STATISTICS

• Statistics: A set of tools for collecting, organizing, presenting and analyzing

numerical facts or observations.

1. Descriptive Statistics: Procedures used to organize and present data in a

convenient, useable and communicable form.

2. Inferential Statistics: procedures employed to arrive at broader generalizations or

inferences from sample data to populations.

• Data: Characteristics or numbers that are collected by observation.

• Population: A complete set of potential observations.
• Parameter: A number describing a population characteristic; typically inferred from
sample statistic.
• Sample: A subset of population selected according to some scheme.
• Random Sample: A subset selected in such a way that each member of the
population has equal opportunity to be selected.
• Variable: A phenomenon that may take different values.
STATISTICS
• Measures of Central Tendency
- Mean: The point in a distribution of measurements about which the summed
deviations are zero
N
1 1 n

N
x
i 1
i (Population Mean) x   xi
n i 1
(Sample Mean)

-Weighted Mean: Sum of a set of observations multiplied by their respective weights,

divided by the sum of the weights
G

w x
i 1
i i

w
i 1
i

-Median: Observation in a set that divides the set so that the same number of
observations lie on each side of it.
- Mode: Observation that occurs with the greatest frequency
STATISTICS
• Measures of Dispersion

- Sum of squares: deviations from the mean, squared and summed

SS   (x i   x )2

- Variance: The average of square differences between observations and their mean

1 n
1 N s   ( xi  x) 2
2

2 
N
 (x
i 1
i   ) 2 (Population Variance) n i 1 (Sample Variance)

- Standard Deviation: Square root of the variance

N
1
 
N
 (x
i 1
i   )2

N
1
- Covariance: For a bivariated distribution Co 
N
 (x
i 1
i   x ) * ( yi   y )
STATISTICS
Probability

Probability of occurrence of event A

p(A) = # of outcomes of the event A / #total outcomes

Type of Events:

1. Exhaustive: two or more events are said to be exhaustive if all possible outcomes
are considered. p (A y B) = 1
2. Mutually exclusive: Events that cannot occur simultaneously.
p (A y B) = 0, p (A o B ) = p(A) + p(B)
3. Non mutually exclusive: Events that can occur simultaneously.
p (A o B ) = p(A) + p(B) – p(A y B)
4. Independent: Events whose probability is unaffected by occurrence or
non occurrence of each other
5. Dependent: Events whose probability changes depending upon the occurrence
of each other
STATISTICS
Histogram
Is an easy way to analyze the data. In this example the sample is divided in classes,
where a class is defined as a range of values. The number of measures that fall in
a class is called Class Frequency and the graph is called Histogram. The addition of
frequencies represents the cumulative frequency.

Class Frequency % Cumulative

0.015 1 0.01

0.030 6 0.05

0.045 13 0.14

0.060 20 0.28

0.075 11 0.35

0.090 4 0.38

0.105 6 0.42

0.120 16 0.53

0.135 12 0.62

0.150 22 0.77

0.165 19 0.90

0.180 9 0.97

And
greater
... 5 1.00
STATISTICS

Cumulative Frequency Distribution (cdf): Is analogous to cumulative relative

frequency distribution. Is the probability that a value of a random variable is less or
equal than a certain value.

F(x) = prob (X<x)

According with the graph there is about

50% probability that a variable X random
chosen will be less than zero.
STATISTICS

One of the most famous cdf in geology is this figure that shows the distribution of shale
Lengths in various geological environments.
STATISTICS
• Probability Density function (pdf):
We consider next the probability distribution (or density) function or pdf, probably the
most familiar way of presenting the distribution of a random variable. This forms the basis
for interpreting f(x) as a probability of a value of x in the neighbourhood of x.
A typical continuous pdf is shown below

Specific Distributions - There are more than 100 probability distribution functions observed
in nature. We deal with only a few of these. When we speak of a frequency or probability
distribution function (a pdf or histogram) the first thing that naturally comes to mind is the
normal distribution in which a variable x is distributed with probability f according to the
normal distribution .
STATISTICS
• Probability Density function (pdf) for normal distribution:

The quantities  and  are the mean and standard deviation of the distribution; this is a
two parameter distribution since it can be completely specified with only  and  . The
normal distribution is also known as a Gaussian distribution.
STATISTICS
Valores Z para Distribucion Normal
Desv Std Probabilidad Desv Std Probabilidad
desde el promedio Acumulada desde el promedio Acumulada
z


1  2
F (z ) 
-3.0 0.0014 0.0 0.5000
-2.9 0.0019 0.1 0.5398 z
e 2 dz
 2p
-2.8 0.0026 0.2 0.5793
-2.7
-2.6
0.0035
0.0047
0.3
0.4
0.6179
0.6554

