PDF of article - Earth Science Associates

John D. Grace, Scott Morris and Tony Dupont
Earth Science Associates
Long Beach, CA
In the Gulf of Mexico, there are hundreds of millions of barrels of
oil and trillions of cubic feet of natural gas “forgotten” by lessees
in reservoirs shut-in before 2010. A study by Earth Science
Associates (ESA), based on decline curve analysis of 59,025
completions and 27,328 reservoirs, located and evaluated all
abandoned accumulations and high-graded opportunities that
would be attractive today. Additional “data-mining” analyses of
well test records and 35 years of independent reserve estimates
for all fields and sands in the Gulf, made by the US Minerals
Management Service/Bureau of Ocean Energy Management
(MMS/BOEM), identified further prospects for new production.
The study generated over 2,000 significant leads – all within the
bounds of producing or abandoned fields in the Gulf. These
were filtered out of databases with up to 7 million records,
identifying five key types of opportunities:
1. Over 500 shut-in oil reservoirs with more than 100,000
barrels (bbls) of recoverable oil remaining and 29 with
more than 1 million bbls estimated by decline curve
analysis. For gas reservoirs, decline curves show more
than 1,400 reservoirs with over 1 billion cubic feet (bcf)
remaining and 57 with more than 10 bcf.
2. Hundreds of reservoirs shut-in before production began
to decline and whose final production rates were higher
than the average for today’s on-line reservoirs.
3. Over 100 completions with prospective initial test rates
that never produced.
4. Dozens of fields where the MMS/BOEM wrote down
large reserve volumes in the past but some of which
represent possibly profitable targets at today’s prices,
costs and technology.
5. A dozen sands booked by the MMS/BOEM when they
were discovered, which never produced and were
subsequently taken out of the government’s inventory
of accumulations.
The study required development of automated batch
estimation of tens of thousands of decline curves, statistical
identification of significant breaks in production data to refine
decline curve estimates and machine processing operatorassigned reservoir names to define reservoirs usable in the
analysis. These techniques were also applied to finding
discovered sands that slipped from inventory and promising
completions that never produced. Both time-series and spatial
analytic techniques were focused on mining the government’s
reserve histories of the Gulf’s fields and sands.
The heavy computing was done in R, a widely respected opensource statistical package; the spatial analysis and integration of
input data and outputs was accomplished in ESA’s GOM
system, which is the industry standard for geographic
information systems (GIS) in the GOM and is written on ESRI’s
ArcGIS platform. Only data publically released by MMS/BOEM
was used. As the study was of “forgotten” oil and gas, only
targets abandoned before 2010 were evaluated.
Batch Decline Curve Analysis
Decline curve analysis of the production from oil and gas wells
entered petroleum engineering nearly a century ago. Most of
the mathematical foundations and practice were formalized by
J.J. Arps in his seminal 1945 article (1). The basic principle is
that, after production peaks, future output can be forecasted by
the rate at which post-peak production declines. Originally,
these estimates were made manually: an engineer plotted
production rate versus time (or cumulative production), “eyeballed” a line best fitting the data and drew it in with a straightedge from the peak through the last production observation.
Extending that line predicts future output and remaining
producible volumes can be calculated from the predicted
production rates (Figure 1).
Eventually, computers replaced graph paper and rulers.
Software developed to estimate the four mathematical classes
of decline models: exponential and hyperbolic functions of
production rate versus time; a linear model of rate versus
cumulative production and finally, for gas wells, adjusted
pressure (P/Z) versus cumulative production. Because of
insufficient quality pressure data for the GOM, the P/Z model
could not be run in the study. Typically, estimation is
deterministically reported, not including confidence intervals on
either future production or remaining resources. Also, decline
curve analysis is almost always run by an engineer studying one
well or reservoir at a time.
