The progressive role of Quantitative Seismic Interpretation Unlocking subsurface opportunities From qualitative to quantitative Bruno de Ribet, Technology Global Director Peter Wang, Technical Sales Advisor
Messaging Improve productivity with faster and more interconnected workflows Focus main attention on most critical workflows 1. Re-rank and re-qualify prospects you thought you knew well Refine your older assessment of recoverable volumes for a more accurate evaluation of your assets worth 2. Revisit assets you don t know well: Quicker reinvestigation of assets with limited data Don t dismiss assets because limited data from qualification 3. Recover, process and visualize all available data from seismic for better imaging and property distribution New seismic imaging enhancements focus on reservoir formation fabric and properties, including Orthorhombic and Q (De-absorption) compensation and migration 4. Interdisciplinary workflows in HD Increase productivity Increase efficiency when using existing resources
Agenda Contents QSI: Definition Challenges Workflows A Case Study Ultra-thin oil-saturated reservoirs Log Prediction Fracture Detection Analysis Conclusion Excellence in Geoscience
A Definition Quantitative Seismic Interpretation = qualify and quantify reservoir rock properties from seismic; i.e. lithology, fluid content Opportunity: Expand tasks beyond seismic interpretation for better prospect validation & controlled risk management Avalanche of available pre-stack data Require a common infrastructure (pre-stack and post-stack) Require the needed technology, based on an integrated software suite concept
Challenges Both product and problems passed down the line Line pace does not vary No communication along the line No cross-training or learning Processing Geophysicist QSI Expert Interpreter Quality? Quality?
E&P Operators Preparing for the Upturn Unprecedented challenges in budget and resources Asset portfolios being high-graded G&G Integrated Interpretation QSI Facies Modeling Uncertainties Automated Interpretation Velocity Modeling Attributes Time-depth conversion
A Progressive Series of Technology Where to frac? Fracture Detection 4 Analysis Fast-track colored inversion 1 AVO inversion 2 Elastic inversion 3 US EIA offshore-mag.com How uncertain? Stochastic refinement 5 approach geology.about.com Rocks? Fluids? Facies classification Rock physics is a foundational technology 6
3D Environment Implementation Dynamic gathers display & scrolling On-the-fly gathers preconditioning Chained workflows On-the-fly AVO & PMLI inversions Amplitude display on pre-stack picks Analysis tools: AVO plots, frequency spectrum, crossplots Gather conditioning engine (Interactive, batch/cluster) P15.5: Multi-volume attribute extraction and blending Improved management extraction along horizons Addition HSV/RGB merge mode along seismic horizons Data Courtesy of AWE Limited
Quantitative Seismic Interpretation CRP Condition? No Data Preconditioning AVO Inversion 1 Petrophysics & Rock Physics Background Model Building Wavelet Analysis Elastic Impedance Inversion Fracture Detection Analysis VVAZ/AVAZ Fast-track colored inversion Log Prediction Neural Network based Technology No Well Tie? Well Tie? Stochastic Refinement No Lithology and Fluids Classification
Gradient Qualitative AVO Analysis at Well Synthetic AVO NI Reflectivity, Gradient at the well Extrapolate to the 3D seismic volume P15.5: Improved well to seismic analysis Easier log data analysis in stratigraphic units Interval selection from markers and use of classification table for lithology logs makes crossplot analysis much easier Normal Incidence
Porosity and Fluid Substitution Modeling Scenarios to predict change in seismic response due to change in rock properties; i.e. porosity and fluid content Porosity substitution, based on Raymer s equation Fluid substitution is based on Gassmann equation P15.5: Wedge Modeling as a key technology to understand seismic amplitudes Porosity and fluid substitution Synthetic gather from scenarios