-2.5 0.0062 0.5 0.6915
-2.4 0.0082 0.6 0.7257
-2.3 0.0107 0.7 0.7580
-2.2 0.0139 0.8 0.7881
-2.1 0.0179 0.9 0.8159 Z is called a unit normal
-2.0 0.0228 1.0 0.8413
-1.9 0.0287 1.1 0.8643
variable with mean of zero and
-1.8 0.0359 1.2 0.8849 a variance of one.
-1.7 0.0446 1.3 0.9032
-1.6 0.0548 1.4 0.9192
-1.5 0.0668 1.5 0.0332
-1.4 0.0808 1.6 0.9452
(x  )
-1.3 0.0968 1.7 0.9554
z

-1.2 0.1151 1.8 0.9641
-1.1 0.1357 1.9 0.9713
-1.0 0.1587 2.0 0.9773
-0.9 0.1841 2.1 0.9821
-0.8 0.2119 2.2 0.9861
-0.7 0.2420 2.3 0.9893
-0.6 0.2743 2.4 0.9918
-0.5 0.3085 2.5 0.9938
-0.4 0.3346 2.6 0.9953
-0.3 0.3821 2.7 0.9965
-0.2 0.4207 2.8 0.9974
-0.1 0.5602 2.9 0.9981
0.0 0.5000 3.0 0.9987
STATISTICS
The normal distribution is rather uncommon in nature. For example, while porosity
seems to be reasonably normally distributed, permeability distributes in a manner far
removed from the symmetry demanded by the normal distribution. Permeability
seems to be distributed commonly as log-normal. The following figure shows both
the cdf and pdf for a log-normal distribution

The variable X being characterized cannot be less than zero. There are a few very large
values of X; most of the values are small. Because of this the mode (the most likely
value) is less than the median which is less than the mean.
STATISTICS AND GEOSTATISTICS

Heterogeneity

Dispersion or Variability - Heterogeneity is the spatial variation of properties.

All reservoir properties are heterogeneous, but we focus on flow properties here,
especially permeability.
Heterogeneity measures fall into two categories:

1. Static
Coefficient of variation
Dykstra-Parsons coefficient
Lorenz coefficient

2. Dynamic
Channeling factors
Dispersivities
STATISTICS AND GEOSTATISTICS

Heterogeneity: Coefficient of Variation (CF)

The most direct measure of heterogeneity is the variance and standard deviation

Statistical Measures of Heterogeneity . . .

Coefficient of Variation (CF) = Standard deviation/ Mean

The standard deviation is, of course, the positive square root of the variance.
Both quantities have units but the coefficient of variation, CF, does not.
STATISTICS AND GEOSTATISTICS
Heterogeneity: Dykstra-Parsons Coefficient (VDP)
A common measure of permeability
variation used in the petroleum
industry is the Dykstra-Parsons
coefficient VDP.

The Dykstra–Parsons coefficient is

computed from a set of k data
arranged in increasing or decreasing
value. The values to be used in the
definition are taken from a "best fit"
line through the data when they are
plotted on a log–probability plot.
VDP takes values between 0 and 1.
STATISTICS AND GEOSTATISTICS

Dykstra-Parsons Coefficient

10000

1000
V DP = 0.84

100
k/Phie

10

0.1
1 2 5 10 20 30 50 70 80 90 95 98 99

Probability, % less than

STATISTICS AND GEOSTATISTICS
Heterogeneity: Lorenz Coefficient (LC)

A less well–known, but more general

measure of variability is the Lorenz
Coefficient LC. If A is the area under
the F–C curve, the Lorenz coefficient
is defined as

LC = 2( Area between F–C curve

and 45o line)

Just as does VDP, LC is 0 for

homogeneous reservoirs and 1 for
infinitely heterogeneous reservoirs.
GEOSTATISTICS

• Geostatistics: A set of tools for collecting, organizing, presenting and analyzing

numerical facts in the space

In Geostatistic we look for:

• Unbiasedness (estimated value = real value)

• Minimum variance (low dispersion)

Only the Normal Distribution has these characteristics

and optimises the geostatistical results
 GEOSTATISTICS
We need to transform the original data to have a normal distribution, in the case of data
with lognormal distribution, we only have to apply the logarithm.

Forecast: Yacimientos Zapotal-Merina

10.000 Trials Frequency Chart 0 Outlie
,33 3339

,25

,16

,08 834,7
Mean = 62,34

,00 0
0 139 278 418 557

Cumulative production 36 months

Histogram of Original data Log Cumulative production

Histogram of Transformed data

GEOSTATISTICS
This tool allow to analyze the spatial behaviour of any
Semivariogram property over a particular zone.
Let us express the variance of N samples
measured a distance h apart as
The semivariance is, in effect, the variance of a
property Z at successively larger spacings . . .
h, 2h, 3h, etc. When the semivariance is plotted
Vs. the spacing, a semivariogram results.
When h=0 the covariance is equal to variance.

When h is large enough, the data are poor

correlated so the covariance will tend to
zero. The range is the larger distance of
correlation. Semivariogram
A semivariogram try to quantify the Covariance

variation of the property with the distance.

