lable at ScienceDirect

Journal of Loss Prevention in the Process Industries 24 (2011) 138e145

Contents lists avai

Journal of Loss Prevention in the Process Industries

journal homepage: www.elsevier .com/locate/ j lp

Dispersion modeling approach for quantification of methane emission ratesfrom natural gas fugitive leaks detected by infrared imaging technique

Anisa Safitri, Xiaodan Gao, M. Sam Mannan*

Mary Kay O’Connor Process Safety Center, Artie McFerrin Department of Chemical Engineering, Texas A&M University, College Station, TX 77483-3122, USA

a r t i c l e i n f o

Article history:Received 22 July 2010Received in revised form20 October 2010Accepted 28 November 2010

Keywords:Leak detectionInfrared imagingEmissions rateGas dispersion

* Corresponding author. Tel.: þ1 979 862 3985; faxE-mail address: mannan@tamu.edu (M.S. Mannan

0950-4230/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.jlp.2010.11.007

a b s t r a c t

Recently, infrared optical imaging has been applied in the oil and gas industry as a method to detectpotential leaks in pipelines, components and equipment. The EPA suggested that this impending tech-nique is considered as a smart gas LDAR (leak detection, monitoring and repair) for its rapid recognitionof leaks, accuracy and robustness. In addition, compared to the conventional method using Total VaporAnalyzer (TVA) or gas sniffer, it has several other advantages, such as the ability to perform real-timescanning and remote sensing, ability to provide area measurement instead of point measurement, andprovide an image of the gas which is not visible to naked eye. However, there is still some limitation inthe application of optical imaging techniques; it does not give any measurement of gas emissions rates orconcentrations of the leaking gas. Infrared cameras can recognize a target gas and distinguish the gasfrom its surrounding up to a certain concentration, namely the minimum detectable concentration. Thevalue of the minimum detectable concentration depends on the camera design, environmental condi-tions and surface characteristics when the measurement is taken. This paper proposed a methodology topredict gas emissions rates from the size of the dispersed gas plume or cloud to the minimum detectableconcentration. The gas emissions rate is predicted from the downwind distance and the height of thecloud at the minimum detectable concentration for different meteorological conditions. Gas release anddispersion from leaks in natural gas pipeline systems is simulated, and the results are presented.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Natural gas is one of the major energy resources in the UnitesStates. The use of natural gas as a source of energy is very crucial forall the sectors of the US economy, including residential, commer-cial, industrial, utility and transportation. According to the AnnualEnergy Report 2008 by Energy Information Administration (EIA)natural gas accounts for 24% of primary energy consumption in theUS by source, of the total of 98 Quadrillion Btu. (EIA, 2009a, 2009b)EIA also predicted that the consumption of natural gas will increaseby 25% in 2030.

Based on EIA data, there are more than 450,000 natural gasproducing wells in the U.S. operated by more than 6300 producers(EIA, 2009a, 2009b). The natural gas distribution system consists ofapproximately 305,000 miles of interstate and intrastate trans-mission pipelines, more than 1400 compressor stations, 11,000delivery points, 5000 receipt points and 1400 interconnectionpoints throughout the United States. This distribution network is

: þ1 979 845 6446).

All rights reserved.

capable of transferring over 148 Billion cubic feet (Bcf) of gas perday. There are 400 underground natural gas storage facilities in theUS, with storage capacity of 4059 Bcf and the capability to deliver85 Bcf of gas per day, in addition to 8 LNG (Liquefied Natural Gas)import facilities and 100 LNG peaking facilities (EIA, 2009a, 2009b).

Natural gas is considered one of the cleanest energy sourcesbecause the combustion of natural gas results in low SO2 and NOx

emission and almost no particulate or ash. However, methane,which is the main constituent of natural gas, is one of the mostpotent greenhouse gas precursors. In natural gas systems, methaneis constantly released during production, processing, and trans-mission as well as distribution. EPA reported the emission fromnatural gas system accounts for 104.7 Tg CO2 Eq. (4985 Gg) of CH4 inthe U.S. Inventory of Greenhouse Gas Emissions and Sinks 1990/2007 (EPA, 2009). More than half of the methane emissions comefrom fugitive leaks, which occur during normal operation, routinemaintenance or system upsets.

In order to reduce methane emissions from natural gas system,EPA had developed a program called Leak Detection and Repair(LDAR), which requires gas producers to regularly monitor anypotential leaks from pipeline components or equipment at regularintervals using protocol called EPA Method 21. This protocol used

A. Safitri et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 138e145 139

the Vapor Analyzer (TVA) or the Organic Vapor Analyzer (OVA) tofind any potential leak sources. (EPA, 1999) However, finding leaksusing this method is labor intensive, time consuming and surveysmust be conducted point by point in close proximity with pipes,valves and other volatile organic compounds (VOC) carryingcomponents. In addition, vapor analyzers do not actually measurethe emissions rate of the leak, rather they use empirical equation toconvert the concentration (parts per million by volume) into massflow rate (Robinson & Luke Boone, 2003).

