Click to view presentation in PDF format.
Almond Formation Lithostratigraphic Genetic Units, Greater Wamsutter Field, Southwest Wyoming: Phase III: From Iterative Geostatistical Approach to High-Grading Well Locations*
Natasha M. Rigg1 and Jeffrey M. Yarus2
Search and Discovery Article #50159 (2009)
Posted January 30, 2009
*Adapted from oral presentation at AAPG Annual Convention, San Antonio, TX, April 20-23, 2008. See companion articles, Search and Discovery Article #50157 (2009), Phase I: A Visual Twist on an Old Approach: The Middle Main Almond Unit, and Search and Discovery Article #50158 (2009), Phase II: What to Do without a Flooding Surface: Mapping the Coastal Plain.
1 Anadarko Petroleum Corporation, Denver, CO ([email protected])
2 Landmark, Houston, TX.
3 Wexpro, Salt Lake City, UT ([email protected])
In order to continue an economic drilling program in greater Wamsutter field, better prediction of reservoir-quality sands is imperative. Our objective is to use geostatistical analysis of net sand distribution to high-grade well locations in order to augment economics in the field and prepare for future increased density spacing.
Previous work established eight genetic lithostratigraphic intervals through correlation of regional flooding surfaces within the middle marginal marine unit of the Main Almond. The total available wells in the field were divided into training and testing sets in order to set up an iterative process for achieving convergence around accurate predictions. Training wells were used to create hand-drawn, gross depositional environment (GDE) maps and for simulation. Variograms were constructed from the GDE maps to ensure the “human” element was included in the models for each interval. Net sand data were then subjected to a Markov-Bayes collocated cosimulation. Test wells were used to measure the uncertainty of the final models.
Initial correlation between predicted and actual net sand data for the test wells was not high for most intervals. A series of refinement steps, with improvements to the log normalization process, well top correlations, and the GDE maps, greatly improved the prediction of net sand within each genetic interval and allowed well locations to be high-graded based on total net sand thickness.
Geostatistical modeling of gross and net sand data prove to be an efficient, cost-effective method to high-grade well locations, given an established geologic model. Additionally, geostatistics can and should be used as part of an iterative process to highlight potential sources of error that may be corrected or improved, ultimately creating a more reliable geologic model.