Seismic Synthetic Pitfalls

Data preparation = Stage before inversion preparation Inversion preparation = First step involves understanding the relationship between seismic data times and well borehole depths Well-to-seismic tie process = Correlating well depths to seismic times and generating Depth-Time curves Deterministic wavelet extraction = Step critical for stable inversion wavelets

Sonic logs = Provide information about seismic waves velocity in the subsurface Density logs = Give data on density variations in the subsurface Resistivity logs = Indicate electrical resistivity of subsurface formations Checkshot surveys = Measure travel times of seismic waves from a near-surface source

Correlating well depths to seismic times = Establish a reliable time to depth conversion for seismic data interpretation Generating Depth-Time curves = Provide information about the relationship between depth information from well logs and seismic reflection events Deterministic wavelet extraction = Relies on good well-seismic match to generate stable inversion wavelets Checkshot surveys = Measure travel times of seismic waves to provide precise information about travel time from surface to depth points in borehole

<p>Full Wavelet method = Determines phase and amplitude by solving for the time-domain operator Constant Phase method = Calculates amplitude spectrum using autocorrelation and approximates phase spectrum as a linear constant value Roy White Algorithm = Extracts wavelet amplitude and phase through cross-coherence analysis between seismic and synthetic reflectivity Roy White Wavelet estimation algorithms = Optimally estimates seismic wavelet by calibrating and correlating well log data and seismic data</p> Signup and view all the answers

<p>Proportion of Energy Predicted (PEP) = Measure of predictability or goodness-of-fit, ranging from 0 to 1 Root Mean Square Error (RMSE) = Measure of accuracy instead of goodness-of-fit, estimate of expected error ranging from 0 to 1 Normalized MSE (NMSE) = Alternative name for RMSE, a measure of accuracy Expected RMSE value = Estimate of expected error arising from noise propagating into synthetic seismogram</p> Signup and view all the answers

<p>Extract single wavelets at each well = To compare them and ensure similar phase and amplitude spectrum for reliable average wavelets Remove unstable wells = If there are many wells and some are not stable enough Compare computed operators = To control quality between wells, stacks, multi-well/single-well estimates, and computed versions Ensure wavelet is close to constant phase = If amplitude spectra does not resemble zero-phase spectra, suggest retrying with zero-phase or constant phase wavelets</p> Signup and view all the answers

<p>Poor well logs quality = Generates inaccurate well synthetics No check shots available = Well tops and horizons can help match correct events Synthetic seismograms only calculate zero offset seismic trace = Seismic data composed of stacked traces from range of offsets Synthetic seismograms do not generally account for wavefront dispersion and absorption effects = Wavelet is frequently unknown</p> Signup and view all the answers

<p>The wavelet is the filter through which subsurface geology is viewed = Important to know phase and polarity of seismic wavelet Most inversion algorithms use one representative wavelet per seismic stack = Implies a stationary assumption Statistical estimation approach for wavelet extraction = Uses seismic data alone to extract the wavelet and assumes zero phase Deterministic estimation approach for wavelet extraction = Uses wells, gives good estimate of amplitude and phase spectra</p> Signup and view all the answers

<p>Extract a zero-phase statistical wavelet from the seismic over a representative area = Optimize well-seismic tie to minimize time-depth relationship uncertainties Extract a deterministic wavelet using calibrated wells = Account for possible phase shifts and changes in frequency content Extract a wavelet using logs after optimum calibration = Wavelet must already be known for proper calibration Selecting statistical or deterministic wavelet during inversion = Update well-seismic tie accordingly</p> Signup and view all the answers

<p>Convolution theory = Seismic assumed to be convolution of earth reflectivity with a wavelet Well log reflectivities derived from P-Impedance or P-Velocity, S-Velocity, Density in angle stacks or gathers = Aim to find convolutional operator that maximizes match with seismic Deterministic wavelet extractions conducted by frequency = Resulting wavelets bear full description of amplitude and phase spectrum</p> Signup and view all the answers

<p>Identifying and removing bad points = Checking input information for errors Trying different interpolation methods = Increasing smoothing to remove sharp variations Checking for possible errors in units of measurement or reference points = Using full stack or partial-offset stack to estimate wavelet Assuming seismic data is zero phase = Limiting number of tie points and using strong events</p> Signup and view all the answers