GEOSTATISTICS Semivariogram

 1 n
 ( h)  
2n i 1
(V ( xi )  V ( xi  h)) 2

1 N
Co C ( xi, xi  h)   (V ( xi )  V ( x)) * (V ( xi  h)  V (x))
N i 1

The semivariance does not, in general, go through the origin, an effect known as the
nugget effect. Represents the variance at a range below the lag distance. This effect is
generally attributable to:
1. Measurement error.
2. Miscorrelation on a distance smaller than the smallest lag distance.

The semivariogram is the most powerful geostatistical tool. It forms the basis of Kriging
which is itself the basis of geostatistical estimation, and it is unbiased to estimates of
the sample mean, although it tends to lose precision when autocorrelation is large.
The most singular feature of the semivariogram is that it is a combined measure of
heterogeneity (sill and nugget) and autocorrelation (the range).
GEOSTATISTICS
Semivariogram
1
Experimental Variogram  ( h) 
*

2 N h 
 ( z ( xi )  z ( x j )) 2
xi  x j  h

How to build an experimental Semivariogram?

Choose a direction 

Chose a distance or lag h

experimental variogram
6

Calculate  * for values of h,2h, 3h,...,nh 5

Graph  * vs. h,2h, 3h,...,nh 3

Variograma
2 experimental

1
0
0

4
0 .4
0 .8
1 .2
1 .6

2 .4
2 .8
3 .2
3 .6
Distance
GEOSTATISTICS
Experimental Variogram: Direction

For each direction  we define a tolerance 

and we use the point that lie in    and   

   

   

b
Used points

Unused points

b = band width
GEOSTATISTICS
Anisotropy
Semivariograms in different directions, perpendicular to each other, can be used to
demonstrate anisotropy. This could be particularly useful where the trends are not
visually obvious.
The anisotropy could be Geometrical and Zonal

GEOMETRIC ANISOTROPY 3.5

3
R2 R1 2.5
R1=R2

SemiVariogram
 (hy) 2

R1 1.5

 (hx) 1

0.5
R2
0
0 0.94 1.99 3.04 4.09 5.14 6.19 7.24 8.29 9.34 10.4 11.4

Lag

Geometric Anisotropy Zonal Anisotropy

GEOSTATISTICS
Geometric Anisotropy

2,5

Variograma
1,5 N-S
E-O
1

0,5

0
0,0 0,9 2,0 3,0 4,1 5,1 6,2 7,2 8,3 9,3 10,4 11,4

Distancia
GEOSTATISTICS
Geometric Anisotropy
GEOSTATISTICS
Behavior short distances

Pure Nugget Effect

0 si h  0
  h    S

  s si h  0


Semivariogram

lag

This model represents a random effect where there is no spatial correlation, no

matter how close the values are.
 Comportamiento discontinuo
GEOSTATISTICS

Nugget effect

 h   var [ Z ( x)  Z ( x  h)]
1
2
 0  0

At a distance close to zero the

value of the semivariogram is
different to zero.
GEOSTATISTICS

Nugget effect
The causes could be:

1) Z(x) = Z(x+h) h0

2) Z obs x   Z x    x 
real

3) Presence of structures or missing

values at distances lower than the lag
GEOSTATISTICS
Semivariogram Models

Most common Semivariogram Models:

Spherical Model

Exponential Model

Gaussian Model

Potential Model

Nugget Model
GEOSTATISTICS
Semivariogram Models

Models with Sill: The semivariogram exhibits a constant value after a certain distance

• Spherical
• Exponential
• Gaussian
Semivariogram

Co

Spherical
Exponential
Gaussian

Lag
GEOSTATISTICS
Models without Sill: The semivariogram does not reach a constant value

• Power
• Cosine
• Log normal

Power Model

 h   s h
Semivariogram
p
s=2.5, p=0.4

s=0.4, p=1.8

s scale factor s=1.15, p=1

0 p2

lag

Represents no stationary phenomenon .The behavior in the origin depends of

the value of p
GEOSTATISTICS
Data can be much more complex than shown by certain idealized correlation behaviour.
Autocorrelation in clastic formations tends to be scale-dependent, with different scales
being associated with laminae, bedforms and formations.
GEOSTATISTICS

Cross Semivariograms
The cross-semivariograms represent and quantify the spatial relationship between 2
variables. These cross semivariograms are necessary in cokriging where we use the
information of one variable to estimate the other. Mathematically it can be expressed:

1 n
 c ( h)  
2n i 1
[V ( xi )  V ( xi  h)] * [W ( xi)  W ( xi  h)]

To estimate the cross semivariogram we need to have the data of both variables in
xi y (xi+h).
The model for the cross semivariogram has to be equal to the single semivariograms for
the variables, the range values have to be the same but not the Sill.
GEOSTATISTICS

Estimation vs. Simulation

Estimation
• Honour the data
• Smooth (good to see trends)
• Inappropriate where the extreme values are important
• Just produce one output
• Does not quantify uncertainty

Simulation
• Honour the data
• Realistic heterogeneity
• Very flexible (can match geology)
• Honour the spatial variability (Semivariogram)
• Produces equi-probables outcomes
• Uncertainty prediction
 Kriging
Kriging

Lets consider Z u ,   1,2, N as data points in the field

and the estimation of Using the data points Z u 

Z u 
Kriging

There are several methods to get

• Closest Point
• Least Squares
• Projected Slopes
• Weighted Average
• Distance
• Isopach
• Bounded range
• ……

These methods do not take into account the semivariogram !

