Вы находитесь на странице: 1из 16

SPE-181024-MS

Recovery Factor Prediction for Deepwater Gulf of Mexico Oilfields by


Integration of Dimensionless Numbers with Data Mining Techniques
Priyank Srivastava and Xingru Wu, University of Oklahoma; Amin Amirlatifi, Mississippi State University;
Deepak Devegowda, University of Oklahoma

Copyright 2016, Society of Petroleum Engineers


This paper was prepared for presentation at the SPE Intelligent Energy International Conference and Exhibition held in Aberdeen, United Kingdom, 6 8 September
2016.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
The objective of this paper is to integrate Big Data concept with petroleum engineering knowledge for
the prediction of recovery factor in Deepwater Gulf of Mexico (dGOM) oilfields. Recovery factor is
affected by many geological and engineering factors; as a result, there is no explicit approach to accurately
calculate the recovery factor. This is particularly true for deepwater development as the parameters
associated with the recovery factor estimation have significant uncertainties and usually extremely costly
to obtain. Typically, the recovery factor of a field is estimated using analogs, material balance, decline
curve or numerical simulation. These deterministic approaches requires representative geological models.
However, enough information is often not available to capture the realistic flow. In addition, the estimated
recovery factor can be very different using different methods. Reservoir engineers are faced with the
challenging task of estimating recovery factor by optimizing a large number of parameters with limited,
sometime inaccurate information. This dilemma calls for an alternative approach in handling the noisy
data.
Data mining and classification identify hidden patterns in unstructured data and tend to be fairly robost
in the presence of noisy data. Using a database of 395 Deepwater Gulf of Mexico (dGOM) oilfields with
84 attributes, a set of dimensionless numbers are calculated for 59 oilfields with water drive. This helps
in dimensionality reduction and scaling of reservoir models for comparison. Based on the distribution of
dimensionless numbers, data mining techniques like K-means clustering followed by principal component
analysis (PCA) are used for classifying oilfields into four categories. Subsequently, partial least square
(PLS) regression is used for relating dimensionless numbers to recovery factor from sparse data in dGOM,
which gives good coefficient of correlation for some clusters.
This paper shows that dimensionless numbers together with data mining techniques provides a new,
easy to implement method for predicting recovery factor for large datasets where application of other
methods are limited due to requirement of high computational cost and time.

SPE-181024-MS

Introduction
Gulf of Mexico (GoM) basin encompasses a well-established petroleum system from which the first
deepwater project (water depth 1000 ft.) was delievered in 1979 (Lach, 2010). Conventional hydrocarbon production is becoming increasingly focused on deep and ultra-deep water GOM (Disenhof,
Mark-Moser, and Rose 2014). Developers of offshore energy face great challenges to ensure safety of
environmental system in harsh working environments; thus, it becomes imperative to have a proper risk
and uncertainty assessment for these deepwater fields. The success of drilling, completion and production
strategies depends on the quantifiable accuracy of reservoir characterization. Traditionally, performance
prediction is accomplished using either analytical material balance equations or numerical reservoir
simulation. However, both methods adopt a bottoms-up workflow that suffers from a drawback: the
requirement of accurate representation of subsurface system.
With the digitization of information and rise of inexpensive sensor technologies, the petroleum industry
has entered a new era of computing big data where the acquired data can be used to show hidden patterns
and trends. This method of computing is very efficient and effective in solving inverse problems in which
parameters affecting system behavior are not completely known. Hydrocarbon reservoir provides a classic
case of a natural system where, engineers have limited control on the system they work with; thus, they
have to rely on indirect measurements to determine the properties of the system/reservoir and use these
properties to predict future trends of production behavior. Data mining, on the other hand, provides an
alternative top-down intelligent modeling approach which uses the measured reservoir properties as the
basis for modelling (Mohaghegh 2000). Data analytics process provides an alternate method to enhance
reservoir modelling by integrating structured data with unstructured data. Data mining workflow is free
from issues inherent in the traditional deterministic reservoir simulation models.
Different methods of analytics from data mining deliver efficient and flexible workflows for reservoir
managment along with the quantification of uncertainty and risk assessment of various parameters used
in the analysis (Holdaway 2014). The workflow of data analytics as described by Kikani (2013) can be
summarized as follows:

According to this workflow, data mining model is a precursor to analysis, models shows the hidden
patterns and trends in dataset on which analysis is done. The workflow of data mining provides an
alternative solution to deepwater development, and here we use Gulf of Mexico (GoM) as an example to
demonstrate its application. In this example, we show that data mining methods provide a novel ways to
use all the data from geoscientists and engineers to find and produce hydrocarbons, to automate simple
decisions & guide harder ones, ultimately reducing the risk and resulting in finding and producing more
oil & gas.