Recently, a new gas detection method using infrared opticalimaging techniques became popular in oil and gas industry. Thistechnique allows the visualization of several hydrocarbon gases.This method is much more effective compared to the conventionalmethod for its ability to (Benson et al., 2006; EPA, 2008):

� visualize leaks in real-time� pinpoint leak locations� scan large areas rapidly� scan components that are hard to reach by contact measure-ment tools

� detect leaks without interrupting processes� monitor potentially dangerous leaks from a distance� record and documents leaks� detect the gas remotely

The utilization of infrared imaging systems potentially simplifiesthe EPA LDAR program and therefore it is currently recognized inindustry as a Smart-LDAR approach. Optical imaging techniquesmust meet several requirements to qualify as a smart approachequivalent to the traditional DLAR approach including (Epperson,Lev-On, Tabback, Siegell, & Ritter, 2007):

� It should be able to significantly reduce monitoring cost.� It performs more frequent monitoring of process equipment.� It should be able to locate leak form highly leaking componentswithout having to monitor every individual piping component.

� It provides better control effectiveness.� It results in better emission reduction.� It gives better environmental protection.� Based on API report, 90% of all emission comes from 1% of theleaks. Smart LDAR is able to detect and locate the high leakersand therefore it is more cost-effective.

Table 1 provides the comparison between smart LDAR usingoptical imaging instruments compared to traditional LDAR usingsingle point measurement.

Although infrared camera imaging systems possess severalleading characteristics compared to the traditional method, this

Table 1Comparison of traditional and smart LDAR.

Measurement characteristics Traditional LDAR

Sampling capability Single point measurementSampling of components Only small number of compon

duration of sampling.Mass emission rate measurement Inaccurate because it measure

concentration adjacentto the leaking components. Poof the actual emission rate.

Location of instrument during measurement Must be in the vicinity of theleaking components

Ability to pinpoint leaks No

Multiple components monitoring No

Monitoring frequency 90 days for every piping comp

technique is unable to measure the concentration, nor the leak rateof the gas; hence, it does not give any information on the amount ofgas released. It is important to know how much gas being emittedin order to quantify the total methane emissions resulted from thefugitive leaks, as well as to eliminate hazards due to the presence offlammable material in the environment. Therefore, this research iscarried out in order to troubleshoot the shortcoming in this rela-tively new technology.

2. Potential use of infrared imaging technique forfugitive emissions control

Infrared optical imaging has been tested in oil and gas indus-tries, including refineries, and has demonstrated promising resultsfor fugitive emissions detection and survey. This technology allowsthe user to see leaking gas from equipment as a real-time videoimage.

Infrared camera can be used to detect the absorptive andemissive characteristics of several gases. Gas molecules absorba photon and transitions from one state to another due to the dipolemoment, which enables the molecule to oscillate in the samefrequency as the incident photon, and transfer electromagneticenergy. Each molecule has a specific absorption range within theinfrared spectrum. In a particular type of infrared camera, the focalplane arrays of the camera detector can be tuned to a very narrowspectral region where the gas has a strong absorption rate, andtherefore, the gas can be visualized.

Infrared imaging systems detect the radiation emitted by thetarget and background. Detection can occur only if the targetsignature can be distinguished from the background. The basicinput parameter for detection is the transmitted radiation contrastbetween the target and background which generates a detectoroutput voltage. The infrared camera can see certain gases, mostlyhydrocarbons, because the camera’s detector is equipped witha filter which is sensitive to only a certain wavelength. The CeHbond has an absorption characteristic at 3.46 mm; therefore, theinfrared camera for hydrocarbon gas visualization generally has anabsorption window within this range i.e. mid-wave IR range. Sincethe camera detectors are sensitive to only certain wavelengths,detectors which are active at different wavelength will havedifferent spectral responses. The radiation intensity of the targetobject and background depends on their temperatures. It ispossible to visualize the gas against the background only if thebackground and the target gas cloud are not in thermal equilibrium.If the gas temperature is uniform with the background tempera-ture, the gas cannot be discriminated against the background. Inthis application, the infrared camera can visualize gas which haslower temperature from the surrounding due to the expansion

Smart LDAR

Area scanning of a facilityents in 1-h Able to scan up to 3000 components per hour, 20 times

faster than traditional method.s the ambient

or correlation

Leak detection threshold is identified from severallaboratory and field tests which was measured in g/hr.

Capable to monitor the leaks remotely.

Yes. It can visualize gas leaks thus able to exactlylocate the location of the leak.Yes. Able to simultaneously identify and controlmultiple components.

onents 60 days for valve, 45 days for pumps and 30 daysfor flanges.

A. Safitri et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 138e145140

process when the gas is released from the high pressure. Differentcamera systems generally have different amplifier. The combina-tion between the detector and amplifier will result in different gainand level and thus different noise pattern (Holst, 2000).

The utilization of infrared imaging for fugitive methane gasdetection and visualization had been demonstrated by Gross et al.In their work, the methane gas from buried, leaky pipelines wasmonitored using a staring, high resolution, PtSi Schottky barrierfocal plane array infrared camera with 256 � 256 pixels and cooledto 77 K. The optical camera was operating at 3.32 mm wavelength,and a narrow band (0.16 mm) interference filter was placed in thecamera detector in order to reduce the spectral range and enhancethe sensitivity. The temperature resolution of the infrared camerawas 0.1e0.01 K at 300 K background temperature. Methaneemanating from a buried pipeline was simulated, a halogen lampwas used to irradiate the object region and a diffuse reflector wasplaced behind the target to reflect the infrared radiation and geta better contrast of the methane gas and the environment. Theabsorption of infrared radiation by the methane gas caused thereduction in radiation intensity, and the gas was detected as flick-ering dark cloud in front of uniform background. The test by Grosset al. was conducted under a controlled environment, thereforewind-force and moisture level was considered negligible. A cloudwith a diameter of 30 cm was observed and a methane concen-tration as low as 0.03% can be detected. During the test, the gas flowrate was gradually decreased, and a gas mass flow rate as low as0.1 L/min could still be recognized (Gross et al., 1998).

