Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Tools for Assessing Stray Gas Migration: A Case Study in
Pennsylvania
Seth Pelepko, P.G. & Stew Beattie
PADEP: Bureau of Oil and Gas Planning & Program Development
Stray Gas Incidence & Response Forum
July 25, 2012
Cleveland, OH
• Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
• Transitioning from a One-Dimensional Analysis to a Three-Dimensional Analysis: Evaluating Data Trends and Building a 3D Site Conceptual Model
• GIS Applications: Topographic Position Index (TPI), Surface Modeling, and the ArcScene Environment
• Lessons Learned and Improvements Moving Forward
Presentation Outline
Dependent Variable: Predrill Methane for Years Three and Four (n = 307) Add %’s
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Site Size: Background Area: 172 sq. mi. Stray Gas Site: 10.5 sq. mi.
(81.0%)
(9.0%)
(1.0%)
(2.0%)
(6.5%)
(0.5%)
Independent Variables Considered
• Surface Elevation
• Bottomhole Elevation
• Topographic Position Index (TPI)
• Gas Well Distribution
• Gas Show Density (0 to 500 feet and 0 to 1000 feet)
• Depth to Tully
• Longitude
• Latitude
• % Sand (0 to 500 feet and 0 to 1000 feet)
• Sample Date
• Gas well production for first period of reporting
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Surface Elevation and Bottomhole Elevation (n = 307)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Topographic Position Index (TPI) (USGS 10 m DEM) (Weiss, 2001)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Gas Well Distribution (n = 245)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Gas Show Density (0 to 500 feet, n = 89; and 0 to 1000 feet, n = 122)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Depth to Tully Limestone (n = 229)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
N-S Extent of Water Supplies
N-S Extent of Gas Wells
A (S) A’ (N)
meters
met
ers
Depth to Tully Limestone (n = 229)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
A (S) A’ (N)
meters
met
ers
?
?
Longitude and Latitude (n = 307)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
S
W E
N
% Sand (0 to 500, n = 192; and 0 to 1000 feet, n = 192)
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Model Results: Under-Predicting Methane
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
Model Results (Stepwise Regression)
• Most important variables for predicting dissolved methane are TPI (X1), latitude (X2), and bottomhole elevation (X3)
• Regression equation is:
ln(methane concentration + 1) = -130.923 - 1.192(ln(TPI+300) + 3.323(latitude) - 0.001(bottomhole elevation)
• Model was statistically significant and no assumptions were violated
• Goodness-of-fit was less than desirable – only 36% of variance explained
Where do we go from here?
• Consider only confirmed thermogenic methane
• Consideration of other parameters or models (non-linear)
• Improved quality control/consistency for existing parameters
Conclusion:
• If 36% of variance can be explained with limited time and resources, there seems to be potential with this approach
Defining Background in Absence of Pre-Drill Samples: A Multiple Regression Model
• 1D: Simple linear regression model to examine dissolved- and free-phase methane time-series trends
• 2D: Shallow groundwater quality spatial and temporal trends
• 3D: Integrating gas well construction and MIT data with subsurface structure, water quality trends, and the time dimension
Transitioning from a One-Dimensional Analysis to a Three-Dimensional Analysis: Evaluating Data Trends and Building a 3D Site Model
Affordable and Accessible Tools:
• Topographic Position Index (TPI)
• Interpolation
– Natural Neighbor
– Kriging
• 3D Modeling and ArcScene
GIS Applications
Valley vs. Ridges
• Is there a methodology for exploring variability in the concentration of methane as a function of location, i.e., in valleys versus ridges, and how do we classify these zones?
GIS Applications
Topographic Position Index (TPI)
• The topographic position index reflects the difference in elevation between a focal cell and all cells in the neighborhood
• Geospatial tool developed by Andrew Weiss in 2001
GIS Applications
TPI
• Positive values indicate the cell is higher than its average surroundings while negative values mean it is lower
GIS Applications
Defining a Surface
• Interpolation
– Estimating values within the range of available data
• Extrapolation
– Predicting values of locations outside the range of available data
GIS Applications
Interpolation: Natural Neighbor
• Natural Neighbor interpolates a value based on the closest subset of samples and weights them based on their proportionate area to the value being interpolated (Sibson,1981)
GIS Applications
• Kriging is an advanced, statistically-based procedure that estimates spatial variables and predicts surface trends using a technique in which the surrounding measured values are weighted to derive a predicted value for an unmeasured location (ESRI, 2012)
– Semivariogram: allows examination of spatial relationships between measured points
– Provides Cross-Validation: “how well” the model predicted unknown values
GIS Applications
3D Modeling and the ArcScene Environment
• The 3D environment permits visualization of a complex subsurface geological problem, e.g., stray gas migration
• By converting data to or compiling it in a suitable format it becomes easy to integrate surface and subsurface data to get a more complete picture
GIS Applications
3D Modeling and the ArcScene Environment
• ArcScene is a 3D visualization application that allows you to view GIS data in three dimensions (ESRI 2012)
GIS Applications
Lessons Learned and Improvements Moving Forward
A Proactive Approach and Progressive Thinking Matters • Improvements in data quality and management will facilitate the
construction of robust statistical and geological models • Some well-construction and operational practices are believed to have
exacerbated conditions at residential water supplies considered in the case study
• Data trends for a dynamic compound like methane may better be inferred
by examining monthly or quarterly averages • Passage of new well-construction regulations in February 2011 has been
accompanied by a decrease in the number of stray gas incidents • Multiple discussions and partnerships with research agencies, academia,
and industry are leading to improved operations throughout the Commonwealth of Pennsylvania
Thank You – Questions?
Seth Pelepko, P.G., LPG Well Plugging and Subsurface Activities Division 717.772.2199 ([email protected]) Stew Beattie, GIS/Information Specialist Well Plugging and Subsurface Activities Division 717.772.2199 ([email protected])