Quantitative Seismic Interpretation 2 CRP Condition? No Data Preconditioning AVO Inversion Petrophysics & Rock Physics Background Model Building Wavelet Analysis Elastic Impedance Inversion Fracture Detection Analysis VVAZ/AVAZ Fast-track colored inversion Log Prediction Neural Network based Technology No Well Tie? Well Tie? Stochastic Refinement No Lithology and Fluids Classification
Fast-track Colored Inversion A single P Impedance output for a single post-stack amplitude volume Operator based on the frequency of seismic and well data With or without well data With or without a background model Relative impedance Operator in time Seismic CI operator AI from well Seismic data P Impedance
Post-stack Inversion Methods Colored Inversion Unconstrained Maximum Likelihood Inversion (MLI) Sparse Spike Inversion L1 Norm Inversion Constrained Inversion Background Model Building Wavelet Analysis Single output Absolute impedance
Quantitative Seismic Interpretation 3 CRP Condition? No Data Preconditioning AVO Inversion Petrophysics & Rock Physics Background Model Building Wavelet Analysis Elastic Impedance Inversion Fracture Detection Analysis VVAZ/AVAZ Fast-track colored inversion Log Prediction Neural Network based Technology No Well Tie? Well Tie? Stochastic Refinement No Lithology and Fluids Classification
AVO Inversion Require flat pre-stack gathers (AVO Preconditioning RMO correction) Require amplitude preserved seismic (AVO Preconditioning Amplitude Correction) Require accurate Velocity Model for Ray Incidence Angle determination AVO Inversion Generate AVO attributes from Angle stacking, Shuey, and/or A&R approximation (2 or 3 terms) AVO Inversion Interpretation Validate the prospect with pre-stack data
Automatic Residual Moveout Analysis Velocity Errors and AVO Response Under Corrected Correct velocity Over Corrected Residual velocity Normal Incidence Gradient AVO response biased by non-flatness of the gather Velocity Autopicker based on AVO Criteria (Herbert W. Swan, 1991)
Prospect Validation through Cross-plot Cross-plotting angle stacks can provide indicators for the presence of hydrocarbons or unusual lithology. Near Angle Stack Far Angle Stack
Quantitative Seismic Interpretation CRP Condition? No Data Preconditioning AVO Inversion 4 Petrophysics & Rock Physics Background Model Building Wavelet Analysis Elastic Impedance Inversion Fracture Detection Analysis VVAZ/AVAZ Fast-track colored inversion Log Prediction Neural Network based Technology No Well Tie? Well Tie? Stochastic Refinement No Lithology and Fluids Classification
Background Model Initial guess Low frequency to respect structure and stratigraphy Geostatistical volume creation or reservoir modeling workflow P15.5: SKUA model saved as formation volume Impedance model within simplified framework Impedance model within SKUA grid
Modeling the simple and the complex Validate the structure by modeling any kind of structure and stratigraphy: Any type of fault can be modeled: X, Y, λ, listric, dying faults Compressive and extensive environments Salt, rafts, thin layers Damien Thenin et al, 2013
Modeling the simple and the complex Validate the structure by modeling any kind of structure and stratigraphy: Any type of fault can be modeled: X, Y, λ, listric, dying faults Compressive and extensive environments Salt, rafts, thin layers Impedance from simplified background model Impedance from geologically constrained model
Pre-stack Seismic Inversion When post-stack inversion fails to sufficiently differentiate geologic features with similar P-impedance signatures Simultaneous inversion solves for S-impedance and density, in addition to P-impedance as many geologic features can express similar P-impedance
Pre-stack Inversion Methods Colored Inversion Unconstrained Maximum Likelihood Inversion (MLI) Sparse Spike Inversion L1 Norm Inversion Constrained Inversion Pre-stack Constrained Stratigraphic Inversion (IFP) Pre-stack Maximum Likelihood Inversion (PMLI) Background Model Building Wavelet Analysis Global wavelets P-Impedance S-Impedance Vp/Vs Poisson s ratio Lambda*Rho & Mu*Rho Synthetics, Residuals
