Professional Documents
Culture Documents
Dr David Stewart
Date: 4 July 2005
for M Valente
Managing Director
EXECUTIVE SUMMARY
During the period between February 2004 and February 2005, NEL in collaboration
with OPUS PLUS Ltd carried out a Joint Industry Project (JIP) entitled “Oil-in-Water
Analysis Method” (OIWAM). The JIP was initiated in response to an OSPAR
Recommendation to implement a new oil-in-water analysis reference method based
on using a modified version of the ISO 9377-2 method (ISO Mod). The Project was
sponsored by the DTI in the UK, the State Supervision of Mines (acted on behalf of
the Ministry Economic Affairs and NOGEPA) in the Netherlands, and eight UK
offshore operators including BG Group, BHP Billiton, CNR International,
ChevronTexaco, Kerr-McGee, Marathon Oil, Shell UK, and Talisman Energy (UK).
The proposed new reference method is a gas chromatography (GC) based method,
and is significantly different from the current infrared (IR) methods. Also GC
instruments used in the new reference method in general require much more skill to
operate than the IR instruments used in the current approved methods. As a result of
this, it was thought that there would be many challenges in implementing the new
reference method.
The objective of the study was to prepare both the regulators and the industry for a
smooth transition from the current reference method to the new GC based reference
method. Specifically the project aimed to achieve the following:
• To identify the best practical means to implement the new reference method
• To establish best practice guidelines for sample taking and handling
• To develop a realistic set of acceptance criteria for alternative methods
• To advise on how to relate data from the new reference method to the current
approved method
The study has concluded that the best practical way forward to implement the new
ISO Mod reference is to use alternative methods which are calibrated against the
new reference method. Acceptance of these alternative methods shall be primarily
based on having a valid calibration established either by analysing calibration
standards prepared in a shore based laboratory or by analysing replicate produced
water samples taken from fields. Different calibration approaches have been
suggested for alternative methods using laboratory bench-top analysers and online
monitors respectively. However these approaches will need field trials to confirm their
practicality. Calibrating an alternative method against the ISO Mod reference method
for oil-in-water analysis, and ensuring the calibration is valid over time will not be
easy. Indeed operators should be prepared to expect more effort in implementing the
ISO Mod method than the current IR methods.
Eight general guidelines have been established for oil in produced water sample
taking and handling. Proper adherence of these guidelines will minimise errors which
may be resulted from improper sample taking and handling. With regard to data set
continuity, for specific installations, simple correction factors may be established to
convert data from the ISO Mod method to IR method. However it is not realistic to
expect to develop a universal correlation or correction factor between the new ISO
Mod method and current IR methods across all the fields in the North Sea, because
there are significant differences between the current IR and the new ISO Mod
methods and also because there are variations in the current IR methods used in
different North Sea countries.
CONTENTS
Page
1 INTRODUCTION
1.1 Background ................................................................................................... 5
1.2 Objectives ...................................................................................................... 6
1.3 Outline of Technical Programme ................................................................... 6
2 REVIEWS
2.1 Current Oil-in-Water Analysis Practices in the UK and the Netherlands ....... 9
2.2 Current Progress in Identifying Alternative OIW Analysis Methods .............. 12
2.3 Use of the ISO 9377-2 GC-FID Method Offshore ......................................... 14
2.4 Availability of Suitable GC-FID Instruments .................................................. 16
2.5 Onshore Laboratories with IR and GC-FID Analytical Capability .................. 18
2.6 Possibility of Establishing a Secondary Reference Method .......................... 20
6 ACCEPTANCE CRITERIA
6.1 Basis of Acceptance Criteria ......................................................................... 46
6.2 General Calibration Issues ............................................................................ 46
6.3 Instrument Availability and Dealing with Failure ............................................ 48
6.4 Data for Reporting ......................................................................................... 49
6.5 Recommended Calibration Approach for Laboratory Based Bench
Top Analysers ............................................................................................... 49
6.6 Recommended Calibration Approach for Online Monitors ............................ 53
9 CONCLUSIONS ............................................................................................. 66
ABBREVIATION......................................................................................................... 70
GLOSSARY ............................................................................................................... 71
REFERENCE ............................................................................................................. 72
1 INTRODUCTION
1.1 Background
Oil in produced water is a key parameter for the offshore industry both for process
control and for the purpose of reporting to the authorities. The determination of oil in
produced water for reporting has been traditionally carried out using approved
reference methods [1-4] based on solvent extraction followed by an infrared
quantification. Whilst there are differences between the exact IR reference methods
used in the different North Sea countries, these methods have been used for offshore
oil in produced water analysis for the last two decades. However due to the phase-
out of Freon-113, and concerns over health and safety of its replacement solvent
Perklone (Tetrachloroethylene or TTCE), there has been a drive to develop and
adopt a new reference oil-in-water analysis method by both the authorities and the
offshore industry.
Following the establishment of the ISO 9377-2 method [5] (also see Appendix 1) for
the determination of hydrocarbons in water in 2000, there have been several studies
[6-8] aimed at establishing its suitability as a new reference method for the
determination of oil in produced water for the offshore industry. However all the
studies have indicated that there were differences between the results from the ISO
method and the current infrared based methods. As a result a modified version of
the ISO method was considered and a large comparison study [9] involving samples
taken from 58 installations was carried out under the auspices of the OSPAR OIC
(Oslo-Paris Commission Offshore Industry Committee). Despite the significant
difference observed between this method and the current Infrared methods from the
latest study, the modified ISO method was recommended in 2003 [10]. Modifications
to the ISO method are detailed in an OSPAR document [11] (also see Appendix 2).
At the 2003 OSPAR OIC meeting it was formally agreed that the modified version of
the ISO 9377-2 GC-FID (Gas Chromatography and Flame Ionisation Detection)
method will become the new reference method for the determination of dispersed oil
in produced water offshore. It was also agreed that the method will enter into force
from 1 January 2007.
As the new method is inherently different from the current infrared based methods,
and more importantly the new method is not ideally suited for offshore use, it is
recognised that there will be many practical challenges in the implementation of this
new reference method.
The challenges in implementing the new reference method result from the following
issues:
• The new reference method is very different from the current IR methods and will
produce different results (this has been proved by a number of different studies)
• Results obtained from all the comparison studies have indicated that no
generalised correlation could be established across the spectrum of the offshore
installations between the ISO based method and the current methods
• There may be a need to determine a correction factor on an installation by
installation basis
• The practical limitations of the new reference method may necessitate the use of
alternative methods in the offshore environment.
It is on this basis that NEL proposed that a Joint Industry Project (JIP) be conducted
to prepare for the implementation of the new reference method and to ensure a
smooth transition from the use of the current infrared method to the new reference
method.
1.2 Objectives
The overall objective of the JIP was to prepare the authority and the industry for a
smooth transition from the infrared based oil-in-water method to a method based
around the modified version of the ISO reference standard method. More specifically
the JIP aimed to achieve the following:
• To identify the best practical means to implement the new oil-in-water analysis
reference method
• To establish best practice guidelines for oil-in-water sample taking and sample
handling
• To develop a realistic set of acceptance criteria for alternative oil-in-water
analysis methods including both bench-top and on-line methods. This will
encourage the use of alternative methods for both reporting and process
management purposes.
• To advise on how to relate results from the new reference method, and
alternative methods, to the current approved method in order to maintain the
continuity of data and to facilitate practical comparison of new and historical data
by the regulators.
The aim here was to establish an overall strategy as to how the new reference
method can be implemented smoothly. This involved reviews on the following
issues:
• Current practices
• Availability of suitable GC-FID instruments that could be used for offshore
applications in either permanent or portable form
• Availability of the required skills for operating GC-FID instruments on offshore
installations
• The possibility of establishing a secondary reference method that is calibrated
against the current method and/or the new reference method with the view that it
can then be used offshore for calibrating other alternative methods
• Experience and lessons learned from using GC-FID methods for the
determination of oil-in-water analysis offshore
• Basic requirements for implementing the method such as frequency and number
of samples to be taken
• Availability of onshore laboratories capable of analysing oil-in-water samples
according to the modified ISO GC-FID method
• Current progress in identifying alternative oil-in-water analysis methods by
offshore operators in the North Sea or elsewhere
Results obtained from reviewing the above should provide a clear indication of the
challenges of implementation and the guidance required to ensure a smooth
transition to the use of the new reference method. Three scenarios were considered
while the review process is carried out. These were:
Sample taking and sample handling has been recognised as probably the most
significant factor that contributes random errors and uncertainties in the analysis of
offshore oil in produced water. There are some general guidelines available through
a number of studies that have been carried out on this subject. These were reviewed
so that best practice guidelines suited for the implementation of the new reference
method could be established.
Depending on what methods were acceptable and where these were to be used,
different requirements resulted in terms of sample taking and sample handling.
A separate set of acceptance criteria were established for both laboratory bench-top
analysis methods and on-line monitoring methods. These criteria were produced
with detailed consideration of the following issues:
What are the basic requirements for oil-in-water analysis instruments and methods to
be considered for acceptance? These included parameters such as accuracy,
measurement range, reproducibility, and explosion-proof rating etc.
Which data should be reported? In what format should the data be presented?
These will be particularly important when considering the acceptance of on-line oil-in-
water monitors.
As the two reference methods were considered to be inherently different and the
recent OSPAR comparison project failed to establish a generalised correlation
between the two methods, there will be difficulties in relating data from the new
method to the current one.
In order to relate results from the new reference method to that of the current
method, it may be necessary that conversion factors be used to compensate for the
differences between the two methods for individual installations. Advice is therefore
given on how to establish such conversion factors and on what basis.
2 REVIEWS
UK
The vast majority of samples are taken from the overboard discharge lines (all
approved by the DTI). For installations that discharge continuously samples are
taken twice a day at equal time intervals. All samples are collected using 500 ml
bottles with caps; however not all of the caps have inert liners.
Sampling technique varies, with some using a “quill”. Sampling tube size ranges from
¼ inch to ½ inch and in most cases made of stainless steel. Samples are often taken
after running the sampling line from ½ minute to 5 minutes, with the majority around
a couple of minutes. There are also a number of installations where the sampling line
runs continuously.
The vast majority of samples are acidified with hydrochloric acid; acid is added to
sample bottles either prior to sample taking or after the sample bottle is filled.
Samples are usually analysed within minutes, normally after the samples have
cooled to room temperature.
The Netherlands
16 samples are analysed per month for those that discharge produced water
continuously. Samples are taken using ½ inch to ⅜ inch stainless steel tubes with
some using a “quill”. Both measuring cylinders and bottles are used for collecting
samples. Sample volumes vary from 250 ml to 1 litre. Samples are often collected
after running the sampling line for ½ minutes to 5 minutes. Some of the sampling
lines run continuously.
Approximately half of the samples analysed offshore are acidified. All samples are
analysed within 48 hours (as legally required).
UK
Analysis instruments used for oil-in-water include Miran 1FF, InfraCal, Buck Scientific
HC404 and Horiba OCMA-310. Apart from the Horiba instrument, the calibration and
analysis procedures specified by the DTI are followed. For Horiba instruments, a
Horiba S-316 solvent is used. Tetrachloroethylene (TTCE) is used for all other IR
based methods.
When the samples are sent to the onshore labs, information put on the label varies;
this may include date, time, platform, sample point / location, acid added or not,
destination lab, analysis required, bottle unique identifier etc.
The Netherlands
Analytical instruments used include Thermo Nicolet FT-IR, Perkin Elmer Spectrum
one, BOMEN MICHELSON MB 100 and OMS-1. Two different approaches have
been noticed. One of the operators adopts a method based on using the OMS-1
system. The method is then correlated to the NEN 6675 modified method [3]. Results
from the OMS-1 system and the reference method are then compared statistically.
The others follow the NEN 6675 modified method directly. Hexadecane is used for
calibration. For the OMS-1 based method, zero calibration is done daily and routine
calibration checks are made weekly. Correlation between the OMS-1 and NEN 6675
mod is re-established if it is found that the difference between the OMS-1 results and
the reference method results is statistically significant. For the others that follow the
NEN 6675 modified method, calibration is done once or twice a year.
Approximately 10% of installations in the UK have GCs available onsite and are used
for gas composition analysis. The carrier gas used is usually helium. Bottled
hydrogen or on-site generated hydrogen is used as fuel gas for around 50% of the
GCs. Skilled GC personnel are available at these installations, but not 100%. In the
Dutch sector only one out of 22 installations has got a GC instrument.
UK
Operators in the UK sector have not been actively seeking to use alternative
methods. Only one operator is actively working on alternative laboratory-based
methods. Some sponsors consider the current JIP as a way of looking into alternative
methods. One operator has carried out several trials with different lab-based and
online methods, but did not proceed any further as no reasonable correlation could
be established between these methods and the current DTI approved method.
The Netherlands
Out of 22 installations, only one installation has an on-line oil-in-water monitor, which
has been used so far for testing/trial purpose. A few trials have been conducted in
the past by two other operators, but none of them intends to install such a device in
the near future.
None of the five operators is currently seeking / testing an alternative method (non-IR
based) in preparation for the implementation of the ISO 9377-2 Mod method.
However the operator who uses the OMS-1 system, which is considered as an
alternative to the current approved Dutch method, may use the same instrument and
approach (of SPC) for implementing the new ISO 9377-2 Mod method.
A new ASTM infrared based oil-in-water analysis method was made official in May
2004. This method has been designated as ASTM D 7066-04 and is based on using
dimer/trimer of chlorotrifuoroethylene (S-316) [12], also called Flon.
The method is similar to ASTM 3921-85 method [13] in which oil is extracted by
Freon-113 and then quantified using an IR instrument at a fixed wavelength of 2930
cm-1 or 3.4 microns. This method is also very similar to the present DTI approved IR
method [1].
S-316 was one of a few solvents included in a previous UKOOA /DTI study [7]
carried out by NEL/OPUS. Its main composition is 1,2,3,4-tetrachloro-1,1,2,3,4,4-
hexafluorobutane. The solvent seems to have been developed specifically for the use
of oil-in-water analysis by Horiba in Japan who supply both laboratory and online IR
based oil-in-water analysers. Further information on the solvent is given below:
According to the Chair person of the ASTM Committee who developed the method,
the solvent S-316 is class “0“ (minimal) for health and protection (safety glasses
required for safety), it has little Global Warming Potential (GWP). Also it has a high
boiling point (above 200 ºF), which makes it less likely to enter into the atmosphere
than Freon. More importantly it is not on the Montreal Protocol list of solvents to be
discontinued.
In Norway Statoil have been using the S-316 solvent in their IR method for the
determination of gas / condensate in water for reporting. Prior to the use of S-316, a
study was carried out to find out if S-316 could be used to directly replace Freon in
their IR method. In the study tests were performed by taking two spot samples on
various installations, which were then analysed by the IR method using either Freon
or S-316 as an extraction solvent. The IR instrument was calibrated using a mixture
of crude oils from 7 installations. IR results were obtained both before and after the
clean-up of polar components. These results are shown in the Figure 1 and Figure 2.
A seven percent difference between the use of Flon and Freon is probably within the
IR method reproducibility. On this basis, the use of Flon in the IR method may be
accepted without even applying the correction factor.
A key issue with S-316 is the cost. In the UK currently the price is about 10 times that
of TTCE. For a discharge point, assuming two samples are taken and analysed per
day, (730 samples per year), it is estimated that about 58 litres of the S-316 solvent
is required, which may cost about £10K (compared to about £1k using TTCE).
However this may be reduced if the used solvent can be recycled. A solvent
reclaimer named Horiba SR-300, which costs £12k in the UK is available from
Horiba. According to Horiba the solvent can be re-used 3 or 4 times using the
reclaimer.
In the UK, Talisman Energy (UK) Ltd has carried out field evaluation studies
comparing an Arjay UV fluorescence based oil-in-water analysis method against the
DTI specified IR method, and also the ISO 9377-2 method (both original and Mod
version) [15, 55]. Field trips were initially made to six installations where samples
were taken and analysed using the respective methods.
Results showed that the UV fluorescence based method correlated well with the DTI
approved method for five out of the six installations. Repeat work (covering the
calibration procedures used) later carried out on the single installation giving a poor
initial correlation, showed good agreement between the IR and Arjay results. All
installations are oil producing with API gravity ranging from 30 o to 39 o (see more in
Section 8).
In Norway UV fluorescence based methods have already been used for monitoring
oil in produced waters, e.g. ConocoPhilips for some of their Ekofisk platforms.
Also both in Norway and Denmark InfraCal HATR (Horizontal Attenuated Total
Reflection) instruments have been used. In this method, pentane is used as a solvent
to extract oil from water samples. A fixed volume of the extract is then placed into a
HATR platform, which has an infrared transmitting crystal made of cubic zirconia or a
sapphire window. After evaporation of the solvent, an IR beam fixed at certain
wavelength is internally reflected down the ATR crystal. At each of the reflection point
the IR energy actually penetrates into a thin oil film left after the evaporation of the
solvent. IR energy is then absorbed each time the IR is reflected. The amount of IR
energy absorbed is proportional to the amount of the oil in the original sample. A
schematic diagram showing the principle of the InfraCal HATR is given in Figure 3.
However the method is not thought to be suitable for analysis of gas / condensate in
water samples.
Figure 3 A Schematic diagram showing the working principle of the InfraCal HATR
As far as offshore oil-in-water analysis using the ISO 9377-2 GC-FID method is
concerned, only operators in the Norwegian sector have any experience. Norsk
Hydro is by far the most active operator in implementing the ISO GC-FID method for
produced water oil-in-water determination. Nine of their installations have had the
method implemented. ConocoPhilips has also implemented the method at their
Ekofisk field.
(i) General
The GC-FID method was validated by West Lab Services in Norway in terms of
linearity, discrimination, reproducibility and sensitivity before the method was
implemented.
QC Level 1: Monthly water samples with known amount of reference oil are prepared
at the Department of Analytical Chemistry at Hydro’s Research Centre and sent
offshore to be analysed. The labs get feedback on the deviation from the “true” value.
• Telephone
• E-mail
• Hydro’s Intranet using Windows NetMeeting
Hydro thinks that the combination of telephone and Intranet NetMeeting gives the
supporting person at the Research Centre in Porsgrunn the opportunity to operate
the offshore GC instrumentation from onshore. The supporting person can control all
aspects of the GC instrument with the exception of physically changing components.
By adding web cameras at both ends, the support can be enhanced.
Norsk Hydro considers that the ISO Mod method is based on on-column injection,
which Hydro would not recommend for offshore use.
According to Norsk Hydro, the implementation of ISO 9377-2 has been easier than
originally anticipated. This was due to the following:
(i) Summary
ConocoPhilips implemented the ISO 9377-2 GC-FID method on its Ekofisk field
installations in 2002. The field has 6 platforms and two central laboratories. Both
laboratories were equipped with GCs for the analysis of oil-in-water. However due to
the long analysis time, only one GC is now in use. Samples taken from three
platforms are analysed using the ISO method; samples from the other three
platforms are now analysed using an UV fluorescence based handheld instrument.
15k NOK per person approximately £10k in total. Running cost: 10k NOK
(approximately £8K) per year for maintenance and operation
• Quality control – (a) one sample per month is sent to WestLab for check. (b)
there is also the yearly verification requirement by SFT, which is carried out by
WestLab
• Installation and maintenance by SAMSI - a separate small service company in
Norway
According to the Norwegian Petroleum Directory (NPD) website [20], the average
concentration of oil in discharged produced water in Norway had been reduced from
21.6 mg/l in 2002 to 16.9 mg/l in 2003, about 22% reduction in a year. Whilst this
may partly due to the cleaning measures implemented by the industry, it has been
acknowledged that the transition to a new reference method in 2003, i.e. from an
infrared based to the ISO 9377-2 GC-FID method, may have contributed to the
results. The site also confirmed that the impact of the change of the analytical
method on the results is being investigated.
If one looks at Norwegian historical oil in produced water data – 24.9 mg/l in 1999,
25.8 mg/l in 2000, 24.6mg/l in 2001 [21], between year 1999 to 2002, the reduction of
oil concentration was 3.3 mg/l (from 24.9 mg/l to 21.6 mg/l). Therefore an average
reduction of 4.7 mg/l in a single year from 2002 to 2003, coinciding with the
implementation of the ISO 9377-2 GC-FID method, raises the suspicion that the
transition from the old IR method to the ISO method has an impact on the results.
According to the ISO 9377-2 Standard [5], a GC must be equipped with a non-
discriminating injection system and a flame ionisation detector.
