Draft proposal. Building an Intelligent Water Information System Basin-Scale Pilot. Prepared by. Steven Glaser 1 and Roger Bales 2

Transcription

1 Draft proposal Building an Intelligent Water Information System Basin-Scale Pilot Prepared by Steven Glaser 1 and Roger Bales 2 University of California Center for information Technology Research in the Interest of Society (CITRIS) January Professor of Engineering, UC Berkeley (glaser@berkeley.edu, ) 2 Professor of Engineering, and Director of the Sierra Nevada Research Institute, UC Merced (rbales@ucmerced.edu; )

2 Building an Intelligent Water Information System Basin-Scale Prototype Issue. With better management, California s existing water supply could go further to meeting the needs of the state s urban and agricultural uses. Currently, almost all operations of California s hydropower and water-supply water reservoirs are controlled and regulated using forecasts based on historical snowpack and runoff data. In the face of global climate change, these forecasts will becoming increasingly inadequate to manage water resources in optimal and equitable ways. We propose implementing an intelligent water infrastructure system that leverages the newest frontiers of information technology. The immediate need is for further development of currently operating basin-scale prototype information systems. As the next step toward broader implementation a partnership involving the Department of Water Resources, the University of California and key stakeholders is required to evaluate, refine and implement this technology. Technology opportunities. Relatively small investments in satellite remote sensing data, groundbased measurements and cyber-infrastructure have the potential to dramatically improve hydrologic information, reduce uncertainty in water availability, and improve water-supply reliability. These improvements will inform decision making by offering more accurate and timely information for predicting precipitation, runoff and water conditions across catchments. Satellite and aircraft remote sensing offer the only practical means for basin-wide measurement of snow properties, soil moisture, and other watershed conditions. A strategically deployed ground-observation system compliments satellite measurements and provides continual and accurate estimates of snowpack, soil moisture, vegetation state and energy balance across watersheds. The proposed real-time intelligent water infrastructure system is a cyber-physical system exhibiting both computational and physical elements (Figure 1). The link between these cyber, and physical components is achieved through a real-time sensing backbone. The system management is designed as a network of interacting elements with physical inputs and a coordinated suite of informationdriven outputs, rather than a disparate set of models, control policies, and physical infrastructure. Figure 1. Information system components and interdependencies.

3 Such systems will improve the link between computational and physical elements of water networks, dramatically increasing the adaptability, autonomy, efficiency, functionality, reliability, safety, and usability of these systems. Advances in two areas offer this linkage: a new-generation sensing system and decision support that includes modeling and analysis. Such sensor networks were developed at the Southern Sierra Critical Zone Observatory, involving more than 300 sensors over a 1.5-km range. Basin-scale progress. We are currently installing a basin-scale ground-based sensor network focusing on measurements of snowpack, solar radiation, temperature, rh and soil moisture across the American River basin using $2 million of research funding from the National Science Foundation (lower left component in Figure 1). The basin is being instrumented with more than 1800 sensors spread over 20 key locations (Figure 2). Although this is a research network, it also provides core elements of a full ground-based operational system. A full-scale operational system will require sets of sensors placed at an additional 12 to 25 locations across the basin. Conceptual planning is also taking place for prototype systems in the Feather and upper San Joaquin basins, in cooperation with local stakeholders. DWR previously initiated sensor placements at four locations in the American River basin under ongoing UC investigations in headwater catchments in the basin, which will be integrated into the proposed ground sensor network. The current spatial snow-water equivalent product is based on a very limited set of ground data; and although it shows great potential, is limited by the paucity of representative ground measurements of snow depth and water equivalent in the Sierra Nevada. Beginning pieces of the proposed remote sensing system has been initiated by DWR, which is working with researchers at the Jet Propulsion Laboratory (JPL) to start providing spatial snow products from the MODIS satellite (upper left component in Figure 1). These prototype valueadded data include estimates of both snow-covered area and snow-water equivalent (Figure 3). These SWE products will continue to undergo evaluation and improvement, and even with limited Figure 2. Current and planned sensor locations for American River basin research network.