Pre-stack Maximum Likelihood Inversion Invert every trace in a gather On-the-fly preprocessing of input gathers From depth or time gather and from offset or angle gathers Single or multiple wavelets Residuals/Synthetics for QC and results validation Minimization of the error function, using the conjugate gradient algorithm J = Jb * Wb + Js * (1-Wb), where: Jb Js Wb = diff(background model, output impedances) = geologic term = diff (synthetic seismic, input seismic) = seismic term = background model weight-to-seismic weight ratio
High Level of Regular Noise (SNR <= 1) Input gathers before PMLI Synthetic gathers after PMLI Residual gathers West Siberia Sandstone reservoir
Amplitude vs. Synthetic PMLI forward modeling result has better apparent resolution Refine horizon interpretations. Correlation = 0.832 Amplitude cube (all offsets) Synthetic amplitude cube (angle range 0 ⁰-32⁰)
Sandstone Reservoir P-Impedance P-Impedance PMLI P-Impedance Angle Stacks Observations: - Better match with the target log curve - Lower level of regular noise
Sandstone Reservoir S-Impedance S-Impedance PMLI S-Impedance Angle Stacks Observations: - Better match with the target log curve - Lower level of regular noise
A Case Study from Belarus 1 RUP PO Belorusneft 2 Paradigm Moscow
Project Objective Detect and detail zones of improved reservoir properties in thin oilsaturated layers (about 5 meters) Structure: Uplifted area bounded on N, W, S, SE sides by faults with throws 180, 170, 120, 200m respectively Divided into 2 blocks (western, eastern) by a 70m fault throw Stratigraphy: Reservoirs are sub-salt carbonates: Voronezh, Semiluksky, Sargaevsky Argillaceous dolomite to mudstone to dense, fine-grained brecciated dense marl Porosity due to cavernous secondary porosity in dolomites, fractured in some areas Belarus oil & gas production concentrated in Pripyat Trough area (Gomel region, SE Belarus)
Survey Characteristics 3D survey: Size 120 km2 Fold 100 Bin size 40 m x 40 m Маximum offset 4000 m Gather quality: Signal/Noise ratio 1.5 Usable bandwidth 10 80 Hz Маximum usable angle 32⁰ at 4300m Fragment of a structural map of the subsalt terrigenous formation surface and vertical section across the deposit area
On-the-fly Preprocessing Coherent noise suppression (including multiple waves) Wavelet stability & amplitude behavior along offsets 10-80Hz seismic spectrum S/N increased from 1.5 to 5.3 Resolution, Flattening improved Full stack Extracted wavelet for seismic inversion
Seismic Inversion Ip Is P-Impedance (left) correlates with the oil reservoir better than S-Impedance (right) P-Impedance extracted at the top of the formation of interest
AVO versus Seismic Inversion Poisson s ratio extracted at the top of the formation of interest P-Impedance extracted at the top of the formation of interest
Opportunity Original seismic data P-Impedance Prospective zone Oil well
Measured Depth Challenge: PMLI Missing Thin Beds Correlation = 90% Inversion results are limited to the frequency range of the seismic signal From Well From PMLI Oil pay zones
Vertical Resolution Resolution from the deterministic inversion dh = ~5-9 m Estimates consistent with formula used to evaluate vertical resolution in the time domain dh = 0.125 *Vint/(Fb - Fh) Kondratyev, 1995 Fb Fh is the width of the seismic spectral spectrum Increase the vertical resolution of the seismic impedance to 1-2 meters using stochastic refinement and well logs
Quantitative Seismic Interpretation CRP Condition? No Data Preconditioning AVO Inversion Petrophysics & Rock Physics Background Model Building Wavelet Analysis Elastic Impedance Inversion Fracture Detection Analysis VVAZ/AVAZ Fast-track colored inversion Log Prediction Neural Network based Technology No Well Tie? 5 Well Tie? Stochastic refinement No Lithology and Fluids Classification
Stochastic Refinement Reservoir Modeling workflow Leverage the deterministic inversion result for thin bed characterization Use well logs, structural framework, stochastic tools Generate as suite of realizations