In addition, a data system should be suitable for integrating the total area of the gas
chromatogram, compensating for ‘column bleeding’ and for re-integrating after
drawing a new baseline.
Concentration of the extract to 1.0 ml may be omitted if a high hydrocarbon oil index
is expected or if a large quantity, e.g. 100 µl of the partially or non concentrated
extract is injected by means of the so called “large-volume injection systems”.
2.4.2 GC-FIDs Available for the ISO Method from Leading Suppliers
There are many GC-FID instrument suppliers available. Some of the key suppliers
have been contacted, including ATAS, Agilent, PerkinElmer, Shimadzu, Thermo and
Varian. Summary information obtained from each of these suppliers regarding
suitable GC-FID systems is given below:
ATAS
Agilent
Agilent [23] is the leading supplier of GCs in the GC market. However it has not itself
worked on the ISO 9377 GC-FID method, but instead provided suitable GCs for
others, such as ATAS, to develop complete packages for analysing oil-in-water using
the ISO method.
PerkinElmer
PerkinElmer was one of the earliest GC suppliers who were involved in supplying GC
instruments for the analysis of oil-in-water using the ISO 9377-2 method. Before the
ISO Mod method was recommended by OSPAR, PerkinElmer instruments were also
used in the Norwegian OLF study [6], comparing the GC method against the IR
based oil-in-water analysis method.
The system allows the injection of up to 100 microlitres of sample into a special large
volume liner. The bulk of solvent can be purged off through the split vent. Use of the
PreVent technique removes the entire concentration step, and simplifies the basic
procedure. According to the supplier, calibration of the system is done on a weekly to
monthly basis, with a run time for one sample of just 15 minutes.
development involved using two columns and two FIDs, where the light end (C7-C10)
is characterised and separately quantified from the C10 to C40 [24].
Shimadzu
A relatively small player in the European GC market, Shimadzu has also worked on
the ISO 9377 GC-FID method for the analysis of oil-in-water [25]. Its systems include
the following:
Thermo (Finnigan)
Thermo claim, that only their GC system enables a complete analytical cycle (sum of
sampling time, run time, and cooling time) to be completed for the ISO 9377-2
method in less than 5 minutes.
Varian
A lot of the GC instruments currently used offshore are supplied by Varian. Varian
has also carried out its own evaluation of the ISO 9377-2 method. It has been
suggested that a basic GC system for the ISO 9377-2 method from Varian would
comprise the following:
Varian Model 3800 GC fitted with Model 1079 PTV Injector and a Flame Ionisation
Detector and optional 8400 Auto-sampler [27].
2.4.3 GC Costs
Whilst a basic GC that can be used for the ISO 9377-2 GC-FID method may cost
about £11K, those like Anature TPH-1 from ATAS and Clarus / AutoSystem XL GC
from Perkin Elmer may cost as much as £27K.
Many laboratories in the UK have been contacted; these labs may be grouped into
three main categories
• Others
It is clear that apart from Opus, ITS and SGS who have experience with the ISO
9377-2 and its Mod method, no other laboratories who have close links to the oil and
gas industry have analysed oil-in-water using these methods.
The following labs have been contacted. Their oil-in-water analysis capability is
summarised:
• Scottish Water: Two of its 5 labs are equipped with oil-in-water analysers. Each
lab has an FTIR and GC-FID (PerkinElmer or Varian). Blue Book method is used
for IR analysis and again TNRCC Method 1005/1006 has been loosely followed.
ISO 9377-2 method has not been followed
2.5.3 Others
It is now at least four and half years since the ISO 9377-2 method was made
available for oil-in-water analysis, however from information gathered so far there is
clear indication that this method has not been widely adopted by the land based
industries such as the water and petrochemical. Even laboratories at the government
agencies are not familiar with the method.
Overall there are only a limited number of organisations who have experienced the
use of the ISO 9377-2 method for oil-in-water analysis. However many laboratories
do have GC-FID capability for the analysis of oil-in-water that quantifies
hydrocarbons with a different window from the ISO 9377-2 or the Mod method.
One of the possible considerations for implementing the ISO 9377-2 Mod method is
to use a secondary reference method, which is “calibrated” against the new reference
method, to calibrate other alternative methods for offshore use.
A commercial SFE-IR system for oil-in-water analysis has been recently made
available [29]. An example of such a system is shown in Figure 4. Both portable IR
instrument such as the Wilks InfraCal and FTIR can be connected to the SFE
system. SFE-IR effectively uses liquid CO2 as an extraction solvent for an IR based
method. A recent evaluation study [30] supported by BP, DTI and Statoil has
confirmed that SFE-IR is more than capable of analysing oil in simulated produced
water in the concentration range of interest to produced water discharges. Extraction
efficiency has been found to be generally in line with that from the DTI approved
method.
The method is environmentally friendly. Also its long term use is assured and it
allows the continued use of the IR methods without the use of harmful solvents. The
method is still relatively new and has not been tested offshore in the fields. In the
recent evaluation study, difficulties were also experienced in establishing calibration
and analysis for oil-in-water samples made using heavy crude oils. Cleaning of heavy
oil residue in the SFE-IR system was found to be time consuming. Some operators
have also expressed safety concerns in using a high pressure CO2 vessel offshore
and more importantly on CO2 asphyxiation in the event of a leak. In addition there are
no standardised procedures for such a method.
Both the current IR-TTCE and the new ASTM D 7066-04 methods have approved
standards. Some operators are already using the IR + TTCE method as a reference
method for checking alternatives such as UV fluorescence, e.g. Talisman and
ConocoPhilips. Therefore it would be natural to consider such a method for use as a
secondary reference method. The key issue will be how results correlate between the
secondary method and the new reference method.
In terms of measurement uncertainties, there are two main aspects – systematic and
random, which have also been demonstrated in the OSPAR Comparison study.
Calibration was considered as an important source for systematic uncertainty while
sampling was thought to be a significant source for random uncertainty.
In conclusion provided that data from IR based methods can be correlated to the new
reference method, or alternatively that regulators will continue to recognise the old IR
based method and accept results from using the old reference method, an IR based
method (either using TTCE or S-316 as an extraction solvent) could be considered
as a possible secondary reference method.
Since the recommendation was made by OSPAR on using the ISO Mod method as
the new reference method for the determination of dispersed oil in produced water,
there have been a number of interesting developments and observations which lend
support for a continued use of the current IR based reference methods. These are:
As detailed in Section 2.2, a new ASTM IR reference method (ASTM D 7066-04) has
been established. The method is similar to the current DTI approved IR method and
also the current OSPAR IR method [31]. The key difference is the use of a solvent,
S-316, in the ASTM method compared to TTCE in the OSPAR and DTI approved IR
methods.
The approval of the new ASTM IR method would allow the continued use of IR based
method for oil-in-water analysis. As the solvent S-316 is C-H free, it could also be
used to directly replace TTCE in the current DTI / OSPAR IR methods.
Survey results indicate that UK offshore operators are reasonably happy continuing
with the current IR based method. The issue of TTCE being a suspected carcinogen
has been appropriately addressed by implementation of proper HSE procedures by
the offshore operators both in the UK and in the Netherlands [32,33] .
Comparison studies have clearly shown that the ISO Mod method is different from
the IR based methods. As a result it would be much easier for both operators and
regulatory bodies to implement the 40 mg/l (30 mg/l by 2006) performance standard
and also to interpret historical data if IR methods continue.
Commercial SFE-IR systems are now available. One such system has been recently
tested and evaluated independently [30]. Results obtained have clearly shown that
SFE-IR is capable of measuring oil-in-water, in particular samples prepared using
gas/condensate and light crude oils. Further trials of the SFE-IR system are currently
being carried out by Statoil where field oil in produced water samples are collected
and analysed.
In addition, when the new reference method is implemented, there may be a need to
establish a link between results from the new reference method to that from the
Therefore it is felt that there is a good case for continued use of the IR based
reference methods. A continued acceptance of the IR based methods after year
2006, will cause no disruption and will have no problem in interpreting historic data.
Therefore it would be prudent for both the DTI and offshore operators to ask OSPAR
to re-examine the case of continued acceptance of the IR based oil-in-water analysis
reference methods after year 2006.
From the UK point of view, if OSPAR continues to accept the IR based reference
methods, there is little which requires to be changed in terms of implementation.
Nevertheless the following points should be considered to help reduce the exposure
of operators to TTCE and also to improve the current oil-in-water analysis practice:
To implement the ISO 9377-2 Mod method, four possible options have been
considered,
Direct implementation of the ISO Mod method by using GC-FIDs offshore would
seem the most straightforward from a regulatory point of view. However there are
potential difficulties which exclude it from further consideration. The reasons for
rejecting this option are presented in the following arguments:
- Whilst the original ISO method has been successfully implemented offshore by
ConocoPhilips, Norsk Hydro and also others in the Norwegian sector, there is
little support from the Norwegian’s offshore operators’ experience for
implementing the ISO Mod method directly offshore
- The ISO Mod method is less robust than the original ISO method
- All operators in the UK and the Netherlands will have to purchase new GC-FIDs
suitable for the analysis of oil-in-water using the ISO Mod method. In addition to
the capital costs, there will also be significant running and training costs
- Analysis of oil-in-water using GC-FIDs requires technicians with considerably
more skills compared with the currently approved IR methods offshore
- GC-FIDs will require the use of a pressurised carrier gas and also the use of
pressurised hydrogen; hence HSE issues need to be addressed
- Analysis of oil-in-water using GC-FID based methods is slow. For offshore
operators, oil-in-water analysis is also required for process control and
optimisation in addition to regulatory compliance monitoring
- Both the DTI and the State Supervision of Mines have indicated that they were
not so keen to take on this option
Advantages:
- Onshore laboratories would be usually much better equipped and will have skilled
GC specialists
- Avoids the offshore use of GC-FIDs for oil-in-water analysis and hence the
additional costs associated with the purchase and running of these instruments
- Easier to solve problems and maintain the GC-FID instruments onshore than
offshore
Disadvantages
- Need to send all samples to onshore laboratories, which can be expensive and
logistically difficult
- There will be a delay in obtaining the results, taking days or even weeks
- It is most likely that the operators will still need a monitor / analyser for routine
analysis offshore for the purpose of process control and monitoring
• Reduce the number of samples from currently two per day to one a day or even
less.
• All samples send to a “central laboratory” where a procedure can be developed in
which a large number of samples can be analysed automatically. This will reduce
the overall analysis costs and the errors which would result from the uses of
different laboratories.
• As offshore routine analysis is still expected to be carried out by offshore
operators, one should encourage the establishment of a correlation between the
method used for routine analysis and the ISO Mod method.
With this option a non GC method is calibrated against the ISO Mod method and
then used to calibrate other alternative methods offshore. Two IR based methods, i.e.
the current approved IR and also the newly available ASTM IR methods are
potentially suitable for such a purpose. Both methods have a relevant standard.
One of the key considerations in using a secondary reference method should be that
once it is calibrated against the ISO Mod method, it can be used to calibrate other
alternative methods regardless of the types of installations. There is no point in
having a secondary reference method that can only be used to calibrate an
alternative method for a specific installation. To achieve this it may be necessary to
develop a commonly accepted correlation between the secondary reference method
and the ISO Mod method. One will probably have to use calibration standards made
from a synthetic oil for both the ISO Mod method and the secondary reference
method.
Provided that the correlation between the secondary reference method and ISO Mod
method is accepted by the industry and regulators, this secondary reference method
could then be used to calibrate other alternative methods offshore.
In this approach, an alternative method is directly calibrated against the ISO Mod
method established onshore. Once the alternative method is calibrated, it can then
be used to determine oil in produced water offshore. This is considered as probably
the best option for the following reasons:
The various implementation options are summarised in the diagram shown in the
Figure 5.
Implementation
Options
Direct Indirect
ISO Mod method ISO Mod method Alternative method Alternative method
GC-FID offshore onshore; samples sent calibrated against calibrated against the
/ analysed onshore a secondary ref. ISO Mod
For routine oil-in-water analysis offshore, the following four methods are considered
as the leading alternative bench-top methods, which may be calibrated against the
new reference method and used for reporting.
• UV fluorescence
• InfraCal HATR
• IR using TTCE and / or S-316
• Horiba OCMA-310
For the InfraCal HATR, the principle of operation is briefly described in section 2.2.2.
A picture of the InfraCal HATR is shown in Figure 7(a).
For IR methods using TTCE and / or S-316 solvent, there are a number of portable
instruments, which have already been used in the UK sector of the North Sea. These
include the Wilks InfraCal (CVH model), Miran 1FF, Buck HC 404. Portable IR
instrument is also available from Duratech (Duratech Infrared Analyser). Pictures of
these instruments are shown in Figure 7(b) and Figure 8.
In the North Sea there are a number of installations that have been equipped with
online monitors using either light scattering and / or UV fluorescence. In the last few
years the use of UV fluorescence systems appears to have gained popularity.
UV fluorescence based instruments are generally more sensitive to the types of oils,
the presence of production chemicals, and the droplet size, but both dissolved and
dispersed oils can be measured. Light scattering based instruments however are
more sensitive to the droplet size, the presence of gas bubbles and solids. However
they can not measure dissolved hydrocarbons. Hence selection of an online monitor
will largely depend on the characteristics of the produced water stream.
Having considered the various implementation options, it looks that the best option
for implementing the ISO Mod method is to use alternative methods which are
calibrated against the ISO Mod method. This is based on the following
considerations:
• Direct implementation of the ISO Mod method offshore has practical difficulties
and also implication on costs
• Direct implementation of the ISO Mod method by sending all samples onshore is
logistically difficult, but more importantly, due to the delay in obtaining the oil-in-
water results, operator will most probably need to have an oil-in-water analysis
method offshore for routine monitoring anyway
• Using a secondary reference method that is calibrated against the ISO Mod
method, and then use the secondary reference method to calibrate other
alternative methods will involve two stage calibrations. Whilst it will make offshore
calibration possible, it will also introduce additional uncertainties to the oil-in-
water analysis results
Photoacoustic By focusing a pulsed laser light into small volume of sample, oil absorbs the energy and Prototype Planned Still to be proven, in particular for low
sensor creates high frequency pressure wave. This is detected and correlated to the amount of conc. OIW analysis. Develop commercial
oil. type on-going
Spectral SFS is based on measurement of both excitation and fluorescent spectra instead of 1 Commercial No Complex software, more capability than
Fluorescent using a fixed wavelength in the conventional UV fluorescence method. In addition to just measuring oil concentration
Signatures concentration measurement, it can be used to distinguish substances and sources of
(SFS) oils.
Laser Induced In many ways this is similar to the photoacoustic sensor. The differences are that in LIF, Prototype On Trial Still to be proven. Development on going.
Fluorescence UV is focused into a small volume of water sample, also instead of detecting the high
(LIF) frequency wave, it detects the UV fluorescence intensity to quantify oil concentration.
UV Fluorescence intensity is measured and correlated to the concentration of the total or At least 4 Yes May be affected by gas and chemicals.
fluorescence aliphatic hydrocarbons provided that the ratio of aromatic to the total hydrocarbons commercial Already in use. Trend of increasing use
remains relatively constant.
3+
UV absorbance is correlated to the concentration of the total or aliphatic provided that At least 2 No Complex and may be affected by Fe and
UV absorption the ratio of aromatic to the total hydrocarbons remain relatively constant. commercial chemicals
Notes: Costs for offshore on-line systems can vary from £25K to £100K. In comparison bench-top ones cost about £3K to £5K
5.1 Background
How samples are taken and how they are then subsequently handled, is generally
recognised as one of the most significant factors in contributing to errors associated
with the analysis of offshore oil in produced water. Studies have been carried out on
this issue in the past and several papers and documents published. The issue is
also referenced in publications and procedures relating to onshore and offshore
sampling, so it is clearly recognised as an important factor in establishing reliable
results.
sampling locations
sampling techniques
sample handling
sample shelf life
labelling and documentation
A wide range of literature has been examined in helping to build the guidelines for
good sampling practice, which are presented later in this section. The references
examined include analytical methods, technical papers, technical guidelines and
technical standards [1,5,31,35-40]. Examination of these references revealed that
most of the applicable guidelines are already present in the literature, but are not
specifically written to comprehensively cover all the issues relevant to offshore
sampling.
Each of these steps has a set of errors associated with it, whilst this section will
primarily focus on the collection aspect, other sample handling aspects such as
sparging, the use of florisil etc will also be covered.
The potential errors in the steps are not all of the same magnitude. The collection of
the produced water sample has the potential to introduce the largest error in the
entire process, although quantifying the potential error is difficult.
The quantification step is likely to be much less variable than the other two steps, as
it is less prone to human error. If so, then differences in results when comparing the
analysis of replicate sets of samples, one resulting from the defining method for oil
and the other resulting from a potential new measurement technology may result
from sampling and extraction errors not from differences in the methods.
In the UK sector, the discharge of oil to sea is currently regulated through the
'Prevention of Oil Pollution Act 1971', which is shortly to be superseded by the
'Offshore Petroleum Activities (Oil Pollution Prevention and Control) Regulations
2005'. POPA defines that water is to be sampled and analysed for oil content twice
daily, to DTI approved standard. (In the Netherlands, sampling and analysis of
produced water is required every 48 hours on manned installation in line with
OSPAR). Further, the legislation states that the sample point to be used must be
approved by the DTI. In other countries, similar guidance is provided in the relevant
legislation. Whilst the methods in themselves are often described in detail, less
emphasis is often placed on the sampling process. Yet this is one of the main factors
influencing the quality of any samples taken for analysis.
As an example, Section 2 of the DTI method [1] provides some brief guidance on the
method of sampling and storage, as follows:
2.1 "Samples should be taken at least twice each day (24 hour period) such that
they are taken at equal intervals of time".
The siting of the sample point must be approved by the Secretary of State
and may not be moved subsequently without permission".
2.3 "Care should be taken to ensure that the sample is of adequate size and
representative of the effluent. The sampling line/point should run continually
or at least be flushed for one minute before a 500 ml or 1 litre sample is
taken".
2.4 "The bottle should be scrupulously clean and should not be rinsed with
effluent prior to sampling nor over-filled during sampling".
4.1 "If the sample was not acidified at the time of collection, add 5 ml hydrochloric
acid to the sample bottle".
Interestingly, although the DTI method provides some guidance on sample point
location, the sample point design is not considered. Also, whilst the solvent
extraction method is covered in detail, the potential impact of acidification and
storage time is not mentioned.
The operator survey results in Section 2 confirmed a number of issues relating to the
analysis of oil-in-water offshore. Specifically these were that the vast majority of
installations employ an IR spectrophotometric method which uses TTCE solvent.
Further, the vast majority of samples are being taken from the overboard discharge
line via the approved DTI sample point.
Some respondents confirmed that centre line quills were used for sampling, whilst
others were unsure. Sample points were confirmed as being run for the necessary
1-2 minutes prior to sampling, with some operators running the sample points
continuously. The vast majority of samples were confirmed as being acidified with
hydrochloric acid, and analysis was carried out as soon as the samples were cooled
down.
The operator survey confirmed that operators are already following several of the
guidelines specified in the DTI method. This in itself is quite encouraging and the
establishment of more formal guidelines should help improve this further.
It is very difficult to actually quantify errors associated with some aspects of sampling
as the errors will only reveal themselves when the differences between good and bad
practice are examined. Since the actual analysis method produces the final result, if
bad practice in collection of samples continues to be used without any understanding
of the impact on the final result, then no obvious errors will be apparent. It is
therefore essential that good practice is adopted in the collection and handling of
samples, in order to minimise or eliminate potential errors.
The error introduced, for example, by analysing a sample from a line which had not
been flushed, compared to one which had, is very difficult to quantify or even
replicate. However, it is clear that failure to flush a sample line prior to sample
collection can adversely affect the result, so it is thus essential that good practice
guidelines are followed. Where samples may have low concentrations of oil-in-water,
such as sub 40 mg/l, then it is entirely possible that errors of 100% or more may
result from failure to follow best practice.
The current DTI approved method [1] and the PARCOM method [31] provide clear
guidance on the location for where samples should be taken for oil-in-water analysis.
These state that the sample point should be downstream of the final stage of oil-in-
water separation, and it should be from, or just downstream of, a turbulent region.
The POPA legislation also defines that the sample point has to be approved by the
DTI, and that the sample point location cannot be moved without prior authorisation.
Since the objective of this guidance is to obtain a representative sample for analysis
at the point of discharge, there would appear to be no reason to change these
requirements for the transition to the ISO 9377-2 Mod method, or indeed any other
method which has been approved for use.
