Seismic velocity modeling 3D grid based

Objectives
Create a simple 3D grid Create layers in the 3D grid Sample the interval velocity data points into the 3D grid Populate the 3D grid using Ordinary kriging Sample well velocity into the 3D grid Calibrate the interval velocity property anisotropy factor Convert calibrated interval velocities to average velocities Setup Velocity model calibrated seismic velocities Do Checkshot velocity modeling guided by seismic velocities Ordinary kriging with trend Collocated co-kriging - optional Setup Velocity model checkshots guided by seismic velocities Quality control the velocity models

Seismic velocity workflows

Load: Seismic velocities Checkshot surveys DIX conversion

Make seismic velocity surfaces corrected to well velocities

Setup velocity model using velocity surfaces

Set up 3D grid and model seismic velocities together with checkshot velocities

Setup velocity model using 3D grid velocity property

Seismic velocities - 3D grid modeling

Petrel Velocity modeling can use an average seismic velocity field property of the 3D Grid

Recommended workflow: Load SEGY or ASCII seismic Interval velocities into Petrel Sample the seismic interval velocities into a 3D Grid Sample the well interval velocities into the 3D Grid Derive the anisotropy factor from the well data and the seismic data and apply it to the velocity field Quality control the calibrated interval velocity field Convert the interval velocity field into an average velocity field Setup a velocity model using the calibrated average velocity property

Interval velocity versus average velocity

It is recommended to model interval velocities instead of average velocities because they are easier to QC : Interval velocity is a petrophysical property Average velocity shows only smooth velocity variations It is difficult to locate the origin of velocity anomalies shown by the average velocity field

Seismic interval velocities

Average velocities, derived from interval velocities

Setup of a 3D grid - consideration

Choose a grid spacing of approximately the horizontal and vertical velocity data sampling Reasons: The number of grid cells has a big influence on the performance of the chosen gridding algorithm A smaller horizontal grid spacing would make sense if local velocity variations should be captured. However due to their nature seismic velocities generally can only provide a reliable regional velocity trend; therefore strong local variations should be filtered A small vertical layering is not necessary because the velocity field varies smoothly with depth

Setup of a simple 3D grid

Input Data: enter the constant surfaces for top (0 ms) and bottom of the velocity cube. Include the time surfaces that are defined as velocity boundaries Geometry: use Automatic (from input data/boundary) and choose a grid increment of approximately the velocity location distance Tartan grid: allows to create grids with non-uniform refinement

Setup of a 3D grid - Layering

In general the type of layering is not of critical importance For Proportional layering the number of layers defines the layer thickness. Use trial & error to get a layer thickness similar to the vertical velocity sampling spacing: display an Intersection in a 3D window and measure the layer thickness. Then adjust the number of layers

Sampling of seismic velocities into 3D grid

Sampling done via the Scale up well logs process under Property modeling

Sample checkshot interval velocity into 3D grid

Open the Scale up well logs process

Upscaled check shot velocities and wells

Choose Point attributes and select the checkshot survey of interest Select the interval velocity attribute of this data set

Gridding of seismic velocities

Three suitable gridding algorithms: Functional - Moving average - Kriging Important parameters: Functional, Moving average - Use a vertical range that covers 2 or more velocity samples - Use Inverse distance (squared) as point weighting Kriging - Select Ordinary from the Expert tab. It is more suitable for velocity interpolation than Simple because it uses a locally varying mean. Consequently trends in the data are better honored. Note: Kriging is an optimized algorithm introduced in Petrel 2008. It shows very good performance and offers collocated co-kriging and kriging with trend. Kriging by Gslib shows a low performance and should not be used any more! Kriging Interpolation does not provide Ordinary kriging option and should no longer be used.

Gridding of seismic velocities - moving average

Vertical range: 10 Point Weight: Inverse distance quadruple

Vertical range: 10 Point Weight: Equal

Vertical range: 1000 Point Weight: Inverse distance quadruple

Vertical range: 1000 Point Weight: Equal

The artifacts are the upscaled velocity data: Increasing the Vertical range will smooth (average) the velocities vertically Increasing the horizontal distance (from Inverse distance quadruple to Equal) will average the velocities horizontally The upscaled data are not changed and may appear as anomalies. They are adjusted through filtering of the velocity property

Gridding of seismic velocities - moving average after filtering

Vertical range: 10 Point Weight: Equal Vertical range: 10 Point Weight: Inverse distance quadruple

Vertical range: 1000 Point Weight: Inverse distance quadruple

Vertical range: 1000 Point Weight: Equal

Select Smooth from the Property operations Typically a couple of iterations are sufficient for removing artifacts