Geological description of GoM


Gulf of Mexico Outer continental shelf (OCS) planning area is divided into three zones: western, central
& eastern. Each zone is subdivided into a number of protraction areas containing different sands and
blocks. Figure 1 depicts the location of different types of producing fields in GoM. Optimal combination
of rate of sedimentation and presence of abundant organic matter supported formation of GoM hydrocarbons, while impermeable salt and shale provide the seals and base rocks. Majority of depositional
history of the basin is concentrated to period after middle Jurassic (Galloway (2008), Woods, Salvador,
and Miles (1991)). Figure 2 shows the local stratigraphic cross section along with generalized crosssection for the depositional model of GoM continental shelf region. The complex interaction of GoMs
depositional history with its structure and salt tectonics has resulted in a heterogeneous system that
requires varied techniques to evaluate the risk of hydrocarbon extraction (Disenhof, Mark-Moser, and

SPE-181024-MS

Rose 2014). According to Galloway (2008), GoM basin reached present structural configuration in the
early cretaceous period. Buggler and Thomas (1994) discuss the rifting and subsidence in the Mesozoic
era, which led to middle Jurassic deposition of evaporites layers that form Louann salt in the northern
GoM. This region influences modern geography and bathymetry of the GoM basin as well as the
underlying hydrocarbon system in the region (Hudec et al. 2013). Early cretaceous deposition was
dominated by carbonates and evaporites on the shelves and shallow marine clastics on northern basin rims
(McFarlan Jr and Menes 1991), while large volume of clastics were deposited in northern GoM during
Cenozoic period (Salvador 1991).

Figure 1Location of fields producing Oil (light green), Gas (Red) and Condensate (Dark green) in Central and Western GOM scaled
at 1:200000. Eastern GOM is relative unexplored.

Figure 2Local Stratigraphic section for GOM sands (Slatt and Zou 2014). North-south Generalized Cross section for offshore Northern
GOM deposits (Galloway 2008, Slatt 2006). GOM deposits are characterized by Louann salt, which is sandwiched between organic rich
Miocene sands and Plio- Pleistocene deposits.

SPE-181024-MS

Using available data for wells in GoM including information on geology, geophysics, reserves,
production and infrastructure, we applied various data mining and predictive analytics algorithms to
estimate recovery factor. Instead of using traditional deterministic methods such as material balance or
numerical simulation, we use data driven analytics to estimate the recovery factor.

Data Mining Methodology for dGOM dataset


Using the database from BOEM (2013), we classified fields with GOR 9,700 SCF/STB as oil
producers. This exercise resulted in 395 oil fields and 905 gas fields, indicating that GoM is mainly a
gas-prone basin. Figure 3 displays the classification of data present in GoM dataset (the definitions of
acronyms are available in the nomenclature of this paper). Gas volumes (Mscf) is converted to barrels of
oil equivalent (BOE) and is used in calculations. Data for deepwater oilfields (Sand typeO) is selected
for this study.

Figure 3Lists of attributes Used for Data-Mining study on GOM dataset.

This preliminary classification of GoM oilfield dataset yield a subset database consisting 395 oilfields
with 84 attributes. Attributes are grouped together in four classes as shown in Figure 3. While most of the
geological attributes are result of seismic interpretation, the engineering attributes such as reserves,
petrophysical and PVT properties are reported by operators to the federal government. In addition to
correlations for recovery estimation proposed by many researchers (Craze and Buckley (1945), Vietti et
al. (1945), Muskat and Taylor (1946), Guthrie and Greenberger (1955), Gulstad (1995)), while Noureldien
and El-Banbi (2015) used neural networks for estimation of recovery factor. The objective of this study
is to use other data mining techniques to explore the relationship between the recovery factor and these
attributes using historical data in dGOM oilfields. The study result may be particular to a region, but the
framework and working procedure can be exteneded to other regions.
Data Preparation & Exploratory Data Analysis
Figure 4 describes the workflow used in the integration of all the measured and calculated data for
prediction of recovery factor on dGOM reservoirs. Scaling of attributes and removal of outliers is an

