BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Dallas Geophysical Society//Silkstart Technology Inc.// BEGIN:VTIMEZONE TZID:US/Central BEGIN:DAYLIGHT DTSTART;VALUE=DATE-TIME:20260308T020000 TZNAME:CDT TZOFFSETFROM:-0600 TZOFFSETTO:-0500 END:DAYLIGHT BEGIN:STANDARD DTSTART;VALUE=DATE-TIME:20261101T020000 TZNAME:CST TZOFFSETFROM:-0500 TZOFFSETTO:-0600 END:STANDARD END:VTIMEZONE BEGIN:VEVENT SUMMARY:Dallas Geological Society - November Luncheon - Alex Blizzard DTSTART;TZID=US/Central;VALUE=DATE-TIME:20261104T113000 DTEND;TZID=US/Central;VALUE=DATE-TIME:20261104T130000 DTSTAMP;VALUE=DATE-TIME:20261030T172450 UID:6ae4d2e11995c706255da22a DESCRIPTION: Dallas Geological Society November Luncheon This talk is hosted by the Dallas Geological Society - please visit their website for registration information! Comparing Predictive Signals Across the Benches of the Bone Spring Formation\, Deleware Basin: Integrating a 3D Geologic Framework with Machine Learning Alex Blizzard\, Sr. Techincal Advisor - Data &\; Analytics\, S&\;P Global Talk Abstract Development activity in the Bone Spring Formation has expanded both laterally across the Delaware Basin and vertically across the operator-defined first\, second\, and third benches. Although commonly grouped within a single stratigraphic interval\, each bench presents distinct development challenges related to reservoir quality\, pressure\, thermal maturity\, and depletion history. This study integrates detailed geologic mapping with machine learning to identify the primary controls on oil productivity within each Bone Spring bench and evaluates how those controls vary stratigraphically. The foundation of this work is a 3D geologic framework that includes individual bench-level structure\, thickness\, and lithology interpretations derived from more than 700 hand-digitized lithology logs. These geologic attributes were integrated with well-level completion\, spacing\, production\, and pressure-related variables across the Delaware Basin. Additional inputs included stress orientation\, distance to basement-related structures\, and mapped hydrostatic pressure gradients. Data were compiled and analyzed separately for each bench. Following quality control and feature reduction\, machine learning models were trained to predict 12-month average oil rate (barrels per day) normalized to 1\,000 ft of lateral length. Feature selection was performed using Minimum Redundancy Maximum Relevance (mRMR) analysis to identify the most informative variables for each bench-level model. Model performance varied by bench. The first Bone Spring bench model produced the strongest results (R²\; = 0.64)\, followed by the second (R²\; = 0.58) and third (R²\; = 0.45) benches. The stronger first Bone Spring performance may reflect its more confined development footprint and the reduced variability in depth\, thickness\, and lithology over that footprint\, whereas the lower performance of the second and third Bone Spring benches suggests broader heterogeneity and productivity controls not fully captured by the available feature set. Several consistent geologic patterns emerged from the feature rankings. Average true vertical depth (TVD) and true vertical thickness (TVT) were important geologic variable across all three benches\, likely reflecting its relationship to formation pressure\, thermal maturity\, and basin position. Lithologic controls differed by bench. The first Bone Spring bench retained net-to- gross sand and carbonate variables in the model\, indicating mixed lithologic controls on productivity. In contrast\, the second Bone Spring bench retained net-to-gross shale\, while the third Bone Spring bench retained net-to-gross carbonate\, highlighting meaningful differences in reservoir character. Early production ratios also proved highly diagnostic. Peak rate water&ndash\;oil ratio was the most informative production ratio feature across all three benches\, increasing in importance with depth into the Third Bone Spring bench. Peak rate gas-oil ratio had an inverse relationship\, decreasing with depth with highest importance in First Bone Spring\, and lowest importance in Third Bone Spring. These relationships suggest differing influences of maturity\, reservoir quality\, hydrocarbon saturation\, depletion\, and parent-child well interactions among the benches. This study demonstrates the value of combining detailed geologic mapping with machine learning to identify bench-specific controls on Bone Spring productivity. By linking model- selected drivers back to geologic processes\, the analysis provides insights that can improve reservoir characterization and future development planning across the Delaware Basin. Speaker Bio Alex Blizzard is a Senior Technical Advisor for Data &\; Analytics at S&\;P Global\, where he combines his geoscience background with data science to address complex subsurface challenges through machine learning\, Spotfire dashboards\, and data visualization. In a previous role\, he worked as a Technical Advisor on the Petra and Kingdom Geoscience software team. Throughout his career\, Alex has published research applying machine learning to subsurface data across the Haynesville\, South Texas geothermal resources\, and Wolfcamp carbonate gravity-flows. He holds an M.S. in Geology from Stephen F. Austin State University and a B.S. in Geosciences from Texas Tech University. Questions? Contact Vishal Singh\, Dallas Geological President: vishal.singh@terraguidance.com Contact Parker Johnson\, Dallas Geological Vice President: parker@elkrange.com LOCATION:The Clubs of Prestonwood\, 15909 Preston Rd\, Dallas\, TX 75248\, USA PRIORITY:5 URL:https://www.dallasgeophysical.org:443/events/november-luncheon-dallas- geological-society-alex-blizzard END:VEVENT END:VCALENDAR