際際滷

際際滷Share a Scribd company logo

Ad mi floridan-aquiferwls-for-pps

Download as PPT, PDF
John B. Cook, PE, CEO
John B. Cook, PE, CEO

The document describes how Advanced Data Mining used data mining techniques to create an accurate model of an expansive groundwater system in Florida using data from hundreds of wells. Data mining was used to classify wells, assign locations to classes, and create individual models for each class to predict water levels. The resulting groundwater model integrated all class models and could accurately simulate and predict water levels over time. The data mining model was developed much faster than a traditional model and was more accurate at predicting water levels.

1 of 134
Download to read offline
(864) 201-8679, www.advdmi.com Visualize Complexity, Discover Solutions, Shatter Limits
Data Mining anData Mining an Expansive GroundwaterExpansive Groundwater SystemSystem Presents..!Presents..!
Press your Pause key to stop/ restart this presentation at any time. Press your Esc key to end it.
Advanced Data Mining (ADMAdvanced Data Mining (ADMii) has) has developed unique Data Mining technologydeveloped unique Data Mining technology for modeling natural systems. This videofor modeling natural systems. This video demonstrates its application to andemonstrates its application to an expansive groundwater system.expansive groundwater system. Data Mining extracts valuable knowledgeData Mining extracts valuable knowledge from large amounts of data. It employsfrom large amounts of data. It employs advanced methods from several scientificadvanced methods from several scientific disciplines.disciplines.
The groundwaterThe groundwater system of interest issystem of interest is the Upper Floridianthe Upper Floridian Aquifer in theAquifer in the Suwannee River ValleySuwannee River Valley
This system is approximately 100 x 120This system is approximately 100 x 120 miles with a maximum surface elevationmiles with a maximum surface elevation of 220 feet.of 220 feet. The following illustration shows itsThe following illustration shows its topography. Land elevation is indicatedtopography. Land elevation is indicated by the key at left. The path of theby the key at left. The path of the Suwannee River can be readily seenSuwannee River can be readily seen near the center.near the center.
Suwannee River Valley
Ad mi floridan-aquiferwls-for-pps
This groundwater resource isThis groundwater resource is managed by the Suwannee Rivermanaged by the Suwannee River Management District in Live Oak,Management District in Live Oak, Florida.Florida. They maintain a network of severalThey maintain a network of several hundred wells that provide datahundred wells that provide data about the behavior of the aquifer.about the behavior of the aquifer.
The following shows the locations ofThe following shows the locations of wells for which there are significantwells for which there are significant amounts of data.amounts of data. Note that some areas have severalNote that some areas have several wells clustered together and thatwells clustered together and that others have few or none.others have few or none.
Gulf of MexicoGulf of Mexico
Histories for a few wells go back to theHistories for a few wells go back to the 1940s, however, the record prior to1940s, however, the record prior to 1982 is sparse.1982 is sparse. The vertical blue streaks in theThe vertical blue streaks in the following 3D image show the historicalfollowing 3D image show the historical range of individual wells. Together theyrange of individual wells. Together they show the dynamic range of the aquifer.show the dynamic range of the aquifer.
Elevation above Sea Level N W S E Gulf of Mexico
Collectively, these data comprise aCollectively, these data comprise a vast, but unwieldy source ofvast, but unwieldy source of potentially valuable knowledge.potentially valuable knowledge. We researched how Data MiningWe researched how Data Mining could be used to extract knowledgecould be used to extract knowledge about this complex system andabout this complex system and others like it.others like it.
Computer models of groundwaterComputer models of groundwater systems are important tools for learningsystems are important tools for learning how these invaluable resources arehow these invaluable resources are affected by weather, pumping and landaffected by weather, pumping and land development.development. Our goal was to use Data Mining toOur goal was to use Data Mining to create an accurate model of thecreate an accurate model of the aquifers water level.aquifers water level.
The following is a 25 x 30 mileThe following is a 25 x 30 mile detail from near the center of thedetail from near the center of the system. It shows the positions of 22system. It shows the positions of 22 wells and their histories since 1982.wells and their histories since 1982. Note that the two groups of circledNote that the two groups of circled wells clearly behave differently fromwells clearly behave differently from each other.each other.