4 Figure 3. Snow water equivalent estimates for two dates in spring 2012, based on MODIS data and ground- based energy balance calculations. Images prepared by Noah Molotch, JPL and U. Colorado. ground data provide an improvement over previous operational data. DWR is already making an initial set of satellite-based and modeled snow products available. For modeling-driven planning products (lower right component on Figure 1), both DWR and selected other water managers have initiated use of PRMS (Precipitation-Runoff Modeling System), a modeling system that has an ability to optimize the utility of the rich spatial datasets emerging from the proposed information system. PRMS and companion models that include a richer treatment of subsurface fluxes have the potential to greatly improve hydrologic forecasting for the Sierra Nevada and other areas in California, when run with accurate, spatially dense input data. Next steps. Effort is needed in four areas represented on Figure 1, identified as follows and further described by the work plan following. 1. Establishing a spatially and temporally dense wireless network of sensing stations throughout the American and other river basins, as pilots of a broader system. These networks will provide spatially representative measurements of precipitation, snow depth and water equivalent, temperature, relative humidity, solar radiation, soil moisture and temperature, matric potential, and evapotranspiration.

5 2. Creating automated tools to acquire, correct and integrate satellite and LIDAR imaging data to better determine snow presence and distribution, snowpack water content, and more accurately predict runoff. 3. Developing a data-integration and information-management toolbox to synthesize data from remote sensing and wireless-sensor networks involving many thousands of individual sensors into quality-controlled real-time spatial estimates of water- and energy-balance attributes. 4. Producing tools for predictive physical modeling and scenario analysis that incorporate spatial data products and modeling to support more-reliable decision making for operations and assessment. Work plan. The following activities are proposed under the four next steps identified above. 1. Real-time ground-based measurement network. A strategically deployed ground-observation system will complement satellite measurements and provide continuous and accurate estimates of snowpack, soil moisture, vegetation state and energy balance across watersheds. Using new technologies developed by our team, we deliver measurements of hydrologic variables over a multitiered network of wireless sensor arrays, with a granularity of time and space previously unheard of. The wireless sensor network allows placement of thousands of sensors over kilometers of area affordably and reliably. The ground-based measurement network is made up of multiple tiers with a mesh topology, each layer aggregating over a larger area. At the fundamental level sensing takes place at individual extremely low-power wireless sensor stations aggregated to a network manager over a redundant mesh of stations and signal relay stations (Figure 4). In the typical design, data from a local array of sensors nodes distributed over 1-2 km 2 are transmitted to a network manager, allowing data links over multi-km distances The network manager can assemble a metanetwork of other arrays over several-km hops using a point-to-multipoint Ethernet bridge, or communicate to the outside world by GPRS or satellite modem or in-place UHF networks. The data end up in real time at a control point. Figure 4. A cartoon illustrating the concept of a wireless sensor network node (mote). The mote supervises and integrates data from a large palate of sensors and telemeters this information up through a multi- tiered mesh network as readily accessible web feeds. The multi-sensor station arrays are particularly well suited to headwater catchments in forest, grassland and sub-alpine mountain ecosystems where spatial variability in landscape features (elevation, slope, aspect, vegetation, soils) is very high and conventional point

6 measurements fail to provide spatially representative values for components of the water and energy balance. At present our stations measure precipitation, snow depth, temperature, relative humidity, solar radiation, soil moisture and temperature, matric suction, and sap flow, with complete ability to add others. 2. Acquisition and interpretation of remote sensing data. Satellite and aircraft remote sensing provide the only practical means of basin-wide measurement and monitoring of snow properties, vegetation characteristics and other watershed conditions. This program would use data daily from NASA satellites on Sierra Nevada snow-covered area and albedo, and provide value-added products of unprecedented spatial detail and accuracy that are useable on a watershed level. While some satellite and aircraft products are now routinely available, they require further evaluation and processing before they can be used for watershed-scale hydrologic science and applications. We will derive daily spatial estimates of snow water equivalent (SWE) by combining the remotesensing and ground-based data. For example, the ground-based snow data are interpolated to a surface, then total basin SWE is obtained by masking (multiplying) the interpolated SWE product with the MODIS fractional SCA product. Validation of this product will be performed using a SWE reconstruction analysis over the period of the MODIS record. During the snow season, SWE estimates will be updated at least weekly, with estimates provided by sub-basin and elevation band. 3. Data integration and information management. The most critical project component will be the cyber infrastructure that links measurements, data processing, models and users (central box on Figure 1). Distributed sensors and sensor networks will dynamically transmit data directly, providing real-time information for end users. Quality-control procedures and algorithms must be built directly into the system. Current and historical data must be easily accessible. System software must provide flexibility for multiple types of access from user queries to automated and direct links with analysis tools and decision-support systems. The overall system must be designed to guarantee efficient scaling. A coupler will ensure that addition or removal of any of the other services occurs seamlessly, and without disruption the performance of the overall system. This will include, but is not limited to, routing real-time data feeds, automatically registering sensors and clients, forwarding modeling outputs to the control algorithms, and resolving multi-client conflicts. In essence, the coupler is the operating system upon which all of the other components will interact. Development of the coupling system will be based on core operating system and networking theory, and will be expanded to acknowledge the cyberphysical aspect of the system. Such a system demands a coupling component with tight interaction between physical knowledge and computation. Efforts underway within the hydrologic research community provide a template that takes advantage of recent advances. Five key design features of the data and information system are: 1. Distributed sensors and sensor networks dynamically transmit data directly into an information system, thus making data accessible to all users in real time as they are collected.