The objective of the Forgotten Oil and Gas study, however, was
to estimate tens of thousands of decline curves and
Published in Oil & Gas Journal, August 4, 2014
quantitatively evaluate the statistical confidence in the results
of each. The algorithm for batch analysis estimated the three
alternative models for each reservoir and then determined
which curve was the statistically best fit. It also accommodates
a major problem engineers face and solve when performing the
analysis on individual wells – breaks in production. Most wells
experience occasional unusual drops in output or are
temporarily shut-in; sometimes this has little effect on the
overall decline. More commonly, an engineer must adjust the
decline curve, often starting the analysis after the break in
production (i.e., a significant change in slope). Processing tens
of thousands of production histories required finding these
breaks automatically and adjusting the starting (peak) point of
the decline analysis as necessary. A typical run for all GOM
reservoirs takes about 48 hours.
To identify “forgotten” oil and gas worthy of remembering, the
production rate at which interesting reservoirs were abandoned
many years ago must be greater than the rate at which the
same reservoir would be abandoned today. Abandonment rates
for similar reservoirs, on average, have fallen over time (Figure
2 shows abandonment rates for reservoirs with depths <7,000
feet and in <300 feet of water). Some of this change was due to
higher prices, some to improved technology that lowered costs
and made it profitable to let a well flow to a lower rate. Some
was because as infrastructure on the shelf became denser, both
production and transport became cheaper. Working in the
other direction, water depths have increased, as have reservoir
total vertical depths (TVD), raising marginal and average costs
of production for those wells, which boosts the minimum flow
rate at which they are abandoned.
Production Rate (BOPD)
Today’s abandonment rate
adds 21 months of
production & ~ 100,000 bbls
Last Production/
Original Abandonment
Rate: 148 bopd
Rate: 105 bopd
Fig. 1
To estimate “today’s” abandonment rates, a
separate study was made of the 2,065 GOM
reservoirs shut-in between 2010 and 2012.
These newly shut-in reservoirs were placed in
a matrix based on ranges of gas-oil ratio,
water depth and reservoir TVD. This allowed
fair comparisons: the final production rate of
each reservoir shut-in before 2010 was
compared to the final rates from the 20102012 set which shared similar gas-oil ratio,
water depth and reservoir TVD.
With the best-fitting model of the decline
from each reservoir shut-in before 2010 and
an estimate of the production rate at which it
would be shut-in today, the remaining producible oil and gas
was calculated, along with 10%/90% confidence intervals on
that estimate. Additionally, for each reservoir analyzed,
statistics describing the goodness of fit of the best regression
model were reported for assessment of the quality of the
12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63
Number of Months on Production
Preparing Data for Decline Analysis
Before the decline curve algorithm could be run, two
preparatory steps were completed: aggregating completions
into reservoirs and determining the rate at which each reservoir
abandoned before 2010 would be abandoned under today’s
technical and economic conditions.
Only one completion was required to produce 78% of all GOM
reservoirs; the balance needed more than one. For those, the
monthly oil, gas and water from each completion must be
summed to obtain a production profile for the entire reservoir.
If the productive intervals in every well were consistently
named, aggregation would be easy. Unfortunately,
inconsistencies arise due to different practices between
operators, idiosyncratic hand-recording and simple entry errors.
In the easiest cases, a completion in the “A1” reservoir needs to
be joined to completions labelled “A-1”, “A 1” and “A/1”. To
check the reliability of these re-assignments, all completions
grouped into a single reservoir were required to be within
specific lateral and vertical distances of each other.
Results of Decline Curve Analysis
Ignoring statistical confidence and minimum volumes remaining
in each of 27,328 reservoirs, the decline curve analysis found
2.81 billion barrels of oil equivalent (BOE) of remaining
producible oil and gas. That number, however, must be
immediately cut to eliminate estimates made by statistically
unreliable regression models of decline. Using a common
measure for statistical goodness of fit, called R , to eliminate
bad regressions (i.e., R < 0.6), a smaller but still impressive
volume of forgotten resources remained: 304 million bbls of oil
and 6.6 trillion cubic feet (tcf) of gas (or a total of 1.47 billion
BOE). In the analysis that follows, only decline curves with
statistical fits carrying R  0.6 are included.
