Presentation on theme: "Smarter Water Management: A Challenge for Spatio-Temporal Network Databases (Vision Paper) KwangSoo Yang, Shashi Shekhar Jing Dai, Sambit Sahu, and Milind."— Presentation transcript:
1Smarter Water Management: A Challenge for Spatio-Temporal Network Databases (Vision Paper) KwangSoo Yang, Shashi Shekhar Jing Dai, Sambit Sahu, and Milind Naphade Department of Computer Science, University of Minnesota IBM T.J. Watson Research Hawthorne
2The importance of water Water is one of our most important natural resources and water scarcitymay be the most underestimated resource issue facing the world today.Source: World Meteorological Organisation (WMO), Geneva, 1996; Global Environment Outlook 2000 (GEO), UNEP, Earthscan, London, 1999.By 2025 about 3 billion people could face water scarcity due to the climate change, population growth, and increasing demand for water per capita United Nations, The Millennium Project
3Water scarcityDistribution of precipitation is varying in space and time.High rate of evaporation from surface water resourceWater quality- Leakage of saline or contaminated water from the land surface - Water contamination by sewage effluent.Water quality degradation can be a major source of water scarcity.Example) Water PollutionWater scarcity is not only an issue of enough water but also of access to safe water.Developing intelligent water resource management systems is necessary to remedy the problem.
4Current water management Hand grab samplingCrumbling infrastructureLeaks and theft in agriculture areaFew sensors (home water meter and plant meter)Hidden water network (underground water and buried pipelines)Water TheftWater sensorsOne technology to remedy this problem is to implement fully integrated systems which allow monitoring, analyzing, controlling, and optimizing of all aspects of water flows.
5Smarter Water Management IBM Smarter Planet emphasizes smarter water management for planning, developing, distributing, and managing optimal use of limited water resources.Measuring, Monitoring, Modeling, and ManagingSource: IBM Smarter Water ManagementSensingMeteringReal Time Data IntegrationReal Time + Historical dataData Modeling + AnalyticsVisualization+ DecisionsA lot of data are needed to fully understand, model, and predict how water flows around the planet.
6Spatio-Temporal Network (STN) Water network flow is represented and analyzed as spatio-temporal network datasets.Minnesota river network source:Ground water network source:Water pipe network source: WikipediaSpatio-temporal network databases (STNDB) will likely be a key component ofsmarter water management since effectiveness of decision depends on the quality of information
7Challenges for STN of water networks The key issue is the quality of dataset to fully understand, model, and predict water flows in the network.1. STN datasets not fully observable underground natural flows buried pipes may shift2. Assess of STN datasets requires a novel semantics ex) Lagrangian reference frame3. Heterogeneity of real-world STN datasets
8Hidden STN Water Network Tomography Example) Water Cooling System The internal structure and status of the cooling system may not be directly observable due to radioactive emissions from the damaged nuclear reactors.Water Cooling System / Nuclear reactor source:Network tomography is one solution to understand the internal characteristic of the network using end-to-end measurements, without needing the cooperation of internal nodes.water lossflow delay or bottlenecktopology/connectivityGiven end-to-end measurementsinferidentifymonitor
9Hidden STN Y = A X Water Network Tomography Y1 Y2 Y3 Y4 1 1 1 1 X1 X2 111X1X2X3X4=Origin-destination (OD) water network flowY = A XInput: Given nodes X and Y, flows at nodes YOutput: Estimate AObjective: Minimize the errorConstraints: Incorporate the temporal and spatial dependence : Sparse matrix : Tracer dataEx) Cost matrix Flow matrixConnectivity matrixOD matrix contains several components, including origin, destination,and barrier information (e.g., max flow, amount of resistance, speed, direction)N measured data N2 hidden dataHardnessunder-constrainedunder-determined
10STN Non-stationarityWater supply and consumption patterns change over timeWaternetworksupplydemandFlow distribution in water networks is not i.i.d.Time varying factorsEx) Time varying water ConsumptionClimate changeEventGeographyDroughtContaminationPopulationTemperatureLeakageFactories/FarmsPrecipitationDisasterPools
11STN Non-stationaritySpatial location + network connectivity + time-varying propertyDatasets grows massively while the density of the datasets becomes sparse.Example) Water leakage detection systemHistorical and real-time datasetsInappropriate pressure levels and flowsHigh risk areas in networksBurst/corroded pipesWater networks ex) sink, branch points, and water mainsMonitoringGeographical information ex) home address and elevationDetecting anomaliesTemporal information ex) hot moments and weather change
12Complexity of STN datasets 12345678910111213141516InflowoutflowEx) earthquake eventPipes can be damaged which reduces the flow ratesInspections of individual network components such as buried pipes are often impractical due to exceedingly large costs and time.Monitoring dataNetwork flow Pressure Transmission timeOD matrixseismogenic ruptureCost MatrixFlow MatrixStructure MatrixDetect the water leakage Recover the damaged pipe lineEvent dataSpatial LocationTemporal propertyThe challenge here is that the anomaly detection algorithm would be intractable due to the size of the spatio-temporal network datasets.