The location of any sample point clearly has an impact on how representative any
sample may be of the main flow. Sampling from dead legs, for example, cannot give
a true indication of the main flow conditions. Similarly sampling from the top, bottom
or side of a pipe can yield different results, particularly where oil is present in the
water phase. However, a number of other factors can also affect how representative
any sample may be, including:
Assuming the criteria for location of the sample point are met, the design of the
sample point is equally important. The design may vary depending on whether spot
samples are required or an on-line oil-in-water monitor is being installed. The main
objective in designing any sample point is to ensure that it can provide a
representative sample, assuming all other conditions are met. For most pipework (1”
and upwards), this can be achieved by the use of centre line sampling probes, such
as that provided by Jiskoot [41] (see Figure 10 below). Pipewall effects mean that it
is essential to sample at least from the centre 1/3rd of the pipe flow. Ideally,
installation in vertical pipes with up-flow is preferable for oily water samples to
minimise any potential for settlement. However, providing turbulence is maintained,
installation in horizontal pipework should cause no problems.
In terms of sample point design and the minimisation of sampling errors for spot
samples, the use of a completely integrated and automated sampling system should
be considered. Such systems are available and in widespread use for fiscal sampling
and monitoring e.g. crude oil sampling.
The systems can sample using either a fast loop (see Figure 11) or inline sampling
system. For oil-in-water sampling, the use of such a system would have benefits in
minimising the sorts of errors associated with sampling from manual sample points.
However, the use of such a system on balance probably introduces an unnecessary
level of complexity to the collection of oily water samples. This would also increase
the need for maintenance.
Providing some simple rules are followed with respect to the design of the sample
point, then representative sampling should always be attainable. These are detailed
below.
Where practicable, a centre line pitot should always be used. These are
available commercially and can be of such a design that they can be withdrawn
for inspection. Pitots can also be made very easily using stainless steel tubing.
The pitot should be of at least ½” bore and the edges of the pitot should be
smoothed off prior to use. On older installations, where installation of centre line
pitots may prove difficult, side wall sampling may be used together with a suitably
designed sample point. However care needs to be taken with regard to ensuring
adequate rigidity.
For external valving at the sample point, there will normally be a requirement for
double block and bleed valving. Ideally, full-bore sample valves should be used,
such as ball valves. Specifications for valving on offshore installations normally
means that other valve types may be required, such as needle or gate valves.
The distance between the sample probe and the sample valving should always
be minimised, as the longer the sample line, the greater the chance of errors
influencing any subsequent analysis. For example, where a sample point has not
been flushed properly, there is a possibility that oil residue retained within the
sample line will introduce an artificially high result.
Sample tubing requires no special guidelines, other than the material used should
be compatible with the fluids being sampled. Generally, stainless steel is the
preferred material of choice offshore, although special alloys may be required
subject to fluid characteristics.
GATE VALVE
BALL VALVE
1
2 " STAINLESS STEEL
PITOT
(VARIABLE
INSERTION)
FLOW
Some general guidelines can be applied to the installation of on-line systems such
that representative sampling is achieved.
Where on-line systems require a sidestream feed to the monitor, the feed should
be provided by way of a dedicated sample connection. A centre line pitot should
be used to ensure that the sample is representative. Depending on the available
line pressure and the monitor flow requirements, the sample may need to be
pumped to the monitor.
Where on-line systems require the installation of a probe in the main pipe, the
location should be such that there is confidence that the probe is exposed to
representative flow conditions.
Variation in flow conditions can also affect how representative a sample may be of
the main flow. As stated earlier, the sample point should ideally be located within a
turbulent zone. If the flow is too low, then it is possible that separation of the oil in
the fluid will occur, leading to stratification and thus unrepresentative sampling.
Knowledge of the flow regime within the main pipe is therefore important for ensuring
representative sampling.
The condition of the sample point is important, as contamination of the sample point
is likely to lead to contamination of any subsequent samples. At concentrations in
the 0-40 mg/l range, such contamination could have a marked impact on the end
result. Sample points not in continuous use can naturally cause hydrocarbons or
solids or both to collect within the probe. It is therefore essential that adequate
flushing time be provided prior to a sample being taken. As an absolute minimum,
the sample point should be flushed for 1 minute before a sample is taken. For ideal
representative sampling, the sample point should be allowed to flow continuously to a
suitable drain on the installation. The sample point itself should be maintained in
good working order and kept as clean as possible. On offshore installations this can
prove difficult. Two examples of offshore sample points are shown in Figure 13.
Sample bottles
Given some of the issues discussed above with respect to sampling, it is unsurprising
that sample bottle condition and cleanliness are of equal importance if representative
results are to be obtained by either the IR or GC methods, or indeed any other
approved method.
For oil-in-water sampling, Pyrex (Duran) glass narrow necked bottles of 500 ml
should be used. Plastic bottles should never be used, as the hydrocarbons in the
sample can readily diffuse through the bottle. Plasticisers can also leach from the
plastic, resulting in misleading results. A good example of the type of 500 ml bottle
that should be used is shown Figure 14 below.
The DTI method for IR analysis states that sample bottles "should be scrupulously
clean". It is common offshore to clean and re-use sample bottles and it is likely that
errors are introduced into the analysis as a result of this. New bottles are therefore
preferred for each sampling event, but if bottles are to be reused, then they should be
washed thoroughly, rinsed with solvent prior to use and air-dried.
The bottle cap is also important, and teflon lined caps should be used where
possible. Previous studies have shown that incorrect selection of both the bottle
material and cap liner can cause contamination if the initial extraction is done from
the bottle.
How samples are handled during and after collection can also prove very important in
minimising the errors associated with the sample collection and analysis process. As
detailed in 5.5, suitably clean bottles should be made available for collection of the
sample.
Once collection of the sample has started, the bottle should never be allowed to
overflow. If it is overfilled, then a fresh bottle should be used with a new sample
taken. Once the sample collection is complete, the bottle should be sealed with the
cap. One point worth noting is that a single sample should never be subdivided. The
nature of oil droplets is such that the droplets can stick on the internal surface of the
bottle. Therefore attempting to split a single sample from a bottle, could never
guarantee an even concentration split.
If the sample is hot when it is collected, then the sample should be cooled as soon
after collection as possible, prior to any analysis being carried out. The bottle top
should be loosely applied during the cooling process and then tightened securely.
The GC ISO9377-2 method mentions cooling of the sample as part of the procedure.
Whilst the DTI method does not specify, it is common practice offshore to cool the
sample prior to analysis, to prevent any vaporization losses. It is also recommended
that the samples are acidified to a pH less than 2 by addition of 5 ml of HCl 1:1. This
is specified in both the DTI method and the GC ISO9377-2 method, and is
particularly important if there is likely to be a delay in the time between sampling and
anlaysis. The acid should be added after sample collection. Whilst it makes little
difference from a sample preservation point of view if the acid is added before or
after the collection of sample, it is not considered best practice to add water into an
acid. If the sample is to be analysed immediately, then no further special handling
requirements need to be adhered to.
Finally, it is worth pointing out that all subsequent stages in the actual analysis of the
fluids, can have an impact on the end result. These issues are not discussed in
detail in this section of the report. For samples which are to be sent elsewhere for
analysis, these should be stored and transported in the dark or at least out of
sunlight. Certain hydrocarbons can degrade in the presence of sunlight, so storage
in the dark can prevent such degradation.
Sample shelf life is an issue, which has been considered in previous studies to
compare the effect of storage time on end results. This is of particular importance
where duplicate samples may be sent to an onshore laboratory for verification
analysis or for methods comparison studies and there is transit time involved.
It also states that where samples are analysed after the maximum preservation time,
then the results should be accompanied by a statement to the effect that the results
may not be reflective of the concentration at the time of sampling, and that the
maximum preservation time has been exceeded. The standard further states that
"prolonged storage of samples may be appropriate, if the laboratory can demonstrate
that there is no difference between the test result obtained after extended
preservation time and the test result obtained within the preservation time given in
this part of ISO5667".
A study carried out by the Orkney Water Test Centre Ltd (OWTC) in 1994 [42],
investigated the effect of storage time on laboratory prepared oily water samples
made up from different oil types, from condensate through to heavy oil. Oily water
samples were stored in a window exposed to sunlight for a three-month period, and
the replicate samples analysed at 1, 2 and 3 month intervals. The results, which are
shown in Figure 15, showed little or no difference in the end result for medium and
heavy crude samples, even at the three-month interval.
The condensate sample by contrast showed a 10% reduction in the quantity of oil
extracted at the three-month interval. This suggests that lighter crudes may be more
susceptible to degradation in the presence of light than medium or heavy crudes
might be. Whilst there was no degradation apparent after 1 month in this study, it is
recommended that a maximum storage time limit is applied to avoid unnecessary
delay in samples being shipped ashore for analysis.
If samples are to be stored, then ideally they should be stored in a refrigerator at 4°C.
Where this is impractical, they should at least be stored in a cool place out of direct
light. Samples that are being shipped for later analysis should be transported in
suitable containers designed specifically for holding sample bottles. A wooden crate
is ideal and cushioned material should be included to prevent breakage. The boxes
should be marked to maintain them in an upright position during shipping.
100
90
Extraction Efficiency (%)
80
70
60 Flotta Crude
50 Condensate
40 Heavy Oil
30
20
10
0
0 Months 1 Month 2 Months 3 Months
Storage Tim e
In all of the reference materials examined, there were no formal guidelines presented
for the labelling and documentation requirements for field samples. Some guidance
was provided however in References [36] and [40] and this should be adopted as a
formal guideline for any samples which are being collected for immediate analysis or
for shipment ashore.
During sampling the bottles should be labelled with the following information as a
minimum:
Sample reference
Name of installation
For samples that are to be shipped to an onshore laboratory for analysis, it may also
be beneficial to include details of the personnel taking the samples, as well as
including a sample description form, noting any observations made during sampling.
The following guidelines have been developed on the basis of the information
discussed in the preceding subsections. Overall, they aim to highlight the important
aspects of sampling which are discussed in the reference materials which were
reviewed.
General
For installations that discharge dispersed oil in produced water at over two tonnes
per annum, at least two samples per day are required to be taken and analysed to
demonstrate regulatory compliance. These samples will be taken at approximately
equal intervals of time, using the sample points in line with sampling Guideline 1
below.
For installations (manned and unmanned) that discharge produced water containing
not more than 2 tonnes of dispersed oil per annum, according to the OSPAR OIC
Recommendation 2001/1, the frequency and timing of sampling should be such that
samples are representative of the effluent taking into account operational aspects
and logistics. The following suggestions may be considered:
• Unmanned installations: at least one sample is taken every time when the
installation is visited. Should the installation be temporarily manned for more
than 48 hours, it is suggested that one sample is taken in every 48 hours.
• Installations discharging less than 2 tonnes dispersed oil per annum:
produced water samples should be collected and analysed at least once per
month, at approximately equal time intervals.
For unmanned distillations that discharge more than two tonnes per annum, it is
recognised that it would be unrealistic to take two samples per day, an agreement
must be obtained from the DTI regarding an appropriate sampling frequency.
The siting of the sample point should be agreed by the regulator and should not
be moved subsequently without permission.
Spot Samples
Where practicable, a centre line pitot should always be used. These are available
commercially and can be of such a design that they can be withdrawn for
inspection. Pitots can also be made very easily using stainless steel tubing. The
pitot should be of at least ½” bore and the edges of the pitot should be smoothed
off prior to use. On older installations where installation of centre line pitots may
prove difficult, sidewall sampling may be used together with a suitably designed
sample point.
For external valving at the sample point, there will normally be a requirement for
double block and bleed valving, although this may vary according to operator.
Ideally, full-bore sample valves should be used, such as ball valves.
Specifications for valving on offshore installations normally mean that other valve
types may be required, such as needle or gate valves.
The distance between the sample probe and the sample valving should always
be minimised, as the longer the sample line, the greater the chance of errors
influencing any subsequent analysis. For example, where a sample point has not
been flushed properly, there is a possibility that oil residue retained within the
sample line will introduce an artificially high result.
Sample tubing requires no special guidelines, other than the material used should
be compatible with the fluids being sampled. Generally, stainless steel is the
preferred material of choice offshore, although special alloys may be required
subject to fluid characteristics.
Where online systems require a sidestream feed to the monitor, the feed should
be provided by means of a dedicated sample connection. A centre line pitot
should be used to ensure that the sample is representative. Depending on the
available head and the monitor flow requirements, the sample may need to be
pumped to the monitor.
Where online systems require the installation of a probe in to the main pipe, the
location should be such that there is confidence that the probe is exposed to
representative flow conditions.
The distance between the sample point and any external online monitor, should
be minimised as far as practically possible.
Prior to any samples being collected, the sample point should be left flowing for at
least 1 minute.
Sample points not in continuous use can naturally cause hydrocarbons and/or
solids to collect within the probe. It is therefore essential that adequate flushing
time be provided prior to a sample being taken. As an absolute minimum, the
sample point should be flushed for 1 minute before a sample is taken.
For ideal representative sampling, the sample point should be allowed to flow
continuously to a suitable drain on the installation. However one must ensure that
this flow is accounted for as part of the discharge.
Where the main pipe flow is low and there are concerns about possible
stratification, installation of an orifice plate to restrict flow and maintain turbulence
should be considered.
The following guidelines are provided with respect to sample bottle requirements.
Pyrex (Duran) glass bottles of 500 ml should be used for collection of the
samples.
The bottle caps should be teflon lined. This is very important in preventing any
potential contamination of the sample from plastic bottle caps.
All sample bottles should be solvent washed prior to use and left to dry.
Where bottles are reused for produced water analysis, they should be solvent
rinsed and then washed thoroughly and dried after each sampling event.
Plastic bottles should not be used under any circumstances, as plasticizers can
leach in to the sample and affect the analysis results.
The following steps are provided as guidelines for collection and subsequent storage
of oily water samples:
Fill the sample bottle, taking care not to overfill it. Where the sample temperature
is hot, gloves must be worn whilst collecting the samples. If it is overfilled, then a
fresh sample bottle should be used.
Add 5 ml HCl 1:1 to acidify the sample if the sample is not going to be analysed
within 24 hours. Otherwise the acid can be added just prior to the analysis being
carried out.
Cool the sample immediately to 20 oC prior to analysis being carried out. The cap
should be loosened during the cooling process and then tightened securely.
Gas sparging of samples: sample sparging using nitrogen (or any other gas)
should not be used under any circumstances, unless it is an accepted part of the
analytical method and has been discussed and agreed with the DTI.
Use of florisil: the use of florisil to treat the sample is recommended where there
are concerns that polar components may interfere with the analysis/detection
method. This step will ensure that polar components are removed from the
sample extract before quantification. This is in line with the ISO Mod method in
which the use of florisil is specified. It is recommended that florosil is stored in a
dessicator in order to prevent degradation of the florosil. If the use of florisil is not
included, there is no need to report this fact to the DTI
Once the samples have been cooled, they should ideally be stored at 4°C in a
refrigerator. Again, prior to any analysis being carried out, they should be warmed
to 20 oC first.
For sample shipment, sample bottles should be stored and transported in suitable
sealed containers to prevent ingress of light.
Ideally, the maximum storage time between sampling and analysis should be
limited to 1 month, but in any case should be minimised as much as possible,
according to the shipping time of the sample to shore.
Where possible, for samples taken and analysed offshore, analysis should be
completed within 12 hrs of the sample being taken.
For samples which are taken offshore for analysis onshore, the analysis should
be carried out within 7 days from their arrival onshore
Where analysis within 1 month is impractical due to logistical reasons, then the
results should be accompanied by a statement to the effect that the results may
not have been reflective of the concentration at the time of sampling, and that the
maximum preservation time has been exceeded
During sampling the bottles should be labelled with the following information:
Sample number
Name of installation
When field replicate samples are required to be taken, e.g. for validation or
calibration purpose, a “Y” shape piece sampling device could be used.
An alternative method is to mark the replicate sample bottles at 1/3, 2/3 and full
sample volume, then to fill the sample 1/3 at a time in turn until two replicate
samples are obtained. All other sampling aspects as detailed in the early sections
should still apply.
6 ACCEPTANCE CRITERIA
For an alternative method to be accepted, from the regulator’s point of view the
single most important thing that will be examined is whether the alternative method
produces results equivalent to the new ISO Mod reference method. Therefore
acceptance criteria will have to be established based on techniques that allow one to
make a judgement on this. There are two possible methods which one may use:
However, statistical significance tests [43, 44] (see more details in Appendix 4) are in
general limited to a single level (concentration) of results or a narrow (concentration)
range of results (i.e. in general within the same magnitude). For offshore oil-in-water
analysis where oil-in-water concentration from 0 to over 100 mg/l is of interest,
establishing a correlation using linear regression is likely to be the most appropriate
in terms of acceptance.
In addition to having a valid correlation, the following issues may also be taken into
consideration in the acceptance:
• Instrument availability
• Procedures in dealing with an instrument failure
• Data for reporting
Standards [45, 46] are available for establishing linear calibration curves. “A Guide to
Best Practice on Preparation of Calibration” [47], which was prepared as part of the
UK National Measurement System Valid Analytical Measurement (VAM) Programme,
is also available and is given in the Appendix 5.
For online monitor calibration, a code of practice from EEMUA (the Engineering
Equipment and Materials Users Association) [48] is available.
The current performance standard is 40 mg/l, which will be reduced to 30 mg/l by the
beginning of 2006. The maximum concentration that will still be allowed is 100 mg/l.
Also the current DTI method covers calibration concentrations of 3 mg/l to 75 mg/l. It
is therefore recommended that to establish a correlation between a lab bench top
alternative method and the ISO 9377-2 Mod method, a nominal concentration range
of 2 mg/l to 80 mg/l should be used. Within this range it is expected that for both the
alternative methods and the new ISO Mod method, linearity will hold between
For online monitoring, a concentration range covering at least from 0 to 100 mg/l
should be used.
If linearity is not in doubt, but there is a known bias (i.e. a non-zero intercept), a two
point calibration may be used. In this case, two calibration standards are prepared
with concentrations that encompass the likely range of concentrations for test
samples.
For the purpose of calibrating alternative methods against the ISO Mod method, six
calibration points are suggested for calibrating lab based bench top analysers, and
three points for online monitors.
Generally the more replicates that are used for each calibration point, the more
accurate the respective method, and hence the better the calibration curve.
According to both the BS ISO 11095:1996 and the LGC/VAM best practice
document, at least two analyses should be carried out for each calibration point.
However for the purpose of calibrating an alternative method against the ISO Mod
method, for each of the calibration point, a single analysis is suggested for calibrating
lab based bench top analysers, but three analyses are suggested for each calibration
point for online monitor calibrations.
The current DTI requirement is that the instrument / method should be re-calibrated
at least twice a year.
Unless it has been proved that the intercept is consistently near zero, the calibration
line should not be forced through the origin.
It is suggested that once an alternative method has been calibrated and used in the
field, validations are carried out from time to time to confirm that the initial calibration
is still valid. It is suggested that this is done initially once a month.
According to the DTI, two samples must be analysed every day. Therefore for the
purpose of reporting a properly calibrated instrument should ideally be available
every day.
Instrument failures can be very obvious, e.g. loss of the light source in IR or UV
instrument, or a failure of the fuse etc. However a failing instrument is not so easy to
be identified. A useful method of detecting a failing instrument is to use a Statistic
Process Control (SPC) chart, which may be particularly useful for online oil-in-water
monitors used offshore as they tend to develop faults / problems more frequently
than lab based analysers because of the exposure to the harsh offshore environment
and also direct contact with produced water.
If it is found that the instrument is failing to produce correct results, action must be
taken. This might include contacting manufacturer/supplier to identify the source of
the problems, and order and replace parts as required. The instrument may have to
be sent onshore. In any case, once the instrument is repaired, it should be
recalibrated against the reference method.
Daily samples should be collected as usual. These samples will have to be cooled
down to room temperature and then stored in a refrigerator unit, and analysed once a
recalibrated instrument (or a back-up analyser) is available.
For laboratory based bench top methods, data will be reported in the same format as
before. These data will be ISO Mod method equivalent and expressed in units of
mg/l.
It is suggested that two results per day from an online monitor are reported to the
DTI. Each of the results is taken as the average value from the monitor over a period
of 12 hours. For each of the 12 hour periods, an allowance will be given for
conducting online monitor daily maintenance. This allowance may be agreed with the
DTI prior to the online monitor being used for reporting.