SPE-181024-MS

essential step for preparing data as an input to various data mining algorithms. In this step, data quality
is checked by inspecting or removing impaction of dataset. For example, some redundant and unuseful or
missing valves from the database was inspected or removed. Out of the original dataset containing 13303
sands with 90 attributes, based on our objective and prior knowledge of estimation of recovery factor, we
have selected attributes given in Figure 3 as an input to data mining workflows. Table 1 gives number of
missing attributes in GOM sand dataset.

Figure 4 Workflow for prediction of recovey factor in dGOM oilfield dataset


Table 1Number of missing items in dGOM oil sands (See
Nomenclature for definition of attributes)
Attribute name

No. of missing rows

FTRAP1
FTRAP2
BHCOMP
Field production rate

6
197
135
9

Although more data helps to increase confidence in models. Nevertheless, presence of high number of
attributes significantly increases the complexity of the model and degrades the performance (Kotu and
Deshpande 2014). Reducing the number of attributes, without significant loss in the performance of the
model, can be done either by ranking attributes in order of their sensitivity towards the objective function
(target variable). But, this technique is completely mathematical and does not take into account any
physical process.
Figure 5 shows scatter plot of reservoir and fluid properties on which performance of reservoir
classically depends. The scatter of data illustrates complexity of basin where various geological, reservoir,
completions and operating constraints interacts resulting in definite behavior of fields in terms of
production. This behavior is quantified by physical quantities like production rates of fluids and gases.
Figure 5 shows wide variation for reported recovery factor for reservoirs having comparable porosity,
permeability, sand thickness and fluid properties. While data is highly multicolliear and scattered. Data
mining algorithms are likely to show patterns and clusters for reservoir with identical characteristics.

SPE-181024-MS

Figure 5Exploratory data analysis for dGOM oilfields. According to linear model (Red line) Recovery factor (ORF) shows positive
correlation with porosity, permeability, oil API, Boi & Rsi while recovery factor is negatively related with water saturation (SW), net pay
thickness (OTHK). However, due to multidimensionality and interdependence of different variables it is difficult to perceive the
individual effect of each variable on recovery factor.

SPE-181024-MS

Figure 6 K-means clustering on original attributes for 395 deepwater oilfields in PCA space.

PCA is used for reducing dimensionality of system and visualization of clusters for 395 dGOM
oilfields. PCA is run on original attributes as shown in Error! Reference source not found., the clusters
were obtained using distance based K-means algorithm. However, PLS regression on these clusters is
unable to predict recovery factor with accuracy as shown in Figure 7. In addition, first two principal
component (PC1 & PC2) are able to capture only 30% of variance in data. Therefore, reservoir models
needs to be scaled before application of PLS algorithms, which is done for water drive reservoir in next
section.

SPE-181024-MS

Figure 7PLS regression for clusters obtained on 395 dGOM oilfield show very low correlation. Maximum coefficient of correlation is
0.2 for cluster 3. To conclude, original attributes used in PLS are not able to predict recovery factor with accuracy

Generation of Dimensionless numbers


As the datamining is a pure mathematical tool in analyzing data, and the outcome should be meaningful
in physics also. Additionally, if too many attributes are analyzed, the outcome from the analysis can be
meaningless. Dimensionless numbers provides a way to scale and reduce dimensionality of dataset.
Consequently, performance from different fields can be compared. These numbers deliver insights into the
relative importance of driving forces such as viscous, gravity and capillary forces on the fluid flow in
porous media (Shook, 1992). There are many different definitions of dimensionless variables reported in
the literature (Shook et al., 1992; Rapoport, 1955). Dimensionless numbers lead to scaled attributes based
on which reservoir performance from a variety of fields is compared. Dimensionless numbers used in this
paper were developed for an immisible displacement of oil by water in a homogenous reservoir. Hence,
only 59 deepwater oilfields with water drive is used for application of dimensionless group to GoM
dataset.
Four dimensionless numbers generated by Shook, Li et al. (1992) based on forces (gravity, viscous,
capillary & dispersion) controlling the displacement process, described in Table 2 : List of Dimensionless
Numbers, are used in this study with the following assumptions:
Table 2List of Dimensionless Numbers