350000 370000 390000 410000 430000 450000 470000 490000 2360000 2380000 2400000 2420000 2440000 2460000 2480000 2500000 25miles 30 miles SuwanneeRiver
Because the wells exhibited so manyBecause the wells exhibited so many different behaviors, it was necessarydifferent behaviors, it was necessary to group them into classes. Wellsto group them into classes. Wells assigned to a particular class behaveassigned to a particular class behave similarly.similarly. Data MiningData Mining optimallyoptimally determined thedetermined the number of classes and how the wellsnumber of classes and how the wells would be assigned.would be assigned.
The following shows that 12 classesThe following shows that 12 classes were used and how the wells werewere used and how the wells were assigned. The classes are numberedassigned. The classes are numbered 1 to 12.1 to 12. It was surprising how some classesIt was surprising how some classes are distributed over a broad area andare distributed over a broad area and are intermingled with other classes.are intermingled with other classes.
Ad mi floridan-aquiferwls-for-pps
Closer inspection showed that DataCloser inspection showed that Data Mining did indeed optimally assignMining did indeed optimally assign the wells.the wells. The following shows the normalizedThe following shows the normalized histories of wells for two of thehistories of wells for two of the classes.classes. Note the seasonal variability.Note the seasonal variability.
History from April 1982 to October 1998
The next Data Mining task was to assignThe next Data Mining task was to assign aquifer locations to the 12 classes.aquifer locations to the 12 classes. Locations were optimally assignedLocations were optimally assigned based on their topologicalbased on their topological characteristics and proximity to wellscharacteristics and proximity to wells whose classes were known.whose classes were known. Results are shown in the following.Results are shown in the following.
Ad mi floridan-aquiferwls-for-pps
The next Data Mining task was toThe next Data Mining task was to create a water level model for eachcreate a water level model for each class. Every location was assigned toclass. Every location was assigned to a class, and therefore, a model.a class, and therefore, a model. Inputs to each model were theInputs to each model were the characteristics of a location and watercharacteristics of a location and water levels of selected wells. The outputlevels of selected wells. The output was the predicted water level of thewas the predicted water level of the location.location.
The models are very accurate.The models are very accurate. Accuracy can be checked at locationsAccuracy can be checked at locations where there are well histories.where there are well histories. The following compares predictions toThe following compares predictions to actual histories for wells of fouractual histories for wells of four different classes. The water levels aredifferent classes. The water levels are normalized to land surface elevation.normalized to land surface elevation.
History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Actual Prediction Class 1Class 1
History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Class 3Class 3 Actual Prediction
History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Actual Prediction Class 6Class 6
History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Actual Prediction Class 10Class 10
The model of the aquifer is actually aThe model of the aquifer is actually a collection of models, one for each class.collection of models, one for each class. A computer program was created thatA computer program was created that integrates the models, a history database,integrates the models, a history database, and a graphical user interface.and a graphical user interface. The following shows a long termThe following shows a long term simulation of the aquifers water levelsimulation of the aquifers water level generated by the model. Note the colorgenerated by the model. Note the color key at right, and that time is reversed.key at right, and that time is reversed.
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Ad mi floridan-aquiferwls-for-pps
Often multi-dimensional visualizationOften multi-dimensional visualization reveals important information thatreveals important information that would otherwise go unnoticed. ADMwould otherwise go unnoticed. ADMii has world-class capabilities inhas world-class capabilities in advanced visualization technology.advanced visualization technology. The following shows the modelsThe following shows the models prediction of the upper range (ceiling)prediction of the upper range (ceiling) of the aquifer. The vertical scale isof the aquifer. The vertical scale is exaggerated to show details.exaggerated to show details.
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico
Gulf of Mexico Max elevation above sea level ~ 180 feet
The following compares theThe following compares the models prediction of the floormodels prediction of the floor and ceiling of the aquifer.and ceiling of the aquifer.