Ideally these two results are then combined with the produced water volumes which
have been discharged over the respective 12 hour periods to give flow weighted oil in
produced water discharge figures.
b) Onshore laboratories: onshore laboratories that have been linked to the offshore
industry on oil-in-water analysis using both IR and GC methods are listed in the
Section 2.5.1. However it should be noted that not all the laboratories listed are
UKAS accredited.
A series of 12 duplicate samples, i.e.12 x 2, are initially analysed by both the ISO
Mod method and the alternative method. These samples are collected by
sampling at regular intervals during a period of days/weeks. Assuming that the
content of the oil in the samples lies within the linear range of both methods, a
linear regression is carried out with a linear calibration curve obtained. The
calibration curve may be expressed using equation 1:
Y = aX + b [1]
Using basic statistics the 95% confidence limits are also calculated.
After the initial calibration, a limited number, e.g. 4 duplicate samples per month
are taken and analysed using both methods. If the results are within the 95%
confidence, the points are added to the calibration set and the 4 oldest calibration
points are discarded. A new calibration line is re-calculated together with a new
set of confidence limits.
If the 4 new calibration points are outside the 95% confidence limits, 8 duplicate
samples are analysed during the next month and a totally new calibration graph is
calculated using 4+8 points. Further details are given in the Appendix 6.
One of the benefits of using this dynamic calibration approach is that the
instrument used in the alternative method may not need to be sent onshore. For
FPSO operations, this may be particularly advantageous. Also correlations are
validated on at least a monthly basis. With the dynamic calibration method, the
effect of sample matrix on both the alternative method and reference method is
taken into account. One of the key problems will be the concentration range. The
samples taken may have a relatively narrow range if the process has been
relatively steady.
Results from both the alternative method and the reference method are then
plotted. Linear regression is then carried out to establish a correlation with the
95% confidence limits calculated using basic statistics.
Validation of the calibration may be carried out initially once a month (say at the
middle point of each calendar month). Validation can be done by analysing a
standard solution sample, which has been prepared in the onshore laboratory
where the calibration had been conducted, using both the alternative method and
the DTI approved method. The corresponding result is then checked against the
original correlation plot. If the result is within the 95% confidence limits, the
correlation is considered to be still valid. If the result is outside the 95%
confidence limit, a second sample should be analysed by both the reference
method and the alternative method, if the second result is also outside the 95%
confidence limits, then a new correlation should be established.
When an alternative method uses a different solvent from the pentane, and when
it is also felt that making exact concentration of two oily water samples (for the
purpose of preparing two respective standard solutions) is difficult, then it is
probably better to use dynamic calibration procedures as detailed above.
e) More details on linear regression and how to calculate the 95% confidence limits
can be found in the Appendix 5 and also in a book entitled “Practical Statistics for
the Analytical Scientist – a Bench Guide” [44]. Useful information on linear
correlation and regression can also be found on the web [50].
f) Responsibility: For laboratory bench top instruments, it is possibly better that the
onshore laboratory takes the full responsibility in establishing and validating the
correlation between the alternative method and the ISO Mod method.
Calibration approaches for laboratory based bench top analysers are summarised in
the diagram shown in the Figure 16. Traditional calibration method 1 applies to cases
where pentane can be used as an extraction solvent in the alternative method, e.g.
UV fluorescence based. Traditional calibration method 2 applies to cases where a
solvent different from Pentane has to be used in the alternative method.
N
Use solvent Traditional calibration
pentane? method 2
Y
Y
For method 2
A correlation established for use A correlation established for
- Prepare respective stock solutions
- Prepare 6 standard solutions for
respective method from respective stock
N solution - Take two samples per week
Monthly validation due? - Analysis - Analysis by two methods
- Establish a correlation with 95% - 4 data points obtained
Y confidence limits
c) Initial calibration / correlation: To calibrate an online monitor against the ISO Mod
method, a three-calibration points approach is suggested [48]. These three
calibration points shall ideally correspond to high, mid-scale and low readings
within the monitor measurement range, which should at least cover 0 to 100 mg/l.
As guidance, “low” means a reading in the 10-25% of the measurement range,
“mid-scale” in the 40-60% and high in the 75-90% respectively. For each
calibration point, at least 3 parallel analyses / readings should be respectively
taken by the online monitor and also by the ISO Mod method. Correlation should
be obtained by plotting the average reading from the monitor at a time when
samples are withdrawn against the average reference value and then the points
fitted using linear regression. One must realise that establishing this initial
calibration/correlation may take weeks instead of days. Such a calibration may
also include drifting which can occur during the time period in which this initial
calibration is established.
e) Validation: Validation should be done to confirm that the initial calibration is still
valid. Procedures involved in validation and routine calibration checks are similar.
Calibration shall consider to be valid if the difference between its readings (after
converting to the reference equivalent using the calibration curve / correlation)
and the ISO Mod reference method does not exceed the reproducibility of the
ISO Mod method.
If the difference is within the Upper and Lower Warning Limits, no action need be
taken unless four consecutive results are on the same side of the datum,
indicating bias. Any apparent bias should be investigated quickly by two further
tests. If substantiated by the two results in the same band as the original four,
consideration should be given to remove the bias. If the difference falls between
the Warning and Action Limits, then another calibration check need be carried
out, no action should be taken if the new result falls back to within the Lower and
Upper Warning limits, but immediate corrective action / investigation should be
carried out if the second result still falls into the Warning and Action Limits. On
the other hand if a result (monitor error) falls outside the Upper and Lower Action
Limits, a further check should be carried out, if the latest result is also outside the
Action Limits, the on-line monitor should be considered as unfit.
In practice, one may use a control chart. The standard deviation s and arithmetic
mean can be established by initially carrying out say 15 measurements using the
calibration material. A control chart is then constructed based on these. For each
of the routine calibration checks, initially once a week is recommended, this solid
calibration material is inserted into the online monitor, the difference between the
latest reading and the arithmetic mean obtained previously is calculated. The
same criteria as given in (f) shall apply in terms of actions that one should take,
which will depend upon the difference falling into which part of the control chart.
However the procedures outlined here can only be used to check the
performance of the online monitor in terms of drifting, faults etc. Validation must
be still carried out to ensure the initial calibration / correlation is valid.
k) Responsibility: For online monitors, it is thought that it is better that the operator
using the monitor takes the full responsibility in establishing and validating the
correlation between the alternative method and the ISO Mod method, and also
the running of the control chart.
l) Safety issue: An important aspect of fitting an online monitor offshore is the safety
rating consideration. Operators need to tell online monitor suppliers what rate is
required. In general online oil-in-water monitors should be Ex d & p certified,
which means that they are made with flameproof and / or pressurised protection
and suitable for zone 1 & 2 operations.
Calibration approach for online monitors are summarised in the diagram shown in
Figure 18. For online monitors, there is no easy solution for calibrating them against
the ISO Mod method such that they can be accepted for regulatory compliance
monitoring. The difficulties arise from the fact that oily water is unstable, and as a
result, calibration against a reference method can only be carried out after the
installation of the online monitor. At the same time, once an online monitor is
installed, to cover a sufficiently wide range of concentrations will entirely depend
upon the process plant and may take a long time.
Samples parallel to the ones analysed by the online monitor are taken regularly and
analysed by the ISO Mod method. After a period of time, statistical significance tests
are carried out to see if the data obtained from the online monitor are equivalent to
the ones from the ISO Mod Method. If they are statistically the same, then the online
monitor should be allowed for the purpose of reporting. If not, then one should see if
it is possible to have a correction factor established using statistical significance
tests. If such a correction factor can be established, then the online monitor should
still be allowed for regulatory compliance reporting, if not, then the online monitor
should not be permitted for the purpose of reporting.
- 3 points calibration N
- 3 replicate samples for mod method
- Average reading taken from online monitor
- Establish a correlation using linear regression Y Bias
removed?
A correlation established for use (it may be built - Construct a control chart from
into the analyser so that readings give mod previous calibration check data
method equivalent results)
N
Action to
remove bias
If the two new results
Project No. OIW001 Page 57 of 166 still the same side of Y
Report No. 2005/96 datum?
TUV NEL Ltd
As the new ISO Mod reference method is recognised to be inherently different from the
current IR method, it is understood that OSPAR is now considering to have two separate
datasets [51] - one for IR results (i.e. those produced before the year end of 2006) and one
for the new reference method results (those after the implementation of the ISO 9377-2 Mod
method from 1 January 2007). If this happens, there will probably be no need now to
establish a "link" between the new reference method and IR methods, i.e. one does not have
to compare results from ‘the new’ reference to the current / old IR methods.
As the ISO 9377-2 Mod method in general produces lower “dispersed” oil results than the
current IR method, achieving the 30 mg/l new performance from 1 January 2007 may
become artificially easier for many of the installations following its implementation. Therefore
it is possible that OSPAR may want to consider reducing the performance standard further
from 30 mg/l to a lower level to reflect the change of the reference method from IR based to a
GC-FID based.
In order to answer this question, it is necessary to look at what are measured by the two
methods. These are summarised in the Table 4.
From Table 4 it is not difficult to conclude that the two reference methods are very different.
As a result it is inevitable that data obtained from the OSPAR / DTI method will be different
from those obtained from the ISO Mod method.
Comparison studies [8,9,10] have clearly shown the significant differences between results
obtained from using the IR based methods and the ISO Mod method.
Having accepted that the two methods will produce different results, is it then possible to
convert results from one reference method to the others using a simple correction factor
across fields for the North Sea? Clearly if such a conversion factor can be found, there would
have less implications when implementing the new ISO Mod method.
In the OSPAR comparison study [9], the ratio of IR result to ISO Mod result was calculated.
In total 240 data points were available. A huge range from 0.05 to 45.5 of this ratio was
found with an overall mean of 1.33. The contributions to the huge variations in this ratio could
have resulted from a number of factors as given below:
Clearly from the data obtained from the study, it was not possible to derive an universally
acceptable conversion factor that can be used to convert results from the ISO Mod method to
the IR methods. Even when data are grouped for different types of installations, e.g. gas,
condensate and oil, the variation in the ratio remains considerable.
While it is clear from the OSPAR comparison study that a meaningful general correlation
between the results obtained from the Mod method to that from the current IR based
methods is challenging. This does not mean that for a specific installation results from the
two methods can not be linked. For a specific installation, it should be possible to establish a
conversion factor between the currently approved method and the ISO Mod method.
In this approach, nine sets of replicate samples are collected over a period of time (18
samples in total). The samples are then analysed by using two different methods, e.g.
Infrared and the new reference method. F-test and Student-t test are then followed to check if
there is a difference between the two methods.
Provided that the results from both methods come from the same underlying distribution (F-
test), if the Student-t test shows that there is a statistically significant difference between the
paired results from the two methods, a correction factor (CF) could be made. If the Student-t
test shows no statistical significant difference, then there is no need to have a correction
factor.
The correction factor (CF) can be calculated using the equation 2 below:
dm
CF = 1 + [2]
x
where dm is the mean of the differences between the paired results obtained from two
different methods (e.g. Infrared and the new reference method) and x is the mean of a
reference method values, e.g. the old IR method.
Final values are obtained by multiplying all the analytical results from the new reference
method values by CF to obtain the final values (i.e. old IR method equivalent).
This method is simple and straightforward, but according to the developer, it will not be
acceptable if the correction factor is outside the range of 0.9 to 1.10.
In the USA when the Method 1664A [54] was established for the analysis of oil and grease to
replace its predecessor, method 413.1 [55], specific procedures were developed to compare
the method 1664A with method 413.1. The procedures enable one to determine if the
difference between results obtained from the two different methods is significant or not and
also allow one to calculate a conversion factor if the difference is statistically significant.
The method is also called Side by Side Comparison [53]. It involves analysing of three
replicates of each sample by each method on any seven days over a minimum 30-days
period, for a total of 42 analyses (21 by the EPA Method 1664A and 21 by the CFC-113
based Method).
The test of statistical significance uses the root-mean-square deviation (RMSD). If the RMSD
is within the acceptance limit, of which the maximum is defined by equation 3 below, the
results obtained using the different methods are equivalent and a conversion factor is not
warranted. If the RMSD is not within the acceptance limit, the difference is significant.
2 × MSE
RMSDMAX = × 2.36 [3]
3
Where MSE is the mean square error, which is the average of all mean square deviations
calculated from triplicate samples. 2.36 is the 95% F value with 7 degrees of freedom in the
numerator, and 28 degrees of freedom in the denominator.
If the difference is statistically significant, a second step is necessary to test whether the bias
between methods varies significantly from sample to sample (a significant method/sample
interaction). If there is no significant interaction, a conversion factor can be developed.
Otherwise a conversion would not be appropriate.
The conversion factor (CFL) is calculated as the ratio of the average of the method means of
the log-transferred results. A log-transformed EPA Method 1664A value is then multiplied by
this conversion factor before being converting it to an EPA Method 413.1 equivalent result
using the exponential function. (note: before any calculation is carried out on the test results
in the Side by Side comparison, the results are first log-transformed to assure that a single
result does not unduly influence the outcome, which is why the conversion factor is the ratio
of the average of the method means of the log-transformed results, and that when actual
conversion is conducted, one needs to using the exponential function)
This method is obviously time-consuming although an Excel spread sheet is available from
USA EPA that allows all the calculations to be carried out automatically once all the analysis
results are entered.
This section of the report provides an overview of the work carried out by Opus, on behalf of
Talisman Energy (UK) Ltd, in evaluating the Arjay Fluorocheck 2000 on a number of
Talisman installations [56]. A summary of the results obtained is discussed, which should be
useful to those who intend to look at using alternative methods. Also some general
comments are made in relation to how the methodology relates to the guidelines provided in
this document.
Prior to the offshore evaluation of the Arjay fluorimeter, Opus carried out a laboratory
evaluation of the instrument for Talisman in late 2002. This work was carried out entirely with
laboratory prepared oily water samples, made up from Piper and Beatrice crude samples.
The results of the study showed a good correlation between the Arjay and the existing IR
method. Following presentation of the results to the DTI, it was recommended that the
instrument be evaluated offshore to determine performance against actual produced water
samples.
The offshore evaluation of the instrument was carried out in two phases. The first phase,
which included evaluation on Piper, Claymore, Tartan, Beatrice, Buchan, Clyde, was carried
out between August 2003 and January 2004. The second phase included for evaluation on
Northern Producer, Saltire and Clyde (repeat visit) and was completed between November
and December 2004. Total cost of the work was approximately £56k.
Allied with solvent extraction using n-pentane or n-hexane, the Arjay fluorimeter offers a
relatively quick alternative means of carrying out routine oil-in-water analysis. The main
objectives of the offshore evaluation of the instrument and analysis technique were to:
The evaluation of the instrument was based on typically 3-5 days on each installation.
The results obtained from the offshore work are summarised below and discussed in detail in
Appendix 8. Also in Appendix 8 analysis methodology and Arjay fluorocheck method
development are detailed.
• The results of the study showed the Arjay fluorimeter to correlate well with the existing
DTI IR method, on five of the six platforms visited (Piper, Claymore, Tartan, Buchan,
Beatrice). The results from Clyde gave very much higher results with the Arjay, the
reasons for which were not clear at the time. However, laboratory work carried out
onshore on the different oil streams being processed on Clyde, showed that the Arjay
was sensitive to changes in oil composition and that the choice of calibration crude
could, in the case of Clyde, significantly affect the measured result.
• During the offshore sampling, samples were taken from a number of points in the
process, to allow evaluation over a range of concentrations. Most of the results
obtained from the Arjay, were within 5-15 ppm of the result obtained using the IR
method.
• In line with previous study work carried out by Opus Plus Ltd, the cross checks
carried out using the GC ISO9377-2 method, showed, with the exception of results on
Claymore, the GC results to be approximately half those obtained using the IR
method and overall the comparison was poor.
• Comparison of the results obtained from using n-pentane and n-hexane with the
Arjay, showed the use of n-pentane to give consistently better results which
correlated better with the IR method. The extraction procedure also proved better
with n-pentane, with a clearer water phase observed.
• The calibration work carried out on the Clyde, Tartan and Claymore installations,
showed that the crude oil characteristics can have a significant effect on the response
of the Arjay instrument. In cases where crudes from different wells or reservoirs differ
significantly, it is important that a representative sample of the crude likely to be
present in the produced water is used for the calibration. This was particularly evident
from the results obtained on Clyde, where subsequent onshore work showed the
crude characteristics to significantly affect the measured result, depending on the
calibration oil used.
• The method development work on the six installations, identified a number of areas
which it is felt merit inclusion in any agreed method adopted for use of the Arjay
instrument offshore. Specifically these include (1) the collection of 500 or 1000 ml
samples as a minimum (as opposed to 100 ml detailed in the Arjay method), (2)
extraction of the sample with a 50 ml volume of extract in a separating flask, as
opposed to the sample bottle, and (3) at least 5 minutes ‘shake time’ to ensure
adequate extraction of the sample.
• The Turner TD500 instrument evaluated on Buchan, also proved to give good results
which compared well with the IR method. The instrument is also highly portable
(hand held size) and the calibration proved very stable. Overall, the results were
slightly lower than those of the Arjay instrument. However, the instrument seems
ideally suited for rapid process trend monitoring or trouble shooting.
• The results obtained from the visits to Clyde and Saltire showed the Arjay to correlate
well with the IR method. The data gathered on Clyde validated the findings of the
laboratory work carried out following the first visit to Clyde. The good results obtained
in the second visit to Clyde provide further evidence that the Arjay method is sensitive
to the characteristics of the crude used to prepare the calibration standard.
• The data obtained from the visit to Northern Producer (Galley field) showed the Arjay
to give consistently higher results than the IR method. Onshore laboratory work
carried out to investigate the reasons for the discrepancies, has not resulted in any
firm conclusions other than the differences observed offshore were not replicated in
the onshore laboratory work. The differences may result from some feature of the live
samples offshore that could not be replicated in the onshore laboratory work.
The results of the cross checks carried out using the GC ISO9377-2 method, showed
reasonable agreement between the IR, Arjay and the C7-C40 GC results (ISO Mod standard),
for both Saltire and Clyde crudes. The C10-C40 results for Saltire were approximately half
those of the IR and Arjay, though unusually gave comparable agreement with the Clyde
crude.
For the purposes of the statistical comparison, the IR method has been assumed to be the
true measure of oil-in-water concentration. The main conclusions from the statistical review
of the data are detailed below.
• Overall, the Arjay pentane and hexane data showed a negative bias compared to the
IR data. Across all data, the average difference between the Arjay and IR results was
-2.6 ppm for the pentane extractions and –3.1 ppm for the hexane extractions.
• Both the pentane and hexane results from the Arjay instrument, showed good straight
line fits over a reasonable range of values, although all coefficients were significantly
less than the desired level of 1. All pairwise correlations (pentane-IR, hexane-IR,
pentane-hexane) showed close proximity to a straight line relationship, with
correlations of 0.927, 0.964 and 0.971 respectively.
• Based on the results in this study, hexane gave slightly better results than pentane (in
terms of comparison to the IR data), although the differences between pentane &
hexane are not statistically significant.
• There is strong evidence that the Arjay data presents a rescaling of the IR data based
on the correlations and the regression analysis. A slight modification of the pentane
and hexane values could therefore give IR equivalence.
• Excluding outliers in the data, all GC measures showed a significant correlation with
the IR results, with a lowest correlation of 0.735 (pentane extraction-modified method-
crude oil calibration standard) and a highest of 0.926 (hexane extraction-existing
method-crude oil calibration standard).
• The scatter diagrams and regression models show significant deviation from a linear
model with again the hexane extraction-existing method-crude oil calibration standard
results giving the best fit.
• The Clyde data were included in the GC comparison as the GC method uses a
separate calibration standard. In contrast to the Arjay results, the Clyde data seemed
to calibrate well in the GC models. This suggests that the existing GC method is
much less sensitive to the characteristics of the oil in the sample, than the Arjay
technique.
As for phase one, following completion of the offshore visits, the data were analysed
statistically, on an independent basis. For the phase two data, the only comparison carried
out was between the IR and the Arjay results. The Galley data were eliminated from the
analysis due to the poor correlation evident between the Arjay and IR method. The main
conclusions from the statistical review of the data are detailed below.
• The new data, combined with the data from phase 1, has enhanced the evidence that
the new methods are simply a rescaling of the existing IR method.
• The pentane data improved the correlation from 0.927 to 0.940 when the phase two
data were combined with the phase 1 data.
Overall, the statistical analysis supports the findings observed in the phase 1 report, with an
excellent correlation evident between the Arjay method and the IR method.