Published in Oil & Gas Journal, August 4, 2014
Although these are large volumes in aggregate, what is
attractive to producers are the very largest accumulations in
these distributions. On the oil side, setting minimum interesting
reservoir size at 100,000 bbls recoverable, 556 oil
accumulations were found; with the minimum raised to 1
million bbls, 29 reservoirs made the grade, the largest of which
was 10.2 million bbls. For gas, setting the minimum size to 1
bcf, 1,417 reservoirs met that threshold and 57 contained at
least 10 bcf, the largest of which was 67.4 bcf.
Oil (BOPD)
Fig. 2
While 100,000 bbls of oil and 1 bcf of gas may seem low
thresholds, these equal the decadal median sizes of new oil and
gas reservoirs brought on-line in the GOM over the last 40 years
(Figure 3). It is true that the “anchoring” (i.e., largest) reservoirs
of new field discoveries are larger than these medians.
However, over the last 40 years, half of the new reservoir
additions within discovered fields, have fallen between 10,000
and 226,000 bbls of oil and 500 mmcf and 6 bcf for gas. These
are comparable to the ranges of most remaining recoverable
volumes of the 556 oil and 1,417 gas reservoirs highlighted in
the Forgotten Oil and Gas study. The reservoirs assessed in the
study between 100,000 and 1 million bbls contain 127 million
bbls collectively; for gas reservoirs between 1 and 10 bcf, 3.1 tcf
was found. At the richest ends of both distributions, the 29 oil
accumulations > 1 million bbls hold 65.9 million bbls collectively
and the 57 gas reservoirs with remaining recoverable >10 bcf
have 1.1 tcf.
By joining the reservoirs identified in the study with the leases
in which they are contained and allocating the volumes to
lessees by their percent ownership of those leases, the
companies holding the greatest volumes of forgotten oil and
gas were determined (Table 1). Approximately 30% of the
resources in reservoirs > 100,000 bbls and > 1 bcf are under
currently open lease blocks.
Reservoirs Shut-In Before Decline
Considering all reservoirs shut-in before 2010, about 7% of oil
reservoirs and 6% of gas reservoirs were abandoned before
their production began to decline. To be fair, most of them
were not increasing when turned off. These reservoirs made
small volumes monthly and at some point, operators judged
them uneconomic and abandoned them.
Some did, however, exhibit positive production profiles. More
importantly, hundreds had final output (averaging the final two
months) higher than the average daily rate for reservoirs online in the 2010-2012 control period (representing “today’s”
rates). Of those equal or exceeding today’s average output, 75
oil reservoirs had final production of  435 barrels of oil per day
(bopd) and 478 gas reservoirs had final production  1.6 million
cubic feet per day (mmcfpd).
Focusing on the 553 reservoirs with high final production rates,
11 were associated with storm-damaged platforms, requiring
facilities as well as completion/reservoir performance analysis.
For the balance of 542, why apparently strong producers were
shut-in is not known. However, if all these 553 high-rate
abandoned reservoirs could all be turned back on at their final
rates, it would result in 76,525 bopd and 2.7 bcfpd of gas.
Reserve History Analysis
Fig. 3
Each year from 1975 through 2010, the MMS published its
Gas independent estimates of the reserves in all fields in the
of Mexico. Starting in 1995, they also estimated reserves at
the sand level. Each year’s report gives cumulative production,
estimated reserves and their sum, estimated ultimate recovery
Published in Oil & Gas Journal, August 4, 2014
Recoverable Volume Threshold
Recoverable Volume Threshold
> 100,000 bbls
> 1 bcf
1 million bbls
10 bcf
(million bbls)
(million bbls)
Energy XXI
Black Elk
Table 1. Companies whose current lease interests in the Gulf of Mexico cover oil and gas resources identified as
“forgotten” in the study (as determined by decline curve analysis); volumes allocated by fractional interests
held as of April, 2014. (Due to a lack of new BOEM data on record title assignments, only Apache resources in
water depths less than 580 feet were assigned to Fieldwood).
Focusing on the 553 reservoirs with high final production rates,
11 were associated with storm-damaged platforms, requiring
facilities as well as completion/reservoir performance analysis.
