Session 1: Current Applications, Lessons Learned and Challenges
Geological Process-Based Forward Modeling (GPFM) is defined by its forward-in-time calculating component. Modeling starts from a initial condition in the geological past and proceeds to a subsequent condition in the younger geological past or until present times. While forward models may be inverted, they fundamentally differ from inverse modeling approaches like geostatistical modeling that rely entirely on static realizations, e.g. based on variograms.
The foundations for GPFM were laid in the 1960s to 1970s but limited to 1D and simple 2D. Since the late 1980s, GPFM has progressed towards advanced 2D and 3D. Key drivers for this development come from both industry and academia. Interest by industry stems primarily from the objective of predicting reservoir quality and of a quantitative understanding of depositional heterogeneity. Interest by academia originated primarily from the objective to develop genetic, quantitative geological models from descriptive or qualitative data. GPFM also serves as an approach to test existing concepts of deposition, transport and erosion. Predicting future environments from changes in eustatic sea level, sediment distribution and coastal morphology constitute a further important application. However, existing challenges in GPFM have limited its application especially in industry, whether as complimentary approach or as a partial replacement for geostatistical approaches.
The intention of the opening session is to set the framework for the workshop by covering developments in GPFM during the last few years, lessons learned from applications in various fields, discussing reasons for success (and failure) until now and by providing a general outlook on challenges existing today.
Session 2:Multi-Scale Calibration of Geological Processes
The benefit of any modeling approach depends on the best possible calibration of input parameters. This fundamental requirement is especially valid for geological modeling because of the inherent complexity of systems, the number of processes involved and the dependency of input parameters from temporal and spatial scaling. A specific challenge is the requirement for parameter calibration in the geological past, i.e., differentiated by specific time intervals and processed to rates of change. This is in marked contrast to parameter calibration in geostatistical modeling, which relies entirely on direct measurements from well locations, seismic data or outcrops.
The workshop intends to cover a wide range of approaches to calibrate depositional, diagenetic and geomechanical parameters. Modern environments and outcrop-subsurface analogues allow e.g., parameter calibration for depositional geometries, transport velocities from sedimentary structures, the degree of compaction and for depositional vs. diagenetic heterogeneities at both high vertical and lateral resolution. The calibration of diagenetic processes and parameters at sample scale predominantly relies on petrographic, cathodo-luminescence, fluid-inclusion, trace element chemistry, δ13C, δ18O and clumped isotope geochemistry data. Core flooding experiments provide e.g. rates of dolomitization and calcite cementation. 3D stratigraphic architecture in combination with local-regional stress data, burial history and results from e.g. Linear Variable Differential Transducer (LVDT) experiments provide direct input to geomechanical modeling.
An assessment of the limitations of modeling results represents an important part of geological parameter calibration. Multiple realizations from parameter sets with minimum/maximum values or serial variations in parameters are one of several approaches to define acceptable levels of calibration in de-risking elements of petroleum, geothermal and CO2 storage systems.
Session 3: Modeling Approaches and Applications
The session intends to address and discuss numerical approaches in GPFM. Diffusion modeling has been widely applied for depositional processes because of its balance between complexity and computational expense. Challenges exist in defining diffusion coefficients for various sediment types and in modeling the transport of multiple grain size classes. Navier-Stokes or hydraulic modeling allows detailed flow and transport modeling at high-resolution although simplifications and approximations are required as equations are non-linear for four independent key variables. However, hydraulic modeling approaches are computationally intensive. Fuzzy Logic modeling approaches offer an extension to Boolean logic and are able to handle only partially constrained data and parameters by defining a combined degree of “truth”. The advantage is computational efficiency at the cost of resolution and predictive capability. Geometric/volumetric modeling honors mass balance, energy conservation and sediment accumulation equilibrium. It combines computational efficiency and alignment with sequence stratigraphic concepts. Geometric modeling programs are mainly 2D with 3D in development. Mixed or Hybrid modeling approaches, e.g. combining GPFM with geostatistical modeling or Machine Learning/Deep-Learning (e.g. physics-based Machine Learning) represent a recent technology development. Key drivers are assisted parameter calibration, improved resolution of forward models and increased match between models and well data.