HSE issues
The Arjay fluorimeter offers a number of potential benefits to personnel from an HSE
standpoint. These are detailed below.
• Both n-pentane and n-hexane can be used with the method. The latter may be used
if there are concerns regarding the volatility of n-pentane, although n-pentane proved
to be a more effective solvent.
• The Arjay method in itself can help to minimise the exposure of personnel to the
solvent, in that the solvent is added directly to the sample bottle and an extract drawn
directly from it after extraction.
• Charcoal badge exposure monitoring to n-pentane & n-hexane carried out on one of
the installations, gave results well below the recommended long term exposure limits.
The ACTS guideline figure for n-pentane is 1800 mg/m3, whilst the long-term
exposure figure for n-hexane is 72 mg/m3. Exposure concentrations of 28.0 mg/m3 or
less were obtained for n-pentane, whilst concentrations of 23.6 mg/m3 were obtained
for n-hexane. These exposure concentrations were determined for 6 separate days,
using 1 badge per day for each solvent. Average exposure time per day was 9 hrs.
In comparison, the long term exposure limit for TTCE is 345 mg/ m3.
• All installations now have a fume cupboard. It is recommended that all use of the
solvents, pentane or hexane is conducted, as far as practically possible, in or in the
vicinity of the fume cupboard.
Instrument usability
• Overall, the Arjay instrument proved rugged and easy to use. Some drift in the
calibration reference standard was evident although if the instrument was left to warm
up for at least 30 minutes this was found to stabilise.
• Throughout the offshore work, n-hexane and n-pentane were used. The grade used
in this study was HiPerSolv HPLC 97%grade. Relative costs of the two solvents were
£26.11 per 2.5 litres for n-hexane and £25.94 per 2.5 litres for n-pentane. Contact
with the instrument suppliers confirmed no minimum grade solvent requirement, as it
was claimed that the instrument zero would eliminate any influence due to solvent
variability. For comparison, the current cost of TTCE is £47.25 per 2.5 litres.
The work carried out in this Talisman project was essentially an evaluation of both the Arjay
instrument, but also the actual method used for carrying out the analysis. As such, there are
many steps associated with the method which have been examined in greater detail than
would be necessary for approval of any instrument under the guidelines provided in this
document. It is therefore not possible to exactly match the work carried out for Talisman,
against the guidelines that should be followed for acceptance of new laboratory instruments.
However, there are many areas where the principles of the guidelines were applied and the
following comments can be made in relation to the guidelines provided in this report.
The performance of the Arjay instrument was assessed first and foremost by comparison
against the IR method. Although GC analyses were carried out, the Arjay calibrations and
results were mainly compared to the IR method, which was assumed to be the true measure
of the oil-in-water concentration.
All calibrations carried out in this Talisman project were based on the methods normally used
for IR calibrations. Calibration standards were, however, prepared using both a back
extraction technique and an oil dissolved in solvent technique, to compare results from each
method of calibration.
Since the Arjay was evaluated on different installations, the instrument was always
recalibrated prior to each visit (roughly monthly). Validation of the instrument for a specific
installation was therefore not part of this project.
The statistical analysis of the data generated in this project included several statistical tests.
However, the main method of analysis was linear regression of the Arjay data against the IR
data. The results from this showed a high level of correlation between the Arjay and IR.
Linear regression was also carried out on the more limited GC data set, with a significantly
lower level of correlation.
Overall, the OIWAM guidelines provide an excellent basis for the approval of laboratory
based instruments. Although the work on the Arjay instrument was carried out before these
guidelines were fully developed, it is clear that many of the general principles have been
examined throughout the course of the Talisman project work.
9 CONCLUSIONS
Reviews were carried out in the early stage of the project aimed at identifying the best
practical means of implementing the ISO Mod method. The reviews have however indicated
that whilst there are differences in the approved IR methods between the UK and the
Netherlands, most of operators were in general happy to continue to use the currently
approved IR methods. Also a new standard IR method based on using S-316 was made
available in 2004. Only a small number of operators (approximately 10%) currently have
GCs in the UK and Dutch sector, which are used for the analysis of gas compositions.
Few operators are actively seeking alternative methods at present. Direct implementation of
the ISO 9377-2 method is possible as has been demonstrated in the Norwegian sector,
however, the experience in Norway does not warrant direct implementation of the ISO Mod
method offshore.
Overall the reviews shows that there is some support for a continued use of the IR based
methods. Implementing the ISO 9377-2 directly offshore is possible but will require significant
initial investment and considerable commitment on training and maintenance from the
offshore operators.
Several options for implementing the ISO Mod method were also reviewed. It was concluded
that the best way forward is to use alternative methods which are calibrated against the ISO
Mod method. Acceptance of alternative methods will be primarily based on having a valid
calibration between the alternative method and the new reference method although statistical
significance tests could also be used.
For acceptance, different calibration approaches for both laboratory based bench top
analysers and online monitors have been suggested. There is no simple solution in
establishing a valid calibration between an alternative oil-in-water analysis method and the
ISO Mod method, and ensuring the calibration remaining valid over time. The difficulties arise
from the fact that oil-in-water is unstable and method dependent, and also that the ISO Mod
method needs to be used in an onshore laboratory. Therefore the approaches suggested in
this report will need field trials to confirm their practicality so that they may be refined and
improved for eventual acceptance and implementation of the ISO Mod method.
Sampling and sample handling has also been reviewed and eight general guidelines have
been established. Proper adherence to these guidelines will minimise the errors which may
result from sample taking and handling.
Regarding data set continuity, whilst a generalised correlation between the new ISO Mod
method and the current IR based methods across the North Sea fields does not look feasible,
this should be possible for individual installations.
Specific Conclusions
• Oil-in-water analyses using IR (or FTIR) and TTCE methods have been generally
accepted by most operators in the UK and the Netherlands, who are reasonably happy to
continue their use.
• GC-FID instruments are currently being used on some of the installations - approximately
10% in the UK and 5% in the Dutch sector - for gas composition analyses.
• The vast majority of operators have not sought to use alternative oil-in-water analysis
methods at this time.
• A new ASTM infrared based oil-in-water analysis method was made available in May
2004. The method is similar to the OSPAR / DTI method but using a solvent called
dimer/trimer of chlorotrifluoroethylene, or S-316.
• Offshore direct use of the ISO 9377-2 GC-FID method has been carried out by some of
offshore operators in the Norwegian sector. With careful planning together with training
and capital investment in the new GC-FID instruments, implementation of the ISO
method has been possible. However, the overall experience in Norway does not fully
support the use of ISO Mod method offshore. This is due to the fact that the Mod method
is less robust and requires the use of higher resolution GCs. In addition GC-FID based
methods take longer to produce results.
• GC-FID instrument suppliers were contacted and potentially suitable instruments were
reviewed. In addition onshore laboratories who have had experience in oil-in-water
analyses have been reviewed.
• Options for implementing the ISO Mod method have been reviewed. It is thought that the
best way forward to implementing the new reference method is to use alternative
methods which are calibrated against the new reference method available onshore.
• Both laboratory based bench top analysers and online monitors are available. Provided
that an alternative method is calibrated against the ISO Mod method with a valid
correlation, its use should be allowed in regulatory compliance monitoring and reporting.
• For laboratory bench top based alternative methods, two different calibration approaches
can be considered, with one based on the traditional method where calibration standards
are prepared and analysed, and one based on using dynamic calibration where field
replicate samples are collected and used.
• Using the traditional calibration approach, the established correlation will be able to cover
as wide a concentration range as desired. The disadvantage is that the effect of sample
matrix on analysis is not fully taken into account. In addition validation of the correlation
is required. Validation is suggested initially once a month by analysing a standard
solution.
• Using the dynamic calibration approach, the correlation established may not cover a
desired concentration range. However, as field samples are used, the effect of sample
matrix on analysis is taken into account. Also the calibration procedures will ensure that
the correlation is kept to date with new correlation established at least every two months.
• For online monitors, no easy way to calibrate it against the ISO Mod method exists.
Calibration can only be achieved over a period of time after proper installation of the
instrument on site.
• When an online monitor calibration is carried out, samples parallel to the ones analysed
by the online monitors should be taken and analysed by the ISO Mod method. Ideally
three points with each point corresponding to low, mid-scale and high readings within the
online monitoring range are made available. For each of the three calibration points, three
replicate samples are taken and analysed by the ISO Mod method. Average results are
then plotted against the average value of the online monitor readings for the period when
the Mod samples were taken.
• Once a calibration is established between the online monitor and the ISO Mod method,
routine calibration checks are suggested, initially once a week. For each check, typically
three samples are taken for the ISO Mod method, the difference between the average
from the online monitor and ISO Mod method is calculated, which is then plotted onto a
control chart. The control chart is used to identify instrument drifting, and fault
development. One of calibration check points may also be used for validation.
• For UV fluorescence based on-line monitors, calibration checks may be carried out using
a secondary reference standard. Again a control chart is used to identify drifting and fault
developing, If the calibration checks are performed in this manner, validation is suggested
initially once a month.
• Due to the difference in analysis principles and the fact that there are differences in the IR
methods currently used in the North Sea in terms of quantification wavelength, calibration
oil and procedures, establishing a generalised correlation between the ISO Mod method
to the IR method across the North Sea fields looks challenging.
• For individual installations, however, establishing a correction factor between the ISO
Mod method and the IR method should still be possible. Two approaches found in the
literature have been summarised in the report.
• Sampling and sample handling remains a key issue in accurately determining oil in
produced water. Current practices and requirements have been reviewed.
• Eight general guidelines have been established covering sample point location, design of
sample point, sample flow rate & sample point condition, sample bottle requirements,
sample handling, sample storage, sample labelling & documentation requirements, taking
field replicate samples.
The current project has focused on identifying the best practical way to implement the ISO
Mod method and also on establishing acceptance for using alternative methods as well as on
establishing best practice guidance on sample taking and handling.
With regard to future work the following two areas are suggested:
The need for further standardisation of the ISO Mod method was highlighted when the
OSPAR comparison project was carried out. In the OSPAR project four laboratories
performed a ring-test in which two samples with known dispersed oil content were analysed
using the ISO Mod method. Significant differences in the analysed concentrations were
observed between the laboratories.
Once the above are clarified, further inter-laboratory trials should be carried out to establish
precision data. The trials should be carried out using samples containing oil level(s) that
reflect those found in produced water offshore.
The best approach for implementing the ISO Mod method for installations that discharge
more than two tonnes per annum is to use alternative methods which are calibrated against
the new reference method. Calibration approaches have been proposed for both laboratory
based bench-top analysers and also on-line monitors. As these have not been used in real
applications, trials are recommended to test their practicality. By carrying out field trials, the
amount of effort and associated costs involved can be firmly established.
Issues such as the number of calibration points, and the number of analyses for each
calibration points, as well as frequency of validation and calibration checks need to be
examined with a view to refining and improving the recommendation of this report.
ABBREVIATION
GLOSSARY
This glossary gives explanations for some of the terms and words used in the report.
Significance Tests: allows one to compare two sets of results in an objective and unbiased
way. Student t-test is a statistical procedure used to compare mean values. F-test is used to
compare the spread of results in two data sets.
TTCE: Tetrachloroethylene, also known as perklone
Validation: to confirm that initial calibration has not changed and is still valid
Variance: a measure of the dispersion of a set of measurements, giving a measure of the
distribution of values in a Normal probability distribution
Uncertainty (of Measurement): quantified doubt about the result of a measurement
REFERENCE
[1] The DTI Approved Oil In Water Measurement Method, ‘Methods of sampling &
analysis of production & displacement water discharges for exemptions from Section
3 of the Prevention of Oil Pollution Act 1971 issued by the Department of Trade &
Industry for Offshore Installations’
[2] NEN 6675 – Determination of mineral oil content by infrared spectroscopy - (three
wavelength) (1989).
[3] NEN 6675 modified – the NEN 6675 whereby the extinction at wavelength 3030 cm –1
is set zero, which gives aliphatic oil concentration.
[4] Norsk Standard NS 9803, “Vannundersøkelse, Bestemmelse av olje i vann, Infrarød
spektrofotometrisk metode”, 1993. (former Norwegian Standard method for detection
of oil in water, equal to Swedish Standard Method SS02 81 45, 1991).
[5] ISO 9377-2:2000 – “Water quality – determination of hydrocarbon oil index – Part 2:
Method using solvent extraction and gas chromatography”.
[6] “Comparative study of Freon-IR (NS9803) and GC-FID (ISO 9377-2) for analysis of oil
in produced water on 33 platforms in the Norwegian sector”, OLF Working Group on
oil-in-water, 2001
[7] “An evaluation of current, pending and alternative oil-in-water monitoring methods”, a
report prepared for the UK Offshore Operators Association (UKOOA) by NEL and
ERT(Orkney), NEL Report No: 042/2002, April 2002.
[8] “Results of a “Quick Scan” study – Comparison of OSPAR Agreement 1997-16 with
GC method based on ISO 9377-2”, presented by the Netherlands at OSPAR OIC
meeting, Cadiz, 11-15 February 2002
[9] M.G.D. Smit, K.I.E. Holthaus, C.C. Karman and P. Frintrop, “Comparison of OSPAR
analytical methods for the determination of dispersed oil in produced water”, TNO-
report R2003/025, 24 February 2003
[10] C.C. Karman, M.G.D. Smit, and P. Frintrop, “ISO 9377-2 Mod; a GC-FID based oil-in-
water analysis method as OSPAR Reference method”, R2003/027, February 2003
[11] “Analysis and Determination of the Oil Content in Produced Water - Specific issues in
relation to the modified ISO 9377-2 method”, a document submitted to OSPAR OIC
meeting in London, 10-14 March 2003
[12] ASTM D 7066-04: Standard test method for dimer/trimer of chlorotrifluoroethylene (S-
316) recoverable oil and grease and nonpolar by Infared determination, 2004
[13] ASTM D 3921 – 96: Standard test method for oil and grease and petroleum
hydrocarbons in water, 1996
[14] Private communication with Statoil
[15] S. Tulloch, M. Tait, J. Rusin and M. Orme, “Field evaluation of alternative methods for
oil in water analysis”, a paper presented at NEL’s 5th oil-in-water monitoring
workshop, 22-23 September 2004, Aberdeen
[16] Private communication with Jorn Paus at Norsk Hydro, Norway, May 2004
[17] J. Paus, “Offshore experiences with ISO 9377-2 for oil-in-water in Norsk Hydro”, a
paper presented at NEL’s 2003 OIW Workshop, 21-22 May 2003, Aberdeen
[18] J. Paus and E. Odden, “Offshore experiences with ISO 9377-2 for oil in water in
Norsk Hydro – Part II, a paper presented at NEL’s 2004 OIW Workshop, 22-23
September 2004, Aberdeen
[19] Private Communication with Frode Flaaten at ConocoPhilips, Norway, May 2004
[20] http://www.npd.no/Norsk/Frontpage.htm
[21] “Emissions and discharged from the Norwegian Petroleum Industry 2000”, an OLF
report prepared by Novatech a.s., June 2001
[22] J. Fardon R. Perkins and D.Turner, “Analysis of total petroleum hydrocarbons in river
and waste water using an automated approach to the ISO 9377 GC-FID standard
method”, a paper presented at NEL’s 2003 OIW Workshop, Aberdeen 2003
[23] http://www.agilent.com/about/index.html
[24] P. Lassen, “Development of large volume injection GC-FID for oil-in-water analysis”, a
paper presented at NEL’s 2003 OIW Workshop, Aberdeen 2003
[25] Private communication with Donald Inglis at Shimadzu, May 2004
[26] T. Aasoldsen, F. Pigozzo and A. Cadoppi, “Determination of oil-in-water using the ISO
9377-2 Method with Ultra Fast GCs from Thermo Finnigan”, a paper presented at
NEL’s 2004 OIW Workshop, 22-23 September 2004, Aberdeen
[27] Private communication with Alan Lewis at Varian, 2004-07-27
[28] http://www.tnrcc.state.tx.us/permitting/analysis.htm
[29] E. Eamsey, “Further developments of SFE-IR method for the analysis of oil-in-water”,
a paper presented at NEL’s 2003 OIW Workshop, Aberdeen 2003
[30] “Testing and evaluation of the Supercritical Fluid Extraction and Infared (SFE-IR) oil-
in-water analysis method”, a final report for BP, DTI, and Statoil, NEL report no.
2003/223, December 2003
[31] OSPAR Agreement 1997-16 on “Sampling and analysis procedure for the 40 mg/l
target standards”
[32] Pauline Shone, “The Case for Perklone as a Freon Replacement for Offshore Oil-in-
Water Analysis”, a paper presented at NEL’s 2000 oil-in-Water Monitoring Workshop,
May 2000, Aberdeen, UK
[33] J. Jager, “Occupational Risk Assessment of using tetrachloroethylene instead of
Freon for oil in water analysis”, Lasmo Nederland B.V. Document, December 2001
[34] M.Yang, “An Overview of Freon replacement solvents and alternative oil-in-water
monitoring and analysis methods”, a paper presented at the NEL’s 5th OIW Monitoring
Workshop, 21-22 May 2003, Aberdeen
[35] ASTM D3370-82: “Standard practices for water sampling”
[36] E. Butler and P. Shone “Produced Water Best Practice – ‘Sampling guidelines,
analytical methods & requirements’”, a paper presented at the UKOOA Condensate in
Water Workshop, 6-7 September 2000, Aberdeen
[37] ISO 5667-3 “Water Quality – Sampling Part 3: Guidance on the Preservation and
Handling of Water Samples”
[38] ASTM D3325-90: Standard practice for preservation of waterborne oil samples
[39] D. Caudle, “Sources of error in oil-in-water measurements and their impact on
comparing oil-In-water monitors”, a paper presented at NEL’s 2002 OIW Workshop,
Aberdeen, 22-23 May 2002.
[40] Guidelines for sampling & analysis of produced water, OLF Guideline Document
[41] Comparison of in-line and fast loop sampling systems, Jiskoot Website
[42] “Review & comparison of oily water analytical techniques’, final report, Dr J M. Tait,
Orkney Water Test Centre Ltd 1994
[43] S. Burke, "Statistics in context: Significance testing", VAM Bulletin, 17, 18-21, Autumn
1997
[44] T. Farrant, “Practical statistics for the analytical scientist – a bench guide”, published
for LGC by the Royal Society of Chemistry, 1997
[45] BS ISO 11095: 1996(E) Linear calibration using reference materials
[46] ISO 8466-1:1990 Water quality – Calibration and evaluation of analytical method
estimation of performance characteristics – part 1: Statistical evaluation of the linear
calibration function
[47] Preparation of Calibration Curves – A guide to best practice, September 2003,
LGC/VAM/2003/032
[48] “Code of practice for calibration and checking process analysers” - EEMUA
Publication no. 175: 1995
[49] “Reference method for dispersed oil in produced water”, OIC 04/2/9-E, OSPAR OIC
meeting, Dublin, 15-19 March 2004
[50] http://www.fao.org/docrep/W7295E/w7295e08.htm
[51] A. Laurie, “Update on oil in produced water issues”, a presentation made at NEL’s
OIW Club meeting, December 2004, Aberdeen
[52] “Method 301 – Field validation of pollutant measurement methods from various waste
media”, http://www.epa.gov/ttn/emc/promgate/m-301.pdf
[53] W. A. Telliard, “EPA effort to replace Freon-113”, a paper presented at NEL’s first oil
& Grease Measurement seminar 2002, 15 March 2002, Houston, USA
[54] EPA Method 1664 – 1994 “N-Hexane extractable material (HEM) and silica gel
treated N-hexane extractable material (SGT-HEM) by extraction and gravimetry (oil
and grease and total petroleum hydrocarbons)”. EPA-821-B94-004, issued October
1994.
[55] EPA Method 413.1 “Standard test method for oil and grease using gravimetric
determination”
[56] Offshore evaluation of an Arjay fluorimeter for produced water analysis, final report
(issue no. 2), by Opus for Talisman Energy (UK) Limited, June 2004
APPENDIX 1
Acknowledgement
APPENDIX 2
ANNEX 10
(Ref. § 4.13a)
OSPAR CONVENTION FOR THE PROTECTION OF THE MARINE ENVIRONMENT OF
THE NORTH-EAST ATLANTIC
MEETING OF THE OFFSHORE INDUSTRY COMMITTEE (OIC)
LONDON (SECRETARIAT): 10-14 MARCH 2003
Pretreatment
1. Since the integration of the gas chromatogram starts at n-heptane, and since this
peak should be well separated from the peak of the extraction liquid, the sample shall be
extracted with n-pentane. This extraction liquid falls within the scope of ISO 9377-2.