A similar camera system had been applied to detect andmonitormethane leaks from pipelines under realistic condition by Schulzet al. The infrared optical imaging was used to monitor leaks froma natural gas pipeline buried 1.2 m underground. The camera wasable to recognize a leak as low as 2 L/min, under moderate weatherconditions, and with a 2 m/s wind speed. The cloud diametermeasured during the observation was 50 cm (Schulz, Gross, &Scheuerpflug, 2008).

The sensitivity of the infrared camera to detect methane gas wasalso performed by Benson et al. They employed a cell with infraredoptical imaging to evaluate the IR absorption characteristics ofmethane gas. The test was conducted in a laboratory where wind-force is negligible. The study showed that methane gas concen-tration as low as 275 ppm should be distinguished by the infrareddetector (Benson, Madding, Lucier, Lyons, & Czerepuszko, 2006;Benson, Panek, & Drayton, 2006).

In the real oil and gas industry, optical infrared imaging tech-niques have been applied for the survey of VOC releases fromequipment and pipelines. Total E&P Indonesie had utilized aninfrared camera system to monitor and repair major and minorreleases of benzene, toluene, ethyl benzene and xylene. The camerahad a spectral range of 1e5 mm, a 30 � 30 InSb detector with320 � 240 pixel array, and operated at near the liquid nitrogentemperature. The integration time of this infrared camera was5e16.5 ms, and the Noise Equivalent Temperature Difference(NEDT) was 18 K. The surveys were carried out for the purpose ofprotecting the operators from toxic chemical releases and mini-mizing the environmental impact from hydrocarbon leaks in thefacility. Leaks fromvalves, flanges, connectors, stuff boxes and otherequipment were exhaustively detected within the 1500 km2

Table 2Summary of previous tests of infrared imaging application for methane gas detection an

Type of test Weather condition Minimum detec

Gross et al. Laboratory e 0.1 L/minSchulz et al. Real field Mild, 2 m/s 2 L/minBenson et al. Laboratory e e

operation. Using this method, they could identify 500 unknownleaks from 2100 fugitive emissions sources. Depending on windconditions, leaks between 3 and 6 g/h can be detected using thissystem (Plisson-Saune et al., 2008).

Basedon the available data on the applicationof infrared imagingtechniques for gas detection, the capability of an infrared camera indetecting minimum flow of methane gas and concentration issummarized in Table 2. The table shows that that locations andprevailing weather conditions are among the most determiningfactors in an infrared camera’s performance. The test, which wascarried out indoors, with a controlled environment, resulted inhigher sensitivity, whereas the sensitivity drops by a factor of two inthe tests conducted under real weather conditions.

Although infrared imaging can identify leaks accurately, it doesnot have the capability to provide quantitative measurementsof the emissions rate or the gas concentration. Quantitativemeasurements of emission rates and concentration are importantin order to estimate the total emissions released from the leak, aswell as to determine the hazardous distance, because methane isflammable within the range of 5e15%-v/v in air.

A semi-quantitative approach to determine the mass emissionsrate of a hydrocarbon gas release has been developed by Lev-Onet al. The work focused on obtaining new emission factors forhydrocarbon gas releases when an optical imaging technique wasused for leak detection. This new emission factor was derived toreplace the emission factors from EPA Protocol 1995. Field data andMonte Carlo statistical simulation techniques were applied toobtain the emission factor for valve, pump, connector, flange andinstrument leaks ranging from 3 to 60 g/h. Components werecategorized into “leaker” and “non-leaker,” based on the gas massflow rate relative to the lowest flow rate, which can be detected byinfrared camera. The emission factors were calculated for 3, 6, 30and 60 g/h detection threshold (Lev-On, Epperson, Siegell, & Ritter,2007).

3. Methodology to estimate methane emissionsby dispersion modeling approach

The semi-quantitative method to determine emission factors fordifferent components was a suitable approach to estimate the totalmass emission rate of fugitive gas released from leaking compo-nents. However, this method evaluated the emission factor ofa component by comparing the leak rate to the lowest threshold, orsensitivity level, where the camera detector responded. The emis-sion factor was calculated by summing the total emissions dividedby the total count of the components. Furthermore, this semi-quantitative measure was general for all hydrocarbons which aninfrared camera could detect. This method was not able to identifythe correlation between the sizes of detected gas plumes and theemissions rate for a specific gas. Different gases have differentcharacteristic when released to the atmosphere. Some gases whichare denser than air will have a tendency to stay on the groundbefore eventually warming up and lifting upwards, whereas, somegases are lighter than air and rise when released into the atmo-sphere. Accurate quantitative measures for specific gases areimportant, and therefore, this paper proposes a solution for

d visualization.

table flow rate Minimum detectable concentration Size of cloud

300 ppm 30 cme 50 cm275 ppm e

Fig. 2. Flow process of a gas leak in a pipeline.

