TO:	EPA Docket No. EPA-HQ-OAR-2003-0146

		

FROM:	Bob Lucas, EPA/SPPD

DATE:	July 27, 2007

SUBJECT:	Equipment Leaks:  Control Options and Impact Estimates

I.	Purpose

This memorandum documents the methodology used to identify the equipment
leak control options and estimate the costs and environmental impacts
for those options.

II.	Background

Section 112(f) of the Clean Air Act Amendments (CAA) directs the U.S.
Environmental Protection Agency (EPA) to assess source categories
regulated under Section 112(d) of the CAA and determine whether any
human health or environmental risks remain from the continued emissions
of hazardous air pollutants (HAP) following implementation of maximum
achievable control technology (MACT) standards.  The CAA further states
that if the MACT standards do not reduce lifetime excess cancer risk to
the most exposed individual to less than one in one million, EPA must
set additional standards to protect human health and the environment, in
accordance with the interpretation set forth in the Benzene NESHAP. 
Additionally, the EPA is required to review these technology-based
standards and to revise them “as necessary (taking into account
developments in practices, processes and control technologies)” no
less frequently than every 8 years, under CAA section 112(d)(6).  The
Refinery MACT 1 (40 CFR Part 63 Subpart CC) was promulgated over 8 years
ago and is now being reviewed.

III.	Summary of Existing Refinery MACT 1 Equipment Leak Control
Requirements

Refinery MACT 1 requires sources to follow a leak detection and repair
(LDAR) program for components (e.g., valves, pumps) in HAP service. 
Existing sources as of July 14, 1994 (proposal of Refinery MACT 1) have
the option of complying with the New Source Performance Standards (NSPS)
for equipment leaks in the synthetic organic chemical manufacturing
industry (SOCMI) (40 CFR part 60, subpart VV; petroleum refineries are
subject to 40 CFR part 60, subpart GGG, which references subpart VV) or
a modified version of the hazardous organic NESHAP (HON) (40 CFR part
63, subpart H).  New sources must comply with the modified HON. 
Modifications to the HON for Refinery MACT 1 sources include leak
definitions of 1,000 ppm for valves and 2,000 ppm for pumps as well as
less restrictive requirements on when regularly scheduled monitoring
events may be skipped based on the number of leaks found.

IV.	Identification of Control Options

EPA recently proposed amendments to the NSPS for equipment leaks in the
SOCMI and petroleum refineries (71 FR 65302, November 7, 2006).  These
amendments included monitoring of valves at 500 ppm and pumps at 2,000
ppm as well as monitoring of open-ended lines.  For this analysis,
Option 1 was identified as baseline, or no proposed changes to the
current MACT requirements.  Option 2 was identified as requiring both
new and existing sources subject to comply with the proposed amendments
to the NSPS subparts VV and GGG.

V.	Impact Estimates for Equipment Leaks

The general methodologies for determining the emission reductions and
the costs of both options are documented in two memoranda:  Data and
Assumptions used in the Equipment Leaks Cost Analysis for Petroleum
Refineries, or “NSPS Refinery Impacts Analysis” (Parrish et. al.,
2006) and Analysis of Additional Data on Leaks from Open-Ended Lines, or
“Open-Ended Line Analysis” (Randall and Parrish, 2007).  Option 1
and Option 2 in this memorandum correspond to Baseline and Option 1 in
the NSPS Refinery Impacts Analysis, respectively.  This memorandum
focuses on the assumptions and calculations that were developed
specifically for this analysis or revised from the previous analyses.

Equipment Leak Models

The NSPS Refinery Impacts Analysis describes four models, two of which
group sources that are subject to a consent decree and sources subject
to Refinery MACT 1.  Upon further review, these sources are subject to
different requirements and utilize different monitoring frequencies.  In
addition, this analysis considers existing sources subject to MACT
whereas the NSPS Refinery Impacts Analysis considers only sources
constructed, reconstructed, and modified in the next five years. 
Therefore, this analysis includes the following assumptions in addition
to those described in the NSPS Refinery Impacts Analysis:

If 60 percent of the 10 large and 5 small refineries’ worth of process
units estimated in the NSPS Refinery Impacts Analysis are new and 40
percent are reconstructed, then 6 large and 3 small refineries’ worth
of process units will be constructed in the future.