 Kriging
Bases for Kriging estimation:

Consider the estimation of as linear combination of weighted averages of all

the sampled values

To get minimum Variance (low dispersion) and Unbiasedness (estimated

value = real value)


var Z u   Z * u  
** In the process of minimizing error variance kriging creates smooth distributions
GEOSTATISTICS
Ordinary kriging: To estimate specific values.
Block kriging: To estimate blocked values. This technique is extremely useful for
estimation of static properties for a grid blok: Phie, Sw, Log(k). These variables are
sensitive to arithmetic average therefore suitable for the block kriging approach.
Universal kriging: if the average of the sample varies in the stationary region, means
that estimates the point values in presence of a trend.
Conventional Cokriging: Honor the sampled data but use both semivariograms:
seismic and wells.
Collocated Cokriging: each grid point is calculated using the closest seismic data
Conditional Simulation: is the process of creating a set of maps, all of which honor
the sparse data (wells) exactly, and at the same time display the spatial continuity
properties implicit in the semivariogram. These maps are equi-probable in the sense that
they are all consistent with the known information. However, they are not strictly random,
because they are constrained by the semivariogram model, as well as the secondary
dense data (optional). The simulation maps differ from the kriged or cokriged maps in that
they contain the possibility of large deviations or outliers. Kriged maps are smoother –the
bias in the kriging algorithm is toward as little variation as possible, consistent with the
hard data. Simulation maps may show the extreme possibilities that are still consistent
with the hard data.
GEOSTATISTICS
Conditional Simulation
Simulation is the process of creating a set of maps, all of which honor the
sparse (well) data exactly, and at the same time display the spatial continuity
properties implicit in the semivariogram.
While simulation maps are interesting in themselves as displays of possible
outcomes, they are most useful when analyzed to see distributions of
features. For this purpose, we create a series of maps, and then analyze
those maps to display the properties.

Conditional Simulation Advantages . . .

1. Realistic heterogeneity
2. Use of statistics for matching
3. Great flexibility
4. Uncertainty prediction
5. Equivalent use of conditioning data
6. Information management
Conditional Simulation
Gross Thickness
Sand G57 B Nilam Field
Demarcating the Polygon Boundary for G53B Tk #1

Polygons G53B
Make Surface (1/2)
 Make Surface (2/2)

Repeat the same procedure for the lower TZ or marker

Make Horizons (1/1)

Surfaces
MarkersTops
Make Zones (1/1)
 Layering - Settings

1. Select the Zone Division (four different types).

2. Specify the number of layers (Proportional), cell thickness

(Follow top/Base) or relative proportions (Fractions).

3. Optional input of a Reference Surface (e.g., Clinoforms)

Proportional

Follow Base

Follow Top

Fractions

Follow Base
with Reference
 Layering - Results

Types of zone division:

Follow base

Proportional

Follow top

Fractions

Follow top + using a

depositional surface

Note: It is always a good idea to display

edges in 3D to see all updates before
applying the Make Zones and Layering
Processes.
 Layering Concepts (1/2)
What is the appropriate model of layering? Parallel to top/base of marker or Proportional?
The choice of correct representation will have considerable impact on modeling phase and finally
on flow simulations, since It defines the spatial architecture of depositional units in the reservoir.
Regional Marker 1 (TZ)

Nilam-165 G53B Nilam-87 G53B Woodchoppers’ yard outcrop

(well
NILAM -core)
87
LEVEL G53B
1(A)
12805
17 FLOOD-DOMINATED
NILAM - 179
LEVEL G61 / H13
12810 17 - 18 FLUVIAL DEPOSITS
with c oal 18
NILAM - 115 12815
c lasts
LEVEL G51 13055 5-7
(distributary system ) Sand body 3 12820
m ostly 12 FLOOD-DOMINATED
(shoal)
13060
with large 12
12825 ESTUARINE SHOAL
17 to 20 ophiomorpha
11770 13065
(flood- 12830 (flood-dom in. 5-7
dom inated shoals)

overall bac kstepping

11775 13070
distributary 12835
m ainly system ) 6
11780 13075
fac ies 12840
TIDAL DEPOSITS
flooding surf.
overall forestepping