SPE-181024-MS

1. The dimensional group was developed for immisible displacement of oil by water for a homogenous reservoir. Relative permeability function and contact angle for all the reservoir are considered same. Each well is centered in a cylindrical drainage area, surface area and thickness of which
are known.
2. Fluid flow velocity is estimated through dividing the average daily production rate for the block
by the surface area. Average daily production rate is defined as the ratio of total cumulative
production to number of days the field has produced.
3. Viscosity of water in reservoir condition is assumed constant as 0.5 cp.
4. Relative permeability is obtained from Coreys relationship (BrooksRH, 1964):

5. Interfacial tension between hydrocarbon and water system is calculated using (Firoozabadi, 1988):

Kw 11.7 (Watson acterization factor)


Reservoir temperature is calculated using the static temperature gradient reported in the database.
K-Means Clustering & Predictive Modelling
59 oilfields having water drive were selected for recovery factor prediction after field classification using
K-means and PCA. K-means distance based clustering technique is used to obtain clusters based on
dimensionless numbers (Figure 8), while number of clusters are optimized based on elbow plot Figure
9. This figure shows cluster 2 (Red triangles) as having low recovery, which corresponds to lower aspect
ratio (Rl) and high capillary (Npc), gravity (Ng) numbers. Cluster 1 (Black circles) have intermediate
recovery factor. While cluster 3 (Green plus) and cluster 4 (Blue crosses) have high to intermediate
recovery factor range. These clusters correspondingly relates with dimensionless numbers. Cluster 4
with smaller Ng values displays higher magnitude of density number and aspect ratio implying these
reservoirs to be dominated by magnitude of residual oil saturations and relative permeability. However,
Cluster 4 have only four data-points therefore is statistically insignificant. Cluster 1 have highest
recovery factor owning to lower magnitude of capillary number, density number and aspect ratio. The
figure also shows that at intermediate values of Npc clusters overlap each other in recovery factor. The
reason for this behavior is not completely known, but it can attributed to difference in petro-physical
(heterogeneity) characteristics of reservoirs, which dimensionless groups does not capture. Figure 10
shows the statistical distribution of dimensionless numbers for different clusters. It is discovered that when
a cluster is having higher Npc, its Ng is also higher. Another way to visualize clustering is through
biplots in principal component (PCA) space. This helps in reducing the system to two dimensions
therefore; helping in visualization of clusters is easier.

10

SPE-181024-MS

Figure 8 Scatter plot for dimensionless numbers according to K-means clustering. (npc is Capillary no., ng is gravity no.,denn is
density no.,rl is aspect ratio). Picture on right is Elbow Plot to optimize number of clusters.

SPE-181024-MS

11

Figure 9 Selecting number of clusters for K-means

Figure 10 Box & whisker plot for two clusters obtained from clustering.

12

SPE-181024-MS

Figure 11 illustrates biplot, as well as the clusters obtained in PC space.

Figure 11Visualization of K-means clustering on dimensionless numbers in PCA space. Angle between the arrows reveals the
correlation between two variables. Thus, gravity no. (Ng) is closely correlated with capillary no. (Npc) while aspect ratio (Rl) and density
number (denn) behavior is similar.

To determine the correlation between recovery factor and the aforementioned dimensionless numbers,
partial least square regression (PLS) is applied on dimensionless numbers. This technique was chosen due
to multicolliearity of dimensionless data points. PLS finds the predictors (dimensionless numbers) which
are relevant to target variable (recovery factor (RF)) in contrast to PCA in which principal component
explains only predictors, here we use univariate PLS as we have only one target variable. Figure 12
demonstrates accuracy of match for different clusters between reported RF and predicted RF by using
PLS regression on dimensionless numbers. Cluster 3 & 4 does not show a good correlation. Reason
for this anomaly is not completely known. However, it can be attributed to inability of used dimensionless
groups to scale reservoirs for these classes properly, as present dimensionless group assumes all reservoirs
to be petrophysically (relative permeability and wettability) similar.