2. ISO 9377-2 applies 2 internal standards (C10H22 and C40H82), which are added to the
extraction liquid. Both internal standards serve for the reproducible establishment of start and
end time of integration. In the modified ISO 9377-2 method (hereafter referred to as ISO
9377-2 mod), the internal standard of n-decane cannot be applied and n-decane shall
therefore not be added to the extraction liquid n-pentane.
3. ISO 9377-2 states the necessity of concentrating the extract, in order to realize a limit
of detection of approximately 0,1 mg/l. To this end, the extract is concentrated “offline”;
“large-volume injection” may serve as an equivalent alternative. It is not known whether this
will also be valid for hydrocarbons between n-heptane and n-decane. For the analysis of
produced water is no compelling need to realize a limit of detection of 0,1 mg/l with ISO
9377-2 mod – a limit of 0,5 mg/l is sufficient for the goal of this project. Therefore the pentane
extract shall not be concentrated off-line. Injection of 50 microliters of extract into a large
volume injection system has proven to give good results.
Gas chromatography
4. ISO 9377-2 does not require a high peak resolution. Since the whole gas
chromatogram is integrated, including non-separated peaks, there is no need for a high
resolution. On the contrary: a high resolution may result in a long gas chromatogram, making
it difficult to apply an adequate baseline correction.
5. ISO 9377-2 mod requires a higher resolution, especially for the range C7H16 - C10H22.
The reason for this is that the peaks of TEX should be adequately separated from the other
aliphatic hydrocarbons in this range, and that the integration starts at the right time at the
beginning of the peak of n-heptane.
7. With ISO 9377-2 mod the start of integration is at the beginning of the peak of n-
heptane, whereas with ISO 9377-2 the integration starts at the end of the peak of n-decane.
8. When applying ISO 9377-2, it is common practice to divide the integration of the gas
chromatogram in different parts, e.g. after-C10-C14, after-C14-C18 etc. The sum of these parts
equals the stretch of C10-C40, which would represent the oil content. The same approach is
foreseen when applying ISO 9377-2 mod, where as a minimum the stretches C7-C10 and
after-C10-C40 should be distinguished. Thus the oil content according to ISO 9377-2 is
established, as well as the part of oil that is “missing” at the end of the volatile fractions.
9. Subsequently, the area of the peaks of TEX should be integrated and subtracted from
the area of C7-C10. It is assumed that a non-polar analytical column is used, so the peak of
benzene will elute before the peak of n-heptane and is thus not included. Should this,
however, be the case, then the peak areas of BTEX should be integrated.
10. When the resolution is not sufficient to integrate the area of the peaks of TEX within
the gas chromatogram, then the determination of TEX should take place in the same extract
under more favorable conditions. This means that the determination of TEX will take place
according to ISO 11423-2 (Determination of benzene and some derivatives - Part 2: gas
chromatographic method after liquid-liquid extraction).
APPENDIX 3
Questions have been prepared and contact with operators has started. It is intended to seek
answers to the following questions on an individual installation basis:
(a) General
Analysis:
(c) In the case where samples are sent for onshore analyses (most likely gas
condensate fields where samples are taken only when people visit the installation)
Is there a GC onsite?
What is it used for?
Manufacturer and model?
What is the carrier gas used?
Is hydrogen supplied or generated onsite?
Do you always have skilled GC operators onsite?
Would you consider to use GC for oil in water analysis?
Have you been actively seeking alternative laboratory based methods with the view that they
can be evaluated and used for reporting when the new GC-FID reference method is
implemented?
Do you plan in the near future to test and evaluate such alternative methods?
If so, are any particular types or techniques being considered?
If yes:
If no:
Ming Yang, NEL (email: myang@tuvnel.com; Tel: +44 1355 272 682)
(a) General
Installation name:
Type of installation (oil or gas?):
Discharging produced water continuously or in intermittently?
Have you carried out analyses in the past year on aromatics, aliphatics and total
hydrocarbons? If so by which method – three peak IR analysis or GC? By whom?
Analysis:
If “yes”
Synthetic oils?
Using platform specific crude oil dissolved in TTCE or another solvent
By adding crude oil in clean water then back extracted using TTCE or other
solvent
If “no”
(c) In the cases where samples are sent for onshore analyses (most likely for some
gas condensate fields where no offshore analysis capability is available)
Have you been actively seeking alternative laboratory based methods with the view that they
can be evaluated and used for reporting when the new GC-FID reference method is
implemented?
Do you plan in the near future to test and evaluate such alternative methods?
If so, are any particular types or techniques being considered?
If yes:
Roughly what has the instrument availability been like in the past year?
How often is the monitor calibrated?
Do you ever use an on-going calibration chart?
What are the key issues in making on-line monitors more widely accepted?
If no:
APPENDIX 4
SIGNIFICANCE TESTING
(i) Basic statistics for significance testing (as given in the 3rd OIWAM JIP bi-monthly report)
Acknowledgement
Permission to reproduce Dr. Burke’s article has been given by Valid Analytical Measurement
(VAM). The article was published in VAM Bulletin, Issue No 17, 18-21, Autumn
Sometimes it may be desirable to be able to compare the results of two methods. The
following is involved in significance testing
Where x and y are respectively the mean of results from method A and B
(iii) Critical value: the tabulated value of the test statistic, generally at the 95% confidence
level. If the calculated value is greater than the tabulated value, then the test results indicate
a significant difference, i.e. x ≠ y
(2) F-Test
F-Test is used to compare two variances, s12 and s 22 , to see if the two sets of data come from
a same underlying distribution. The test statistics is given as
s12
F=
s 22
Where s12 is the larger of the two variances, which makes sure that F ≥ 1, F is then compared
with Fcrit value at υ1 = n1-1 and υ1 = n1-1 degrees of freedom, denoted Fνν1 2 at the 97.5%
confidence level. If F is found to be smaller than Fνν1 2 , s12 and s 22 are considered to be
statistically equal. Otherwise they are not.
• A sample mean with a stated value, e.g. average of measured oiw values against 40
mg/l or 30 mg/l performance standards
• The means of two independent sets of data
• The means of two matched samples (called a paired comparison test)
The calculated test statistic, t, is compared with the tabulated value with υ degree of freedom,
at various percentages of confidence level, for either one or two tailed tests. A two-tailed test
occurs when one of the null or the alternative hypothesis contains an equal sign whilst the
other has a not equals sign. The other possibility is a one –tailed test in which case the
alternative hypothesis is an in equality (i.e. a limit value ≤, >, < or ≥). For an example, when
testing if A and B are different, it is a two tailed tests. When testing A is higher (or lower) than
B, it is a one sided test.
Assume that there are two sets of data with means x and y and variances s12 and s 22 , one
wishes to test the hypothesis that the two means are statistically equal; i.e.
Ho: x = y
H1: x ≠ y
Depending on the nature of the two sets of data, the means of the sets can be compared for
bias. The test procedures will depend on whether s12 is considered to be statistically equal to
s 22 , which can be tested using the F-test.
t=
(x − y )
1 1 s12 (n1 − 1) + s 22 (n2 − 1)
+ ×
n1 n2 (n1 + n 2 −2)
This t is then compared to critical t value from the Student-t table which is obtained at 95%
confidence for υ degree of freedom (υ = n1 + n1 -2). If the t calculated is less than critical t
from the table, Ho is not rejected, i.e. the means of the two sets data are equal, otherwise the
means will be different.
x− y
t=
s12 s 22
+
n1 n1
2
s12 s 22
+
ν= n1 n2 −2
(
s12 n1
2
) (
+
s22 n2 )
2
n1 + 1 n2 + 1
Two methods (Method 1 and Method 2) have been used for determining the concentrations
of selenium, results are shown in the table below. These results are compared to find out if
the means are the same. Both F-Test and Student-t Test are carried out, these are detailed
below:
Avg S
Method 1 4.2 4.5 6.8 7.2 4.3 5.4 1.471
Method 2 9.2 4.0 1.9 5.2 3.5 4.76 2.75
First the variance of the two samples must be examined to determine whether there are any
differences between them (i.e. do they come from the same underlying distribution). This is
done by comparing the variances using the F-Test.
The null hypothesis is Ho: s12 = s22 , i.e. there is no difference between the variances of the two
methods.
To compare if the means are the same, the null hypothesis Ho: x = y , i.e. there is no
difference between the means of the two methods, against the alternative hypothesis H1:
x≠ y
Calculate the t
t=
(5.40 − 4.76) = 0.459
1 1 1.4712 (5 − 1) + 2.750 2 (5 − 1)
+ ×
5 5 (5 +5 −2)
The tabulated 95% critical value is 2.306 for υ1 = n1 + n2 -2 = 8 degree of freedom. This
exceeds the calculated value of 0.459, therefore Ho is not rejected and there is no difference
between the means of results obtained.
On the basis of these tests there is no reason to suspect that there are any significant
differences between the two methods.
To ascertain whether there is a difference in the results obtained by two methods the
difference between the results is used and not the individual values. The best way to
demonstrate how this is done is through an example.
Example [44] In the example 8 paired samples (e.g. replicate sample or duplicate samples)
are analysed by two methods A and B, the results are tabulated in the table below.
Sample 1 2 3 4 5 6 7 8
Method A 2.52 3.13 4.33 2.25 2.79 3.04 2.19 2.16
Method B 3.17 5.00 4.03 2.38 3.68 2.94 2.83 2.18
Diff (d) 0.65 1.87 -0.30 0.13 0.89 -0.10 0.64 0.02
The null hypothesis is: Ho there is no difference between results from the two methods, i.e.
d = 0 against the alternative: H1: d ≠ 0. The test is a two tailed test as we are interested in
both d > 0 and d <0.
The mean is d = 0.475 and the sample standard deviation of the paired difference is sd=
0.700.
d× n 0.475 × 8
t= = = 1.918
sd 0.700
The tabulated value of t (with υ=8-1=7 degree of freedom at the 95% confidence limit is
2.365). As the calculated value is less than the tabulated critical value, then Ho is not
rejected. Therefore there is no difference between the results.
APPENDIX 5
Acknowledgement
Permission to reproduce this edited version of the guide has been given by LGC Limited
September 2003
Contact Point:
Liz Prichard
Tel: 020 8943 7553
Prepared by:
Vicki Barwick
Approved by:
______________________________
Date:
______________________________
Contents
1. Introduction 119
2.6 Using the calibration function to estimate values for test samples 132
2.7 Estimating the uncertainty in predicted concentrations 132
2.8 Standard error of prediction worked example 134
3. Conclusions 136
1. Introduction
Instrument calibration is an essential stage in most measurement procedures. It is a set of
operations that establish the relationship between the output of the measurement system
(e.g., the response of an instrument) and the accepted values of the calibration standards
(e.g., the amount of analyte present). A large number of analytical methods require the
calibration of an instrument. This typically involves the preparation of a set of standards
containing a known amount of the analyte of interest, measuring the instrument response for
each standard and establishing the relationship between the instrument response and
analyte concentration. This relationship is then used to transform measurements made on
test samples into estimates of the amount of analyte present, as shown in Figure 1.
80
y = 52.357x + 0.6286
70
r = 0.9997
60
Instrument response
50
40
30
20
10
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
During August 2002 a benchmarking exercise was undertaken, which involved the
preparation and analysis of calibration standards and a test sample using UV
spectrophotometry. The aim of the exercise was to investigate the uncertainties associated
with the construction of a calibration curve, and with using the calibration curve to determine
the concentration of an unknown compound in an aqueous solution. In addition, it was
hoped to identify any common problems encountered by analysts undertaking calibration
experiments.
Members of the Environmental Measurement Training Network (EMTN) and the SOCSA
Analytical Network Group (SANG) were invited to participate in the exercise. Five members
of EMTN, six members of SANG and three organisations who are members of both EMTN
and SANG submitted results. Some participants submitted results from more than one
analyst, giving 19 sets of results in total. Full details of the protocol and results sheet
circulated to the laboratories can be found in Appendix 1 (not reproduced here – MY)
Appendix 2 (not reproduced here – MY) contains an ideal set of results from the
benchmarking exercise to illustrate how the report should be presented.
The results of the benchmarking exercise were interesting. Although the exercise initially
appeared relatively straightforward, a number of mistakes in carrying out the experiments
and analysing the data were identified. Since a number of the mistakes occurred in more
than one laboratory, it is likely that other laboratories carrying out similar exercises may
make the same errors.
The aim of this guide is to highlight good practice in setting up calibration experiments, and
to explain how the results should be evaluated. The guide focuses on calibration
experiments where the relationship between response and concentration is expected to be
linear, although many of the principles of good practice described can be applied to non-
linear systems.
With software packages such as Excel, it easy to generate a large number of statistics. The
guide also explains the meaning and correct interpretation of some of the statistical terms
commonly associated with calibration.
For an initial assessment of the calibration function, as part of method validation for example,
standards with at least seven different concentrations (including a blank) should be prepared.
The standard concentrations should cover, at least, the range of concentrations encountered
during the analysis of test samples and be evenly spaced across the range (see
Section 2.3.1). Ideally, the calibration range should be established so that the majority of the
test sample concentrations fall towards the centre of the range. As discussed in Section 2.7,
this is the area of the calibration range where the uncertainty associated with predicted
Having decided on the number and concentrations of the calibration standards, the analyst
needs to consider how best to prepare them. Firstly, the source of the material used to
prepare the standards (i.e., the reference material used) requires careful consideration. The
uncertainty associated with the calibration stage of any method will be limited by the
uncertainty associated with the values of the standards used to perform the calibration – the
uncertainty in a result can never be less than the uncertainty in the standard(s) used.
Typically, calibration solutions are prepared from a pure substance with a known purity value
or a solution of a substance with a known concentration. The uncertainty associated with the
property value (i.e., the purity or the concentration) needs to be considered to ensure that it
is fit for purpose.
The matrix used to prepare the standards also requires careful consideration. Is it sufficient
to prepare the standards in a pure solvent, or does the matrix need to be closely matched to
that of the test samples? This will depend on the nature of the instrument being used to
analyse the samples and standards and its sensitivity to components in the sample other
than the target analyte. The accuracy of some methods can be improved by adding a
suitable internal standard to both calibration standards and test samples and basing the
regression on the ratio of the analyte response to that of the internal standard. The use of an
internal standard corrects for small variations in the operating conditions.
Ideally the standards should be independent, i.e., they should not be prepared from a
common stock solution. Any error in the preparation of the stock solution will propagate
through the other standards leading to a bias in the calibration. A procedure sometimes
used in the preparation of calibration standards is to prepare the most concentrated standard
and then dilute it by, say, 50%, to obtain the next standard. This standard is then diluted by
50% and so on. This procedure is not recommended as, in addition to the lack of
independence, the standard concentrations will not be evenly spaced across the
concentration range leading to the problem of leverage (see Section 2.3.1).
Construction of a calibration curve using seven calibration standards every time a batch of
samples is analysed can be time-consuming and expensive. If it has been established
during method validation that the calibration function is linear then it may be possible to use a
simplified calibration procedure when the method is used routinely, for example using fewer
calibration standards with only a single replicate at each level. A single point calibration is a
fast way of checking the calibration of a system when there is no doubt about the linearity of
the calibration function and the system is unbiased (i.e., the intercept is not significantly
different from zero, see Section 2.5.2). The concentration of the standard should be equal to
or greater than the maximum concentration likely to be found in test samples.
If there is no doubt about the linearity of the calibration function, but there is a known bias
(i.e., a non-zero intercept), a two point calibration may be used. In this case, two calibration
standards are prepared with concentrations that encompass the likely range of
concentrations for test samples.
Where there is some doubt about the linearity of the calibration function over the entire range
of interest, or the stability of the measurement system over time, the bracketing technique
may be useful. A preliminary estimate of the analyte concentration in the test sample is
obtained. Two calibration standards are then prepared at levels that bracket the sample
concentration as closely as possible. This approach is time consuming but minimises any
errors due to non-linearity.
All equipment used in an analytical method, from volumetric glassware to HPLC systems
must be fit for their intended purpose. It is good science to be able to demonstrate that
instruments are fit for purpose. Equipment qualification (EQ) is a formal process that
provides documented evidence that an instrument is fit for its intended purpose and kept in a
state of maintenance and calibration consistent with its use. Ideally the instrument used to
make measurements on the standards and samples should have gone through the EQ
process.[1, 2, 3]
Any data sets of equal size can be plotted against each other on a diagram to see if a
relationship (a correlation) exists between them (Figure 2).
y-axis
8
7
Instrument response
6
5
4
3
2
1
x-axis
0
0 2 4 6 8
1
P. Bedson and M. Sargent, J. Accred. Qual. Assur., 1996, 1, 265-274.
2
P. Bedson and D. Rudd, J. Accred. Qual. Assur., 1999, 4, 50-62.
3
D. G. Holcombe and M. C. Boardman, J. Accred. Qual. Assur., 2001, 6, 468-478.
The values plotted on the y-axis are sometimes referred to as the dependent variable,
because their values depend on the magnitude of the other variable. For example, the
instrument response will obviously be dependent on the concentration of the analyte present
in the standards. Conversely, the data plotted on the x-axis are referred to as the
independent variable.
The plot of the data should be inspected for possible outliers and points of influence. In
general, an outlier is a result which is significantly different from the rest of the data set. In
the case of calibration, an outlier would appear as a point which is well removed from the
other calibrations points. A point of influence is a calibration point which has a
disproportionate effect on the position of the regression line. A point of influence may be an
outlier, but may also be caused by poor experimental design (see section 2.1).
Points of influence can have one of two effects on a calibration line – leverage or bias.
Leverage
14
point with high 8
12 leverage
Instrument response
10 6
Instrument response
8
4
6
4
2
2
0 0
0 4 8 12 16 20 24 28 32 0 2 4 6 8 10
Leverage affects both the gradient and intercept of the line with the y-axis.
Bias
Instrument response
6
0
0 2 4 6 8 10
Concentration /mg L-1
8
7
Instrument response
6
5
4
3
2 Observed
Estimated
1 Residual
0
0 2 4 6 8
Figure 5: Least squares linear regression – calculating the best straight line
2.4.1 Assumptions
Basic least squares linear regression relies on a number of assumptions. A ‘best-fit’ line will
only be obtained when the assumptions are met. Significant violation of any of the
assumptions will usually require special treatment which is outside the scope of this guide.
The first assumption was mentioned in Section 2.4, that is that the error in the x values
should be insignificant compared with that of the y values. In addition, the error associated
with the y values must be normally distributed. Normality is hard to test for statistically with
only small data sets. If there is doubt about the normality it may be sufficient to replace
single y values with averages of three or more for each value of x, as mean values tend to be
normally distributed even where individual results are not. The magnitude of the error in the
y values should also be constant across the range of interest, i.e. the standard deviation
should be constant. Simple least squares regression gives equal weight to all points – this
will not be appropriate if some points are much less precise than others. In many chemical
measurement systems the standard deviation increases with concentration, i.e., it is the
relative standard deviation that remains approximately constant rather than the standard
deviation. The general solution to this problem is to use weighted regression, which takes
account of the variability in the y values.[4]
Both the x and y data must be continuous valued, i.e., not restricted to integers, significantly
truncated or categorised (e.g., sample numbers, days of the week, etc.). This assumption
should be met in the case of instrument calibration as both the instrument response and the
concentrations of the standards can, in theory, take any value on a continuous scale.
The equations for carrying out a linear regression are given in Appendix 3, but regression is
usually carried out using software supplied with the instrument or packages such as Excel.
Many software packages allow a regression analysis to be carried out without first plotting
the data, however it is good practice to produce a plot before carrying out the statistical
4
Statistics and chemometrics for analytical chemistry, J. N. Miller and J. C. Miller, 4th Edition, Prentice Hall, 2000, ISBN 0-130-22888-5.
analysis (see Sections 2.3 and 2.5.2). If the option is available, it is also useful to obtain a
plot of the residuals (see Section 2.5.1).
Most software also allows the intercept with the y-axis to be set to zero when carrying out the
regression. This option should not be selected unless it has been proved that the intercept is
consistently not significantly different from zero (see Section 2.5.2). Finally, ensure that the x
and y data have been correctly assigned (see Section 2.4).
Obtaining a plot of the residuals is strongly recommended as it can highlight problems with
the calibration data that may not be immediately obvious from a simple scatter plot of the
data. The construction of a residual plot is illustrated in Figure 6.