11 - 14 14
11785 13080
12845
m ostly 15 -

ABOUT 36 METERS
15
Sand body 2
11790 m ainly 14 13085
12 - 13 17 and FLUVIAL DEPOSITS
12850
with some partly 14 rare 20 (BASE INCISED
(fluvial)
11795 16 / 15 D 13090
ophiomorpha 12855 (valley fill) 17 VALLEY FILL ?)
(floods) (DF lobe)
11800 13095
flooding surf. 12860 6 (tidal 18 DEBRIS FLOW
11805 13100 m ainly 7 - 8 fac ies)
ophiomorpha 12865 14 MOUTH BAR
11 - 14
11810 13105 burrows m ostly 17 -
with some 12870
HCS and 20 and 15
16 - 19
11815 13110 c limb. ripples (valley fill) 6
c oaly sst. 12875 SUBTIDAL
(shelf lobes)
11820 burrowed 13115
siltstone 12880
PRODELTA
11825
Sand body13120
1 12885 3 - 4 with 4
(fluvial) 12890
layers of
8 and 19
(delta front 1(A)
12895 and prodelta)
2 BAY-FILL
12900
SEQUENCE
12905 3

Regional Marker 2 (TZ)

 Layering Concepts (1/2)
What is the appropriate model of layering? Parallel to top/base of marker or Proportional?

Regional Marker 1 (TZ)

Nilam-165 G53B Nilam-87 G53B Woodchoppers’ yard outcrop

(well
NILAM -core)
87
LEVEL G53B
1(A)
12805
17 FLOOD-DOMINATED
NILAM - 179
LEVEL G61 / H13
12810 17 - 18 FLUVIAL DEPOSITS
with c oal 18
NILAM - 115 12815
c lasts
LEVEL G51 13055 5-7
(distributary system ) Sand body 3 12820
m ostly 12 FLOOD-DOMINATED
(shoal)
13060
with large 12
12825 ESTUARINE SHOAL
17 to 20 ophiomorpha
11770 13065
(flood- 12830 (flood-dom in. 5-7
dom inated shoals)

overall bac kstepping

11775 13070
distributary 12835
m ainly system ) 6
11780 13075
fac ies 12840
TIDAL DEPOSITS
flooding surf.
overall forestepping

11 - 14 14
11785 13080
12845
m ostly 15 -

ABOUT 36 METERS
15
Sand body 2
11790 m ainly 14 13085
12 - 13 17 and FLUVIAL DEPOSITS
12850
with some partly 14 rare 20 (BASE INCISED
(fluvial)
11795 16 / 15 D 13090
ophiomorpha 12855 (valley fill) 17 VALLEY FILL ?)
(floods) (DF lobe)
11800 13095
flooding surf. 12860 6 (tidal 18 DEBRIS FLOW
11805 13100 m ainly 7 - 8 fac ies)
ophiomorpha 12865 14 MOUTH BAR
11 - 14
11810 13105 burrows m ostly 17 -
with some 12870
HCS and 20 and 15
16 - 19
11815 13110 c limb. ripples (valley fill) 6
c oaly sst. 12875 SUBTIDAL
(shelf lobes)
11820 burrowed 13115
siltstone 12880
PRODELTA
11825
Sand body13120
1 4
No Data “Hole”
12885 3 - 4 with
(fluvial) 12890
layers of
8 and 19
(delta front 1(A)
12895 and prodelta)
2 BAY-FILL
12900
SEQUENCE
12905 3

Regional Marker 2 (TZ)

 Layering (1/2)
If we have high deviated wells use build along the pilars, if we
have mostly vertical wells use vertical thickness. Anyway, Try
to avoid the deviated /horizontal wells

If use Proportional don’t click use minimum cell thickness because if the cell
thickness is less than the number you put, it will create “holes” in maps.

If use parallel to the top/base (better TZ1 than top/base of SS) then you can select
use minimum cell thickness but make sure that this numbber be higher then the cell thickness
 Facies modeling methods - overview

Deterministic Learning system

Estimation Direct Addressing Artificial

Indicator Kriging Assign values Interactive Neural Net

Discrete Choose from Allows the user to Uses the
distribution of the undefined, paint facies classification model
property honoring constant, other directly on the 3D made in the Train
the pre-defined property, surface model. Estimation Model.
histogram and vertical
function.
 Facies modeling methods - overview

Deterministic Learning system

Estimation Direct Addressing Artificial

Indicator Kriging Assign values Interactive Neural Net

 Facies modeling methods - overview
Stochastic
Pixel based Object based
Sequential Truncated Truncated Multi-point Object
Indicator Gaussian Gaussian Facies Modeling
Simulation Simulation Simulation Simulation
with trends
Distributes the Used mostly Distributes the The variogram Allows to
property, with facies based is replaced by populate a
using the carbonates on a transition a training discrete facies
histogram. where facies between image giving model with
Directional are known to facies and both the different
settings, such be sequential, trend facies and the bodies of
as variogram it deals with direction. The relative various
and large amounts trends are position to geometries,
extensional of input data, converted into each other, facies and
trends, are such as global probabilities to describing the fraction
also honored. fractions and then run TGS. spatial
trends. correlation
from one-to-
multiple
points.
 Facies modeling methods - overview