The growth rate of 3 refineries’ worth of process units per year can
be extrapolated back 13 years to the date of promulgation of Refinery
MACT 1.  Therefore, 26 large and 13 small refineries’ worth of process
units have been constructed or modified since 1994.

There are approximately 150 operating refineries; using the assumptions
in the NSPS Refinery Impacts Analysis, there are 100 large refineries
and 50 small refineries.  Therefore, 74 large and 37 small refineries’
worth of process units were constructed or last modified prior to 1994.

About 36 percent of the total components and process units in volatile
organic compound (VOC) service process non-HAP VOC, so the remaining 64
percent are estimated to be in HAP service.  Therefore, as an example,
there are 75 process units in HAP service that will be constructed in
the next five years, 325 process units in HAP service that have been
constructed or modified since 1994, and 926 process units in HAP service
that were constructed or last modified prior to 1994, for a total of
1,326 process units.  Similar estimates were calculated for each type of
component as well.

These assumptions were used to develop five models, all of which include
process units and components that process HAP:

“New Future CD” – This model represents new components and process
units built during the next five years at petroleum refineries that are
subject to a consent decree.  On a refining capacity basis, 76.5 percent
of the petroleum refining industry is subject to a consent decree;
therefore, this model includes the 76.5 percent of the components and
process units in HAP service that are subject to a consent decree.  As
an example, the total number of process units included in this model is
57, or 76.5 percent of the 75 process units described in the previous
paragraph.

“New Future MACT” – This model represents new components and
process units built during the next five years at petroleum refineries
that are not subject to a consent decree.  Therefore, this model
includes the 23.5 percent of the components and process units in HAP
service that are not subject to a consent decree.  As an example, the
total number of process units included in this model is 18, or 23.5
percent of the 75 process units described in the previous paragraph.

“New MACT” – This model represents the components and process
units built during the last 13 years (i.e., since July 1994) at
petroleum refineries that are not subject to a consent decree and that
process HAP.  As an example, the total number of process units included
in this model is 77, or 23.5 percent of the 325 process units described
in the previous paragraph.

“Existing MACT” – This model represents the components and process
units constructed or last modified or reconstructed prior to 1994 at
petroleum refineries that are not subject to consent decree.  As an
example, the total number of process units included in this model is
218, or 23.5 percent of the 926 existing process units described in the
previous paragraph.

“Existing CD” – This model represents the existing components and
process units at petroleum refineries that are subject to a consent
decree, regardless of whether the process unit was constructed prior to
or after 1994.  Therefore, this model includes 76.5 percent of the
existing components and process units in HAP service.  As an example,
the total number of process units included in this model is 957.

See Attachment A for the numbers of components characterized by each
model.

Model Unit Emission Reductions 

As in the NSPS Refinery Impacts Analysis, emissions estimates were based
on data that the American Petroleum Institute (API) summarized in the
1997 report Analysis of Refinery Screening Data (API, 1997).  The
calculation of the revised of leak frequency estimates can be grouped in
five general steps:

Developing leak frequencies for valves at 1,000 ppm

Developing leak frequencies for valves at 500 and 1,000 ppm for
semiannual and annual monitoring

Revising the leak frequency for valves at 10,000 ppm and developing leak
frequencies for other monitoring frequencies

Developing leak frequencies for monthly monitoring of pumps at 1,000 ppm

Revising the leak frequency for pumps at 10,000 ppm and developing a
leak frequency estimate for monthly monitoring

The results of these calculations are presented in Table 1.

Valves – 1,000 ppm

To account for the cost and emission reduction for Refinery MACT 1 new
source requirements, leak frequencies and emission factors were
developed for valves at a leak definition of 1,000 ppm.  For quarterly
monitoring of valves at 1,000 ppm, the leak frequency values were
selected directly from the API data.

