Honoring Uncertainty in Mapping and Interpreting Large Volumes of Digital
Spatial Data*
By
John D. Grace1
Search and Discovery Article #40197 (2006)
Posted July 3, 2006
*Oral presentation at AAPG Annual Convention, Houston, Texas, April 9-12, 2006
Click to view presentation in PDF format (0.7 mb).
1Earth Science Associates, Long Beach, CA ([email protected])
Abstract
The development of
very large digital
collections of spatial oil and gas data has permitted the
easy extraction of information on individual variables (e.g., depth to top of a
unit) or spatial interaction between variables (e.g., sediment accumulation rate
and pore pressure). Yet improving access to
digital
data and the ease of mapping
software have often created a confidence in the results that are justified
neither by the underlying data nor the (usually deterministic) gridding and
contouring algorithms applied to them.
Geostatistical techniques afford an opportunity to reflect not only the anticipated value of a variable in space, but to estimate the probability of its occurrence and confidence intervals surrounding the estimate. A simple procedure is introduced for creation of masks. Their purpose is to eliminate map areas where confidence in the estimated surface falls below a threshold and to characterize the certainty for the remaining, mapped area. Estimated probabilities associated with individual surfaces in a multi-layer analysis can be consistently propagated through to final result maps.
Systematic
inclusion of uncertainty in mapping and spatial analysis is critical where risk
arising from errors in assumptions is an explicit part of decision making. The
procedure is applied within a geographic information system (GIS), the most
common tool used for manipulation and analysis of large, digital
, spatial data
stores. Illustrations are based on very large data bases covering geologic and
engineering variables from the offshore Gulf of Mexico.