A. Safitri et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 138e145 141

estimating the gas mass flow rate of methane emissions based ona dispersion modeling approach.

The methodology for estimating methane gas emissions ratesfrom natural gas distributions and pipeline systems is given inFig. 1. In the event of a leak occurs in a natural gas pipeline, a certainamount of gas will be released, mixed and diluted with air. Theleakage rate from the emission source depends on the differentialpressure between the upstream pressure and ambient pressure, thesize of the leak, and the temperature of the gas inside the pipeline.The concentration of the gas as it is dispersed can be estimatedusing Gaussian dispersion modeling. Sensitivity analysis of the gasconcentration at different meteorological conditions is simulated.Using a Gaussian dispersion model, the correlations betweenleakage rates and plume size, i.e. distance and plume height tominimum detectable downwind concentration for methane gas,can be predicted.

For the discharge or gas emission rate calculation, smooth/fric-tionless converging nozzle for the hole is assumed. It is alsoassumed that the condition in the reservoir is stagnant. Most ofpressure energy is converted to the kinetic energy due to theassumption of frictionless nozzle and therefore the assumption ofisentropic behavior is generally valid. The calculation of gas flowrate will be different for a converging-diverging nozzle or a slitcrack. For a slit crack or slot hole, a throttling process with largefrictional loss occurs and little of the energy inherent to the gaspressure will be converted into kinetic energy. The discharge ratewill be significantly different for a release from frictionless roundconverging nozzle and a slit which eventually affect the concen-tration downstream.

For relatively short pipe, the gas flowing through a pipeline isgenerally calculated with the assumption of adiabatic flowbehavior, where no heat transfer occurs between the system and itssurroundings (Perry & Green, 2008). Fig. 2 illustrates the process ofgas being released and dispersed from a leaky pipeline.

The gas release rate is calculated using equations for flow ofcompressible gas through frictionless nozzles, which are readilyavailable in the literature (Crowl & Louvar, 2002; Perry & Green,2008). It is important to specify the critical conditions prior tocalculating the gas flow rate coming from the leak. The criticalcondition of the gas flow is achieved if the exit velocity of gas isequal to the speed of sound. For ideal gas, the speed of sound isgiven as (Perry & Green, 2008):

c ¼ffiffiffiffiffiffiffiffiffiffikRTMW

r(1)

where, k is the ratio of the specific heats, Cp/Cv; R is the gas constant(3.314 kJ kgmol�1 K�1); T is absolute temperature (K) andMW is themolecular weight (kg kgmol�1).

The term Mach number is introduced as a dimensionlessparameter which compares the velocity of gas to the velocity of

Gas emissions rate prediction based on the size of the gas plume

Gaussian dispersion model

Ambient environmental condition

Methane gas flowing through a leak in a pipeline

Fig. 1. Methodology to estimate methane gas emissions rate using dispersion simu-lation of methane gas leak from natural gas pipeline system.

sound. When the flow is critical, or choked, the Mach number isequal to one.

Ma ¼ uc

(2)

The criticality of the flow is determined by following equation(Crowl & Louvar, 2002):

PchokedP0

¼�

2kþ 1

�k=k�1

(3)

where P0 is the pressure of the gas inside the pipeline.If the downstream pressure, P2, is higher than the choked

pressure, Pchoked, the flow is non-critical and the exit pressure, P1,will be the same as the surrounding pressure, P2. The mass flux rateis calculated from the equation for subsonic flow, as given in thefollowing (Crowl & Louvar, 2002):

G ¼ P0

ffiffiffiffiffiffiffiffiffiffiffiffikMWRT0

sMa1�

1þ k�12 Ma1

�ðkþ1Þ=ðk�1Þ (4)

For subsonic flow, the Mach number can be obtained from thepressure correlation given in the following equation (Crowl &Louvar, 2002):

P0P1

¼�1þ k� 1

2Ma21

�k=k�1

(5)

The property relations for isentropic process are given as (Perryand Green, 2008):

r

r0¼

�PP0

�1=k(6)

and:

TT0

¼�PP0

�ðk�1Þ=k(7)

If the downstream pressure, P2, is lower than the choked pres-sure obtained from Equation (3), the gas flow becomes sonic, orchoked. If this occurs, the exit pressure, P1, is equal to Pchoked. Theexit velocity is equal to the speed of sound, and the maximum flowrate is attained when choked flow occurs. The maximum mass fluxof gas discharged from the leak source is given as (Crowl & Louvar,2002):

G ¼ P0

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffikMWRT

�2

kþ 1

�ðkþ1Þ=ðk�1Þs

(8)

Natural gas at an ambient temperature is considered as light orbuoyant gas, because the density is lower that density of air. The

Table 4PasquilleGifford dispersion coefficients for Plume dispersion (Briggs, 1974; Griffiths,1994).