SPE-181024-MS

13

Figure 12Result of PLS regression on dimensionless variables for prediction of recovery factor. It gives R2 of 0.92 (cluster-1), 0.61
(cluster-2), 0.1 (cluster-3), 0.4 (cluster-4)

Figure 13 shows how various attributes used in PLS regression are related to each other, where Y is target
variable (RF). It is observed from Figure 13 that Npc, Dn lay close to each other and thus are positively
related with each other, while Dn and Rl lay in different quadrants, implying these two groups oppose each
other and the resulting recovery factor is a combination of interaction of all the dimensionless numbers.

Figure 13Correlation circle for Cluster-1 showing relationship of various dimensionless attributes used in PLS regression. Y is
recovery factor

14

SPE-181024-MS

Using PLS on dimensionless numbers leads to following equations for different clusters:
Cluster 1 PLS equation
Cluster 2 PLS equation
Cluster 3 PLS equation
Cluster 4 PLS equation

Conclusions
By analyzing the database from GoM using data mining techniques, this paper shows an integrating
procedure by including the physics based dimensionless numbers with mathematically techniques for
recovery factor prediction. Even though further research is needed to describe what kind of correlation
may imply causality, by using dimensionless numbers we make system unit and scale independent;
therefore, this type of correlation is more close to causality to be applied to any reservoir in general.
From the effort, we can draw the following conclusions:

We successfully classified the main influencing factors on recovery from a database with wide
assortment of reservoir and geological properties. K-means clustering on dimensionless numbers
was able to classify oilfields into four classes.
Dimensionless numbers is extremely useful in reducing multidimensional dataset to characteristic
features that can be used for recovery factor prediction. These dimensionless variables are well
defined in petroleum literature and have physical meanings.
Through application of PLS regression, we were able to get high correlation coefficient for some
of the clusters in deepwater GoM oilfields with water drive. However, cluster 3 & 4 shows very
low R2, reason for this is not completely known. Higher coefficients of aspect ratio (Rl) and
density number (Dn) suggest strong influence of reservoir fluid characteristics (Dn) and reservoir
geometry (Rl) in estimation of recovery factor. While gravity and capillary numbers may become
important for shallower reservoirs with higher dip angles.
The obtained dimensionless variable based correlation can be used in recovery prediction with an
improved confidence. The coefficients of the correlation highlight the importance of a particular
dimensionless variable. Therefore in field development, the uncertainities associated with the
dimensionless variables that have high impact on recovery should be managed.

Acknowledgements
The authors thank the University of Oklahoma for providing financial support for the completion of this
study. We would also like to acknowledge Earth Science Associates for donating GOM3 software for
academic purposes.

Nomenclatures
Symbols
Kx
Kz
Ut

:
:
:
:
:

Average horizontal permeability of reservoir, md


Vertical permeability of reservoir, md
Total fluid velocity (oilwater), ft./day
Relative mobility of residual phase-2
Density of non-wetting liquid phase

SPE-181024-MS

Dn
Krw
l
Ng
Npc
Rl
hw
H

15

:
:
:
:
:
:
:
:
:
:
:

Dimensionless density number


Relative permeability to water
Length of reservoir, ft.
Dimensionless gravity number
Dimensionless capillary number
Dimensionless aspect ratio
Interfacial tension of hydrocarbon-water system, dynes/cm
Net Pay Thickness, ft.
Density difference between oil-water densities
Dip-angle,
Porosity

Acronyms
API
BHCOMP
BOI
dGOM
DISCOIL/BOE
FCLASS
FSTRU
FTRAP1
FTRAP2
GOR
OAREA
OIP
ORECG
ORECO
ORECO_AF
ORF
ORP
OTHK
P_CUMOIL/BOE
P_RECOIL/BOE
PC1
PI
RSI
SDPG
SDTG
SS
SW
TCNT
TI
UCNT
WDEP