Valves – 500 and 1,000 ppm

To estimate leak frequencies for valves at monitoring frequencies other
than quarterly for leak definitions of 500 and 1,000 ppm, the numbers of
components leaking more than once in four consecutive quarters as
provided in the API Report (API, 1997) were used to determine the number
of unique leaks found during a particular monitoring event.  This
examination helped to determine that one leak is not counted twice
during a semiannual or annual monitoring event.  The calculations can be
described by Equations 1 and 2:

 	Equation 1

where 

VFS 	= Leak frequency for semiannual monitoring, percent

VFQ	= Leak frequency for quarterly monitoring, percent

VLQ	= Number of valves leaking per quarter

VR2	= Number of valves leaking twice in four consecutive quarters

VR3	= Number of valves leaking three times in four consecutive quarters

VR4	= Number of valves leaking four times in four consecutive quarters

 	Equation 2

where

VFA	= Leak frequency for annual monitoring, percent

VFQ	= Leak frequency for quarterly monitoring, percent

VLQ	= Number of valves leaking per quarter

VR2	= Number of valves leaking twice in four consecutive quarters

VR3	= Number of valves leaking three times in four consecutive quarters

VR4	= Number of valves leaking four times in four consecutive quarters

Valves – 10,000 ppm

The NSPS Refinery Impacts Analysis estimated annual leak frequencies of
1.18 percent for gas valves and 1.38 percent for light liquid valves
based on data from the U.S. EPA National Enforcement Investigations
Center (NEIC) (EPA, 1999).  Further consideration of those data, as well
as information provided by industry representatives (Dee, 2006),
suggests that the average monitoring frequency for valves subject to
NSPS subpart GGG is semiannual rather than annual.  Therefore, this
analysis assumes that the values of 1.18 and 1.38 percent for gas and
light liquid valves, respectively, correspond to semiannual monitoring. 
The ratio of the leak frequencies for quarterly monitoring to semiannual
monitoring at 1,000 ppm was assumed to be equal to the ratio of leak
frequencies for quarterly monitoring to semiannual monitoring at 10,000
ppm.  Likewise, the ratio of the leak frequencies for annual monitoring
to semiannual monitoring at 1,000 ppm was assumed to be equal to the
ratio of leak frequencies for annual to semiannual monitoring at 10,000
ppm.

Pumps – 1,000 ppm

For quarterly monitoring of pumps at 1,000 ppm, the leak frequency value
was selected directly from the API data.  Similar to valves, the numbers
of components leaking more than once in a year as provided in the API
Report (API, 1997) were used to estimate the leak frequencies for pumps
at monitoring frequencies other than quarterly.  However, since the
desired monitoring frequency of monthly for pumps is more frequent than
quarterly rather than less frequent as for valves, the API data were
examined to include pumps assumed to be leaking more than one month in a
quarter.  Assuming repeat leakers are distributed evenly throughout the
consecutive four-month period, Equation 3 was developed:

 	Equation 3

where

PFM	= Leak frequency for monthly monitoring, percent

PFQ	= Leak frequency for quarterly monitoring, percent

PLQ	= Number of pumps leaking per quarter

PR2	= Number of pumps leaking twice in four consecutive quarters

PR3	= Number of pumps leaking three times in four consecutive quarters

PR4	= Number of pumps leaking four times in four consecutive quarters

Pumps – 10,000 ppm

The quarterly leak frequency value for pumps at 10,000 ppm is calculated
using the assumption that the ratio of the leak frequency for valves
leaking greater than 10,000 ppm at a leak definition of 10,000 ppm to
the leak frequency for valves leaking greater than 10,000 ppm at a leak
definition of 1,000 ppm is equal to the same ratio for pumps (all
monitored quarterly).  Specifically, this value was calculated using
Equation 4:

 	Equation 4

where all variables are for components leaking greater than 10,000 ppm
and monitored quarterly and

	PF10,000	= Leak frequency for pumps with a leak definition of 10,000
ppm, percent

	PF1,000	= Leak frequency for pumps with a leak definition of 1,000 ppm,
percent

	VFG;10,000	= Leak frequency for gas valves with a leak definition of
10,000 ppm, percent

	VFG;1,000	= Leak frequency for gas valves with a leak definition of
1,000 ppm, percent

	VFLL;10,000	= Leak frequency for light liquid valves with a leak
definition of 10,000 ppm, percent

	VFLL;1,000	= Leak frequency for light liquid valves with a leak
definition of 1,000 ppm, percent

The ratio of leak frequency values for pumps at a leak definition of
10,000 ppm monitored monthly to those monitored quarterly was assumed to
be equal to the ratio of pumps monitored at those frequencies with a
leak definition of 1,000 ppm.

