streamlining project delivery through predictive modeling · associate, applied technology...
TRANSCRIPT
Streamlining Project
Delivery Through Predictive
Modeling
Idaho National Laboratory and
GeoEngineers, Inc
ESRI International
User Conference
GeoEngineers – Applied Technology Business Unit
14 offices nationwide
GIS/CAD/IT Experts
Subject matter expertise in earth sciences and engineering
We are focused on: GIS/IT Strategic planning
Application development
System Integration and Implementation
GIS Analysis and Spatial Modeling
CAD/GIS Cartography
CAD Technologies and Integration
From one-off projects to Repeatable Solutions
Patterns of needs
Leveraging technology frameworks (Esri, Equis, etc)
Responsible Engagements
Consistent approach to our Services
Keeping the best interests of our clients in mind
Stop wasting everyone’s time
GeoEngineers Services
Applied Technology - Solutions
GeoPredictWatershed Prioritization
Environmental Data Viewer
Mobile
Avian
Protection
Idaho National Laboratory (INL) Overview
INL is a US Department of Energy (US DOE) facility
History of reactor development, fuel reprocessing and nuclear research
Yearly monitoring requirements for the radionuclide Cs-137
Cs-137 is a fission product distributed in/on site soils
GeoEngineers:
Joanne Markert –Project Manager
Tonya Kauhi – GeoProcessing Lead
Stan Miller - GeoStatistics
Rob Smith – Technical Architect
Mike Mills - Programmer
Shawna Heilman – Installation
Gene Lohrmeyer – GeoProcessing
Programmer
GeoPredict – Team
INL:
Chris Oertel - Research and Development
Scientist
John Giles - Research and Development
Scientist
Kara Cafferty - Research and
Development Scientist
Boedre Reynolds - Research and
Development Scientist
Business Problem
Measure at specific locations
Large Area to Monitor
Need to predict Cs-137 at
unmeasured areas
Dynamic Contamination
Patterns (wind, activities at
the site)
With thousands of acres, how can the monitoring be optimized
and defensible?
Data Business Needs
Use measured data to predict the Cs-137 values at
unmeasured locations across the site
Defendable to highest scientific level
Defend site baseline
Alleviate public concerns when trying to attract new missions
Compare post event release levels to radiological baseline
In emergencies, this data is the INL radiological baseline
condition
Previous INL Modeling Efforts
Ordinary Kriging
Standard Technique provided via
GeoStatistical Analyst
Not Robust Enough
Incorrectly predicted high Cs-137
(ground-truthed areas)
Results showed spatial gaps in error surfaces
Disjunctive Kriging with Declustering
Available via GeoStatistical Analyst
Declustering Method DID NOT reflect field
sampling conditions
Declustering by cell or polygon method was not accurate
Models underestimated variability of predicted values
INL Modeling Efforts – Why a new Approach?
Requirements for Predictive Modeling at INL:
Declustering by nearest neighbor techniques
Prediction surface with low uncertainties
Spatial dependencies across entire site
Environmental Parameters need to be used in predictions
Ability to defend DOE/INL positions regarding siting of new projects
Monitoring methods at or beyond EPA scientific protocols
More efficient and dynamic soil monitoring
Combine data from multiple years for predictions
State-of-the-Art defense for key environmental media
Provide a monitoring baseline necessary prior to any new large onsite projects
GeoPredict: Computational model created to best predict the
probability of something occurring…
GIS (Spatial Relationships)
GeoStatistics (Stan Miller)
Unique approach
to GeoEngineers
GeoPredict - What is it?
Archaeology, Contaminants, Environmental, Landslides, Weapons Caching, Human Terrain,
Earthquakes
Open and Scalable Solution Scientific Peer Reviewed Reusable Adaptable
ESRI GIS Application interface(Task Assistant Manager) Computational engine
GeoEngineers created a solution framework called GeoPredict
that leverages a number of data inputs displayed in a map output
GeoPredict
Parameters
Elevation
Slope Percent
Aspect
Wind
Geology
Study Area (A)
A1
A2
A3
A4
A5
Bayesian Calcs.
(B)
B1
Known Areas
C1
C2
C3
C4
C5
Exposure
Planning
Risk
Defensibility
Project Cost
Adjustable Constraints
GeoPredict - The Framework
GeoPredict – What Does it Look Like?• Very complicated science rolled into easy to use interface
Known Areas
Existing GIS DataPhysical Environmental
Elements
GeoPredict - Knowing the Parameters
•Bayesian:=P[(E1/A)(E2/A)(E3/A)(E4/A)(E5/A)]*P(A)
P[E1*E2*E3*E4*E5]
•Kriging with Exhaustive Secondary Information
Kriging with External Secondary Information
ESI = bayesian calculations based on environmental parameters
Kriging with ESI –
Includes environmental parameters (the external secondary information as Bayesian) known to be part of the contamination patterns (wind, soil, slope, etc.) combined with measured locations
Better predictions with lower errors (more confidence in the results)
What’s next? Additional modules, scalable, etc.
GeoPredict - Model Output
Optimize monitoring locations
Error surface demonstrating accuracy of method
Spatial Information with high probability
Spatial product regenerated with new input data
Exposure to
Risk/
Prioritize
Work Effort
Red indicates the optimal
locations for monitoring
GeoPredict - Model Output
• Managing Client Risk• Due Diligence
• Defensible Data
• Responsible monitoring
• Project Delivery• Efficient planning
• Minimize surprises
GeoPredict- Summary
Risk Management Air Monitoring
Contaminants
Archaeology
Landslides
Weapons Caching
Human Terrain
Homeland Security
GeoPredict- Other Configurations
Joanne Markert
Associate, Applied Technology
GeoEngineers, Inc.
253.831.3217
Chris Oertel
Research and Development Scientist
Idaho National Laboratory
208.533.7122
John Giles
Research and Development Scientist
Idaho National Laboratory
208.533.7088
Thank You- Questions