7 2. Quality assurance and quality control procedures and algorithms are built directly into the information system. 3. Metadata clearly document data at all levels, from raw data to processed, mature products. 4. The system has the flexibility for multiple types of access, from user queries to automated and direct links with analysis tools and decision support systems. 5. Current and historic data from satellite and aircraft observations, ground-based networks and models are easily accessible. 4. Hydrologic modeling. Together the satellite and ground-based data make possible the updating of forecast tools, and routine use of physically based hydrologic models. During snowmelt runoff, SWE and snowmelt runoff estimates would be updated at least weekly with SWE estimates provided by sub-basin and elevation band. The decision-support framework will provide tools to extract and visualize information of interest from the measured and modeled data, to assess uncertainties, and to optimize operations. The second and further application that we are developing is the SCA and SWE products to constrain snowmelt-runoff modeling using models such as PRMS (Precipitation Runoff Modeling System). PRMS is currently being used by DWR, albeit without the benefit of distributed snow and water-balance information. In addition, the PRMS modeling effort can provide an evaluation tool to increase the accuracy of the snow measurement network. This evaluation will be performed using synthetic tests developed from available historical data sets, leading to a cost/benefit analysis of the system. The goal is to enable more spatially explicit modeling of snowmelt, evapotranspiration and runoff, through spatial data cubes that capture landscape and temporal variability (Figure 5). Figure 5. Spatially explicit model domain for hydrologic forecasting.

8 Project plan. This proposal is for CITRIS to implement a pilot system for the American River basin. For the development phase CITRIS is prepared to provide project management and coordination. Part of this effort will involve a transition plan to hand off operations to a third party, with our training and consultation. This project is part of the Intelligent Water Infrastructure for California initiative at the Center for Information Technology Research in the Interest of Society (CITRIS), a research institute created with support from the State, and University of California (UC) in CITRIS has established a multidisciplinary team to provide technical and scientific leadership to address California and the world s increasingly severe water supply problems. The team draws on world-leading expertise at UC Berkeley, UC Davis, UC Merced, and UC Santa Cruz; research that invented and pioneered the essential building blocks to enable intelligent infrastructure, including smart dust ; wireless sensor networks. This proposed project also plans to draw on expertise from UC Santa Barbara, UCLA, and JPL.

