Recommendations to Develop Intelligent Watersheds

MAPPING INTELLIGENCE: A CASE STUDY OF THE UPPER COLORADO RIVER BASIN Project team: Carly Hansen*, Vincent Tidwellǂ, Faisal Ashraf*, Brittany Davisǂ, Christopher DeRolph*, Amelia Green ϵ, Natalie Griffiths*, Shih-Chieh Kao*, Sara Larsen ϵ, Kyle Larsonǂ, Maruti Mudunuruǂ, Kyle Onda ϵ, Esther Parish*, Debjani Singh*, Faith Sternlieb ϵ, Sean Turner*, Nathalie Voisinǂ, Andrew Whiteǂ * Oak Ridge National Laboratory ǂ Pacific Northwest National Laboratory ϵ Internet of Water Coalition In the context of watershed intelligence, the role of technology and its networking across the data-todecision spectrum has been explored, as well as the necessary coordination of this information network across disparate agents in watersheds. This section examines how information networks actually operate in a real watershed. Of particular interest are: • what decisions are made in the watershed and by whom? • what information is used to inform each decision? • how is the flow of information coordinated between different agents? • how is information communicated? The third phase of stakeholder engagement endeavored to address these questions by conducting a series of focus-group discussions. The Upper Colorado River Basin (UCRB) served as the site for the case study, where 24 participants representing 16 different entities were involved in 6 virtual sessions. INFORMATION FLOWS IN THE UPPER COLORADO RIVER BASIN With the focus on a single watershed, and centering discussions around how various decisions are made, the primary interest for Phase 3 was to follow the entire process – from gathering information to responding and making decisions – rather than focusing on individual technologies or parts of the process. Three broad observations were clear from this work. First, management of our nation’s watersheds is highly complex. Second, operational decisions are supported by extensive informational flow networks. Third, the lines of communication are largely defined and well-practiced. Below we discuss these insights in more detail in the context of an intelligent watershed. A key step toward interpreting the feedback that was received across the six discussion sessions was the preparation of a simplified map of the informational flows that trace the path from data to decision. The constructed map is given in Figure 1. Note that this is a very high-level view (simplified) of a very complex and multifaceted process. The intent is to capture the basic structure rather than unique details. To explain the flow of information throughout the watershed, consider first the types of decisions being made. Individual decisions are marked by large bold black print (Figure 1). Here the decision structure is drawn from the perspective of a single reservoir operator (e.g., Navajo Reservoir, Flaming Gorge Reservoir) each faced daily with the decision of when and how much water to release. That decision must consider pending decisions by agents that rely on the released water—hydropower operators make decisions concerning the scheduling of power production, water districts schedule deliveries to their customers, while environmental managers schedule flows to meet environmental objectives. Note that all these scheduling decisions are interdependent. Decisions follow a generalized process where target releases are established each day by the reservoir operator, then the dependent water agents post desired schedules that the reservoir operator integrates to formulate a final release schedule. In a similar fashion, each reservoir operator must interact to develop a coordinated schedule of operations across all basin reservoirs (e.g., upstream reservoirs impacting operations of Lake Powell and Mead). Not shown are equally dependent decisions at the local level, e.g., scheduling of a farmer’s irrigation deliveries or diversion by a powerplant. Overall, decisions address many different and competing needs that are all interdependent and build upon one another from the local to basin scale. Now consider the mapped flows of information, the black arrows, feeding each decision (Figure 1). Generally, the flows of information can be classified as policy (green boxes), water supply (blue boxes), or water demand (orange boxes) oriented. Water policy sets the broad operational framework that the system must perform under, e.g., environmental flows, flood control, and water delivery priorities. Water supply defines the basic condition of the watershed, e.g., streamflows, reservoir storage. Water demand represents the water delivery requests by various agents within the watershed, e.g., irrigation, power, industry, recreation. Water policies are established to govern how water is managed, allocated and protected for the benefit of its user community. Policies range in perspective from local, addressing the operations of a single asset, to national, defining coordinated operations over many and varied assets. These policies are established and enforced by a wide range of agencies using a variety of legal instruments, e.g., original authorizing language, environmental impact statements, interstate compacts, treaties, water rights. Policies are operationalized as a set of rules. These rules can take on a variety of forms including rating curves (flood control), instream flows (timing, intensity, duration of particular flows to manage habitat), maximum releases (downstream flood control), water delivery priorities, and others. The complex set of rules governing a particular project are organized into an operational plan. Plans are developed for both normal and extreme (drought and flood) conditions. These policies and operational plans are generally developed in an open and public engagement process. This is the primary vehicle for voices to be heard and influence operations. Other key informational flows relate to the supply of water available to the watershed (blue boxes in Figure 1). This includes information concerning the current state of the system as well as future forecasted conditions that extend out a few hours, days, seasons, and even years. The state of the system includes real-time data on such factors as weather conditions, snowpack, soil moisture, gauged runoff across the basin, water diversions, water quality, and the stored volume of water in reservoirs. These data are largely accessed through public databases maintained by various federal, state, and local agencies (e.g., National Weather Service, NRCS, USGS). Forecast projections endeavor to predict how water supply (i.e., the system state variables noted above) will evolve in the near future. Forecasts are generally issued in probabilistic terms using data and/or physical models of the watershed. For the UCRB, the Colorado Basin River Forecast Center (CBRFC) is the primary source of river forecasting, although local entities will often augment projections with other resources. Water demands (orange boxes in Figure 1) in the form of delivery requests round out the primary informational flows informing