PasquilleGiffordstability class

sy(m) sz(m)

Rural conditionsA 0.22x(1 þ 0.0001x)�1/2 0.2xB 0.16x(1 þ 0.0001x)�1/2 0.12xC 0.11x(1 þ 0.0001x)�1/2 0.08x(1 þ 0.0002x)�1/2

D 0.08x(1 þ 0.0001x)�1/2 0.06x(1 þ 0.00015x)�1/2

E 0.06x(1 þ 0.0001x)�1/2 0.03x(1 þ 0.0003x)�1

F 0.04x(1 þ 0.0001x)�1/2 0.016x(1 þ 0.0003x)�1

Urban conditionsAeB 0.32x(1 þ 0.0004x)�1/2 0.24x(1 þ 0.0001x)�1/2

D 0.22x(1 þ 0.0004x)�1/2 0.2xD 0.16x(1 þ 0.0004x)�1/2 0.14x(1 þ 0.0003x)�1/2

EeF 0.11x(1 þ 0.0004x)�1/2 0.08x(1 þ 0.00015x)�1/2

2500

3000

3500

4000

4500

(k

g m

-2

s

-1)

A. Safitri et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 138e145142

distribution profile of the methane gas plume can be predictedusing a generalized Gaussian dispersion model. The equation forcontinuous point source plume model is given in the following(Beychok, 1994):

Ciðx; y; zÞ ¼ Qm;i

2psyszuexp

�� 12

�ysy

�2�(exp

"� 12

�z� Hsz

�2#

þ exp

"� 12

�zþ Hsz

�2#)

ð9Þ

where: Qm,i is source emission rate (m s�1); H is effective height ofrelease (m); sy is diffusion coefficient in y direction (m); sz isdiffusion coefficient in y direction (m).

The diffusion coefficients used are obtained from the stabilityclassification by Pasquill and Gifford and determined from theamount of incident solar radiation and wind speed. Table 3 givesthe PasquilleGifford criteria to determine the stability class andTable 4 provides the equations to calculate the diffusion coeffi-cients. The mixing between methane gas and air highly depends onthe magnitude of wind speed and the surface roughness. Thesurface roughness for rural condition is used in this paper.

A dispersed gas can be recognized by the infrared camera toa certain threshold concentration. The concentration of thedispersed gas at a downwind distance from the release depends onthe meteorological conditions including wind speed, wind direc-tion and ambient temperature. Since the atmospheric conditionsare constantly changing, it is impossible to assign a single value tothose parameters and therefore the atmospheric parameters aregenerally presented in time-average value. By examining the windspeed and the solar radiation condition during measurement, thePasquilleGifford stability class can be assigned to the sitewhere thescanning will be performed. The wind speed and direction as wellas ambient temperature on site can be observed and recorded usingweather station at real-time.

Using infrared optical imaging, the gas plume can be visuallyobserved. Based on several published works listed in Table 2, thesensitivity of the camera and the capability of the camera to detectthe lowest flow rate and lowest concentration have been described.In a controlled environment the infrared imaging system can detectup to 300 ppm. However, the sensitivity slightly decreases bya factor of two to 600 ppm when tested under the realistic condi-tions of mild weather with 2 m/s wind speed. Using a Gaussiandispersion approach, the plume size of theminimum detectable gasconcentration can be used to predict the release rate for the pre-determined environmental condition.

The concentration profile of the dispersed gas in air can becalculated using Equation (9). The concentration at a specificdownwind distance depends on the discharge rate and the pre-vailing meteorological conditions at the point of release. Sensitivityanalyses for different variables were carried out in order to obtainthe significance of each variable. The magnitude of the dischargerate depends on the upstream pressure and leak diameter. Themass release rate was calculated for an upstream pressures range

Table 3PasquilleGifford stability class (Gifford, 1976).

Surfacewindspeed

Daytime solar radiation Nighttime conditions

Strong Moderate Slight Thin overcastor >4/8 low cloud

�3/8cloudiness

<2 A AeB B F F2e3 AeB B C E F3e4 B BeC C D E4e6 C CeD D D D>6 C D D D D

120e810 kPa and a leak diameter of 2.5e25 mm. This pressurerange represents the typical upstream pressure in natural gastransportation and distribution pipelines. The leak sizes are thetypical fugitive leak sizes in the oil and gas industry which aregenerally used for consequence and risk assessment purposes.

4. Results and discussion

4.1. Simulation of gas release from a leak in a pipeline

The leak rate of methane gas into the atmosphere at differentupstream pressures and temperatures is calculated using Equations(4) and (8) described in the previous section. At upstream pressureshigher than 187 kPa, the gas becomes choked and the gas velocityreaches sonic velocity. The gas mass flux keeps increasing at sonicconditions because the density of the gas is increased. A slighttemperature decrease of the natural gas inside the pipeline will notaffect the amount of gas discharged from the leak source. The gasmass flux as a function of upstream pressure is given in Fig. 3.

Gas is usually transported at ambient conditions, except forunderground pipelines in which the temperature of the gas isslightly lower than the ambient temperature during warmweatherand higher than ambient temperature during cold weather. Fig. 4shows the mass discharge rate of the gas from 2.5 mm leak atvarious pressures and temperatures. From the figure, it can beconcluded that the mass flow rate highly depends on the upstreampressure, but the effects of a slightly increasing or decreasingtemperature is insignificant to the gas mass flow rate.

0

500

1000

1500

2000

1.0E+05 1.0E+06 1.0E+07

Ga

s M

as

s F

lu

x, G

Upstream pressure (Pa)

Fig. 3. Methane gas mass flux at various upstream pressures at 293 K.

0.0E+00

1.0E-03

2.0E-03

3.0E-03

4.0E-03

5.0E-03

6.0E-03

7.0E-03

1.0E+05 1.0E+06

Mass release rate, Q

m (kg

s

-1)

Upstream pressure, P0

(Pa)

T = 273 K

T = 283 K

T = 293 K

T = 303 K

Fig. 4. Mass discharge rate of methane gas at various pressures and temperatures from2.5 mm leak source.

