To balance these competing objectives and provide a robust probabilistic foundation for reservoir operations decision-making, the New York City Department of Environmental Protection (DEP) recently completed initial development of an Operations Support Tool (OST), a state-of-the-art decision support system to provide computational and predictive support for water supply operations and planning. This paper describes the technical structure of OST, including the underlying water supply and water quality models, data sources and database management, types and sources of reservoir inflow forecasts used by OST, and functionalities required to meet the needs of a diverse group of end users.
Key features of interest that distinguish OST from other reservoir operations simulation platforms include: the ability to run both long-term simulations and short-term probabilistic simulations on the same model platform; automated data acquisition and data quality control for a wide variety of near-real-time data sources; application of ensemble reservoir inflow forecasts to support look-ahead operational simulations; and integration of water supply and water quality models to account for the feedback between supply and quality objectives.
OST allows users to execute two types of simulations on the same model platform. Conventional long-term planning simulations consist of running the system model over an extended historical record (e.g. 80+ years) that covers a broad range of hydrologic conditions. These simulations are used to develop and test reservoir operating rules, analyze alternative release policies, or evaluate system infrastructure modifications. This mode is also used to evaluate climate change scenarios, with inflows derived from downscaled GCM data.
Short-term operational or position analysis (PA) simulations are used to provide near-term operational guidance. Operational simulations consist of multiple (e.g. 80+) short (e.g. one year) simulations, all starting from the same initial storage, hydrologic, and water quality conditions. Typically, the starting conditions for an operational run represent those for the current day, and ensemble reservoir inflow forecast traces are used to drive the model for the duration of the simulation period.
The result of these simulations is a distribution of potential future system states based on system operating rules and the range of input ensemble inflow predictions. DEP managers analyze the output distributions and make operations decisions using risk-based metrics such as probability of refill or the likelihood of a water quality event. Results can be post-processed within the OST interface to compare the results of alternative operating decisions. Use of the same model for both planning and operations simulations ensures that both applications are based on a consistent set of system operating rules. This approach also eliminates the need to maintain two separate code bases and model platforms.
For operational simulations, the OST data management system acquires and processes a broad array of near-real-time data from DEP internal sources, including SCADA system operations data on reservoir elevations and flows, water quality monitoring data at critical system locations, automated in-stream water quality measurements at key tributaries, automated in-reservoir water quality depth profiles, and measurements from a network of meteorological stations and snowpack monitoring sites. OST also acquires streamflow and reservoir elevation data from USGS, and ensemble inflow forecasts from the National Weather Service (NWS). Incoming data passes through an automated flagging process to check for nulls and out of range values. Null values are filled based on station-specific rules, and all data is presented to operators for provisional approval prior to use as model input. NRT data is used to initialize the model with current reservoir storage and water quality states, and to support the generation of ensemble inflow forecasts.
OST allows the user to drive operational simulations with two types of ensemble reservoir inflow forecasts. Statistical forecasts are based on historical inflows that are conditioned on antecedent hydrology. The statistical algorithm is relatively simple and versatile and is useful for longer-term projections, but lacks short-term skill critical for water quality and spill management. To improve short-term skill, OST will rely on meteorologically-based ensemble hydrologic forecasts provided by the National Weather Service (NWS). A post-processor that runs within the OST framework will provide bias correction for the ensemble streamflow predictions.
The underlying reservoir operations simulation model is based on the OASIS platform (HydroLogics). The OASIS model is linked with mechanistic reservoir water quality models (CE-QUAL-W2) for four reservoirs in DEP's Catskill system, where periodic major storm events can increase turbidity levels and may lead to alum treatment to protect the quality of NYC's unfiltered drinking water supply. OST is a key component of DEP's strategy for managing reservoir diversions and releases to mitigate the impacts of major storm events. The OASIS and W2 models exchange information on a daily timestep, thus explicitly accounting for the feedback that exists between reservoir water quality and reservoir operations. OST is used by system operators to support the timing of reservoir release and diversion decisions during turbidity events, estimate the potential duration of adverse water quality conditions, and support decision-making regarding alum treatment.
Additional OST applications to date have included routine short-term operational projections to support release decisions, facility outage planning and emergency management, development of new operating rules and release policies, long-term water supply/infrastructure planning, and climate change adaptation planning.
The structure and capabilities of OST are expected to be a useful template for drinking water utilities and water system managers seeking to balance competing objectives in the context of both near-term operations and long-term planning.