Oil API degree


Total number of bottom-hole completions in well
Initial formation volume factor
Deepwater Gulf of Mexico
Discovered oil/ bbl. oil equivalent, bbls
Field class
Field structure code
Field primary trap type
Field secondary trap type
Gas-Oil ratio, scf/bbl
Oil zone area, acre-ft
Oil initially in-place, bbls
Oil recoverable from gas reservoir, bbls
Oil recoverable from oil reservoir, bbls
Recoverable oil/acre-ft
Oil recovery factor
Produced GOR from oil reservoir, Mscf/stb
Oil zone thickness, ft
Proved cumulative oil/ bbl. oil equivalent, bbls.
Proved recoverable oil/ bbl. oil equivalent
First Principal component
Initial Pressure, psi
Solution Gas-Oil ratio, scf/bbl
Static Pressure Gradient, psi/ft
Static Temperature Gradient, degF/100 ft.
Subsea depth, ft.
Water Saturation
Total number of sand intersected by well
Initial Temperature, degF
Number of under-saturated sand encountered in the well
Water depth, ft

16

SPE-181024-MS

References
Bascle, B. J.; Nixon, L. D.; Ross, K. M. Atlas of Gulf of Mexico Gas and Oil Sands as of January 1, 1999; OCS Report;
U.S. Department of Interior: New Orleans, LA, 2001.
BOEM, 2013a, Deepwater Natural Gas and Oil Qualified Fields. http://www.data.boem.gov/homepg/data_center/other/
tables/deeptbl2.asp
BOEM, 2014, Resource Evaluation Program, Bureau of Ocean Energy Management, U.S. Department of the Interior.
http://www.boem.gov/Resource-Evaluation-Program/
Craze, RC, and Stuart E Buckley. 1945. A factual analysis of the effect of well spacing on oil recovery. Drilling and
Production Practice.
Disenhof, Corinne, MacKenzie Mark-Moser, and Kelly Rose. 2014. The Gulf of Mexico Petroleum SystemFoundation
for Science-Based Decision Making. Journal of Sustainable Energy Engineering 2 (3):225236.
Galloway, William E. 2008. Depositional evolution of the Gulf of Mexico sedimentary basin. Sedimentary basins of the
world 5: 505549.
Gulstad, Rick L. 1995. The determination of hydrocarbon reservoir recovery factors by using modern multiple linear
regression techniques. Texas Tech University.
Guthrie, RK, and Martin H Greenberger. 1955. The use of multiple-correlation analyses for interpreting petroleumengineering data. Drilling and Production Practice.
Holdaway, Keith. 2014. Harness Oil and Gas Big Data with Analytics: Optimize Exploration and Production with Data
Driven Models: John Wiley & Sons.
Hudec, Michael R, Ian O Norton, Martin PA Jackson, and Frank J Peel. 2013. Jurassic evolution of the Gulf of Mexico
salt basin. AAPG bulletin 97 (10):16831710.
Kikani, Jitendra.. 2013. Reservoir surveillance.
Kotu, Vijay, and Bala Deshpande. 2014. Predictive Analytics and Data Mining: Concepts and Practice with RapidMiner:
Morgan Kaufmann.
McFarlan, Edward, Jr, and L Silvio Meness. 1991. Lower Cretaceous. The Gulf of Mexico Basin: Boulder, Colorado,
Geological Society of America, The Geology of North America, v. J:181204.
Mohaghegh, Shahab. 2000. Virtual-intelligence applications in petroleum engineering: Part 1Artificial neural networks. Journal of Petroleum Technology 52 (09):64 73.
Muskat, M, and MO Taylor. 1946. Effect of Reservoir Fluid and Rock Characteristics on Production Histories of
Gas-Drive Reservoirs. Trans. AIME 165 (78):8.
Noureldien, Darhim M, and Ahmed H El-Banbi. 2015. Using Artificial Intelligence in Estimating Oil Recovery Factor.
SPE North Africa Technical Conference and Exhibition.
Salvador, Amos. 1991. Origin and development of the Gulf of Mexico basin. The gulf of Mexico basin:389 444.
Slatt, Roger M. 2006. Stratigraphic reservoir characterization for petroleum geologists, geophysicists, and engineers. Vol.
61: Elsevier.
Slatt, Roger M, and Fuge Zou. 2014. Turbidite Petroleum Geology in the Deepwater/Subsalt Gulf of Mexico.
Vietti, WV, OF Thornton, JJ Mullane, and AF Van Everdingen. 1945. The Relation Between Well Spacing and
Recovery. Drilling and Production Practice.
Woods, RD, Amos Salvador, and AE Miles. 1991. pre-Triassic. The Gulf of Mexico Basin: Geological Society of
America, The geology of North America, v. J:109 130.

Вам также может понравиться