Stochastic
Pixel based Object based
Sequential Truncated Truncated Multi-point Object
Indicator Gaussian Gaussian Facies Modeling
Simulation Simulation Simulation Simulation
with trends
 Facies Modeling: Scale up Well Logs: NtG_disc

1
2

As lines
 Facies Modeling (NtG_disc) for determine Polygon

In settings use “NO seed”

Same variogram and same Method (SIS)

applied for Res and Non Res
Map of AVG most of NtG_disc from 10 realizations

1
4
Making the AVG Map Most of NtG_disc
Determining the Polygon from AVG Map most of NtG_disc

Polygon for Probability

AVG map

Polygons G53B

“Use esc to edit any point in polygon”

 Make/edit Surface: Areal Trends Maps
How to create a Surface Probability Map for input in Facies modeling

Don’t use smooth iterations

Areal Trends Maps

Z value inside polygon

P1

P0
 Facies Modeling (NtG_disc) using Probability Map

Use variograms and proportions in “res” and “non res”

Because impossible to assign values to “non res” or “res” by separate
 Facies Modeling (NtG_disc) using Probability Map

In settings use “NO seed”

Use trend in res Don’t use trend in non res

 Facies Modeling (NtG_disc) using Probability Map

3
4

2
Results in Facies Modeling (NtG_disc) using Probability Map
 Petrofacies: Winland/Pittman Method

Equations to Calculate Permeability & Swi

Porosity vs. Permeability
G52A, G53A, 53B, G55B, G56A, G57B,G61
Swi = ( 1/ Phie m ) 1/n Wells NLM 225, 179, 62, 65, 77D, 30, 36
Cte 1 1000.0000

100.0000

ka = (Cte 2 * Phie m *(1-Swi) ) 2

Permeability, mD @ 800psi
Swi 10.0000

1.0000

Equations to Calculate Pore throat Radius 0.1000

ka@ 800# = 0.0046e0.6564*Phie@800#
R2 = 0.7489
log(R10 ) = 0.459 + 0.500 * log(K aire ) - 0.385 * log(f  0.0100

log(R15 ) = 0.333 + 0.509 * log(K aire ) - 0.344 * log(f 

0.0010
log(R20 ) = 0.218 + 0.519 * log(K aire ) - 0.303 * log(f  0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0

log(R25 ) = 0.204 + 0.531 * log(K aire ) - 0.350 * log(f  Porosity, % @ 800psi

log(R30 ) = 0.215 + 0.547 * log(K aire ) - 0.420 * log(f 

Pittman log(R35 ) = 0.255 + 0.565 * log(K aire ) - 0.523 * log(f 
Equations log(R40 ) = 0.360 + 0.582 * log(K aire ) - 0.680 * log(f 
log(R45 ) = 0.609 + 0.608 * log(K aire ) - 0.974 * log(f 
log(R50 ) = 0.778 + 0.626 * log(K aire ) - 1.205 * log(f 
log(R55 ) = 0.948 + 0.632 * log(K aire ) - 1.426 * log(f 
log(R60 ) = 1.096 + 0.648 * log(K aire ) - 1.666 * log(f 
log(R65 ) = 1.372 + 0.643 * log(K aire ) - 1.979 * log(f 
log(R70 ) = 1.664 + 0.627 * log(K aire ) - 2.314 * log(f 
log(R75 ) = 1.880 + 0.609 * log(K aire ) - 2.626 * log(f 
 Calculating Petrofacies (R35) for use in modeling: Pittman Method

Equations to Calculate Pore throat Radius

Apex Graph log(R10 ) = 0.459 + 0.500 * log(K aire ) - 0.385 * log(f 
G Sandstones
Wells N60, N62, N77, N87, N225
log(R15 ) = 0.333 + 0.509 * log(K aire ) - 0.344 * log(f 
0.03 log(R20 ) = 0.218 + 0.519 * log(K aire ) - 0.303 * log(f 
0.025 log(R25 ) = 0.204 + 0.531 * log(K aire ) - 0.350 * log(f 
0.02 log(R30 ) = 0.215 + 0.547 * log(K aire ) - 0.420 * log(f 
SHg/Pc

0.015 log(R35 ) = 0.255 + 0.565 * log(K aire ) - 0.523 * log(f 

0.01 log(R40 ) = 0.360 + 0.582 * log(K aire ) - 0.680 * log(f 
0.005 log(R45 ) = 0.609 + 0.608 * log(K aire ) - 0.974 * log(f 
0 log(R50 ) = 0.778 + 0.626 * log(K aire ) - 1.205 * log(f 
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
SHg log(R55 ) = 0.948 + 0.632 * log(K aire ) - 1.426 * log(f 
log(R60 ) = 1.096 + 0.648 * log(K aire ) - 1.666 * log(f 
Porosity vs. Permeability vs. Pore Throat log(R65 ) = 1.372 + 0.643 * log(K aire ) - 1.979 * log(f 
Radius (R35) log(R70 ) = 1.664 + 0.627 * log(K aire ) - 2.314 * log(f 
G sandstones
1000 log(R75 ) = 1.880 + 0.609 * log(K aire ) - 2.626 * log(f 
10
Petrofacies
100 5
Megapores
ROCK TYPES
Air permeability, mD