13Lagrangian reference frame 2012345678TimeWater quality index910Measurement methodsWater Quality Index : (bad) – 0 (good)Eulerian : stationary pointLagrangian : moving pointMoving sensorTime1234Water quality index20304040Environmental Forensics: Where did contaminant come from ?: What are hotspots and hot moments ?Introduction to Environmental Forensics Brian L. Murphy, Robert D. Morrison2012345678TimeWater quality index40910
14Lagrangian reference frame Access of STN datasets requires a Lagrangian frame of reference which coordinates STN datasets with STN connectivity.12345678Time91012345678Time9091012345678Time9091012345678Time910(Source:
15Lagrangian reference frame Need new STN database systemsExample) A moving fluid (ABD)STN database systemsSnapshot modelData typesStorage modelsAccess MethodsIndexesTime expanded graphQueries
16General frameworks to analyze STN datasets STN models for complex real-world networksNovel network analysis modelsMathematical approach - Real-world network model is unknown or too complex to be mathematically described.Heterogeneous multi-modal networksExample) Nokia water supply contaminationWater supply networkGround water networkRiver NetworkSource:Sewerage networkSTN data integration problem - Heterogeneous multi-modal networks - Time-varying properties - Correlation propertiesPopulation map
17ConclusionWater resource management is one of most important part for our survival.We need a Spatio-temporal network databases and analysis tools to monitoring, analyzing, controlling, and optimizing of all aspects of water flows.We poses 3 main challenges to handle spatio-temporal network datasets for water flows.
18AcknowledgementThis work was supported by NSF and USDOD.
19References1. Water: How can everyone have sufficient clean water without conflict?(2010), United Nations, Retrieved Mar 17, 2011.2. Smarter planet, Wikipedia, Retrieved Mar 17, 2011.3. Water Management, Wikipedia, Retrieved Mar 17, 2011.4. Let’s Build a Smarter Planet: Smarter Water Management, Dr. Cameron Brooks, Sep 22, 2010, IBM,5. IBM Smarter Water Keynote(2010), IBM,6. Water supply network, Wikipedia, Retrieved Mar 17, 2011.7. Nokia water supply contamination, Wikipedia, Retrieved May 30, 2011.8. Batchelor, G.: An introduction to fluid dynamics. Cambridge Univ Pr (2000)9. Chartres, C., Varma, S.: Out of Water: From Abundance to Scarcity and How to Solve the World’s Water Problems. Ft Pr (2010)10. Chen, F., et al.: Activity analysis based on low sample rate smart meters In: Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining(to appear). ACM (2011)11. Dai, J., Chen, F., Sahu, S., Naphade, M.: Regional behavior change detection via local spatial scan In: Proceedings of the 18th SIGSPATIAL International Conference on Advances in Geographic Information Systems. pp. 490–493. ACM (2010)12. Marien, M.: Jc glenn, tj gordon and e. florescu, 2010 state of the future, the millennium project, washington, Futures (2010)13. Molden, D.: Water for food, water for life: a comprehensive assessment of water management in agriculture. Earthscan/James & James (2007)14. Perry, W.: Grand challenges for engineering. Engineering (2008)15. Vardi, Y.: Network Tomography: Estimating Source-Destination Traffic Intensities from Link Data Journal of the American Statistical Association 91(433) (1996)
20Source: International Water Management Institute
21Spatio-Temporal Network (STN) Water network flow is represented and analyzed as spatio-temporal network datasets.Minnesota river network source:Ground water network source:Water pipe network source: WikipediaSpatio-temporal network databases (STNDB) will likely be a key component ofsmarter water management since effectiveness of decision depends on the quality of information
22A(t) 1. Hidden STN Time varying water network tomography Issues : X1Y3t1t2t3t4TimeIssues :Time varying water flowsDependency of each nodesUneven distribution of flowsData quality: missing, uncertain, unstable sample rate323Time dependent network flowA(t)Computationally expensiveX1(t)Y3(t)How can we estimate network topology & properties ?(?)
23Water Pipelines closed open open closed Flow direction Pressure Flow rate