8
7
Instrument response
6
5
4
3
2
1
0
1.5
0.5
Residual
0
0 2 4 6 8
-0.5
-1
0.5 1
0.3 0.5
Residual
Residual
0.1
0
-0.1 0 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8
-0.3 -0.5
9
-0.5 -1
Concentration /mg L-1 Concentration /mg L-1
a) Ideal - random dist ribut ion of residuals about zero b) St andard deviat ion increases w it h concent rat ion
1 1
0.5 0.5
Residual
Residual
0 0
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
-0.5 -0.5
-1 -1
Concentration /mg L-1 Concentration /mg L-1
A typical output from a regression analysis is shown in Figure 8. The output shown is from
Excel, but similar information is obtained from other software. The different parts of the
output are described in more detail in the following sections.
Regression Statistics
Multiple R 0.999955883
R Square 0.999911768
Adjusted R 0.999889709
Square
Standard Error 0.005164622
Observations 6
ANOVA
df SS MS F
Significanc
eF
Regression 1 1.2091 1.2091 45330.79 2.93x10-9
Residual 4 0.00010669 2.67x10-5
Total 5 1.2092
• correlation and linearity are only loosely related. The coefficient r is a measure of
correlation not a measure of linearity;
• it is relatively easy to generate data with apparently good correlation. However, a plot of
the data may well reveal that the data would be unsatisfactory for the purposes of
calibration (see Figure 9);
• for predictions made from the calibration curve to have small uncertainties, r needs to be
very close to 1 (see Section 2.7);
• A low r value does not necessarily mean that there is no correlation. There could be a
relationship between the y and x values, but not a linear one (see Figure 9).
For these reasons, it is essential to plot calibration data, and not just rely on the statistics,
when assessing the fitness-for-purpose of a calibration curve. Figure 9 shows some
examples of how the correlation coefficient can be misleading.
800 400
r = 0.985 r=0
600 300
400 200
200 100
0 0
0 1 2 3 4 5 0 1 2 3 4 5
60 200
r = 0.989 r = 0.957
40 150
20 100
10 50
0 0
0 1 2 3 4 5 0 1 2 3 4 5
80
r = 0.998
60
40
e) Poor experiment al design
20
0
0 1 2 3 4 5
shows any significant correlation. Although there are 12 data points, the distribution of the
points indicates that this is in effect a two-point calibration (which will always give r = 1).
The question which analysts often ask is, “How close to 1 does the correlation coefficient
have to be for a ‘good’ calibration curve?”. What the analyst is really after is a calibration line
that will result in a satisfactory level of uncertainty in the values predicted from the fitted line.
The particular value of r that shows a statistically significant correlation between y and x
depends on the number of data points used to calculate it. Figure 10 shows the value of r
that would indicate statistically significant correlation for different numbers of data points.
Absolute values of r within the shaded area indicate a statistically significant correlation, at
the 95% confidence level. Remember that statistical significance only indicates some
evidence for correlation. It does not necessarily mean that the data would be appropriate for
calibration. For example, with 10 data points a value of r = 0.6 would be statistically
significant. However, it is highly unlikely that a calibration curve with a correlation coefficient
of 0.6 would be of any use, as the uncertainties associated with predicted values obtained
from such a line would be prohibitively large (see section 2.7).
1.0
0.8
0.6
r
0.4
0.2
0
10 20 30 40 50 60 70
Figure 10: Statistically significant values of r (shaded area) at the 95% confidence
level
The parameters related to r are r2 and adjusted r2. r2 is often used to describe the fraction of
the total variance in the data which is contributed by the line that has been fitted. Ideally, if
there is a good linear relation, the majority of variability can be accounted for by the fitted
line. r2 should therefore be close to 1.
The adjusted r2 value is interpreted in the same way as r2 but is always lower. It is useful for
assessing the effect of adding additional terms to the equation of the fitted line (e.g., if a
quadratic fit is used instead of a linear fit). The r2 value always increases on the addition of
an extra term to the equation, but this does not mean that the extended equation is
necessarily a better fit of the data. The adjusted r2 value is more useful in such cases as it
takes account of the reduction in the degrees of freedom which occurs each time an
additional term is added to the equation of the line (see section below on the residual
standard deviation), and therefore does not automatically increase on addition of extra terms.
This guards against ‘overfitting’, which occurs when the equation fitted has more terms than
can be supported by the amount of data available (i.e., there are insufficient degrees of
freedom).
ANOVA table
Software such as Excel produces an analysis of variance (ANOVA) table for the regression.
The sum of squares terms (SS) represent different sources of variability in the calibration
data. Figure 11 illustrates the origin of these terms. The regression term represents the
variability in the data that can be accounted for by the fitted regression line. Ideally this
should be large; if there is a good linear relationship, the fitted line will describe the majority
of the variability in response with concentration. The residual term is the sum of the squared
residuals (see section 0). This value should be small compared to the regression sum of
squares terms because if the regression line fits the data well, the residuals will be small.
df SS MS F Significance F
Total 6 16.60
8
7 1
6 0.5
Residual
Response
5
4 0
3 -0.5
2
1 -1
0 0 1 2 3 4 5 6 7 8
0 2 4 6 8 Concentration /mg L-1
Concentration /mg L-1
use the significance F value should be extremely small (see Figure 8). This value is also
known as the p-value.
Regression coefficients
The final section of the regression output shown in Figure 8 gives information about the
regression coefficients m (the gradient of the line) and c (the intercept of the line with the y-
axis). The first column of numbers gives the values of the coefficients. In Excel, the gradient
is described as “X Variable 1”. The next column gives the standard errors (also know as the
standard deviations) for each coefficient. These values give an indication of the ranges
within which the values for the gradient and intercept could lie. Related to these values are
the lower and upper confidence limits for the gradient and intercept (final two columns of the
table). These represent the extremes of the values that the gradient and intercept could
take, at the chosen level of confidence (usually 95%). The equations for calculating these
values are given in Appendix 3.
The t-stat and p-value relate to the significance of the coefficients, i.e. whether or not they
are statistically significantly different from zero. In a calibration experiment we would expect
the gradient of the line to be very significantly different from zero. The t-value should
therefore be a large number (for a calibration with 7 data points the t-value should be much
greater than 2.6, the 2-tailed Student t value for 5 degrees of freedom at the 95% confidence
level) and the p-value should be small (much less than 0.05 if the regression analysis has
been carried out at the 95% confidence level). Typical values are shown in Figure 8.
Ideally, we would like the calibration line to pass through the origin. If this is the case then
the intercept should not be significantly different from zero. In the regression output we
would expect to see a small value for t (less than 2.6 for a calibration with 7 data points) and
a p-value greater than 0.05 (for regression at the 95% confidence level). Whether the
calibration line can reasonably be assumed to pass through zero can also be judged by
inspecting the confidence interval for the intercept. If this spans zero, then the intercept is
not statistically different from zero, as in the example shown in Figure 8.
2.6 Using the calibration function to estimate values for test samples
If, after plotting the data and examining the regression statistics, the calibration data are
judged to be satisfactory the calibration equation (i.e., the gradient and the intercept) can be
used to estimate the concentration of the analyte in test samples. This requires each sample
to be analysed one or more times, under the same conditions that the calibration standards
were measured. It is also useful to obtain an estimate of the uncertainty associated with the
predicted concentration values for test samples. This is described in Section 2.7.
Instrument response
6
0 2 4 6 8
8
7
Instrument response
6
5
y0
4
3
2
prediction interval
1
0 xpred
0 2 4 6 8
( y o − y )2 Eq. 2
s (r ) 1 1
s x0 = + + n
m N n
m 2 ∑ (xi − x )
2
i =1
Where:
A confidence interval is obtained by multiplying s x0 by the 2-tailed Student t value for the
appropriate level of confidence and n-2 degrees of freedom.
Table 1 and Figure 14 show a set of calibration data which will be used to illustrate the
calculation of a prediction interval.
1.6
1.4 y = 0.1054x + 0.0533
1.2 r 2 = 0.9999
Absorbance
1.0
0.8
0.6
0.4
0.2
0.0
0 2 4 6 8 10 12 14
-1
Concentration /mg L
∑ (y − yˆ i )
n 2
x y ∑ ( xi − x )
i =1
i =1
i
8.317x10 −5
s (r ) = = 0.00408
7−2
Applying Eq. 2, the prediction interval for a sample which gives an instrument response of
0.871, is:
0.00408 1 1 (0.871 − 0.886 )
2
s x0 = + + = 0.0414 mg L-1
0.1054 1 7 0.1054 2 × 65.985
0.871 − 0.0533
x pred = = 7.76 mg L-1
0.1054
The uncertainty in predicted values can be reduced by increasing the number of replicate
measurements (N) made on the test sample. Table 3 shows how s x0 changes as N is
increased.
N s x0 /mg L-1 Uncertainty
/%relative
1 0.041 0.53
2 0.031 0.40
3 0.027 0.35
4 0.024 0.31
5 0.023 0.30
3 Conclusions
The benchmarking exercise highlighted a number of problems associated with carrying out
instrument calibration. Some common pitfalls encountered in calibration studies include:
• the concentration range covered by the calibration standards does not adequately
cover the range of concentrations encountered for test samples;
• the concentrations of the calibration standards are not evenly spaced across the
calibration range;
• the uncertainty associated with the concentrations of the calibration standards is too
large (e.g., inappropriate glassware is used to prepare the standards, the material
used to prepare the standards is not of an appropriate purity);
• the wrong regression is carried out (i.e., regression of x on y rather than y on x);
• the calibration line is fitted through zero when the intercept is, in fact, significantly
different from zero;
• instrument software is used to carry out the regression and automatically calculate
the concentration of test samples but a plot of the calibration data is not obtained;
• the residual standard deviation is used as an estimate of the uncertainty in predicted
concentration values, rather than carrying out the full standard error of prediction
calculation;
• the performance of the instrument used to make the measurements is not within
specification.
Following the steps listed below should avoid these problems:
• plan the calibration study so that the concentration range of interest is covered and
the concentrations of the calibration standards are evenly distributed across the
range;
• include a standard with zero analyte concentration (i.e., a blank);
• ensure that appropriate materials and apparatus are used to prepare the calibration
standards;
• ensure that the instrument used to make the measurements is fit for purpose (i.e.,
carry out equipment qualification);
• plot and examine the results;
Appendix 3:
∑ {(x
i =1
i − x )( y i − y )}
m= n
∑ (x
i =1
i − x)
2
∑ {(x
i =1
i − x )( y i − y )}
r= 1
n 2
n
2
2
∑ ( x i − x ) ∑ ( y i − y )
i =1 i =1
Residual standard deviation, s(r) n
∑ (y i − yˆ i )
2
s (r ) = i =1
n−2
Standard deviation (error) of the gradient, sm s (r )
sm = 1
n 2
2
∑ (x i − x )
i =1
Standard deviation (error) of the intercept, sc 1
n
2
∑ x 2
i
s c = s (r ) n i =1
n ∑ ( x − x )2
i =1 i
i =1
APPENDIX 6
Introduction
This document contains a brief description of a method that may be used for the calibration
of the results of an alternative method against a reference method. The description is of a
general nature; it may be used e.g. for the calibration of mineral oil measurements by infrared
spectrophotometry (the alternative method) against mineral oil measurements by gas
chromatography (the reference method).
Principle
For the initial calibration a series of 12 duplicate samples is analysed with both methods. This
series of samples is obtained by sampling of regular intervals during a period of n
days/weeks, depending on the expected variability of the mineral oil content.
Assuming that the content of mineral oil in the samples lies within the linear range of both
methods, a regression line is calculated :
Y = ax + b
where Y is the alternative measurement and x is the reference measurement.
In addition the curves for the upper and lower 95%-confidence limit (= UCL and LCL) are
calculated, using common statistics.
From the obtained graph a limit value of 30 mg/l may result in a reading of e.g. 40 mg/l for
the alternative method with e.g. UCL = 50 mg/l and LCL = 30 mg/l.
Alternative measurements < 30 mg/l may be considered as below the limit value; alternative
measurements > 50 mg/l may be considered as above the limit value.
Alternative measurements between 30 and 50 mg/l need special attention because it is not
clear whether or not the limit value has been exceeded. It is up to parties involved to decide
how the risk is shared.
After this initial “static” calibration, a limited number of e.g. 4 duplicate samples per month is
taken and analysed with both methods. If these 4 calibration points are within the UCL and
LCL, they are added to the calibration set and the 4 oldest calibration points are discarded. A
new calibration line is now calculated together with new curves for the UCL and LCL.
However if these 4 new calibration points are outside the UCL – LCL range, 8 duplicate
samples are analysed during the next month and a totally new calibration line is calculated
using the 4 + 8 points.
Limitations
This method works well if the content of mineral oil shows sufficient fluctuation during the
time; i.e. the calibration points should have a more or less even distribution over the x-axis. If
this is not the case the UCL and LCL curves will diverge to such an extent that the “grey-
area” i.e. the area between UCL and LCL becomes too large to be useful. In such a case
spiking of some the duplicate samples may be used to overcome this. Especially spiking to a
content near the limit value shall be considered.
Discussion
The number of samples per month that shall be analysed in duplicate should have some
relation to the number of alternative measurements in the same period. The example of 4
duplicate samples per month was proposed in relation to a situation where 16 samples per
month are required.
There are statistical software packages that will do all the necessary calculations in a user-
friendly way.
If the selected alternative method is based on different principles compared to the reference
method, the grey area may be relatively large. This is no problem as long as the content of
mineral oil is well below the limit value.
Agreements are needed about how to deal with results lying in the grey area. Shared risk
doesn’t necessarily mean a 50 – 50 sharing.
The calculation of the total amount of oil discharged may be calculated on the basis of the
results of the analysis of samples used for the calibration, if these samples provide
assurance that an agreed maximum statistical variance is met. The statistical variance might
differ between different types/categories of platforms.
APPENDIX 7
Summary of method
The sample is acidified to pH <2 and extracted with a single volume of solvent (n-pentane or
n-hexane). The UV-fluorescence of the sample extract is measured in an instrument
calibrated with known concentrations of the oil in solvent.
Sampling
Ideally, the sample point should be located on a vertical section of pipe fitted with a centre-
line pitot tube at, or just downstream of, a turbulent region. The sample line should be
flushed for at least one minute. The sample volume should be minimum of 500 ml to ensure
that the sample is representative.
The sample is collected in a glass bottle with a pressure-sealing cap with inert liner. The
sample bottle should be clean and should not be rinsed with sample prior to sampling, or
overflowed during the sampling. The sample is acidified with 5 ml 1:1 hydrochloric acid: this
acid should preferably be added to the bottle prior to sampling in order to prevent formation
of iron oxyhydroxides and/or sulphides. Cool the sample to less than 20 ºC prior to further
analysis.
Apparatus
Note: To ensure all glassware is clean and free from oil, rinse with clean solvent prior to use.
Reagents
• Hydrochloric acid, 1:1. Mix equal amounts of conc. HCl and deionised/distilled water.
CAUTION. Slowly add the acid to the water in a conical flask whilst gently swirling the
flask.
• N-pentane (or n-hexane), Merck HiPerSolv or UniSolv or equivalent purity grade. (Each
batch of solvent should be checked for purity prior to use by carrying out a blank
extraction on deionised/distilled water).
• Reference oil: sample of crude from the same production stream as the oily water
sample.
Stock reference standard: Accurately add about 150 mg (or 180 µl) of oil from a 250 µl
glass syringe to a 50 ml volumetric flask containing a small volume of n-pentane. If a
balance giving 4-decimal place stable reading is available weigh the syringe with the oil and
again after the oil has been added to the flask: this will give an accurate weight of oil added
to the solvent. If a balance is not available then the oil is added by volume and the
concentration can be given as ppm (v/v) or converted to mg/l if the density of the crude is
known.
The instructions in this procedure relate to use of n-pentane as the extraction solvent and the
Arjay FluoroCheck 2000 Style A instrument. The solvent n-hexane can be used instead of n-
pentane. If other UV fluorescence photometers are used then reference should be made to
the manufacturer’s instructions and this procedure modified accordingly.
• Switch on instrument and allow the lamp to warm up for at least 30 minutes. When the
instrument is in regular use select from the Main Menu, 2> Setup then 4> More, 5>
Autoshut and select 1> off. This disables the lamp automatic shut-off. If the instrument is
to be idle for some time, eg overnight, then select 5> Autoshut, 2> on so that the lamp will
shut off after one hour of no keypad activity. This will help to extend the life of the UV
lamp but still retains any calibration set points.
• From the Main Menu, 2> Setup select 1> Prompt, 1> off and 2> Units, 2> none.
• Zero the instrument. Fill a clean cuvette about ¾ full with neat n-pentane (or n-hexane).
Slip the black cuvette holder onto the cuvette and wipe the glass with a clean tissue.
Place the cuvette holder (dot facing toward front) into the well of the FluoroCheck, close
the lid and press the ZERO key.
• Add the lowest concentration calibration standard. Fill a clean cuvette with the prepared
dilution sample and record the reading displayed. Repeat the measurement by pressing
ENTER and continue until a stable reading is obtained for successive measurements.
• Continue the above procedure to record fluorescence readings for all prepared dilution
standards.
• Place the Calibration Standard Cuvette (dot facing toward front) into the instrument and
record the stable fluorescence reading.
the straight-line equation. Note the limit of fluorescence covered by the straight line: any
samples reading above this limit should be diluted and the measurement repeated.
• Use the straight-line equation, y=mx (fluorescence=gradient x concentration), to calculate
the equivalent concentration for the Calibration Standard Cuvette and proceed to
calibrate the Arjay instrument using this concentration value, as follows below.
Calibration of Arjay FluoroCheck 2000 using derived concentration value for Calibration
Standard Cuvette
• If the sample was not acidified at time of collection, add 5 ml 1:1 hydrochloric acid per
500ml water sample to the sample bottle.
• After mixing the sample, check the pH by touching pH-sensitive paper to the inside of the
cap. Add more acid if necessary to achieve pH of 2 or less.
• Transfer the sample to a 500ml graduated separating funnel. Note the sample volume to
the nearest 5ml.
• Rinse the sample bottle and cap with 25ml of solvent and transfer to the separating
funnel. Repeat with a further 25 ml. A sample:solvent ratio of 10:1 should be maintained
for different volumes of sample collected,i.e. if 1000ml sample is collected then use a
total of 100ml solvent.
• Shake the separating funnel for at least five minutes. Allow the layers to separate for a
few minutes. The pentane or hexane will now form the upper layer. Drain off the water
then collect the solvent in a volumetric flask, passing the solvent through a funnel
containing a phase separator filter paper. Rinse the inside of the separating funnel with
small portions of solvent and collect the rinsings in the volumetric flask. Adjust the extract
volume up to 50ml, stopper the flask and mix the contents well. Proceed to UV
fluorescence analysis of the solvent extract without delay.
ENSURE THAT THE UV LAMP HAS BEEN ON FOR AT LEAST 30 MINUTES PRIOR TO
THIS MEASUREMENT.
• Use a fresh clean pipette and remove some of the solvent from the volumetric flask. With
this sample fill a clean cuvette to about ¾ full.
• Slip the black cuvette holder onto the cuvette and wipe the glass with a lint free tissue.
• Place the holder (dot facing front) into the well of the instrument and close the lid. Note
the reading. Repeat the measurement by pressing ENTER and continue until a stable
reading is obtained for successive measurements. Record the final reading.
• If this reading is higher than the accepted straight line calibration range (determined
above) then a dilution is required. Record the reading for the diluted sample
Oil content of water sample = Arjay reading of solvent extract x dilution x (solvent
volume/sample volume)
• Report results in ppm (v/v) or mg/l or ppm (w/w), as required, if the density of the oil and
produced water is known.
APPENDIX 8
1 Analysis Methodology
1.1 Sampling
The objective of the work was to compare the results from the Arjay fluorimeter with the DTI
IR method but also the ISO9377-2 GC method. Pentane and hexane solvents were also
assessed for both the Arjay and GC methods.
In order to compare results, samples were taken on each platform either in triplicate or
quadruplicate at each sampling event. Two of the samples were extracted using pentane
and hexane, respectively. Sub samples of these two extracts were then analysed using the
Arjay fluorimeter, whilst two additional sub samples were retained for GC analysis onshore.
The third sample was analysed by the normal DTI solvent extraction/IR spectrophotometry
method. Typically 12-16 sets of samples were analysed during each installation visit.
To minimise the degree of variation between samples at each sampling event, the sample
bottles were part filled, consecutively in 25% increments until full. Since all the samples
analysed were field samples, this method of filling was used to even out any variations
resulting from changes in the quality of the produced water.