water operations. These requests often represent many different users whose demands are largely in competition. Demands include water for hydropower generation, irrigation, municipal/industrial supply, recreation, and environmental management. Requests are issued to the project operator in the form of desired delivery schedules (quantity and time). The operator then balances the demands against the available water supply subject to the policies governing operations to yield a final operations schedule. Note that there is an added layer of governance when it comes to water use. Governance follows a hierarchy that begins with the Colorado River Compact that allocates water among states. The states then allocate water among water cooperatives (e.g., water districts, irrigation districts) that then distribute the water to individual users. Throughout the process, operational decision making is largely assisted by some type of numerical modeling (denoted in red type). This includes operations from the local irrigation district to large Federal projects as well as most planning exercises. In the UCRB, this most often involves the use of Excel and RiverWare-based simulation tools. Similarly, Western Area Power Administration (WAPA) has developed personalized tools to aid in scheduling hydropower generation based on market and grid forecasts. Modeling also aids coordination across the entire river basin. Each UCRB project reservoir schedule is incorporated into the 24-Month Study Model to inform operations planning for the Lower Colorado River Basin; specifically, operations at Lake Powell and Lake Mead. Figure 1: Informational flows within the Upper Colorado River Basin. Primary operational decisions are in bold black print; arrows map out primary information flows; boxes represent key decision resources colored according to whether they pertain to policy (green), water supply (light blue), or water demand (dark blue). GAPS From the discussion above, it is clear that lines of communication are largely defined and well-practiced in the UCRB. However, this is not to say that there is no room for improvement. Below, gaps in the informational flow network are identified. These gaps are organized broadly according to technology, digital coordination, and digital communication. Several opportunities for technological improvement were voiced during the discussions. The clear message was that particular benefit would be realized by improved measurement and forecasting of water supply. Specifically, noted were the need to improve the measurement of snowpack and modeling of snow melt dynamics. Closely related was the need for a higher density of stream gauging networks along with improving and standardizing gauge calibration. Similarly, there was a need for expanding the gauging of water diversions and return flows. Many of these same needs were expressed in earlier phases of our stakeholder engagement. Discussions included representation of both basin resource managers (practitioners) and research scientists/engineers. During these discussions, evidence of clear and defined avenues for the diffusion of new technology from research to practice were lacking. While there were numerous examples of practitioners participating in research projects, there was little opportunity for technology transfer to occur during the limited timeframe of a standard research project. Other challenges included the lack of broad awareness of new technologies; sufficient head-to-head comparisons between new and existing technologies to build practitioner confidence; staff training and familiarity with the new technology; and funding to implement new technologies. The discussions revealed considerable coordination across the diverse agents managing the water resources of the UCRB. Also clear is that this coordination follows well defined channels. While the network is well practiced, there are opportunities to improve efficiencies. Most evident was the fact that each resource manager largely “recreates the wheel” when it comes to developing their decision-making platform. Specifically, each agent accesses the same data from the same sources and then takes steps to sanitize, rectify, format, and automate the process—in each case there is the risk of doing things in slightly different ways. In some cases, data are being shared across this network; however, there is no established protocol or database to capture this information and make it broadly accessible outside the network. There is also a proliferation of models in the basin to aid in operational decision making. Given that many different decisions are being made at many different scales, this abundance of models is expected. Nevertheless, adoption of a standardized hydrofabric (a dataset containing a network of connected representations of rivers, lakes, and catchments) and model workflows would help with modeling errors and improve operational efficiency. Such steps would also provide opportunity for the benchmarking of models across the simulation spectrum. Over this complex water-management network, the principal lines of communication are on a one-to-one basis by way of phone and email, this includes critical real-time data, proposed operational schedules, water requests, and others. Lacking is a centralized clearing house where these data and requests are broadly shared and critical information is archived. While there is the clear flow of information to those making day-to-day operational decisions, there is a gap in that decision being communicated back to the network or the public; that is how much water was released from a reservoir, how much water was diverted into an irrigation system, or how much hydropower was generated. There are also gaps in how information relevant to long-term planning is shared with the public. Additionally, communications across the network are not uniform. Often, particular water users or water-management entities will choose to limit communications within the network due to ongoing litigation, stressed relationships, or just due to general distrust. RECOMMENDATIONS •Create a community of practice between researchers and practitioners to improve lines of communication and facilitate transfer of knowledge/technology to practice. Connect researchers and practitioners •Establish a platform for broadly sharing data, hydrofabrics, workflows, and models supporting UCRB water operations decision making. Central to this is developing strategies for long-term support. Facilitate basin-specific information and tool sharing •Establish a portal for real-time water management. The portal would facilitate real-time communication of critical data, water requests and proposed operational schedules providing a centralized clearinghouse for information to all water managers. This portal would also provide archiving capabilities of past operations decisions. •Create a public facing water data portal that publishes water operation decisions in a broadly accessible format. Tackle real-time water management challenges.