Scribd
Upload
50 views

Conditional Simulation

Kriging and conditional simulation both produce smooth property maps that honor data points and the variogram, but differ in key ways: - Conditional simulation can generate multiple realizations, while kriging produces a single map. - Conditional simulation honors the data mean, variance and variogram, while kriging only honors the mean and variogram. - As the number of conditional simulation realizations increases, their mean converges to the kriging solution.

Uploaded by

Afc Hawk
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
50 views

Conditional Simulation

Kriging and conditional simulation both produce smooth property maps that honor data points and the variogram, but differ in key ways: - Conditional simulation can generate multiple realizations, while kriging produces a single map. - Conditional simulation honors the data mean, variance and variogram, while kriging only honors the mean and variogram. - As the number of conditional simulation realizations increases, their mean converges to the kriging solution.

Uploaded by

Afc Hawk
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Conditional Simulation

Kriging Vs. CS

Kriging Conditional simulation


Input -> Input ->
•Variogram •Variogram and normal gaussian
distribution
•Values @ certain points
•Values @ certain points
Results ->
Results ->
•Smooth property map
•Smooth property map
•Honors data points
•Honors data points
•Honors data mean
•Honors data mean, variance and
•Generated map has a different
variogram
variogram and variance but minimizes
the error variance. •Several realization can be obtained
•As the number of realizations
becomes very large heir mean
become closer to kriging
Conditional Simulation

CS based on MCS

% % % Funciones de probabilidad

φ1 k1 VR
P90
Probabilidad de
P50
que el VHCIP
sea menor que
el indicado P10

VHCP
I

Valores aleatorios dependiendo de la


función de probabilidad de cada parámetro

Frecuencia acumulada Tipo de valor Nomenclatura


90% Valor probado P90
50% Valor probado + probable P50
10% Valor probado + probable + P10
posible
Conditional Simulation

CS base on MCS
Sampling each parameter according to its input pdf
•Continuous parameter
•Discrete random variables

Field A OOIP prob function Field B OOIP prob function


f(x) f(x)

Field A+B OOIP prob function

f(x) Full independence assumption

Full dependence assumption


Conditional Simulation

Generating samples of parameters that are correlated with each other

MCS classical approach assume full independence between parameters


Sampling correlated parameters through Regression function+ Random
interval

Property A Prob. Function y

f(x)

Property A

Property B
Sampling 1 parameter
st
Regresion + Random interval
(regression reliability)
Conditional Simulation

SGS (Sequential Gaussian Simulation)

CS base algorithm
Generalization of MCS for 3D modelling

Draw random locations between wells Estimate kriged value and SD

17 % 17 % 17 % 18 % 17 %

19 % 21 % 19 % 21 %

Mean &
Sample from varianc Merged sample
Gaussian e value and draw
histogram. MCS 17 % new location
19 % 17 %

x 19 % 21 %
19 %
Conditional Simulation

Conditional simulation for Geological Facies Modeling


1. Lithologies coded by integers
Pixel-based approach
2. Indicator variables (presence or
absence of facies of interest along the
Well A wells)
0 0 3. Indicator variogram
1 1
1 1 4. SIS.- Similar to SGS
1 1 Channel
Coded 1 1
Lithology 1 1 Indicator
1 1 Variable
1 1
0 0
2 2 γ(h)
2 2
0 0
Range
0 Shale 0 No Channel
1 Sandstone 1 Channel
2 Anhidrite Well B

0 0
1 1
1 1 h
1 1 Channel
Coded 1 1
Lithology 1 1 Indicator
1 1 Variable
1 1
0 0
2 2
2 2
0 0
0 Shale 0 No Channel
1 Sandstone 1 Channel
2 Anhidrite
Conditional Simulation

Object based approach

1. Facies associated with well-defined geometries. Size is random and quantified


2. Constrain
3. Well spacing

You might also like