Current clastic diagenetic modeling follows either a rule-based approach or focuses on burial depth, temperature and rock texture-related parameters to model the cementation kinetics of quartz and illite. The relatively most widespread approaches to carbonate diagenetic are either rule-based modeling or Reaction-Transport Modeling (RTM). RTM covers a comprehensive set of physical-chemical parameters and processes with high predictive capability for porosity. However, computational requirements are high and, with sufficient resolution, areas of interest do not exceed the field to prospect scale. Geomechanical modeling of stress-/strain distribution, fracture and fragmentation most commonly relies on Finite and Discrete Element Modeling. Input parameters include gravitational, lithostatic and hydrostatic loads as well as pore pressure and temperature. Today, GPFM usually runs on high-performance workstations. Few software packages offer full multi-processor support and high-performance computing (HPC) technology for multi-node clusters is rare to completely missing.
The workshop aims to cover a wide range of GPFM approaches which include the defining forward-in-time calculation component including methodologies for calibration, verification and uncertainty analysis.
Session 4: Integration of Geological Process-Based Forward Modeling
Current GPFM approaches and studies focus on individual subgroups of geological processes, e.g. depositional, diagenetic or geomechanical processes. However, the full predictive power of GPFM will only be achieved when individual modeling approaches can be linked and integrated.
Examples for challenges in integration include: i) incompatible numerical modeling approaches; ii) temporal and spatial dependencies between key geological processes must be honored; iii) connections to/from public domain or industry databases are required because of data volumes across individual modeling approaches.
For instance, the integration of depositional and diagenetic forward modeling is essential for industry. Porosities derived from depositional modeling reflect textural (“matrix”) porosities from grain size distribution and physical compaction. However, they do not reflect current subsurface porosities after diagenetic overprint at shallow to deep burial depths. The majority of diagenetic overprint in clastic rocks occurs at medium to deep burial depths and, with considerable simplification, may be numerically modeled for a stratigraphic interval at its current burial depth only. However, diagenetic overprint in carbonate rocks starts at deposition (at geologic time scales), tends to peak in shallow burial depths and may re-occur at deeper burial depths. As a result, any comprehensive numerical modeling approach capable of predicting current subsurface porosities will have to follow a process timeline of concurrent deposition, compaction, cementation, solution etc. While the temporal interdependency of other processes differs, similar challenges exist for integrating depositional-diagenetic and geomechanical processes. Very few existing studies have tried to link different geological processes and modeling approaches, e.g. diffusion-based depositional and RTM-based diagenetic modeling. Efforts have also been made to combine process-based and geostatistical modeling for various purposes, whether to increase the resolution of process-based models or to address challenges in matching well data.
Session 5: Case Studies and Applications
The session envisions to include a wide range of case studies and applications from industry, consulting and academia. The best strategy for the implementation of GPFM in all three realms are proven success stories with measurable benefit to operations, whether in oil & gas exploration and production, geothermal exploration and CO2 storage or in research. Because of its importance, this session will cover the entire Day 2 of the workshop and will be subdivided in two sub-sessions:
5.1 Siliciclastic and Mixed Systems
5.2 Carbonate and Evaporite Systems
Examples for applied studies from industry and consulting include but are by no means limited to: i) prediction of depositional environments (GDE maps) for reservoir, source and seal units; ii) sedimentary architecture and heterogeneity; iii) prediction and identification of stratigraphic-diagenetic traps; iv) rock physics; v) geomechanics and fractured/tight reservoirs; vi) porosity and permeability prediction at inter-well to basin scale; vii) prediction of organic matter for conventional and unconventional systems; viii) input to 3D hydrocarbon system and basin models; ix) geological evaluation of depleted reservoirs for CO2storage; x) geothermal exploration, especially for Enhanced Geothermal Systems (EGS); xi) benchmarking GPFM modeling vs. geostatistical predictions in operations and, xii) successes (and failures) of GPFM models.
Examples for studies with a fundamental or applied research background include but are by no means limited to: i) sedimentary basins as archives for depositional, environmental and structural processes; ii) reconstruction of subsidence/uplift, eustatic sea-level, sediment input/production and erosion histories; iii) prediction of future environments due to changes in sea-level, erosion, sediment input, coastal morphology and current dynamics; iv) testing geological observations from surface and subsurface data for controlling processes and advanced sequence stratigraphy concepts.
Session 6: Perspectives and Way Forward
Advanced development of Geological Process-Based Forward Modeling and implementation in standard workflows will be a long-term effort best approached in a step-wise fashion and, as much as possible, in collaboration between industry and academia.
The final session of the workshop will try to: i) summarize learnings and conclusions from the five sessions and related break away meetings; ii) suggest most promising directions in future development of GPFM; iii) serve as opportunity to initiate future research and development collaboration between participants.