Ceiling Gulf of Mexico
Floor Gulf of Mexico
Ceiling Gulf of Mexico
Floor Gulf of Mexico
The following shows the predictedThe following shows the predicted aquifer level for the period fromaquifer level for the period from January 1995 to October 1998.January 1995 to October 1998. Note the spatially asynchronousNote the spatially asynchronous motions caused by variability inmotions caused by variability in rainfall and the Suwannee Riversrainfall and the Suwannee Rivers stage.stage.
Date: 01/01/95 Gulf of Mexico
Date: 02/01/95 Gulf of Mexico
Date: 03/01/95 Gulf of Mexico
Date: 04/01/95 Gulf of Mexico
Date: 05/01/95 Gulf of Mexico
Date: 06/01/95 Gulf of Mexico
Date: 07/01/95 Gulf of Mexico
Date: 08/01/95 Gulf of Mexico
Date: 09/01/95 Gulf of Mexico
Date: 10/01/95 Gulf of Mexico
Date: 11/01/95 Gulf of Mexico
Date: 12/01/95 Gulf of Mexico
Date: 01/01/96 Gulf of Mexico
Date: 01/31/96 Gulf of Mexico
Date: 03/01/96 Gulf of Mexico
Date: 03/31/96 Gulf of Mexico
Date: 04/30/96 Gulf of Mexico
Date: 05/30/96 Gulf of Mexico
Date: 06/29/96 Gulf of Mexico
Date: 07/29/96 Gulf of Mexico
Date: 08/28/96 Gulf of Mexico
Date: 10/01/96 Gulf of Mexico
Date: 11/01/96 Gulf of Mexico
Date: 12/01/96 Gulf of Mexico
Date: 01/01/97 Gulf of Mexico
Date: 02/01/97 Gulf of Mexico
Date: 03/01/97 Gulf of Mexico
Date: 04/01/97 Gulf of Mexico
Date: 05/01/97 Gulf of Mexico
Date: 06/01/97 Gulf of Mexico
Date: 07/01/97 Gulf of Mexico
Date: 08/01/97 Gulf of Mexico
Date: 09/01/97 Gulf of Mexico
Date: 10/01/97 Gulf of Mexico
Date: 11/01/97 Gulf of Mexico
Date: 12/01/97 Gulf of Mexico
Date: 01/01/98 Gulf of Mexico
Date: 02/01/98 Gulf of Mexico
Date: 03/01/98 Gulf of Mexico
Date: 04/01/98 Gulf of Mexico
Date: 05/01/98 Gulf of Mexico
Date: 06/01/98 Gulf of Mexico
Date: 07/01/98 Gulf of Mexico
Date: 08/01/98 Gulf of Mexico
Date: 09/01/98 Gulf of Mexico
Date: 10/01/98 Gulf of Mexico
This Data Mining-based model requiredThis Data Mining-based model required about 10 weeks to develop.about 10 weeks to develop. A conventional finite-difference model ofA conventional finite-difference model of the same natural system was developedthe same natural system was developed by a government agency. It took over 3by a government agency. It took over 3 years to complete! It is much lessyears to complete! It is much less accurate at predicting water level.accurate at predicting water level. ConclusionConclusion ss
Data Mining is incredibly powerful forData Mining is incredibly powerful for extracting knowledge about complexextracting knowledge about complex natural systems from databases.natural systems from databases. The models can be more accurateThe models can be more accurate than traditional approaches, andthan traditional approaches, and require much less time to develop.require much less time to develop. ConclusionConclusion ss
(864) 676-9790, info@AdvDataMining.com Visualize Complexity, Discover Solutions, Shatter Limits

More Related Content

Ad mi floridan-aquiferwls-for-pps

  • 1. (864) 201-8679, www.advdmi.com Visualize Complexity, Discover Solutions, Shatter Limits
  • 2. Data Mining anData Mining an Expansive GroundwaterExpansive Groundwater SystemSystem Presents..!Presents..!
  • 3. Press your Pause key to stop/ restart this presentation at any time. Press your Esc key to end it.
  • 4. Advanced Data Mining (ADMAdvanced Data Mining (ADMii) has) has developed unique Data Mining technologydeveloped unique Data Mining technology for modeling natural systems. This videofor modeling natural systems. This video demonstrates its application to andemonstrates its application to an expansive groundwater system.expansive groundwater system. Data Mining extracts valuable knowledgeData Mining extracts valuable knowledge from large amounts of data. It employsfrom large amounts of data. It employs advanced methods from several scientificadvanced methods from several scientific disciplines.disciplines.