-50

-40

-30

-20

-10

0

10

20

1.0E+05 1.0E+06

Exit tem

ep

eratu

re (

C)

Upstram pressure, P0

(Pa)

T = 273 K

T = 283 K

T = 293 K

T = 303 K

Fig. 6. Exit temperatures of methane gas as a function of upstream pressure at variousgas initial temperatures.

A. Safitri et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 138e145 143

Parameters at the gas exit point can indicate the transition of thegas from critical (choked) to non-critical condition. The exit pres-sures at various upstream pressures are shown in Fig. 5. From thefigure, it can be seen that at upstream pressures higher than 2 kPathe gas becomes choked and therefore, the exit pressure increasesas upstream pressure increases. The exiting gas will experiencea highly non-isentropic series of shocks until it reached theambient pressure.

As the gas reaches critical, sonic, or choked condition, the exittemperature becomes constant although the pressure in theupstream is increased. Fig. 6 shows the gas temperature profile atthe exit (release) point for different initial temperatures. The exittemperature decreases as the upstream pressure increased until itreaches the sonic condition, and it becomes constant afterward. Theexit parameters, i.e. pressures and temperatures, are important todetermine the phase of methane gas being released in order todetermine whether condensation had occurred due to the expan-sion of gas froma pressurized source to atmospheric conditions. Theboiling point of methane gas is at temperature 273 K and 810 kPa is140 K. Therefore, for methane gas released at an upstreamtemperature range between 273 and 303 K and a pressure range of120e810 kPa, the exit pressures and temperatures indicate that thephase of the methane being released is still at 100% gas.

0.0E+00

1.0E+05

2.0E+05

3.0E+05

4.0E+05

5.0E+05

1.0E+05 1.0E+06

Exit p

ressu

re, P

1(P

a)

Upstream pressure, P0

(Pa)

Fig. 5. Pressure of methane gas at release point as a function of upstream pressures ofgas inside a stagnant container (pipeline).

The density of gas as a function of the pressure and temperatureis given in Fig. 7. The gas becomes denser as the pressure increasesdue to compression. However, upon increasing the temperature,the density decreases because themolecules aremore spread out athigher temperature.

The gas mass release rate at various leak sizes with constant gasand ambient temperature is shown in Fig. 8. Themass flow rate wascalculated for leak diameter range of 2.5e25 mm based on thetypical size for fugitive leaks in the pipelines coming from flanges,connections, fittings and valves. The mass release rate increasesrapidly for larger leak areas. In real situations, releases from a largeleak will result in a decrease in upstream pressure. However, in thiscalculation a large inventory of gas inside the chamber is assumed.Therefore, the stagnant upstream condition is applied anda decrease in upstream pressure is not taken into account.

4.2. Simulation of methane gas dispersion

After the source or discharge term is identified, the dispersion ofgas is simulated using the generalized Gaussian model and Pas-quilleGifford stability class to calculate the concentration profilealong the downwind and vertical direction. The PasquilleGiffordstability classification is empirically determined based on the

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

1.0E+05 1.0E+06

Ga

s D

en

sity

(k

g m

-3

)

Pressure (Pa)

T = 273 K

T = 283 K

T = 293 K

T = 303 K

Fig. 7. Methane gas density as a function of pressures and temperatures.

1.0E-04

1.0E-03

1.0E-02

1.0E-01

1.0E+000.0E+00 2.0E+05 4.0E+05 6.0E+05 8.0E+05 1.0E+06

Ga

s m

as

s flo

w r

ate, Q

m (k

g s

-1)

Upstream pressure, P0

(Pa)

dhole = 2.5 mm

dhole = 10 mm

dhole = 15 mm

dhole = 20 mm

dhole = 25 mm

Fig. 8. Gas mass discharge rate at 293 K for various upstream pressures and leak sizes.

0.1

1

10

100

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Clo

ud

H

eig

ht to

m

in

. d

et. c

on

c. (m

)

Qm (kg/s)

Class F, 2 m/s

Class D, 2 m/s

Class D, 5 m/s

Fig. 10. The height of the cloud to minimum detectable concentration as a function ofgas release rate for different stability classes.

A. Safitri et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 138e145144

amount of incoming solar radiation, wind velocity, and terrainsurface condition. In the dispersion calculation, the mass dischargerate is assumed to be constant and the gas in the pipelines is atstagnant condition, so the reduction in the gas inventory upstreamis assumed to be negligible. Using this assumption, the concen-tration downwind and crosswind is merely a function of meteo-rological conditions. In this paper, the dispersion of methane gas iscarried out for three different weather conditions: (i) stableweather condition (Stability Class F) with 2 m s�1 wind speed, (ii)neutral weather condition (Stability Class D) with low wind speed(2 m s�1), and (iii) neutral weather condition (Stability Class D)with medium wind speed (5 m s�1).