Geologic Model & Stochastic Modeling Seismic impedance cube Logs Well markers Seismic Horizon Geologic grid definition: 40 m x 40 x 0.2 m cells 11 million cells Stratigraphic Column Sequential Gaussian Simulation (SGS) Collocated Co-Kriging P-Impedance trend function (60% weight) 50 realizations Exponential variogram model Variogram Range 3.7 km
P-Impedance Distributions Average Меdian 50% case Maximum Optimistic/90% Minimum Pessimistic/10%
Seismic Inversion vs. Stochastic Refinement Final realization of P-Impedance (median): more lateral resolution P-Impedance distribution from PMLI inversion
Measured Depth Final Validation PMLI: Quality seismic impedance (resolution: 5-9 meters) Original log From PMLI Stochastic refinement Oil intervals Correlation 0.90 Correlation 0.96 Resolution impedance after stochastic refinement down to 1-2 meters
Quantitative Seismic Interpretation CRP Condition? No Data Preconditioning AVO Inversion Petrophysics & Rock Physics Background Model Building Wavelet Analysis Elastic Impedance Inversion Fracture Detection Analysis VVAZ/AVAZ Fast-track colored inversion Log Prediction Neural Network based Technology 6 No Well Tie? Well Tie? Stochastic refinement No Lithology and Fluids Classification 45
Reservoir Properties Inference Inference method Supervised Artificial Neural Network Backpropagation ANN Two-step inference process: Training step Network s nodes operators are optimized so as to predict the desired output properties from a given set of input elastic logs Prediction step The network is fed with input seismic attributes (elastic properties) to infer the desired output property volume
Neural Network Inversion Log prediction away from the wellbore Neural Network Porosity from P_Impedance and Poisson's Ratio Support for deviated wells Shale volume, inverted P- Impedance & Poisson s ratio, well Porosity logs No up scaling of input logs required Simultaneous inversion of more than one log
Quantitative Seismic Interpretation CRP Condition? No Data Preconditioning AVO Inversion Petrophysics & Rock Physics Background Model Building Wavelet Analysis Elastic Impedance Inversion Fracture Detection Analysis VVAZ/AVAZ 7 Fast-track colored inversion Log Prediction Neural Network based Technology No Well Tie? Well Tie? Stochastic refinement No Lithology and Fluids Classification 48
Azimuthal Seismic Analysis Confirmed interest in unconventional plays (carbonates/shales) especially in the analysis of vertical fractures/differential stress (HTI/TTI/Orthorhombic) Analyze the azimuthal variations in residual moveout (VVAZ) as well as in amplitude (AVAZ) Extract maximum information about the nature of the azimuthal variations from wide-azimuth, migrated, pre-stack data
Full Azimuth Reflection Angle gathers & Analysis & Analysis Full Azimuth RMO Inversion AVAZ Inversion Estimation of delta2 as a measure of fracture intensity α1 = v rms max /v rms, α 2 = v rms min /v rms δ 2 = α 2- α 1 Fracture Density and Azimuth of axis of symmetry (minimum stress) attributes Delta2 map superimposed with vectors along minimum stress at horizon of interest (Eagle Ford) Eagle Ford Data courtesy of
Unconventional Plays Characterization Cross-plot Normal incidence vs. Gradient 3rd axis PImpedance TOC Land the well Identify Reservoir Quality Brittleness Seismic facies with high TOC zones (black) Brittleness In-situ Stress
Full-azimuth Workflow Enhancements AVAZ enhancements in P15.5 Access to more multi-azimuthal data (OVT, Spiral gathers) New Reliability threshold in FastVel and AVAZ: Feasibility Interactive mute based on continuity criteria (Energy, Semblance, Semblance times Energy) Automatic gather flattening for 2D and 3D gathers (linked to more data) AVO attributes along major axis
Agenda Conclusion Improve Productivity Tighter integration Faster, more interconnected workflows Cross-domain HD workflows Reinforce gains from Paradigm 15 Main attention on critical workflows Excellence in Geoscience Accurate representation of the subsurface
Excellence in Geoscience Thank You