The main steps in analysing a sample with the Arjay instrument are as follows:
Calibration of the instrument was carried out using crude dilution in Pentane (or Hexane). A
dilution series of the stock standard in solvent was used to prepare a calibration curve.
Throughout this study n-hexane and n-pentane were used.
The solvent extraction/IR method normally used was the DTI method.
• Extraction of the water sample using n-pentane (or n-hexane) as the extraction
solvent.
• Removal of the polar substances by cleanup on Florisil.
The following table lists the important details concerning the GC technique used at Opus
Plus Ltd.
Calibration standards for the Arjay instrument were prepared in two ways, namely:
Back extracted standard, in which a known weight of oil was added to seawater and then
extracted by pentane or hexane. Serial dilutions of the extract were then prepared to obtain
the calibration graph.
Oil in solvent standard, in which a known weight of oil was added directly to the solvent
(pentane or hexane). Serial dilutions were then prepared as above to allow its calibration
graph to be determined.
Based on the results obtained the oil in solvent standard gave the best results, compared to
the DTI solvent extraction/IR method. After some of the initial installation visits, the oil in
solvent standard was adopted as the calibration standard used for the remainder of the work.
All calibration standard graphs passed through zero and were not forced through zero.
Calibration procedures for the DTI method and ISO9377-2 method, were as specified in the
written methods.
Whilst there was a recommended operating method for use of the Arjay, covering both
calibration and analysis, one of the objectives of the project was to carry out a rigorous
assessment of the method. The main reason for this was to assess how factors such as the
extraction solvent and extraction procedure might impact on the end result of the analysis. At
the outset, the general plan was to follow as closely as possible the current DTI procedure for
solvent extraction/IR analysis, both in terms of calibration & analysis. Some of the main
areas of the method assessed during the offshore work are described below.
• Stock standard for calibration of Miran IR and Arjay UVF prepared by adding oil to
500 ml seawater then 2 x 25 ml solvent extraction, made up to 50 ml.
• Calibration charts made up for Miran/TTCE and Arjay/n-pentane and n-hexane using
at least four dilutions of stock standard plus zero.
• All produced water samples were analysed by double solvent extraction (500ml
sample into 50 ml solvent, made up to volume).
Same procedures used as for Piper B. IR results consistently about 50% lower than UVF
results. The reason initially was not clear, although later laboratory work identified the cause.
Noted at the outset that the two producing fields, Tartan and Highlander are processed
separately all the way through to overboard discharge of two produced water streams.
The Miran in use onboard was calibrated with Tartan Export (mixed crude). For this study,
calibration charts for the Arjay UVF photometer were prepared with Tartan Export crude,
1500
Fluorescence
y = 2.1267x
1000 y = 1.5826x
500
y = 1.1773x
0
0 100 200 300 400 500 600
Concentration (ppm,v/v)
Tartan-only crude and Highlander-only crude in n-pentane. The sensitivity of the UV-
fluorescence photometer to changes in crude oil composition is clearly illustrated in the
Figure 1.
A small number of single extractions - directly in the bottle, as suggested in the Arjay
recommended procedure, were carried out with n-pentane and n-hexane. The extract was
measured directly from the bottle and therefore no attempt was made to make up the solvent
extract volume to an exact 50 ml. The results from these analyses were generally higher
than for the double extractions.
In consultation with Talisman personnel it was decided to try and adopt the recommended
Arjay procedure for UVF analysis with the exception of the sample volume, which was
maintained at 500 ml. The main differences compared with DTI procedure were:
• Calibration charts for Arjay UVF photometer were made up from stock standards by
direct addition of crude to solvent rather than by back extraction of crude from
seawater.
• Solvent extraction of produced water samples was carried out directly in the sample
bottle by addition of 50 ml solvent (n-pentane or n-hexane) and shaking the capped
bottle for two minutes. Two scenarios were then further explored:
• Decant contents of bottle into a separating funnel, run off water then collect solvent in
volumetric flask and make up to volume.
• Aliquot of solvent was sampled directly from bottle for analysis.
• Not surprisingly, scenario 2 gave slightly higher results than scenario 1.
• The IR analysis procedure was maintained in line with DTI method.
On Claymore there were again two production streams, namely Claymore and Scapa. The
MOL export crude was predominantly Claymore (4:1 Claymore:Scapa). MOL calibrations
proved representative for Claymore stream but slightly better correlation with IR results was
obtained for Scapa produced water if Scapa crude was used for calibrating the Arjay.
Persevered with single extractions for UVF analysis as for Claymore. On average, there was
good agreement between IR and UVF results.
Concentrated on trying to develop the procedure already established on earlier trips to other
platforms.
• Single extraction now performed in separating funnel because some of bottles were
not sealing properly.
• Extraction time increased to five minutes to reflect tests carried out on Gyda that
showed five minutes gave better extraction than two minutes.
The results obtained from the offshore trials are discussed in detail below.
3.1 Phase 1
• Piper B
The results from the visit to the Piper B installation are presented in Figure 2.
200
180
Concentration (ppm vol)
160
140
120
100
80
60
40
20
0
1 2 3 4 5 6 7 8 9 10
Sam ple Num ber
TTCE n-pentane n-hexane
A total of ten sets of triplicate samples were analysed during the visit to Piper B. The Miran
IR results were determined using both the 'in house' platform calibration, and a new
calibration carried out independently by Opus Plus Ltd. Florisil treatment of the IR extracts is
routinely carried out on Piper B. In order to determine the impact on the Arjay extracts,
comparative analysis was carried out for both the pentane and hexane, both with and without
florisil treatment. It should be noted, however, that florisil treatment is not part of the
recommended method for the Arjay fluorimeter. Analysis of the IR extracts before and after
florisil treatment, showed there to be no change in the calculated result. The GC ISO 9377-2
method does employ a florisil treatment stage as part of the standard method, although this
was carried out, in all cases, onshore after return of the extracts.
As the data in the graph show, for nine of the ten sets of samples, there was a reasonably
close correlation between the IR results and the Arjay (for both pentane and hexane
solvents). The data for sample numbers 2, 3, 4, 6, 7 and 9, in particular, give a good
correlation between the IR and the Arjay methods. The reason for the higher IR results for
samples 1 and 5 is unclear, although it may simply be a function of variation in the samples.
The high IR result for sample 10 also cannot be explained, although it appears to be
contamination of the IR sample with n-pentane or n-hexane.
• Clyde
During the visit to Clyde, a total of 15 sets of samples were analysed. Overall, the results
from Clyde were disappointing (see Figure 3), with the Arjay results significantly higher than
the IR method in all cases. In general, the Arjay results for the two solvents were
comparable, although the pentane results were slightly higher than those with hexane in
most cases. Calibrations of both the Infracal and the Arjay instrument were repeated but
gave similar results.
400
Concentration (ppm vol)
350
300
250
200
150
100
50
0
2 3 4 5 6 7 8 9 10 11 12 13 14 15
Sam ple Num ber
In order to try and determine the reason for the difference in the IR and Arjay results, some
additional work was carried out onshore. The Clyde installation exports a blend of oil,
comprising Clyde oil and Orion oil. Since the fluorescing characteristics of the Arjay
instrument are a function of the oil composition, some onshore calibration work was carried
out with the two individual crudes and the combined export crude. Initially, calibration
standards were prepared with the three oil types (Clyde, Orion, MOL export). Laboratory
prepared oily water samples of known concentration (circa 40 mg/l) were then prepared for
the three oil types and extracted in the normal way with pentane. The three extract types
were then measured against the three calibration standards in the Arjay instrument. For
example, an oily water sample made up using Clyde oil was extracted and the extract
compared against the Clyde, Orion and MOL export calibration charts to assess what impact
this had on the end result. This gave nine possible combinations for the sample oil type and
calibration standard comparison. A similar comparison was also carried out for the IR
method.
100
Concentration (ppm vol)
80 Probable
Offshore
60 Condition
IR
UVF
40
20
0
O vs O O vs Ex O vs C Ex vs O Ex vs Ex vs C C vs O C vs Ex C vs C
Ex
Oil type added vs oil type calibration
(C=Clyde, O=Orion, Ex=Export)
Fig. 4 – Clyde Platform oil streams: Oil type added versus Oil type calibration
The results from the laboratory work are shown in Figure 4. The results show major
differences in the measured result for different combinations. A sample made up from MOL
export oil and compared against the Orion crude standard, for example, gave a result of 34
mg/l for the IR and 85 mg/l for the Arjay. In contrast, comparison of the same sample extract
against the Clyde calibration standard gave a result of 68 mg/l for the IR and 37 mg/l for the
Arjay.
On the Clyde installation the bulk of the produced water comes from the Clyde wells. Any
samples taken for analysis from the produced water are therefore likely to contain oil that is
predominantly Clyde in nature. From the onshore laboratory work carried out, the condition
that is most likely to have matched offshore conditions, is that where Clyde samples are
compared to an Export oil calibration. This condition is highlighted in Figure 4 and shows
results from the Arjay which are significantly higher than those from the IR method. This
matches the pattern observed in the offshore visit (see Figure 3). Whilst the onshore work
cannot exactly reproduce the conditions offshore, the results from this work do provide an
explanation for the differences observed. The data also illustrates the importance of using a
representative calibration standard with the Arjay method. Interestingly, because the Clyde
and Orion oils are so different in composition the IR method was also shown to be sensitive
to the choice of calibration standard and sample oil type.
• Tartan
Twelve sets of samples were analysed on Tartan. Again, a similar pattern of analysis was
carried out to the previous two platforms. After the experience with the Clyde results,
however, a number of additional calibrations were carried out on Tartan. The Tartan
installation produces oil from the Tartan reservoir and the Highlander reservoir. Since these
oils have different compositions, the following calibrations were carried out:
The effect of using Tartan-only oil calibration for the Arjay results is shown in Figure 5 for the
Tartan produced water streams (n-pentane extractions).
120
Concentration (ppm vol)
100
80
60
40
20
0
6 7 8 10 11
Sam ple Num ber
Much better agreement with the IR results was obtained by using the oil specific calibration
chart. The same effect was noted with n-hexane extractions.
Whilst on Tartan, some additional work was undertaken to compare extraction of Arjay
samples in a separating flask, compared to a sample bottle. One of the seemingly clear
advantages of the Arjay instrument, is that the solvent can be added to the sample bottle
directly, and the sample extracted. Since both pentane and hexane are less dense than the
water phase, the extract settles out on top and an aliquot can be withdrawn by pipette for the
Arjay analysis. Some comparison was carried out on sample 8, 10 and 11 purely to assess
whether extraction in the bottle gave better results than extraction in a flask. For these
samples, an additional sample was therefore taken, so that four samples were analysed (one
for IR, one for Arjay analysis in bottle, and two for Arjay analysis in a separating flask). The
results are shown in Figure 6. Although only three sets of data are included, the data show
that extraction in the sample bottle gives a higher result. This might have been expected,
given that losses in the extraction process are minimised.
180
160
Concentration (ppm vol)
140
120
100
80
60
40
20
0
8 10 11
Sam ple Num ber
Overall, the visit to Tartan emphasised the importance of using representative crude for the
Arjay calibration. In addition, the work demonstrated that extraction of the Arjay samples in
the sample bottles appeared to give the higher extraction efficiency. Single extraction,
directly in the bottle, formed the basis of all subsequent Arjay analyses.
• Claymore
The results from the visit to Claymore are presented in Figure 7. A total of 12 sets of
samples were analysed during the visit.
70
Concentration (ppm vol)
60
50
40
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12
Sam ple Num ber
As the data in the graph show, a good correlation was obtained between the IR method and
the Arjay method over a range of sample concentrations. In most cases, the IR method gave
very slightly higher results than the Arjay fluorimeter. Comparison of the extraction volumes
used with the Arjay, showed a slight difference when the n-pentane solvent was made up to
50 ml following extraction. As expected, the results where the solvent was made up to the
mark were lower than those where no adjustment was made for volume. In general, the
difference between the results of the IR and the Arjay instruments was of the order of 5 ppm.
On Claymore there were again two production streams, namely Claymore and Scapa. The
MOL export crude was predominantly Claymore (4:1 Claymore:Scapa). MOL calibrations
proved representative for Claymore stream but slightly better correlation with IR results was
obtained for Scapa produced water if Scapa crude was used for calibrating the Arjay. A
summary graph in Figure 8 shows the comparative results for MOL versus Scapa calibrations
for the Scapa produced water streams. As the data show, for the crudes processed through
Claymore, only minor differences were apparent from the different calibrations, suggesting
the oils had broadly similar compositions. The IR calibration charts for MOL and Scapa were
60
50
40
30
20
10
0
4 6 8 10 12
Sam ple Num ber
virtually identical.
• Beatrice
The data from the Beatrice trip are presented in Figure 9. A total of 16 sets of samples were
analysed during the visit.
120
Concentration (ppm vol)
100
80
60
40
20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Sample Number
The data show for most samples a good correlation between the IR method and the Arjay
method. As for Claymore, it was again apparent that when the solvent had been topped up
to the 50 ml mark, following extraction, then a lower result was obtained than with the neat
solvent volume.
• Buchan
The results from the visit to the Buchan installation are presented in Figure 10. A total of 16
sets of samples were analysed during the visit to Buchan.
100
Concentration (ppm vol)
80
60
40
20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Sam ple Num ber
On Buchan an additional instrument, namely the Turner TD500 UV-F, was evaluated in
parallel with the IR method and the Arjay fluorimeter. The instrument uses the same method
of detection as the Arjay fluorimeter and indeed the sample extracts are analysed in exactly
the same type of cuvette as the Arjay instrument. The main differences with the Turner
instrument were that it is much smaller (it can be hand held) and it utilises a single point
calibration only, as opposed to the multipoint calibration which can be used on the Arjay
instrument.
The emphasis of the work carried out on Buchan was to develop the calibration and sample
extraction procedures to be as user friendly as possible. The first modification required to the
single solvent extraction procedure was to transfer the acidified sample from the bottle to a
graduated separating funnel. This was necessitated because it was found that the caps did
not always form a tight seal on the bottle: solvent leaked from the bottle during shaking. A
further change to the extraction procedure was that the separating funnel was shaken for five
minutes instead of two minutes. This change was incorporated as a result of the work
carried out on Gyda. Most of the solvent extracts were made up to 50 ml. The working
range of the UVF photometers was more limited than the Miran so dilutions had to be made
on all extracts for Arjay and Turner analyses.
Generally, the results showed good agreement between the IR method and the Arjay,
particularly for samples 8-16. In terms of the comparison between solvents, the n-pentane
gave the best results, although there was only a minor difference between the two.
Comparison of the performance of the Arjay and the Turner instruments proved a very simple
task, as the same cuvette and extract could be analysed directly in both instruments. A
graph showing the comparative results for the IR method and the Arjay and Turner
instruments (for n-pentane) is presented in Figure 11. As the data show, a good correlation
was obtained both between the Arjay and Turner instruments, and with the IR method. The
results from the Turner instrument were consistently slightly lower than those of the Arjay; the
reason for this was most likely a function of the calibration procedure. Overall, however, the
Turner TD500 proved easy to use and the calibration was very stable from day to day.
90
Concentration (ppm vol)
80
70
60
50
40
30
20
10
0
6 7 8 9 10 11 12 13 14 15 16
Sam ple Num ber
3.2 Phase 2
Following completion of the work in the Phase I installation visits, further visits were made to
Saltire, Clyde (repeat visit) and Northern Producer, to complete the data set.
This work was carried out in the period November to December 2004. Some general
comments on the results from each visit are provided below.
• Saltire
30
Concentration
(ppm vol)
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12
Sam ple No
The data in the above graph show that excellent results were obtained from the visit to
Saltire. Twelve sets of samples were analysed, with the Arjay analysis compared against
both back extracted and oil in solvent standard results gave excellent agreement with the IR
method results.
As expected, the results based on the back extracted standard gave slightly higher results,
but which still correlated well with the IR method.
The data from the GC samples taken on Saltire showed a reasonable correlation between
the IR and Arjay data and the GC C7-C40 result (which is calculated using the modified
method). The C10-C40 results were all significantly lower, which indicates that the modified
GC method provides a better measure of the actual oil in water concentration.
• Clyde
The results from the visit to Clyde are presented in Figure 13.
80
Concentration (ppm vol)
40
0
1 2 3 4 5 6 7 8
Sample No
The repeat visit to Clyde was carried out to validate the findings of the onshore laboratory
work on using different calibration standards with the Arjay (prepared from Clyde, Orion and
export crudes). Twelve sets of samples were analysed offshore, with the Arjay being
compared to the IR method, for both a back extracted and oil in solvent calibration standard.
As the results in the graph below show, a good correlation was obtained between the IR
method and the Arjay. This is particularly true for the oil in solvent results, which gave very
good agreement with the IR method. The results, as a whole, back up the findings from the
laboratory work, which showed that the choice of calibration standard can have a significant
impact on the Arjay results, depending on the characteristics of the oil in the sample relative
to the calibration oil.
The results from the GC checks carried out showed the results obtained from the GC method
were generally lower than those of the IR method, although the correlation appeared better
than the phase 1 results. As with previous GC analyses carried out by Opus, the GC method
exhibits a higher degree of scatter than the IR or Arjay methods. In sampling event 3 for
example, the C7-C40 result was much higher than the C10-C40 result, but was not so in the
other samples.
Northern Producer
The results from the visit to the Galley installation are shown in Figure 14.
100
Concentration (ppm
80
60
vol)
40
20
0
1 2 3 4 5 6 7 8 9 10 11
Sample No
A total of 13 sets of samples were analysed during the visit to Northern Producer. Two wells
were in production at the time of the visit, namely G2 and G3. The combined fluids provide
the export blend from the installation.
As the data in the graph show, in all cases, the Arjay gave substantially higher results than
the IR method. For sample 1, for example, with an IR result of 33 ppm the Arjay gave a
result of 62 ppm. All other samples showed the Arjay result to be at least double that of the
IR result. Despite rechecking the calibration standards whilst offshore, similar results were
obtained. Two samples drawn from wellhead samples of G2 and G3 were also analysed, but
showed similar degree of differences, albeit at much higher oil in water concentrations.
Following completion of the visit and discussions with Talisman, additional onshore
laboratory work was carried out to try and determine reasons for the discrepancies. The
following work was carried out:
• calibrations were carried out on the Arjay and Miran, using samples of G2 and G3 oil
provided by Talisman;
• Flotreat 552 and Dragreducer 2004-031L were added to seawater to determine any
possible influence on the Arjay result;
• samples of produced water from G2 and G3 were analysed by both Arjay and IR for
comparison.
The calibration charts prepared for both the Arjay and Miran showed slight differences in
slope, but were not sufficiently different to explain the offshore results based on the export
mixed crude. The two chemicals were added to seawater, which was then extracted using n-
pentane and measured using the Arjay. The results for both chemicals showed no influence
on the result with the Arjay measuring 0 ppm in both cases.
Analysis of the produced water samples from G2 and G3 were somewhat inconclusive in that
the first two sets gave a higher IR result than the UVF result. One further set gave the IR
lower than UVF.
Sample Arjay
Id Miran IR UVF
n-
TTCE pentane
untreated florisil
trtd. 1 x 50ml
ppm vol ppm vol ppm vol
G2 - 1 97 nd 74
G2 - 2 21 nd 9
G2 - 3 122 104 153
G3 - 1 54 46 44
G3 - 2 54 40 39
Overall, the results from the additional laboratory work do not explain the results which were
obtained offshore. The onshore laboratory work was different in one respect, in that the
samples were not 'live' samples. It is possible, therefore, that other interferences were
present which could not be reproduced in the offshore work.
Due to the poor correlation obtained in the offshore work, no samples were analysed for GC.
3.3 GC results
The GC results for all platforms in Phase 1 are summarised in Figure 15 (for clarity of
illustration, the results for C10-C40 are not included).
140.0
Beatrice Buchan Claymore Clyde Piper Tartan
120.0
100.0
80.0
60.0
40.0
20.0
0.0
3B 4B 9B 3B 8B 11B 6B 8B 5B 7B 2B 4B 7B 5B 6B 8B 4B 8B
In line with previous study work carried out by Opus Plus Ltd [2], the analyses carried out
using the ISO9377-2 method, showed, with the exception of results on Claymore and
Beatrice, the GC results to be approximately half those obtained using the IR method.
Overall the comparison between the GC method and the IR method was poor. In some
cases better results were obtained for calibration against the actual crude standard, although
there was clearly a greater spread in the results relative to the other methods. It should be
noted that the GC method as written does not include the use of crude oil as a calibration
standard.