For the balance of 542, why apparently strong producers were
shut-in is not known. However, if all these 553 high-rate
abandoned reservoirs could all be turned back on at their final
rates, it would result in 76,525 bopd and 2.7 bcfpd of gas.
Reserve History Analysis
Each year from 1975 through 2010, the MMS published its
annual independent estimates of the reserves in all fields in the
Gulf of Mexico. Starting in 1995, they also estimated reserves at
the sand level. Each year’s report gives cumulative production,
estimated reserves and their sum, estimated ultimate recovery
(EUR). In theory, EUR should never fall unless remaining
reserves are explicitly written down. This could happen because
of a change in economics (e.g., drop in price) or realization that
earlier technical assessments were optimistic (e.g., effective
thickness found to be thinner or permeability lower than
originally estimated). EUR can also fall for regulatory reasons
(e.g., a reservoir assigned to one field is later reassigned to a
neighboring field).
Closely examining all 1,296 fields and 14,288 sands evaluated
by the MMS, three classes emerged: 1) fields/sands with EURs
that were never revised downward by the time they were
depleted (or by 2010 for still-producing fields); 2) fields/sands
that had insignificant negative revisions, indicating correction of
entry errors or random fluctuations in the MMS analyses and 3)
fields/sands for which there were large reductions in the MMS
assessments of EUR. The study focused on this last class, for
example, the Ship Shoal 28 field, where approximately 305 bcf
(54 million BOE) of gas reserves were booked in 1978 and
written down in 1984 (Figure 4).
For all fields with large negative revisions, spatial checks were
made to ensure that the resources subtracted from one field or
sand were not simply credited to a neighboring field or sand.
The relationship between revisions and lease holdings was also
examined. From the reserve history analysis, 74 fields were
identified with write-downs of greater than 10 million BOE and
79 sands where the negative changes in EUR exceeded 5 million
BOE. Collectively, these negative revisions sum to 2.9 billion
In all of these cases the presence of producible hydrocarbons is
not in question; so, their geologic risk is minimal. In many cases,
further engineering and geoscience work will show that the
negative revisions were appropriate and would not be revised
today – even in light of different economic and technical
conditions. However, among these 153 fields and sands are
hydrocarbon volumes that were appropriate to write down 510-25 years ago for technical or economic reasons but under
today’s conditions, could be profitable.
Published in Oil & Gas Journal, August 4, 2014
Lost Completions and Sands
Two final study components mined databases on well tests and
production and the MMS/BOEM Sand Atlas series (2) to find
anomalies that might be prospective.
In the first case, all 59,025 completions were checked for initial
tests and those results compared to subsequent production. Of
these, 9% never produced but the rest did. The interesting
subset was those completions with good flow rates on initial
tests that were never put on line. In some cases, further
analysis showed that the oil and gas that would have come
from these completions were produced out of neighboring
wells and additional potential from the unproduced
completions was unlikely.
The core finding of this search was the set of 125 completions
that were never produced but had initial tests that were higher
than the today’s average rates of completion-level production:
350 bopd for oil and 1.2 mmcfpd for gas. The sum of the test
rates of these 125 completions was 28,659 bopd and 526
mmcfpd. Using a general (and high variance) relationship
between completion EUR and initial potential tests, together,
these completions represent a mean of 58 million BOE of
producible oil and gas.
Million BOE
54 mln BOE
Gas Reserves
Cumulative Gas
1975 1980 1985 1990 1995 2000 2005
Fig. 4
The search for “lost” sands centered on the Sand Atlas. It gives
rock and fluid/gas properties and assessment of each sand’s oil
and gas resources. When the first Atlas document was
published by MMS, it was a comprehensive list of all sands in all
GOM fields that either had produced or had qualifying
production but were not yet on-line (e.g., they were waiting on
a pipeline connection or a platform). Every year, the
MMS/BOEM added the sands of newly discovered fields and
newly discovered sands in discovered fields to the Sand Atlas.