2
10 Macropores
0.8
• Megapore Type Rock (R35 >10 µ)
1 Mesopores • Macropore Type Rock (R35 between 5 and 10 µ)
• Mesopore Type Rock (R35 between 2 and 5 µ)
0.1 Micropores
• Micropore Type Rock (R35 between 0.8 and 2 µ)
• Nanopore Type Rock (R35 < 0.8 µ)
0.01 Nanopores
0 5 10 15 20
Porosity, Dec
Calculating Petrofacies (R35) for use in modeling
 Creating new Template for Petrofacies

We generate new template named Untitled, in info tab change the name for
Template Petrofacies (R35), and in colors tab we fill out with nano, micro, etc and it colors
Calculating Petrofacies for use in modeling
 Scale up Petrofacies2

As lines
 Data Analysis: Petrofacies Modeling

Petrofacies
Most of NtG_disc with trend
Nano
Micro
Meso
Macro
Mega

Most of NtG_disc with trend

Copy/paste to all petrofacies

 Petrofacies Modeling

Most of NtG_disc with trend

In settings use “NO seed”

Most of NtG_disc with trend Most of NtG_disc with trend

In non Res: assign value=0 and no trend and no variogram

Petrofacies Modeling (Petrofacies Map)
Petrofacies Modeling (Petrofacies Map)
 Petrophysical Modeling

Stochastic method Deterministic method

Pixel based technique Interpolation

Sequential Gaussian Gaussian Random Moving average

Simulation algorithm Function Simulation
algorithm

Honors well data, input It is faster than SGS, and Based on the input it gives an
distributions, variograms and gives better variogram average value and calculates
trends. The variogram and reproduction. It also has a the weigths according to the
distribution are used to fast collocated co-simulation distance from wells.
create local variations, even option using interactive
away from input data. correlation-coefficient as
slide bar.
Sequential Gaussian Simulation (SGS)

SGS is a krig-based
stochastic method
• Needs a variogram.
• Will honor the distribution
of the input data (upscaled
logs).
• Petrel will automatically do
a Normal score
transformation before the
simulation, and will back-
transform the data.
• The output is distributed in
a blurry manner.
 Petrophysical modeling methods - overview

Deterministic
Estimation Interpolation
Kriging Kriging Kriging by Closest Functional Moving
Interpolation GSLIB average
Honors well It performs It has the It uses the I honors well Based on
data, input fastest. It option of closest well and trend the input it
distributions, has a co- collocated data input data gives an
variograms kriging co-kriging for each creating a average
and trends.It option and and you can unsampled 3D function value and
can work in allows user choose location. (parabolic, calculates
real to choose between simple the weigths
coordinates between ordinary or parabolic, according to
and it’s fast. simple and simple planar or bi- the distance
ordinary kriging. linear) used from wells.
kriging. in the
interpolation
.
 Petrophysical modeling methods - overview

Deterministic
Estimation Interpolation
Kriging Kriging Kriging by Closest Functional Moving
Interpolation GSLIB average
Petrophysical Modeling: Scale up Well Logs, NtG_cont

2
Data analysis for transformations NtG_cont

Most of Petrofacies

Don’t model the shale, only assign value. 1

For properties between 0 and 1

use 0.01 for minimum in input
and output truncation
 Data analysis for Variograms NtG_cont

Most of Petrofacies
2
Copy paste in all petrofacies
 Tips

Scale up: discrete variables (all related with facies) use as lines, continuous variables (petrophysical properties) use as points.
Vertical search radius: Which is the thickness of the zone to be analyzed? Divide the dimensions by 2 to get z.

Use vertical variogram to find the nugget and variogram type of your data.

Number of lags: does not need to be bigger than the number of wells in the analysis direction. I you have 5 wells in the mayor
direction , it is no necessary more than 5 lags.
Lag distance: which is the mean distance between your well? Try to approach the lag distance to this mean.
Vertical lag distance: Try to make this lag distance similar to the data spacing, cell thickness.
Lag tolerance: Use this when you have directional wells, use 25% of the lag distance.

Consider the bandwidth in case of directional wells. For vertical ones, use a small bandwidth. And do not use a big tolerance
angle because it will get large variability. Try to begin with 5 deg.