The sensitivity of infrared imaging to detect methane gasconcentration at real weather conditions is up to 600 ppm, asdescribed in Table 2. Therefore, the gas emission rate is calculatedbased on the size of the cloud up to this minimum detectableconcentration. Fig. 9 shows the downwind distance to theminimum detectable concentration as a function of gas massrelease rate at various weather conditions. From that figure it canbe implied that the downwind distance is highly affected by thewind speed; however, the change due to a different weatherstability is not very significant for Class F and Class D. Based on theprediction using the Gaussian model, the downwind distance to600 ppm of methane gas is increased by approximately a factor oftwo when the wind speed escalates from 2 m s�1e5 m s�1.

1

10

100

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7Do

wn

win

d d

is

tn

ac

e to

m

in

. d

et. c

on

c. (m

)

Gas emission rate, Qm (kg s-1)

Class F, 2 m/s

Class D, 2 m/s

Class D, 5 m/s

Fig. 9. Downwind distance to minimum detectable concentration as a function of gasrelease rate for different stability classes.

The effect of the wind speed is also apparent in the height of thecloud. At 600 ppm concentration, the height of the cloud at lowerwind speeds is higher than the height of the cloud at higher windspeeds, as shown in Fig. 10. This could occur because at higher windspeeds the horizontal momentum from the wind is much largerthan the gravitational momentum and vertical momentum of thegas. However, for the same wind velocity, the distance to minimumdetectable concentration is similar for stable and neutral condi-tions, so the incoming solar radiation or cloudiness has an insig-nificant effect on the travel distance of dispersed gas.

The graphs shown in Figs. 9 and 10 demonstrate that theprediction of the gas emission rate based on the cloud size usingGaussian model works very well for relatively small cloud sizes(less than 30 m in downwind direction and 20 m in height). Forlarger cloud sizes, the prediction of the emission rate might beinaccurate. Therefore, predicting the discharge rate of gas froma leaky pipeline using optical imaging is suitable for fugitive leaksbut not for large leaks or ruptures.

5. Uncertainties evaluation

The application of this methodology carries some uncertaintieswhich come from both the application of infrared imaging and thedispersion modeling. There are parameters affecting the perfor-mance of infrared camera for gas visualization such as the back-ground temperature, environmental conditions, atmosphericconstituents, path length and reflection from the surroundingobjects. The infrared cameras sense the radiation that appears toemanate from target object and its surrounding which also hasradiating power. When the target object reaches a thermal equi-libriumwith the environment, no heat transfer occurs and both gasand the environment are emitting the same amount of radiationpower. The environmental conditions also give a significanteffect to the camera performance. The atmospheric constituentsincluding numerous gases and aerosols absorb and scatter radia-tion as it travels from the target to the infrared camera. The pres-ence of these gases will reduce the target contrast depicted in theimage. The environmental conditions such as relative humidity candegrade the system performance considerably. High humidity cancause aerosol particulate growth due to the molecular absorptionby the water vapor which eventually will reduce the transmittance.Reflection from surrounding objects will affect the image quality inthe thermogram as well because infrared camera cannot separatethe emitted radiations from the reflected radiations. Therefore, anyradiation transmitted to the infrared camera must include theradiation reflected from the surrounding. There parameters can

A. Safitri et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 138e145 145

significantly reduce the image contrast and hence may affectthe ability to recognize the minimum detectable concentrationobserved in the thermogram.

Uncertainties are also generated from the prediction of gasemission rates based on gas cloud sizes calculated by the general-ized Gaussian model. First, the model does not take into consid-eration the changes in local meteorological conditions. In the field,the meteorological conditions are continuously changing and it isalmost impossible to assign a constant value for weather conditionssuch as wind speed, solar intensity and humidity. The generalizedGaussian model has not been able to address these factors in itscalculation; steady weather condition is assumed in the calculationand therefore the simulation result is usually more conservativecompared to the real conditions. The fluctuation in wind directionand wind speed results in the dynamic movement of the dispersedplume and variation in concentration at a certain downwinddistance. This limitation can be resolved by having real-timescanning at the same period when the weather measurements aretaken and simulate the dispersion at a reasonable time interval inorder to get the most accurate simulation results.

Secondly, the Gaussian model estimates the cloud size, shapeand concentration of dispersed gas from a continuous source afterthe gas becomes passive or is at the steady state. The model doesnot take into account the transient condition at the time the gas hasbeen discharged. The generalized Gaussian model also assumesa continuous and constant emission rate; the reduction in the gasinventory, which results in a reduction of the gas emissions rate, isnot taken into account in this model (Beychok, 1994). This meth-odology provides a quantitative prediction of methane emissionrates based on the size of the plume compared to the minimumdetectable concentration observed from the infrared opticalimaging. However, further validation will be required for thismethod, and offer correction factors to the simulation based on theexperiment. Therefore, further study in this area should includea leak simulation experiment to visualize methane gas releasesfrom pipeline leaks. It would be helpful in determining the sensi-tivity of the camera and obtaining a comparison of concentrationprofiles gathered from the infrared imaging technique and themeasurements from total gas analyzer.