9 Appendix 1 - Enabling Technologies Wireless sensor network. Each station is operated by a NeoMote - an industrially hardened, low power, wireless data acquisition controller. Depending on vegetation, wireless transmission range is between 100 to 400 m. The WSN hardware has been aggressively optimized for low power consumption a hopper station with temperature, snow-depth and solar-radiation sensors can operate for three years on a single D-cell battery in mountain conditions. The full sensor stations need a small solar panel and Li-Ion battery to power the many sensors. The station hardware is capable of exciting and reading from up to 48 digital and analog sensors of any sort. The device is built on a 32-bit portable system on a chip (PSoC) allowing for sophisticated computations to be made onboard, and smaller data payloads transmitted. All signal conditioning functions are integrated into the PSoC. All pieces of the PSoC system, and Dust radio transceiver are remotely dynamically reprogrammable so that entire arrays can be reconfigured over radio link. Existing data collection infrastructure, e.g., Campbell data loggers, integrate directly into our WSNs, enhancing both the scientific value of the data by linking to longer-term records, and the longer-term operation by linking to operational sensors. Such sensor networks are in use at our long-standing installation at the Southern California Critical Zone Observatory 3 where the WSN manages more than 300 sensors over a 1.5 km range 4. We are currently installing a prototype of the proposed ground-based measurement network in the American River basin with $2M funding from the National Science Foundation 5. The 4,500 km 2 American River basin is being instrumented with more than 1800 sensors spread over 20 key locations. In several locations data are telemetered out using the cell network of Inmarsat and are available in real time on an iphone. Cyber-infrastructure, We will also develop a middleware (coupling system) layer to keep track of sensors, send out system health reports, real-time data-feeds, and link all components together in a standard fashion. This framework will assure scalability as additional sensors and networks are added to the system. Effective data exchange between components demands a robust interface through which multiple layers of information can be routed. To ensure feasibility of real-time operations, protocols must be developed to facilitate data exchange between sensors, models, and 3 See 4 e.g., 5 e.g.,

10 control algorithms. For example, adding a new sensor to the network must not affect the overall operations of the modeling toolbox, or the decision support. Methods must be developed to dynamically adjust to change since the system. Additionally, the change of any of the other components must not affect the operation of the other services. This will guarantee more robust operations. The overall system must be designed to guarantee efficient scaling. The coupler will ensure that addition or removal of any of the other services occurs seamlessly, and without disruption the performance of the overall system. This will include, but is not limited to, routing real-time data feeds, automatically registering sensors and clients, forwarding modeling outputs to the control algorithms, and resolving multi-client conflicts. In essence, the coupler is the operating system upon which all of the other components will interact. Development of the coupling system will be based on core operating system and networking theory, and will be expanded to acknowledge the cyberphysical aspect of the system. As such, the development of the coupling system demands a tight interaction between physical knowledge and computation. Efforts underway within the hydrologic research community provide a template that takes advantage of recent advances. Five key design features of the data and information system are: 1. Distributed sensors and sensor networks dynamically transmit data directly into an information system, thus making data accessible to all users in real time as they are collected. 2. Quality-assurance and quality-control procedures and algorithms are built directly into the information system. 3. Metadata clearly document data at all levels, from raw data to processed, mature products. 4. The system has the flexibility for multiple types of access, from user queries to automated and direct links with analysis tools and decision support systems. 5. Current and historic data from satellite and aircraft observations, ground-based networks and models are easily accessible.

11 Appendix 2 Project Schedule American River Water Information System year 1 Hold design workshops with American River stakeholders. Refine methodology for siting sensor arrays. Install 10 ground networks and integrate with current system for immediate data feeds. Reassess attributes to be measured, and on which scale; refine as needed. Subset and evaluate current MODIS stream as a corrected value-added product. Develop plan for LiDAR, to provide characterization and ground-truth data. year 2 Complete detailed system design in conjunction with the stakeholders. Document and develop detailed plan for what information needs to be served, when, in what format. Publish a formal Blueprint of the American River water information system. Design a cyber-infrastructure that is scalable across the Sierra Nevada, and elsewhere. Install 15 to 20 wireless sensor networks. Carry out LiDAR flights and data analysis. Design operational tools for blending MODIS, LiDAR and ground data. year 3 Produce and refine tools for blending MODIS and LiDAR data. Develop the information backbone for the blended remote and ground-based sensor network. Initiate the coupling system for real-time operations of a very-large-scale ground-based monitoring system Design tools to enhance the new data streams for use in physically-based models. Form an entity to operate and maintain the system for long-term operation, including scalable features for other watersheds across the Sierra Nevada. Assess prototype operation and provide cost-benefit analysis based on those data. year 4 Perform system assessment and shakedown. Prepare for hand-off of operations to stakeholders. Augment,or relocate, ground networks to fill gaps in information that became apparent during prototype modeling, assessment and decision support.. Deliver completed information integration and control suite, i.e. fully operational sensor network. Offer training of end users and operations staff. Deliver palette of information feeds to the wide variety of stakeholders. year 5 Assess full operation of the system and further refine scalable design. Complete hand off of pilot to operational stakeholders. Smile at a job well done.