Table 1.  Summary of Leak Frequency Values

Monitoring Frequency	Monthly	Quarterly	Semiannual	Annual

Component and Leak Definition	Leak Frequency, percent

Pumps





10,000 ppm	0.88	2.47



1,000 ppm	0.98	2.77



Gas Valves





10,000 ppm

0.61	1.18	2.32

1,000 ppm

0.88	1.69	3.31

500 ppm

1.00	1.91	3.73

Light Liquid Valves





10,000 ppm

0.72	1.38	2.70

1,000 ppm

0.86	1.65	3.23

500 ppm

1.03	1.97	3.83



Emission factors and initial leak frequencies were calculated from these
values as explained in the NSPS Refinery Impacts Analysis.  Emission
factor values were then multiplied by 20 percent to convert the factors
from kilograms of light liquid VOC per hour per component to kilograms
of HAP per hour per component or by 10 percent for components in gas
service (Coburn, 2007).  See Attachment B for these values.

Emission reductions from open-ended lines were calculated assuming
quarterly monitoring.  The numbers of open-ended lines for each model
were presented in Attachment A, and Attachment C shows the emission
reductions and costs for each model.

Model Unit Control Costs 

All of the constant values from the NSPS Refinery Impacts Analysis
remained unchanged except for the labor costs, which were updated to the
values in the May 2006 Bureau of Labor Statistics (BLS) Employment and
Wage Estimates as follows (DOL, 2007):

Wage (refinery maintenance) and overhead factor: maintenance and repair:
 The value of $23.30 per hour is the mean wage rate for Standard
Occupational Classification (SOC) Code 49-9012 (control and valve
installers, except mechanical doors) in NAICS 324 (Petroleum and Coal
Products Manufacturing).  This wage rate is for refinery personnel, not
contractors, so it is likely to be higher than the actual wage rate.

Planning and training labor charge:  The value of $68.78 is an update of
the weighted average for technical, management, and clerical labor.  The
weighted average assumes 5 hours of management time and 10 hours of
clerical effort per 100 hours of technical labor.  The technical rate is
based on the mean wage for environmental engineers ($38.29), the
management rate is based on the mean wage for general and operations
managers ($59.15), and the clerical rate is based on the mean for all
office and administrative support occupations ($17.46).  A 60 percent
overhead factor was added to these rates.

As previously mentioned, the assumptions surrounding the Option 1
monitoring frequency for existing Refinery MACT 1 sources were revised;
therefore, the Option 1 costs for these sources are for semiannual
monitoring rather than annual monitoring.  The leak frequencies for
valves subject to new source MACT requirements are less than 3 percent,
so Option 1 costs for those sources also assume semiannual monitoring. 
For Option 2, the calculated leak frequencies for semiannual monitoring
of gas and light liquid valves at 500 ppm are below 2 percent (it was
previously assumed that these values were likely to be above 2 percent).
 Therefore, semiannual monitoring was assumed for all sources not
subject to a consent decree under Option 2.  Table 2 summarizes the
revised values and parameters defined for each model.  See Attachment D
for detailed costs for each of the five models.

Table 2.  Model Parameters for Each of the Petroleum Refinery Models

Model	New Future MACT	New Future CD	New MACT	Existing MACT	Existing CD

Option 1	Leak definition:

  Pumpsa

  Valves	

2,000

1,000	

2,000

500	

2,000

1,000	

10,000

10,000	

2,000

500

	Monitoring frequency: 

  Pumps

  Valves	

Quarterly

Semiannual	

Monthly

Quarterly	

Quarterly

Semiannual	

Monthly

Semiannual	

Monthly

Quarterly

	Initial leak frequency/ costs	Yes – Program starts at set leak
definitions; initial leak frequency is same as subsequent leak frequency
Yes – Program starts at set leak definitions; initial leak frequency
is same as subsequent leak frequency	No – Initial costs would have
been incurred at construction	No – Initial costs would have been
incurred at construction	No – Initial costs would have been incurred
at construction