Simbox mode vs. real mode: while doing variogram analysis, it is recommended to click simbox mode for horizontal variography
as it will ensure that only samples from equivalent geological layers are compared. But for vertical variography Simbox is not to be
used.
Modeling NtG_cont
 Modeling NtG_cont

1 2

Most of Petrofacies

Repeat this in each Petrofacies except in shales Use trend in each Petrofacies
Final Map of NtG_cont
Petrophysical Modeling: Scale up Well Logs, Sw
 Data analysis for transformations, Sw

Most of Petrofacies

Copy paste in all petrofacies

SW[U]
Data analysis for Vertical Variogram, Sw

Most of Petrofacies

Copy paste in all petrofacies

Data analysis for Horizontal Variogram, Sw

Most of Petrofacies

Copy paste in all petrofacies

Data analysis for Horizontal Variogram, Sw

Copy paste in all petrofacies

 Modeling Sw

1 2

Repeat this in each Petrofacies Use trend in each Petrofacies

Final Map of Sw
Gas Water Contact in G53B
Petrophysical Modeling: Scale up Well Logs, PHIE
Variogram Map from the Scale up PHIE
Data analysis for transformations, PHIE

Copy paste in all petrofacies

Z
Y
X
Data analysis for Vertical Variogram, PHIE

Copy paste in all petrofacies

Data analysis for Horizontal Variogram, PHIE

Copy paste in all petrofacies

Data analysis for Horizontal Variogram, PHIE

Copy paste in all petrofacies

Modeling PHIE

Copy paste in all petrofacies

Repeat this in each Petrofacies

Final Map of PHIE
 QC Results

QC results in a histogram:
• Go to the Settings for the Property 1
and select Histogram tab 2
• Check that the Histogram follows
the distribution from:
Raw logs

Upscaled cells

3D grid

Filter:
1) Use Zone filter
2) Filter on other property values by
pressing the filter button and go to
Property filter in Settings for the
Properties folder.
Volume Calculation
 Principle

Data that can be used: Exact volume calculation

• Boundary, license block Enhanced volume calculation using a
• Zones, Segments triangulation technique.
• Properties (Net/Gross, porosity, Sw)
• Constants (Bo, Bg, GOR oil, GOR gas)
• Contacts (GOC, OWC, GWC)
• Recovery factors (Rfo, RFg)
 Process

Cases pane
Running a volume calculation will create a Case, which will
have a folder in the Cases pane and a volume calculation sub
folder. This will act as a filter for viewing results.

Results pane
Acts as a filter for outputs of the Volume Calculation. The
standard parameters will be calculated for all volume
calculations as long as the appropriate input is supplied.

Note: The Cases and results panes are also where Simulation
cases and results will be stored.
 Process Dialog – Create a New Case
(Properties Tab)

1) Define name for the new case or use an

existing Case name. 1

2) Select which 3D Grid from the drop-down 2

menu to run the Volume Calculation on.

3) Define the Hydrocarbon intervals to use. 3

You may use the Contacts created earlier.
4
4) Select the contacts from the Fluid Contacts
folder in 3D grid and drop them into the
Contacts tab using the blue arrows.
 Process dialog – Create New Case
(Properties Tab)

5) Define Oil and Gas properties. Use either

an already made property (pull down menu)
from the grid, or select the Constant property
box and type in a value.
5
6) Under the General Properties tab, activate
N/G and Porosity properties or use constants.

6
 Process dialog – Create New Case
(Results pane)

7) In the Results pane; select which properties

and volume height maps to generate if needed.

8) When Volume Calculation is run, the

properties are stored in the Properties folder of
the selected grid (useful to create subfolders if
there are many runs).

9) Maps are stored

in a Volume Maps 8
folder in the Input
pane. 7

9
 Process dialog – Create New Case
(Boundaries Tab)

10) Click Apply to save settings and Run to 10

run the Volume Calculation. View output

11) Use filters (A) or Well Boundaries (B) if

desired. Click Run again

11A

11B
Volume Calculation
Volume Calculation
Volume Calculation
 Output – Volume Maps

1 2

1) Volume Maps are stored in the Input pane.

2) Maps can be displayed in a Map window

(e.g. STOIIP).

3) The Volume maps can also be shown in 3D 3a

– draped over a topographic surface:
a) Output a depth Surface from Horizon
b) Copy the STOIIP map as a surface attribute
and paste it onto the new depth surface. 3b

4) Click Apply and view the 4

surface in 3D window.
Uncertainty Analysis

Don’t use average maps

Uncertainty Analysis
Uncertainty Analysis
Uncertainty Analysis

3
 Porosity vs. Permeability correlated with Facies
Porosity vs. Permeability vs. Pore Throat
Radius (R35)
G sandstones 100000
1000

10
10000

Permeability [mD] (mD)

100 5

1000
Air permeability, mD

2
10
100
0.8

1
10
Distributary channels
0.1 1
Distributary channels with
tidal influence
0.01
0.1 Crevasse splays
0 5 10 15 20
Porosity, Dec Tidal bar
0.01
0 5 10 15 20 25
Porosity (%) Pedogenic profiles

Expon. (Distributary
channels)
 Creating new Template for Petrofacies

We generate new template Untitled, in info tab change the name for Template Petrofacies (R35),
and in colors tab we fill out with nano, etc and colors