As the Sand Atlas is a cumulative document, once a sand enters,
it should stay. Some sands ultimately get reassigned to other
sands as additional drilling and geophysical information
indicates that what was earlier judged to be a separate
accumulation was simply an extension of another sand.
However, examination of drilling and completion histories
within a field and spatial analysis of its sands can usually
identify such reassignments.
Eliminating reassignments, 12 sands with EURs greater than 1
million bbls or 10 bcf simply disappeared from the Sand Atlas.
As being added to the Sand Atlas requires detailed geologic and
engineering data, it is unlikely that these represent simply
postulated hydrocarbon accumulations. They collectively
constitute about 43 million BOE of recoverable resources; they
were included in the Forgotten Oil and Gas study because some
exploration/development opportunities.
Two valuable results came from the study of resources
remaining in abandoned reservoirs and wells in the Gulf of
Mexico. First were the empirical findings of the Forgotten Oil
and Gas study: that there are hundreds of millions of barrels of
oil and trillions of cubic feet of gas producible but left behind in
discovered GOM fields. Most of the remaining accumulations
found in the study are too small to economically produce.
However, from tens of thousands of reservoirs, hundreds of the
accumulations identified are as large as or larger than the sizes
of new reservoirs brought on-line in the last 40 years.
These opportunities are very diverse: from completions, to
reservoirs, to sands and to fields themselves. All these volumes
are associated with fields that produced - most with fields that
are currently under production. This minimizes geologic risk and
reduces costs through exploiting existing production and
transportation infrastructure.
The second outcome of the study was a detailed demonstration
that mathematical and statistical methods for “mining” massive
volumes of data – so widely and successfully applied in other
industries – can be usefully employed in oil and gas. This is
enabled by the existence and organization of comprehensive
electronic data bases on geology, drilling, testing, production
and reserves. Invested in all of the study components, where
possible, were explicit evaluations of the statistical certainty
underlying individual results.
While the growth of onshore unconventional oil and gas
production and deep water oil have headlined the fundamental
change in U.S. hydrocarbon supply in the last half dozen years,
both sources have met cost and regulatory obstacles to output
Published in Oil & Gas Journal, August 4, 2014
continuing to increase at recent rates. New resources in
established fields can provide a low-cost/lower-regulation
complement to those supplies and profitable opportunities for
nimble niche producers and particularly for the companies that
own the leases under which most of these accumulations are
Arps, J.J., “Analysis of Decline Curves,” Transactions of
the American Institute of Mining, Metallurgical and
Petroleum Engineers, vol. 160 (AIME, 1945), pp. 228247.
U.S. Minerals Management Service/U.S. Bureau of
Ocean Energy Management, Atlas of Gulf of Mexico
Gas and Oil Sands Data (1995 – 2010). Editions from
1999 to 2010 available as of May 22, 2014 at
Direct access to this article on Oil & Gas Journal’s online version
(subscriber login required):
John D. Grace is the Chief Executive Officer of Earth Science
Associates. He received his PhD in Economics from Louisiana
State University and began his career in the geologic division of
ARCO’s R&D lab then moved to Corporate Planning before
founding ESA in 1991. His specializations are in resource
assessment, basin analysis and mathematical geology. He has
taught geology, geostatistics and mathematics at LSU,
University of Southern California and California State University,
Fullerton. He is a member of SPE. ([email protected])
Tony Dupont is the Chief Operating Officer of Earth Science
Associates, which he joined in 1999. He received an MA in
geography from California State University, Fullerton. His
primarily works in the development of geographic information
system (GIS) technology for ESA’s GOM and GOMsmart
software packages. ([email protected])
Scott Morris earned his bachelor’s degree in Statistics from the
University of California, Riverside and his master’s degree in
Applied Mathematics from California State University,
Fullerton. He joined Earth Science Associates as a Research
Associate in 2012, where he handles various predictive analytic
([email protected])
Earth Science Associates
4300 Long Beach Blvd., Ste. 310
Long Beach, CA 90807
(562) 428-3181
Published in Oil & Gas Journal, August 4, 2014