  • 5. The groundwaterThe groundwater system of interest issystem of interest is the Upper Floridianthe Upper Floridian Aquifer in theAquifer in the Suwannee River ValleySuwannee River Valley
  • 6. This system is approximately 100 x 120This system is approximately 100 x 120 miles with a maximum surface elevationmiles with a maximum surface elevation of 220 feet.of 220 feet. The following illustration shows itsThe following illustration shows its topography. Land elevation is indicatedtopography. Land elevation is indicated by the key at left. The path of theby the key at left. The path of the Suwannee River can be readily seenSuwannee River can be readily seen near the center.near the center.
  • 9. This groundwater resource isThis groundwater resource is managed by the Suwannee Rivermanaged by the Suwannee River Management District in Live Oak,Management District in Live Oak, Florida.Florida. They maintain a network of severalThey maintain a network of several hundred wells that provide datahundred wells that provide data about the behavior of the aquifer.about the behavior of the aquifer.
  • 10. The following shows the locations ofThe following shows the locations of wells for which there are significantwells for which there are significant amounts of data.amounts of data. Note that some areas have severalNote that some areas have several wells clustered together and thatwells clustered together and that others have few or none.others have few or none.
  • 11. Gulf of MexicoGulf of Mexico
  • 12. Histories for a few wells go back to theHistories for a few wells go back to the 1940s, however, the record prior to1940s, however, the record prior to 1982 is sparse.1982 is sparse. The vertical blue streaks in theThe vertical blue streaks in the following 3D image show the historicalfollowing 3D image show the historical range of individual wells. Together theyrange of individual wells. Together they show the dynamic range of the aquifer.show the dynamic range of the aquifer.
  • 13. Elevation above Sea Level N W S E Gulf of Mexico
  • 14. Collectively, these data comprise aCollectively, these data comprise a vast, but unwieldy source ofvast, but unwieldy source of potentially valuable knowledge.potentially valuable knowledge. We researched how Data MiningWe researched how Data Mining could be used to extract knowledgecould be used to extract knowledge about this complex system andabout this complex system and others like it.others like it.
  • 15. Computer models of groundwaterComputer models of groundwater systems are important tools for learningsystems are important tools for learning how these invaluable resources arehow these invaluable resources are affected by weather, pumping and landaffected by weather, pumping and land development.development. Our goal was to use Data Mining toOur goal was to use Data Mining to create an accurate model of thecreate an accurate model of the aquifers water level.aquifers water level.
  • 16. The following is a 25 x 30 mileThe following is a 25 x 30 mile detail from near the center of thedetail from near the center of the system. It shows the positions of 22system. It shows the positions of 22 wells and their histories since 1982.wells and their histories since 1982. Note that the two groups of circledNote that the two groups of circled wells clearly behave differently fromwells clearly behave differently from each other.each other.
  • 17. 350000 370000 390000 410000 430000 450000 470000 490000 2360000 2380000 2400000 2420000 2440000 2460000 2480000 2500000 25miles 30 miles SuwanneeRiver
  • 18. Because the wells exhibited so manyBecause the wells exhibited so many different behaviors, it was necessarydifferent behaviors, it was necessary to group them into classes. Wellsto group them into classes. Wells assigned to a particular class behaveassigned to a particular class behave similarly.similarly. Data MiningData Mining optimallyoptimally determined thedetermined the number of classes and how the wellsnumber of classes and how the wells would be assigned.would be assigned.
  • 19. The following shows that 12 classesThe following shows that 12 classes were used and how the wells werewere used and how the wells were assigned. The classes are numberedassigned. The classes are numbered 1 to 12.1 to 12. It was surprising how some classesIt was surprising how some classes are distributed over a broad area andare distributed over a broad area and are intermingled with other classes.are intermingled with other classes.