6. Conclusion

Infrared optical imaging technology has been used as anadvanced gas leak detection method in the oil and gas industry forits ability to spot leaks not visible to naked eyes. Although theinfrared system can detect the leak sources accurately, it does notgive anymeasurement of the emissions rate or the concentration ofthe leaking gas. A semi-quantitative approach to estimate theemissions rate from hydrocarbon fugitive leaks detected by theinfrared camera has been done previously using a statistical MonteCarlo simulation approach which categorizes leaks from differentequipment and components based on their detectable thresholds.However, this methodology is generalized for all types of hydro-carbon, and the quantification process is only based on a statisticalapproach. In this paper, quantification of the gas emissions rate fromleaks in natural gas pipeline using adispersionmodeling approach ispresented. The release or emissions rate of the gas is predicted basedon the size of the cloud observed from the infrared video image(thermograms) up to a certain concentrationwhich the camera candetect (minimum detectable concentration). The Gaussian model isused to calculate thedispersionof themethanegas clouddischargedfroma leak in a pipeline. The amount of the gas released depends onthe upstream pressure and leak size. Dispersion of gas from a sourceis affected by the amount of the gas released as well as prevailingweather conditions at the time of the dispersion. The infrared

imaging technique has been tested to monitor leaks in natural gassystems under real weather conditions. The results show that thegas emission rate can be accurately predicted at a low gas releaserate (less than 0.1 kg s�1). At a high emissions rate, the growth in thesize of the cloud does not appear to be very significant. Wind speedis the most determining factor in the dispersion of the gas. As thewind speed becomes higher, the horizontal transport due to thewind is dominant compared to the diffusion and gas buoyancy; thusthe downwind distance of the cloud is longer and the cloud shape isnarrower at higher wind speeds. The prediction of the mass emis-sions rate using a generalized Gaussian model carries severaluncertainties. Therefore, the path forward in this research shouldinclude conducting a simulation of a methane gas leak in order tovalidate the developed methodology.

Acknowledgements

This research is funded by Mary Kay O’Connor Process SafetyCenter, Artie McFerrin Department of Chemical Engineering, TexasA&M University.

References

Benson, R., Madding, R., Lucier, R., Lyons, J. & Czerepuszko, P. (2006). Standoffpassive optical leak detection of volatile organic compounds using cooled InSbbased infrared imager. Proceeding of air and waste management association(AWMA) 99th annual conference and exhibition.

Benson, R., Panek, J. A., & Drayton, P. (2006). Direct measurement of minimumdetectable vapor concentrations using infrared optical imaging systems.Proceeding of air and waste management association (AWMA) 99th annualconference and exhibition.

Beychok, M. R. (1994). Fundamentals of stack gas dispersion (3rd ed.).Briggs, G. A. (1974). Diffusion estimation for small emissions. Report ATDL-106.

Washington DC: Air Resources, Atmospheric Turbulence, and Diffusion Labo-ratory, Environmental Research Laboratories.

Crowl, D. A., & Louvar, J. F. (2002). Chemical process safety: Fundamentals withapplications (2nd ed.). Prentice Hall.

Energy Information Administration. (2009a). Section 1: Energy consumption bysector. Annual energy review 2008. DOE/EIA-0384(2008).

Energy Information Administration. (2009b). Section: Natural gas: Data of producingwells in the United States.

Environmental Protection Agency. (1999). Method 21: Determination of volatilecompound leaks.

Environmental Protection Agency. (2008). Environmental and sustainable technologyevaluation (ESTE): Verification of portable optical and thermal imaging devices forleak detection at petroleum refineries and chemical plant.

Environmental Protection Agency. (2009). Section 3.5: Natural gas system (IPCCsource category 1B2b). Inventory of U.S greenhouse gas emissions and sinks:1990e2007.

Epperson, D., Lev-On, M., Tabback, H., Siegell, J., & Ritter, K. (2007). Equivalent leakdefinitions for smart LDAR (Leak detection and repair) when using opticalimaging technology. Journal of Air and Waste Management Association, .

Gifford, F. A. (1976). Turbulent diffusion-typing schemes: a review. Nuclear Safety, 17(1), 68.

Griffiths, R. F. (1994). Errors in the use of the Briggs parameterization for atmo-spheric dispersion coefficients. Atmospheric Environment, 28(17), 2861e2865.

Gross, W., Hierl, T., Scheuerpflug, H., Schirl, U., Schreer, O., & Schulz, M. (1998).Localization of methane distribution by spectrally tuned infrared imaging.InProceeding of SPIE conference on air monitoring and detection of chemical andbiological agents, Vol. 3533 (pp. 234e240). SPIE.

Holst, G. C. (2000). Common sense approach to thermal imaging. SPIE Press and JCDPublishing.

Lev-On, M., Epperson, D., Siegell, J., & Ritter, K. (2007). Derivation of new emissionfactors for quantification of mass emissions when using optical gas imaging fordetecting leaks. Journal of Air andWaste Management Association, 57, 1061e1070.

Perry, R. H., & Green, D. W. (2008). Perry’s chemical engineering handbook (8th ed.).McGraw-Hill.

Plisson-Saune, S., Suripno, S., Cramer, A., Pramono, H., Camps, R., & Lacamoire, H.(2008). Fugitive emissions management through infrared monitoring e fullscale field application. In2008 international SPE conference on health, safety andenvironment in oil and gas exploration and production, Vol. 111584 (pp. 1e10). SPE.

Robinson, D. R., & Luke Boone, R. E. (2003). Identifying natural gas leaks to theatmosphere with optical imaging. ICF Consulting, Inc.

Schulz, M., Gross, W., & Scheuerpflug, H. (2008). High resolution thermophysicalmeasurements using staring infrared detector arrays. In High temperaturesehighpressures. ECTP proceedings, Vol. 32 (pp. 527e536).