average reservoir level for that date from 2008-2022 historical data.

After running the Model for seven different reservoir management scenarios, users can navigate to the Scenario Summary screen, which presents the results in a comprehensive manner. This screen compares the results of the model run for a "Business-as-Usual scenario" along with six other modeled scenarios. The following six scenarios were identified as the most effective in improving reservoir resilience and valuable for implementation into the oneWater portal. Note that these scenarios are shown numerically but are not in priority order.

  1. Demand Increases: Increase demand 10% year-round.
  2. Increase MWRA Purchases: Increase purchase of water from MWRA by 10% year-round.
  3. Partial Watering Restrictions: Enact partial watering restrictions from June through November, reducing demand by 5%.
  4. Full Watering Restrictions: Enact full watering restrictions from June through November with mandatory enforcement, reducing demand by 15%.
  5. Full Restrictions and Throttle Reservoir Use: Enact full watering restrictions from June through November and throttle reservoir use 25% when reservoirs are within four feet of failure.
  6. Spring Pond Fall Restriction: Change the Suntaug Lake/Spring Pond blend ratio at the Walsh WTP to 85:15 from September through November.

The comparisons are presented through charts, graphs, and statistical tables, giving users the insight they need to make informed decisions or recommendations regarding water resource management.

As shown in Figure 17, outputs summarized on this page include the chance of each of Winona Pond, Suntaug Lake, and Spring Pond failing over the next 12 months under a Business-as-Usual approach or one of the six alternative management scenarios, as well as the estimated change in annual revenue associated with each of those six scenarios.

Figure 17: oneWater Scenario Summary screen showing simulated reservoir failure likelihood results for the three terminal reservoirs under a Business-as-Usual approach and six alternative management scenarios, as well as the change in annual revenue associated with each of those six scenarios.

Outputs also include the likelihood that each of the reservoirs will fail over increasingly distant time horizons (e.g., one month, two months, 12 months) and the likely range of the duration of any reservoir failures under a Business-as-Usual approach or any of the six alternatives (see Figure 18, where red means more likely (greater than 67%), orange means likely (33-66%), yellow means less likely (1-33%), and green means unlikely (0% of model simulations).

Figure 18: oneWater Scenario Summary screen showing the likelihood of reservoir failure at 1-month time horizons up to one year and the likely duration of any failures for a Business-as-Usual approach and each of the six alternative management scenarios.

More detailed output is also available for each of the three terminal reservoirs, as shown in Figure 19, in which simulated model results for each of 72 randomized years is compared against a user-selected recent historical year, which was particularly dry or wet.

Figure 19: oneWater Scenario Summary screen showing predicted Spring Pond reservoir levels over the next 12 months for each of 72 randomized model simulations for a Business-as-Usual approach, as compared to a user-selected recent historical drought year (2016)


The City of Peabody’s drinking water supply is comprised of a complex interconnected system of seven waterbodies, three terminal reservoirs, and two water treatment plants, combined with an interconnection to a large regional water system. In light of a history of recent reservoir failures or near failures, the city saw a need to identify appropriate management strategies, changes in operations, and useful capital improvement projects in a defensible way. These strategies would need to not only take into consideration the ever-increasing demand on public water systems, but also the threats to water supply resiliency represented by climate change.

The result is an Integrated Reservoir Model, combined with a web-based guidance tool, that the city can continue to use for years to come that reflects real-time reservoir and watershed conditions. The city can use the tool to understand the likelihood of reservoir failure over the next 12 months, how soon failure might occur, and how long those failures might last under a Business-as-Usual approach.

The city can also use the tool to understand how six different reservoir management strategies might help them avoid or reduce the risk of reservoir failure and how much each of those six strategies will cost the city in terms of lost revenue. Additionally, the oneWater-based model provides objective, science-based information and public outreach graphics when difficult decisions, like water restrictions or bans and increased MWRA purchases, need to be made to protect the long-term viability of the water supply.

Andrew Walker, PH, CFM
Weston & Sampson
Portsmouth, NH 03801

Michael Kastanotis
Weston & Sampson
Reading, MA 01867

Cassandra Albrecht, EIT
Weston & Sampson
Reading, MA 01867

William Paulitz, PE
Peabody Department of Public Services

The authors would also like to thank Davis Scribner and Sandra Howland with the City of Peabody for their contributions to this project, without which this article would not be possible.

Published in NEWWA Journal, March 2024.

Posted in News, Publications and tagged , .