  • 21. Closer inspection showed that DataCloser inspection showed that Data Mining did indeed optimally assignMining did indeed optimally assign the wells.the wells. The following shows the normalizedThe following shows the normalized histories of wells for two of thehistories of wells for two of the classes.classes. Note the seasonal variability.Note the seasonal variability.
  • 22. History from April 1982 to October 1998
  • 23. The next Data Mining task was to assignThe next Data Mining task was to assign aquifer locations to the 12 classes.aquifer locations to the 12 classes. Locations were optimally assignedLocations were optimally assigned based on their topologicalbased on their topological characteristics and proximity to wellscharacteristics and proximity to wells whose classes were known.whose classes were known. Results are shown in the following.Results are shown in the following.
  • 25. The next Data Mining task was toThe next Data Mining task was to create a water level model for eachcreate a water level model for each class. Every location was assigned toclass. Every location was assigned to a class, and therefore, a model.a class, and therefore, a model. Inputs to each model were theInputs to each model were the characteristics of a location and watercharacteristics of a location and water levels of selected wells. The outputlevels of selected wells. The output was the predicted water level of thewas the predicted water level of the location.location.
  • 26. The models are very accurate.The models are very accurate. Accuracy can be checked at locationsAccuracy can be checked at locations where there are well histories.where there are well histories. The following compares predictions toThe following compares predictions to actual histories for wells of fouractual histories for wells of four different classes. The water levels aredifferent classes. The water levels are normalized to land surface elevation.normalized to land surface elevation.
  • 27. History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Actual Prediction Class 1Class 1
  • 28. History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Class 3Class 3 Actual Prediction
  • 29. History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Actual Prediction Class 6Class 6
  • 30. History from April 1982 to October 1998 NormalizedWaterLevelaboveSeaLevel Actual Prediction Class 10Class 10
  • 31. The model of the aquifer is actually aThe model of the aquifer is actually a collection of models, one for each class.collection of models, one for each class. A computer program was created thatA computer program was created that integrates the models, a history database,integrates the models, a history database, and a graphical user interface.and a graphical user interface. The following shows a long termThe following shows a long term simulation of the aquifers water levelsimulation of the aquifers water level generated by the model. Note the colorgenerated by the model. Note the color key at right, and that time is reversed.key at right, and that time is reversed.
  • 68. Often multi-dimensional visualizationOften multi-dimensional visualization reveals important information thatreveals important information that would otherwise go unnoticed. ADMwould otherwise go unnoticed. ADMii has world-class capabilities inhas world-class capabilities in advanced visualization technology.advanced visualization technology. The following shows the modelsThe following shows the models prediction of the upper range (ceiling)prediction of the upper range (ceiling) of the aquifer. The vertical scale isof the aquifer. The vertical scale is exaggerated to show details.exaggerated to show details.
  • 79. Gulf of Mexico Max elevation above sea level ~ 180 feet
  • 80. The following compares theThe following compares the models prediction of the floormodels prediction of the floor and ceiling of the aquifer.and ceiling of the aquifer.
  • 85. The following shows the predictedThe following shows the predicted aquifer level for the period fromaquifer level for the period from January 1995 to October 1998.January 1995 to October 1998. Note the spatially asynchronousNote the spatially asynchronous motions caused by variability inmotions caused by variability in rainfall and the Suwannee Riversrainfall and the Suwannee Rivers stage.stage.
  • 132. This Data Mining-based model requiredThis Data Mining-based model required about 10 weeks to develop.about 10 weeks to develop. A conventional finite-difference model ofA conventional finite-difference model of the same natural system was developedthe same natural system was developed by a government agency. It took over 3by a government agency. It took over 3 years to complete! It is much lessyears to complete! It is much less accurate at predicting water level.accurate at predicting water level. ConclusionConclusion ss
  • 133. Data Mining is incredibly powerful forData Mining is incredibly powerful for extracting knowledge about complexextracting knowledge about complex natural systems from databases.natural systems from databases. The models can be more accurateThe models can be more accurate than traditional approaches, andthan traditional approaches, and require much less time to develop.require much less time to develop. ConclusionConclusion ss
  • 134. (864) 676-9790, info@AdvDataMining.com Visualize Complexity, Discover Solutions, Shatter Limits