Comparison of kriging options

Ordinary kriging: The mean value is locally varying. Therefore Ordinary kriging is less sensitive to the Range Simple kriging: In the case of a small Range the mean value of all input data is influencing the value of the cells between the data points

Ordinary kriging
Horizon

Simple kriging

Range: 700/700/50

Range: 7000/7000/500
Grid spacing: 900ft, 100ms

Range: 700/700/50

Range: 7000/7000/500

Anisotropy handling Workflow

1. Sample the interval velocities of the checkshot surveys into the 3D Grid 2. Calculate the anisotropy factor (using the Property calculator): F_Anisotropy = Vint_Wells / Vint_Seismic 3. In a Function window approximate the anisotropy factor by a polygon 4. Multiply the seismic interval velocity property with the polygon (in the Property calculator) 5. Convert the interval velocity property to average velocities 6. Setup and execute the velocity model based on calibrated average velocity property 7. Address the remaining residual depth error through a second depth correction (depth scaling) model

Calculating anisotropy

Calculation of anisotropy is done with the Property calculator To plot the anisotropy as a function of TWT you need to create the property TWT using the calculator: TWT=Z

Correcting for anisotropy

Cross plot function Use the Property calculator to multiply the velocity cube with the anisotropy polygon (dropped in using the blue arrow)

Property calculator

Anisotropy handling quality control

/
Well velocities Calibrated seismic velocities

=
Residual anisotropy

The residual anisotropy is of random character, it should not show any trend The depth error caused by the residual anisotropy will be addressed through the depth error correction

Converting interval velocities to average velocities

Using the Workflow editor :

You need to create the following properties prior to the application of the Workflow above: Cell height V_Average Use the Property calculator and create dummy properties with these names

Setup of a velocity model based on seismic velocities

Layer cake model set up with a 3D grid average velocity property based on seismic velocities

Velocity model based on seismic velocities well correction

Parameter settings for the depth error correction. Choose an influence radius to limit the range of influence of the depth errors

Velocity field after depth correction based on well tops

Difference between original and corrected average velocity field. Note the effect of the radius of influence

Note: The Interpolation method is of no influence!

Setup of a depth correction model

Residual error after well top correction of a depth surface provided by a seismic velocity model

Set up of a depth correction model for addressing the residual error

Sometimes a residual depth error of several feet still exists after well top correction. This error needs to be addressed by a depth error correction model

Pitfalls in velocity modeling

Influence of selected gridding algorithm Interpolation along layers SimBox gridding Selecting proper gridding parameters

Influence of gridding algorithm

Potential problem: The velocity increase with depth is not correctly handled by the selected gridding algorithm

Layered velocity interpolation

Gridding is done in SimCube: interpolation along layers Moving average with Point Weight: Equal does too much smoothing

Choice of proper gridding parameters

Point Weight: Equal Point Weight: Inverse distance squared

Same velocity within layer

Increasing velocity with depth

Algorithm: Moving average

Modeling of checkshots
General idea : Model checkshot data in a 3D grid guided by seismic velocities

Workflow : Convert the checkshot data to interval velocities Load the point set into the 3D grid Interpolate the data using the seismic velocities as secondary input in Kriging with trend or Moving Average/Functional with trend Convert the resulting interval velocity cube into average velocities Setup a Velocity Model based on the modeled checkshot data Setup a depth correction model for addressing the residual depth error

Note: Collocated co-kriging generally delivers poor results unless you have many wells with checkshot surveys!

Modeling checkshot interval velocities kriging w/trend

Vseismic Seismic interval velocities V seismic

Kriging of checkshotVwells kriging w/trend interval velocities using seismic interval velocities as trend

Vwells - kriging

Deviation of velocities based on Kriging with trend from calibrated seismic velocities: F_Trend (TWT) = V_Kriging_Trend / V_Seismic_Calibrated Kriged interval velocities (checkshot surveys)
Note: Kriging with trend of checkshot velocities deviates from calibrated seismic interval velocities by less then 4%

Modeling checkshot interval velocities collocated co-kriging

Seismic interval velocities Check shot average velocities, Collocated co-kriged with seismic velocities. Correlation coefficient: 0.96 Comparison Co-Kriging with calibrated seismic velocities : F_CoKrig (TWT) = V_Co_Kriging / V_Seismic_Calibrated

Note: with increasing depth, collocated co-kriging delivers too small velocities. Probably there are not sufficient checkshot points to deliver a reliable result

Comparison of depth errors

Depth error (no well top correction) of 3 surfaces using velocity model Kriging with Trend

Depth error (no well top correction) of 3 surfaces using velocity model Collocated co-kriging

The depth error comparison between the two velocity models shows that the model based on checkshots using Collocated co-kriging does not give satisfactory results (at least in this example)

EXERCISE
Seismic velocity modeling 3D grid based