<p>Average seismic character (amplitude and frequency spectrum) = Quality control Convolution with well log reflectivity for well-seismic tie analysis = Deconvolution of seismic for seismic inversion Synthetic gather computation and forward modelling exercises = Spectral analysis of seismic data Estimating wavelet from target interval = Stationarity indication on seismic signal</p> Signup and view all the answers

<p>Computing synthetic seismograms by convolution = Applying residual time shift to maximize correlation Slightly stretching or squeezing the time-depth curve = Checking if corrected sonic is within +-10% of original sonic log Checking polarity of seismic data = Choosing target interval to estimate wavelet Estimating residual phase shift using calibration wells = Analyzing potential phase rotation suggested by well-seismic tie analysis</p> Signup and view all the answers

<p>Starting from a zero-phase assumption = Considering running analysis for different wells and intervals Using correlation function to estimate phase shift = Analyzing residual phase shift using calibration wells Estimating phase shifts based on well-seismic tie analysis = Concluding on seismic phase after consistency check Running analysis for different intervals to check values consistency = Estimating any phase shift between synthetic and seismic trace</p> Signup and view all the answers

<p>Vertical Seismic Profile (VSP) = Involves generating seismic waves at the surface and recording within the borehole Surface Seismic Data = Subject to greater dispersion and absorption than the sonic data recorded in the well Check Shots = Recorded from vertical seismic profiles (VSP) and used to tie the well to the seismic as a first approximation Sonic Logging = Involves transmitting high-frequency sound waves into the formation and recording the travel times</p> Signup and view all the answers

<p>Seismic and log datums are usually different = The first layer velocity is unknown Errors in calculating time-depths from logs accumulate = Interpreted seismic data may be mis-positioned if migrated improperly Seismic data may have time stretch caused by frequency-dependent absorption = Surface seismic data are subject to greater dispersion and absorption than sonic data recorded in the well Loose near surface rocks must be cased before sonic logging tools can be run = Seismic data are generally full stack or near-offset partial angle stack considered in the inversion</p> Signup and view all the answers

<p>Extracting a zero-phase statistical wavelet from a representative part of the seismic = Computing synthetic seismograms by convolution of reflectivities with a zero-phase wavelet Applying global time shifts or controlled Stretch and squeezes to improve match with seismic = Initiating well-seismic tie analysis assuming the seismic is zero phase Considering elastic log quality, data coverage, local geological features, and production effects = Using check shots or VSP data as Time-Depth curves Interpolating a drift curve to measure discrepancy between time-depth curve from sonic log and check shot data = Check shot correction adapts sonic log velocities or log time-depth curve to match time-depth relationship from surface seismic data</p> Signup and view all the answers

<p>Measuring discrepancy between time-depth curve from sonic log and check shot data = 'Check shot correcting' time-depth curve using drift curve Adapting sonic log velocities or log time-depth curve to match time-depth relationship from surface seismic data = Interpolating a drift curve to measure discrepancy between sonic log and check shot data Applying global time shifts or controlled Stretch and squeezes to improve match with seismic = Using check shots or VSP data as Time-Depth curves Extracting a zero-phase statistical wavelet from a representative part of the seismic = Computing synthetic seismograms by convolution of reflectivities with a zero-phase wavelet</p> Signup and view all the answers

<p>To describe frequencies down to 5Hz, we need a wavelet of at least 200ms = 5Hz To describe frequencies down to 2Hz, we need a wavelet of at least 500ms = 2Hz</p> Signup and view all the answers

<p>Have not too many side lobes = Side lobes The phase spectra should be more or less flat and similar for all angle stacks = Phase spectra</p> Signup and view all the answers

<p>Defining the correct wavelet phase is critical so the inversion can position the elastic property contrasts at the true vertical location = Wavelet phase If the phase spectra is not flat it means that the phase varies with frequency, indicating a mis-tie between the well and seismic = Phase spectra</p> Signup and view all the answers

<p>Defining the correct wavelet scaling is critical so the correct reflectivity can be retrieved from inversion = Wavelet scaling Too short a wavelet can lack low-frequency information and generate unstable results in that frequency range = Short wavelet consequences</p> Signup and view all the answers

<p>Statistical approach that uses the seismic data alone = Statistical approach Deterministic approach that uses both wells and seismic data = Deterministic approach</p> Signup and view all the answers

<p>Well-to-seismic tie serves to match geological events in depth with reflections in time and generate stable inversion wavelets = Well-to-seismic tie role Quality Control and comparison between different wells and stacks to find an optimum wavelet for inversion = Wavelet quality control</p> Signup and view all the answers