	Planning and training costs	330 hours per process unit	330 hours per
process unit	330 hours per process unit	320 hours per process unit	330
hours per process unit

	Administrative report costs	95 hours per process unit	95 hours per
process unit	95 hours per process unit	85 hours per process unit	95
hours per process unit

Option 2	Leak definition:

  Pumpsa

  Valves	

2,000

500	

2,000

500	

2,000

500	

2,000

500	

2,000

500

	Monitoring frequency: 

  Pumps

  Valves	

Monthly

Semiannual	

Monthly

Quarterly	

Monthly

Semiannual	

Monthly

Semiannual	

Monthly

Quarterly

	Initial leak frequency/ costs	Yes – Program starts at set leak
definitions; initial leak frequency is same as subsequent leak frequency
Yes – Program starts at set leak definitions; initial leak frequency
is same as subsequent leak frequency	Partial – Only includes initial
costs associated with lowering valve leak definition from 1,000 ppm
Partial – Only includes initial costs associated with lowering leak
definitions from 10,000 ppm	No – Initial costs would have been
incurred at construction

	Planning and training costs	330 hours per process unit	330 hours per
process unit	330 hours per process unit	330 hours per process unit	330
hours per process unit

	Administrative report costs	95 hours per process unit	95 hours per
process unit	95 hours per process unit	95 hours per process unit	95
hours per process unit



VI.	Nationwide Impact Estimates for Equipment Leaks

Nationwide impacts were developed by summing the costs and emission
reductions for each of the five models.  These impacts are summarized in
Table 3 and presented in more detail in Attachment E.  Benzene emission
reductions were assumed to be 1 percent of VOC reductions (Coburn,
2007), or about 2.8 tons per year, for a cost-effectiveness of $360,000
per ton.

Table 3.  Summary of Nationwide Impacts of Refinery Equipment Leak
Control Options

Control Option	Total Capital Investment  (million $)	Total Annual Costs
 (million $/yr)	HAP Emission Reduction from Baseline

(tpy)	HAP Cost-Effectiveness ($/ton)

1.  No proposed changes	2.4	15	0	0

2.  Proposed amendments to NSPS subpart VV/GGG	3.1	16	42	24,000



VI.	References

American Petroleum Institute, Health and Environmental Affairs
Department (API).  1997.  Analysis of Refinery Screening Data.  Prepared
by Hal Taback Company.  Publication Number 310.  November.

Coburn, J.  2007.  Average Refinery Stream Composition

Dee, Norbert.  2006.  NSPS VV GGG industry cost-effectiveness data. 
Email to Karen Rackley, USEPA.  March 16.  Docket ID
EPA-HQ-OAR-2006-0699-0002.

Parrish, K., D. Randall, and J. Coburn.  2006. Data and Assumptions used
in the Equipment Leaks Cost Analysis for Petroleum Refineries.
Memorandum to Karen Rackley, EPA/SPPD.  October 30.  Docket ID
EPA-HQ-OAR-2006-0699-0034.

Randall, D., and K. Parrish.  2007.  Analysis of Additional Data on
Leaks from Open-Ended Lines.  Memorandum to Karen Rackley, EPA/SPPD. 
June 18.  Docket ID EPA-HQ-2006-0699-0074.

U.S. Department of Labor (DOL), Bureau of Labor Statistics.  2007. 
Occupational Employment Statistics:  May 2006 Employment and Wage
Estimates.  http://stats.bls.gov/oes/current/oessrci.htm; accessed in
June.

U.S. EPA, Office of Enforcement and Compliance Assurance.  1999. 
Enforcement Alert: Proper Monitoring Essential to Reducing 'Fugitive
Emissions' Under Leak Detection and Repair Programs.  October.

Technical Memorandum – Equipment Leaks:  Control Options and Impact
Estimates

July 27, 2007

Page   PAGE  10 

Technical Memorandum

